sctp_pcb.c revision 258454
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 258454 2013-11-21 23:00:09Z 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		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
776		vrf->total_ifa_count--;
777		LIST_REMOVE(sctp_ifap, next_bucket);
778		sctp_remove_ifa_from_ifn(sctp_ifap);
779	}
780#ifdef SCTP_DEBUG
781	else {
782		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
783		    ifn_index);
784		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
785	}
786#endif
787
788out_now:
789	SCTP_IPI_ADDR_WUNLOCK();
790	if (sctp_ifap) {
791		struct sctp_laddr *wi;
792
793		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
794		if (wi == NULL) {
795			/*
796			 * Gak, what can we do? We have lost an address
797			 * change can you say HOSED?
798			 */
799			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
800
801			/* Oops, must decrement the count */
802			sctp_free_ifa(sctp_ifap);
803			return;
804		}
805		SCTP_INCR_LADDR_COUNT();
806		bzero(wi, sizeof(*wi));
807		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
808		wi->ifa = sctp_ifap;
809		wi->action = SCTP_DEL_IP_ADDRESS;
810		SCTP_WQ_ADDR_LOCK();
811		/*
812		 * Should this really be a tailq? As it is we will process
813		 * the newest first :-0
814		 */
815		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
816		SCTP_WQ_ADDR_UNLOCK();
817
818		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
819		    (struct sctp_inpcb *)NULL,
820		    (struct sctp_tcb *)NULL,
821		    (struct sctp_nets *)NULL);
822	}
823	return;
824}
825
826
827static int
828sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
829{
830	int loopback_scope;
831
832#if defined(INET)
833	int ipv4_local_scope, ipv4_addr_legal;
834
835#endif
836#if defined(INET6)
837	int local_scope, site_scope, ipv6_addr_legal;
838
839#endif
840	struct sctp_vrf *vrf;
841	struct sctp_ifn *sctp_ifn;
842	struct sctp_ifa *sctp_ifa;
843
844	loopback_scope = stcb->asoc.scope.loopback_scope;
845#if defined(INET)
846	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
847	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
848#endif
849#if defined(INET6)
850	local_scope = stcb->asoc.scope.local_scope;
851	site_scope = stcb->asoc.scope.site_scope;
852	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
853#endif
854
855	SCTP_IPI_ADDR_RLOCK();
856	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
857	if (vrf == NULL) {
858		/* no vrf, no addresses */
859		SCTP_IPI_ADDR_RUNLOCK();
860		return (0);
861	}
862	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
863		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
864			if ((loopback_scope == 0) &&
865			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
866				continue;
867			}
868			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
869				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
870				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
871					/*
872					 * We allow pending addresses, where
873					 * we have sent an asconf-add to be
874					 * considered valid.
875					 */
876					continue;
877				}
878				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
879					continue;
880				}
881				switch (sctp_ifa->address.sa.sa_family) {
882#ifdef INET
883				case AF_INET:
884					if (ipv4_addr_legal) {
885						struct sockaddr_in *sin,
886						           *rsin;
887
888						sin = &sctp_ifa->address.sin;
889						rsin = (struct sockaddr_in *)to;
890						if ((ipv4_local_scope == 0) &&
891						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
892							continue;
893						}
894						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
895							SCTP_IPI_ADDR_RUNLOCK();
896							return (1);
897						}
898					}
899					break;
900#endif
901#ifdef INET6
902				case AF_INET6:
903					if (ipv6_addr_legal) {
904						struct sockaddr_in6 *sin6,
905						            *rsin6;
906
907						sin6 = &sctp_ifa->address.sin6;
908						rsin6 = (struct sockaddr_in6 *)to;
909						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
910							if (local_scope == 0)
911								continue;
912							if (sin6->sin6_scope_id == 0) {
913								if (sa6_recoverscope(sin6) != 0)
914									continue;
915							}
916						}
917						if ((site_scope == 0) &&
918						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
919							continue;
920						}
921						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
922							SCTP_IPI_ADDR_RUNLOCK();
923							return (1);
924						}
925					}
926					break;
927#endif
928				default:
929					/* TSNH */
930					break;
931				}
932			}
933		}
934	} else {
935		struct sctp_laddr *laddr;
936
937		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
938			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
939				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
940				continue;
941			}
942			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
943			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
944				/*
945				 * We allow pending addresses, where we have
946				 * sent an asconf-add to be considered
947				 * valid.
948				 */
949				continue;
950			}
951			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
952				continue;
953			}
954			switch (to->sa_family) {
955#ifdef INET
956			case AF_INET:
957				{
958					struct sockaddr_in *sin, *rsin;
959
960					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
961					rsin = (struct sockaddr_in *)to;
962					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
963						SCTP_IPI_ADDR_RUNLOCK();
964						return (1);
965					}
966					break;
967				}
968#endif
969#ifdef INET6
970			case AF_INET6:
971				{
972					struct sockaddr_in6 *sin6, *rsin6;
973
974					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
975					rsin6 = (struct sockaddr_in6 *)to;
976					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
977						SCTP_IPI_ADDR_RUNLOCK();
978						return (1);
979					}
980					break;
981				}
982
983#endif
984			default:
985				/* TSNH */
986				break;
987			}
988
989		}
990	}
991	SCTP_IPI_ADDR_RUNLOCK();
992	return (0);
993}
994
995
996static struct sctp_tcb *
997sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
998    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
999{
1000	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1001	/*
1002	 * If we support the TCP model, then we must now dig through to see
1003	 * if we can find our endpoint in the list of tcp ep's.
1004	 */
1005	uint16_t lport, rport;
1006	struct sctppcbhead *ephead;
1007	struct sctp_inpcb *inp;
1008	struct sctp_laddr *laddr;
1009	struct sctp_tcb *stcb;
1010	struct sctp_nets *net;
1011
1012	if ((to == NULL) || (from == NULL)) {
1013		return (NULL);
1014	}
1015	switch (to->sa_family) {
1016#ifdef INET
1017	case AF_INET:
1018		if (from->sa_family == AF_INET) {
1019			lport = ((struct sockaddr_in *)to)->sin_port;
1020			rport = ((struct sockaddr_in *)from)->sin_port;
1021		} else {
1022			return (NULL);
1023		}
1024		break;
1025#endif
1026#ifdef INET6
1027	case AF_INET6:
1028		if (from->sa_family == AF_INET6) {
1029			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1030			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1031		} else {
1032			return (NULL);
1033		}
1034		break;
1035#endif
1036	default:
1037		return (NULL);
1038	}
1039	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1040	/*
1041	 * Ok now for each of the guys in this bucket we must look and see:
1042	 * - Does the remote port match. - Does there single association's
1043	 * addresses match this address (to). If so we update p_ep to point
1044	 * to this ep and return the tcb from it.
1045	 */
1046	LIST_FOREACH(inp, ephead, sctp_hash) {
1047		SCTP_INP_RLOCK(inp);
1048		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1049			SCTP_INP_RUNLOCK(inp);
1050			continue;
1051		}
1052		if (lport != inp->sctp_lport) {
1053			SCTP_INP_RUNLOCK(inp);
1054			continue;
1055		}
1056		if (inp->def_vrf_id != vrf_id) {
1057			SCTP_INP_RUNLOCK(inp);
1058			continue;
1059		}
1060		/* check to see if the ep has one of the addresses */
1061		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1062			/* We are NOT bound all, so look further */
1063			int match = 0;
1064
1065			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1066
1067				if (laddr->ifa == NULL) {
1068					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1069					continue;
1070				}
1071				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1072					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1073					continue;
1074				}
1075				if (laddr->ifa->address.sa.sa_family ==
1076				    to->sa_family) {
1077					/* see if it matches */
1078#ifdef INET
1079					if (from->sa_family == AF_INET) {
1080						struct sockaddr_in *intf_addr,
1081						           *sin;
1082
1083						intf_addr = &laddr->ifa->address.sin;
1084						sin = (struct sockaddr_in *)to;
1085						if (sin->sin_addr.s_addr ==
1086						    intf_addr->sin_addr.s_addr) {
1087							match = 1;
1088							break;
1089						}
1090					}
1091#endif
1092#ifdef INET6
1093					if (from->sa_family == AF_INET6) {
1094						struct sockaddr_in6 *intf_addr6;
1095						struct sockaddr_in6 *sin6;
1096
1097						sin6 = (struct sockaddr_in6 *)
1098						    to;
1099						intf_addr6 = &laddr->ifa->address.sin6;
1100
1101						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1102						    intf_addr6)) {
1103							match = 1;
1104							break;
1105						}
1106					}
1107#endif
1108				}
1109			}
1110			if (match == 0) {
1111				/* This endpoint does not have this address */
1112				SCTP_INP_RUNLOCK(inp);
1113				continue;
1114			}
1115		}
1116		/*
1117		 * Ok if we hit here the ep has the address, does it hold
1118		 * the tcb?
1119		 */
1120		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1121		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1122		if (stcb == NULL) {
1123			SCTP_INP_RUNLOCK(inp);
1124			continue;
1125		}
1126		SCTP_TCB_LOCK(stcb);
1127		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1128			SCTP_TCB_UNLOCK(stcb);
1129			SCTP_INP_RUNLOCK(inp);
1130			continue;
1131		}
1132		if (stcb->rport != rport) {
1133			/* remote port does not match. */
1134			SCTP_TCB_UNLOCK(stcb);
1135			SCTP_INP_RUNLOCK(inp);
1136			continue;
1137		}
1138		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1139			SCTP_TCB_UNLOCK(stcb);
1140			SCTP_INP_RUNLOCK(inp);
1141			continue;
1142		}
1143		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1144			SCTP_TCB_UNLOCK(stcb);
1145			SCTP_INP_RUNLOCK(inp);
1146			continue;
1147		}
1148		/* Does this TCB have a matching address? */
1149		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1150
1151			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1152				/* not the same family, can't be a match */
1153				continue;
1154			}
1155			switch (from->sa_family) {
1156#ifdef INET
1157			case AF_INET:
1158				{
1159					struct sockaddr_in *sin, *rsin;
1160
1161					sin = (struct sockaddr_in *)&net->ro._l_addr;
1162					rsin = (struct sockaddr_in *)from;
1163					if (sin->sin_addr.s_addr ==
1164					    rsin->sin_addr.s_addr) {
1165						/* found it */
1166						if (netp != NULL) {
1167							*netp = net;
1168						}
1169						/*
1170						 * Update the endpoint
1171						 * pointer
1172						 */
1173						*inp_p = inp;
1174						SCTP_INP_RUNLOCK(inp);
1175						return (stcb);
1176					}
1177					break;
1178				}
1179#endif
1180#ifdef INET6
1181			case AF_INET6:
1182				{
1183					struct sockaddr_in6 *sin6, *rsin6;
1184
1185					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1186					rsin6 = (struct sockaddr_in6 *)from;
1187					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1188					    rsin6)) {
1189						/* found it */
1190						if (netp != NULL) {
1191							*netp = net;
1192						}
1193						/*
1194						 * Update the endpoint
1195						 * pointer
1196						 */
1197						*inp_p = inp;
1198						SCTP_INP_RUNLOCK(inp);
1199						return (stcb);
1200					}
1201					break;
1202				}
1203#endif
1204			default:
1205				/* TSNH */
1206				break;
1207			}
1208		}
1209		SCTP_TCB_UNLOCK(stcb);
1210		SCTP_INP_RUNLOCK(inp);
1211	}
1212	return (NULL);
1213}
1214
1215
1216/*
1217 * rules for use
1218 *
1219 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1220 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1221 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1222 * NULL.
1223 */
1224
1225struct sctp_tcb *
1226sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1227    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1228{
1229	struct sctpasochead *head;
1230	struct sctp_inpcb *inp;
1231	struct sctp_tcb *stcb = NULL;
1232	struct sctp_nets *net;
1233	uint16_t rport;
1234
1235	inp = *inp_p;
1236	switch (remote->sa_family) {
1237#ifdef INET
1238	case AF_INET:
1239		rport = (((struct sockaddr_in *)remote)->sin_port);
1240		break;
1241#endif
1242#ifdef INET6
1243	case AF_INET6:
1244		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1245		break;
1246#endif
1247	default:
1248		return (NULL);
1249	}
1250	if (locked_tcb) {
1251		/*
1252		 * UN-lock so we can do proper locking here this occurs when
1253		 * called from load_addresses_from_init.
1254		 */
1255		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1256		SCTP_TCB_UNLOCK(locked_tcb);
1257	}
1258	SCTP_INP_INFO_RLOCK();
1259	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1260	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1261		/*-
1262		 * Now either this guy is our listener or it's the
1263		 * connector. If it is the one that issued the connect, then
1264		 * it's only chance is to be the first TCB in the list. If
1265		 * it is the acceptor, then do the special_lookup to hash
1266		 * and find the real inp.
1267		 */
1268		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1269			/* to is peer addr, from is my addr */
1270			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1271			    netp, inp->def_vrf_id);
1272			if ((stcb != NULL) && (locked_tcb == NULL)) {
1273				/* we have a locked tcb, lower refcount */
1274				SCTP_INP_DECR_REF(inp);
1275			}
1276			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1277				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1278				SCTP_TCB_LOCK(locked_tcb);
1279				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1280				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1281			}
1282			SCTP_INP_INFO_RUNLOCK();
1283			return (stcb);
1284		} else {
1285			SCTP_INP_WLOCK(inp);
1286			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1287				goto null_return;
1288			}
1289			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1290			if (stcb == NULL) {
1291				goto null_return;
1292			}
1293			SCTP_TCB_LOCK(stcb);
1294
1295			if (stcb->rport != rport) {
1296				/* remote port does not match. */
1297				SCTP_TCB_UNLOCK(stcb);
1298				goto null_return;
1299			}
1300			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1301				SCTP_TCB_UNLOCK(stcb);
1302				goto null_return;
1303			}
1304			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1305				SCTP_TCB_UNLOCK(stcb);
1306				goto null_return;
1307			}
1308			/* now look at the list of remote addresses */
1309			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1310#ifdef INVARIANTS
1311				if (net == (TAILQ_NEXT(net, sctp_next))) {
1312					panic("Corrupt net list");
1313				}
1314#endif
1315				if (net->ro._l_addr.sa.sa_family !=
1316				    remote->sa_family) {
1317					/* not the same family */
1318					continue;
1319				}
1320				switch (remote->sa_family) {
1321#ifdef INET
1322				case AF_INET:
1323					{
1324						struct sockaddr_in *sin,
1325						           *rsin;
1326
1327						sin = (struct sockaddr_in *)
1328						    &net->ro._l_addr;
1329						rsin = (struct sockaddr_in *)remote;
1330						if (sin->sin_addr.s_addr ==
1331						    rsin->sin_addr.s_addr) {
1332							/* found it */
1333							if (netp != NULL) {
1334								*netp = net;
1335							}
1336							if (locked_tcb == NULL) {
1337								SCTP_INP_DECR_REF(inp);
1338							} else if (locked_tcb != stcb) {
1339								SCTP_TCB_LOCK(locked_tcb);
1340							}
1341							if (locked_tcb) {
1342								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1343							}
1344							SCTP_INP_WUNLOCK(inp);
1345							SCTP_INP_INFO_RUNLOCK();
1346							return (stcb);
1347						}
1348						break;
1349					}
1350#endif
1351#ifdef INET6
1352				case AF_INET6:
1353					{
1354						struct sockaddr_in6 *sin6,
1355						            *rsin6;
1356
1357						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1358						rsin6 = (struct sockaddr_in6 *)remote;
1359						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1360						    rsin6)) {
1361							/* found it */
1362							if (netp != NULL) {
1363								*netp = net;
1364							}
1365							if (locked_tcb == NULL) {
1366								SCTP_INP_DECR_REF(inp);
1367							} else if (locked_tcb != stcb) {
1368								SCTP_TCB_LOCK(locked_tcb);
1369							}
1370							if (locked_tcb) {
1371								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1372							}
1373							SCTP_INP_WUNLOCK(inp);
1374							SCTP_INP_INFO_RUNLOCK();
1375							return (stcb);
1376						}
1377						break;
1378					}
1379#endif
1380				default:
1381					/* TSNH */
1382					break;
1383				}
1384			}
1385			SCTP_TCB_UNLOCK(stcb);
1386		}
1387	} else {
1388		SCTP_INP_WLOCK(inp);
1389		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1390			goto null_return;
1391		}
1392		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1393		    inp->sctp_hashmark)];
1394		if (head == NULL) {
1395			goto null_return;
1396		}
1397		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1398			if (stcb->rport != rport) {
1399				/* remote port does not match */
1400				continue;
1401			}
1402			SCTP_TCB_LOCK(stcb);
1403			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1404				SCTP_TCB_UNLOCK(stcb);
1405				continue;
1406			}
1407			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1408				SCTP_TCB_UNLOCK(stcb);
1409				continue;
1410			}
1411			/* now look at the list of remote addresses */
1412			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1413#ifdef INVARIANTS
1414				if (net == (TAILQ_NEXT(net, sctp_next))) {
1415					panic("Corrupt net list");
1416				}
1417#endif
1418				if (net->ro._l_addr.sa.sa_family !=
1419				    remote->sa_family) {
1420					/* not the same family */
1421					continue;
1422				}
1423				switch (remote->sa_family) {
1424#ifdef INET
1425				case AF_INET:
1426					{
1427						struct sockaddr_in *sin,
1428						           *rsin;
1429
1430						sin = (struct sockaddr_in *)
1431						    &net->ro._l_addr;
1432						rsin = (struct sockaddr_in *)remote;
1433						if (sin->sin_addr.s_addr ==
1434						    rsin->sin_addr.s_addr) {
1435							/* found it */
1436							if (netp != NULL) {
1437								*netp = net;
1438							}
1439							if (locked_tcb == NULL) {
1440								SCTP_INP_DECR_REF(inp);
1441							} else if (locked_tcb != stcb) {
1442								SCTP_TCB_LOCK(locked_tcb);
1443							}
1444							if (locked_tcb) {
1445								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1446							}
1447							SCTP_INP_WUNLOCK(inp);
1448							SCTP_INP_INFO_RUNLOCK();
1449							return (stcb);
1450						}
1451						break;
1452					}
1453#endif
1454#ifdef INET6
1455				case AF_INET6:
1456					{
1457						struct sockaddr_in6 *sin6,
1458						            *rsin6;
1459
1460						sin6 = (struct sockaddr_in6 *)
1461						    &net->ro._l_addr;
1462						rsin6 = (struct sockaddr_in6 *)remote;
1463						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1464						    rsin6)) {
1465							/* found it */
1466							if (netp != NULL) {
1467								*netp = net;
1468							}
1469							if (locked_tcb == NULL) {
1470								SCTP_INP_DECR_REF(inp);
1471							} else if (locked_tcb != stcb) {
1472								SCTP_TCB_LOCK(locked_tcb);
1473							}
1474							if (locked_tcb) {
1475								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1476							}
1477							SCTP_INP_WUNLOCK(inp);
1478							SCTP_INP_INFO_RUNLOCK();
1479							return (stcb);
1480						}
1481						break;
1482					}
1483#endif
1484				default:
1485					/* TSNH */
1486					break;
1487				}
1488			}
1489			SCTP_TCB_UNLOCK(stcb);
1490		}
1491	}
1492null_return:
1493	/* clean up for returning null */
1494	if (locked_tcb) {
1495		SCTP_TCB_LOCK(locked_tcb);
1496		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1497	}
1498	SCTP_INP_WUNLOCK(inp);
1499	SCTP_INP_INFO_RUNLOCK();
1500	/* not found */
1501	return (NULL);
1502}
1503
1504
1505/*
1506 * Find an association for a specific endpoint using the association id given
1507 * out in the COMM_UP notification
1508 */
1509struct sctp_tcb *
1510sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1511{
1512	/*
1513	 * Use my the assoc_id to find a endpoint
1514	 */
1515	struct sctpasochead *head;
1516	struct sctp_tcb *stcb;
1517	uint32_t id;
1518
1519	if (inp == NULL) {
1520		SCTP_PRINTF("TSNH ep_associd\n");
1521		return (NULL);
1522	}
1523	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1524		SCTP_PRINTF("TSNH ep_associd0\n");
1525		return (NULL);
1526	}
1527	id = (uint32_t) asoc_id;
1528	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1529	if (head == NULL) {
1530		/* invalid id TSNH */
1531		SCTP_PRINTF("TSNH ep_associd1\n");
1532		return (NULL);
1533	}
1534	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1535		if (stcb->asoc.assoc_id == id) {
1536			if (inp != stcb->sctp_ep) {
1537				/*
1538				 * some other guy has the same id active (id
1539				 * collision ??).
1540				 */
1541				SCTP_PRINTF("TSNH ep_associd2\n");
1542				continue;
1543			}
1544			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1545				continue;
1546			}
1547			if (want_lock) {
1548				SCTP_TCB_LOCK(stcb);
1549			}
1550			return (stcb);
1551		}
1552	}
1553	return (NULL);
1554}
1555
1556
1557struct sctp_tcb *
1558sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1559{
1560	struct sctp_tcb *stcb;
1561
1562	SCTP_INP_RLOCK(inp);
1563	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1564	SCTP_INP_RUNLOCK(inp);
1565	return (stcb);
1566}
1567
1568
1569/*
1570 * Endpoint probe expects that the INP_INFO is locked.
1571 */
1572static struct sctp_inpcb *
1573sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1574    uint16_t lport, uint32_t vrf_id)
1575{
1576	struct sctp_inpcb *inp;
1577	struct sctp_laddr *laddr;
1578
1579#ifdef INET
1580	struct sockaddr_in *sin;
1581
1582#endif
1583#ifdef INET6
1584	struct sockaddr_in6 *sin6;
1585	struct sockaddr_in6 *intf_addr6;
1586
1587#endif
1588	int fnd;
1589
1590#ifdef INET
1591	sin = NULL;
1592#endif
1593#ifdef INET6
1594	sin6 = NULL;
1595#endif
1596	switch (nam->sa_family) {
1597#ifdef INET
1598	case AF_INET:
1599		sin = (struct sockaddr_in *)nam;
1600		break;
1601#endif
1602#ifdef INET6
1603	case AF_INET6:
1604		sin6 = (struct sockaddr_in6 *)nam;
1605		break;
1606#endif
1607	default:
1608		/* unsupported family */
1609		return (NULL);
1610	}
1611
1612	if (head == NULL)
1613		return (NULL);
1614
1615	LIST_FOREACH(inp, head, sctp_hash) {
1616		SCTP_INP_RLOCK(inp);
1617		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1618			SCTP_INP_RUNLOCK(inp);
1619			continue;
1620		}
1621		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1622		    (inp->sctp_lport == lport)) {
1623			/* got it */
1624#ifdef INET
1625			if ((nam->sa_family == AF_INET) &&
1626			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1627			    SCTP_IPV6_V6ONLY(inp)) {
1628				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1629				SCTP_INP_RUNLOCK(inp);
1630				continue;
1631			}
1632#endif
1633#ifdef INET6
1634			/* A V6 address and the endpoint is NOT bound V6 */
1635			if (nam->sa_family == AF_INET6 &&
1636			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1637				SCTP_INP_RUNLOCK(inp);
1638				continue;
1639			}
1640#endif
1641			/* does a VRF id match? */
1642			fnd = 0;
1643			if (inp->def_vrf_id == vrf_id)
1644				fnd = 1;
1645
1646			SCTP_INP_RUNLOCK(inp);
1647			if (!fnd)
1648				continue;
1649			return (inp);
1650		}
1651		SCTP_INP_RUNLOCK(inp);
1652	}
1653	switch (nam->sa_family) {
1654#ifdef INET
1655	case AF_INET:
1656		if (sin->sin_addr.s_addr == INADDR_ANY) {
1657			/* Can't hunt for one that has no address specified */
1658			return (NULL);
1659		}
1660		break;
1661#endif
1662#ifdef INET6
1663	case AF_INET6:
1664		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1665			/* Can't hunt for one that has no address specified */
1666			return (NULL);
1667		}
1668		break;
1669#endif
1670	default:
1671		break;
1672	}
1673	/*
1674	 * ok, not bound to all so see if we can find a EP bound to this
1675	 * address.
1676	 */
1677	LIST_FOREACH(inp, head, sctp_hash) {
1678		SCTP_INP_RLOCK(inp);
1679		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1680			SCTP_INP_RUNLOCK(inp);
1681			continue;
1682		}
1683		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1684			SCTP_INP_RUNLOCK(inp);
1685			continue;
1686		}
1687		/*
1688		 * Ok this could be a likely candidate, look at all of its
1689		 * addresses
1690		 */
1691		if (inp->sctp_lport != lport) {
1692			SCTP_INP_RUNLOCK(inp);
1693			continue;
1694		}
1695		/* does a VRF id match? */
1696		fnd = 0;
1697		if (inp->def_vrf_id == vrf_id)
1698			fnd = 1;
1699
1700		if (!fnd) {
1701			SCTP_INP_RUNLOCK(inp);
1702			continue;
1703		}
1704		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1705			if (laddr->ifa == NULL) {
1706				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1707				    __FUNCTION__);
1708				continue;
1709			}
1710			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1711			    (void *)laddr->ifa);
1712			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1713				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1714				continue;
1715			}
1716			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1717				/* possible, see if it matches */
1718				switch (nam->sa_family) {
1719#ifdef INET
1720				case AF_INET:
1721					if (sin->sin_addr.s_addr ==
1722					    laddr->ifa->address.sin.sin_addr.s_addr) {
1723						SCTP_INP_RUNLOCK(inp);
1724						return (inp);
1725					}
1726					break;
1727#endif
1728#ifdef INET6
1729				case AF_INET6:
1730					intf_addr6 = &laddr->ifa->address.sin6;
1731					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1732					    intf_addr6)) {
1733						SCTP_INP_RUNLOCK(inp);
1734						return (inp);
1735					}
1736					break;
1737#endif
1738				}
1739			}
1740		}
1741		SCTP_INP_RUNLOCK(inp);
1742	}
1743	return (NULL);
1744}
1745
1746
1747static struct sctp_inpcb *
1748sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1749{
1750	struct sctppcbhead *head;
1751	struct sctp_inpcb *t_inp;
1752	int fnd;
1753
1754	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1755	    SCTP_BASE_INFO(hashmark))];
1756	LIST_FOREACH(t_inp, head, sctp_hash) {
1757		if (t_inp->sctp_lport != lport) {
1758			continue;
1759		}
1760		/* is it in the VRF in question */
1761		fnd = 0;
1762		if (t_inp->def_vrf_id == vrf_id)
1763			fnd = 1;
1764		if (!fnd)
1765			continue;
1766
1767		/* This one is in use. */
1768		/* check the v6/v4 binding issue */
1769		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1770		    SCTP_IPV6_V6ONLY(t_inp)) {
1771			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1772				/* collision in V6 space */
1773				return (t_inp);
1774			} else {
1775				/* inp is BOUND_V4 no conflict */
1776				continue;
1777			}
1778		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1779			/* t_inp is bound v4 and v6, conflict always */
1780			return (t_inp);
1781		} else {
1782			/* t_inp is bound only V4 */
1783			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1784			    SCTP_IPV6_V6ONLY(inp)) {
1785				/* no conflict */
1786				continue;
1787			}
1788			/* else fall through to conflict */
1789		}
1790		return (t_inp);
1791	}
1792	return (NULL);
1793}
1794
1795
1796int
1797sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1798{
1799	/* For 1-2-1 with port reuse */
1800	struct sctppcbhead *head;
1801	struct sctp_inpcb *tinp;
1802
1803	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1804		/* only works with port reuse on */
1805		return (-1);
1806	}
1807	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1808		return (0);
1809	}
1810	SCTP_INP_RUNLOCK(inp);
1811	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1812	    SCTP_BASE_INFO(hashmark))];
1813	/* Kick out all non-listeners to the TCP hash */
1814	LIST_FOREACH(tinp, head, sctp_hash) {
1815		if (tinp->sctp_lport != inp->sctp_lport) {
1816			continue;
1817		}
1818		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1819			continue;
1820		}
1821		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1822			continue;
1823		}
1824		if (tinp->sctp_socket->so_qlimit) {
1825			continue;
1826		}
1827		SCTP_INP_WLOCK(tinp);
1828		LIST_REMOVE(tinp, sctp_hash);
1829		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1830		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1831		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1832		SCTP_INP_WUNLOCK(tinp);
1833	}
1834	SCTP_INP_WLOCK(inp);
1835	/* Pull from where he was */
1836	LIST_REMOVE(inp, sctp_hash);
1837	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1838	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1839	LIST_INSERT_HEAD(head, inp, sctp_hash);
1840	SCTP_INP_WUNLOCK(inp);
1841	SCTP_INP_RLOCK(inp);
1842	return (0);
1843}
1844
1845
1846struct sctp_inpcb *
1847sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1848    uint32_t vrf_id)
1849{
1850	/*
1851	 * First we check the hash table to see if someone has this port
1852	 * bound with just the port.
1853	 */
1854	struct sctp_inpcb *inp;
1855	struct sctppcbhead *head;
1856	int lport;
1857	unsigned int i;
1858
1859#ifdef INET
1860	struct sockaddr_in *sin;
1861
1862#endif
1863#ifdef INET6
1864	struct sockaddr_in6 *sin6;
1865
1866#endif
1867
1868	switch (nam->sa_family) {
1869#ifdef INET
1870	case AF_INET:
1871		sin = (struct sockaddr_in *)nam;
1872		lport = sin->sin_port;
1873		break;
1874#endif
1875#ifdef INET6
1876	case AF_INET6:
1877		sin6 = (struct sockaddr_in6 *)nam;
1878		lport = sin6->sin6_port;
1879		break;
1880#endif
1881	default:
1882		return (NULL);
1883	}
1884	/*
1885	 * I could cheat here and just cast to one of the types but we will
1886	 * do it right. It also provides the check against an Unsupported
1887	 * type too.
1888	 */
1889	/* Find the head of the ALLADDR chain */
1890	if (have_lock == 0) {
1891		SCTP_INP_INFO_RLOCK();
1892	}
1893	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1894	    SCTP_BASE_INFO(hashmark))];
1895	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1896
1897	/*
1898	 * If the TCP model exists it could be that the main listening
1899	 * endpoint is gone but there still exists a connected socket for
1900	 * this guy. If so we can return the first one that we find. This
1901	 * may NOT be the correct one so the caller should be wary on the
1902	 * returned INP. Currently the only caller that sets find_tcp_pool
1903	 * is in bindx where we are verifying that a user CAN bind the
1904	 * address. He either has bound it already, or someone else has, or
1905	 * its open to bind, so this is good enough.
1906	 */
1907	if (inp == NULL && find_tcp_pool) {
1908		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1909			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1910			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1911			if (inp) {
1912				break;
1913			}
1914		}
1915	}
1916	if (inp) {
1917		SCTP_INP_INCR_REF(inp);
1918	}
1919	if (have_lock == 0) {
1920		SCTP_INP_INFO_RUNLOCK();
1921	}
1922	return (inp);
1923}
1924
1925
1926/*
1927 * Find an association for an endpoint with the pointer to whom you want to
1928 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1929 * need to change the *to to some other struct like a mbuf...
1930 */
1931struct sctp_tcb *
1932sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1933    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1934    uint32_t vrf_id)
1935{
1936	struct sctp_inpcb *inp = NULL;
1937	struct sctp_tcb *stcb;
1938
1939	SCTP_INP_INFO_RLOCK();
1940	if (find_tcp_pool) {
1941		if (inp_p != NULL) {
1942			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1943			    vrf_id);
1944		} else {
1945			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1946			    vrf_id);
1947		}
1948		if (stcb != NULL) {
1949			SCTP_INP_INFO_RUNLOCK();
1950			return (stcb);
1951		}
1952	}
1953	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1954	if (inp_p != NULL) {
1955		*inp_p = inp;
1956	}
1957	SCTP_INP_INFO_RUNLOCK();
1958	if (inp == NULL) {
1959		return (NULL);
1960	}
1961	/*
1962	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1963	 * we now place the source address or from in the to of the find
1964	 * endpoint call. Since in reality this chain is used from the
1965	 * inbound packet side.
1966	 */
1967	if (inp_p != NULL) {
1968		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1969		    NULL);
1970	} else {
1971		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
1972		    NULL);
1973	}
1974	return (stcb);
1975}
1976
1977
1978/*
1979 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1980 * find all addresses that the sender has specified in any address list. Each
1981 * address will be used to lookup the TCB and see if one exits.
1982 */
1983static struct sctp_tcb *
1984sctp_findassociation_special_addr(struct mbuf *m, int offset,
1985    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1986    struct sockaddr *dst)
1987{
1988	struct sctp_paramhdr *phdr, parm_buf;
1989
1990#if defined(INET) || defined(INET6)
1991	struct sctp_tcb *stcb;
1992	uint16_t ptype;
1993
1994#endif
1995	uint16_t plen;
1996
1997#ifdef INET
1998	struct sockaddr_in sin4;
1999
2000#endif
2001#ifdef INET6
2002	struct sockaddr_in6 sin6;
2003
2004#endif
2005
2006#ifdef INET
2007	memset(&sin4, 0, sizeof(sin4));
2008	sin4.sin_len = sizeof(sin4);
2009	sin4.sin_family = AF_INET;
2010	sin4.sin_port = sh->src_port;
2011#endif
2012#ifdef INET6
2013	memset(&sin6, 0, sizeof(sin6));
2014	sin6.sin6_len = sizeof(sin6);
2015	sin6.sin6_family = AF_INET6;
2016	sin6.sin6_port = sh->src_port;
2017#endif
2018
2019	offset += sizeof(struct sctp_init_chunk);
2020
2021	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2022	while (phdr != NULL) {
2023		/* now we must see if we want the parameter */
2024#if defined(INET) || defined(INET6)
2025		ptype = ntohs(phdr->param_type);
2026#endif
2027		plen = ntohs(phdr->param_length);
2028		if (plen == 0) {
2029			break;
2030		}
2031#ifdef INET
2032		if (ptype == SCTP_IPV4_ADDRESS &&
2033		    plen == sizeof(struct sctp_ipv4addr_param)) {
2034			/* Get the rest of the address */
2035			struct sctp_ipv4addr_param ip4_parm, *p4;
2036
2037			phdr = sctp_get_next_param(m, offset,
2038			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2039			if (phdr == NULL) {
2040				return (NULL);
2041			}
2042			p4 = (struct sctp_ipv4addr_param *)phdr;
2043			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2044			/* look it up */
2045			stcb = sctp_findassociation_ep_addr(inp_p,
2046			    (struct sockaddr *)&sin4, netp, dst, NULL);
2047			if (stcb != NULL) {
2048				return (stcb);
2049			}
2050		}
2051#endif
2052#ifdef INET6
2053		if (ptype == SCTP_IPV6_ADDRESS &&
2054		    plen == sizeof(struct sctp_ipv6addr_param)) {
2055			/* Get the rest of the address */
2056			struct sctp_ipv6addr_param ip6_parm, *p6;
2057
2058			phdr = sctp_get_next_param(m, offset,
2059			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2060			if (phdr == NULL) {
2061				return (NULL);
2062			}
2063			p6 = (struct sctp_ipv6addr_param *)phdr;
2064			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2065			/* look it up */
2066			stcb = sctp_findassociation_ep_addr(inp_p,
2067			    (struct sockaddr *)&sin6, netp, dst, NULL);
2068			if (stcb != NULL) {
2069				return (stcb);
2070			}
2071		}
2072#endif
2073		offset += SCTP_SIZE32(plen);
2074		phdr = sctp_get_next_param(m, offset, &parm_buf,
2075		    sizeof(parm_buf));
2076	}
2077	return (NULL);
2078}
2079
2080static struct sctp_tcb *
2081sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2082    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2083    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2084{
2085	/*
2086	 * Use my vtag to hash. If we find it we then verify the source addr
2087	 * is in the assoc. If all goes well we save a bit on rec of a
2088	 * packet.
2089	 */
2090	struct sctpasochead *head;
2091	struct sctp_nets *net;
2092	struct sctp_tcb *stcb;
2093
2094	SCTP_INP_INFO_RLOCK();
2095	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2096	    SCTP_BASE_INFO(hashasocmark))];
2097	if (head == NULL) {
2098		/* invalid vtag */
2099		SCTP_INP_INFO_RUNLOCK();
2100		return (NULL);
2101	}
2102	LIST_FOREACH(stcb, head, sctp_asocs) {
2103		SCTP_INP_RLOCK(stcb->sctp_ep);
2104		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2105			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2106			continue;
2107		}
2108		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2109			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2110			continue;
2111		}
2112		SCTP_TCB_LOCK(stcb);
2113		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2114		if (stcb->asoc.my_vtag == vtag) {
2115			/* candidate */
2116			if (stcb->rport != rport) {
2117				SCTP_TCB_UNLOCK(stcb);
2118				continue;
2119			}
2120			if (stcb->sctp_ep->sctp_lport != lport) {
2121				SCTP_TCB_UNLOCK(stcb);
2122				continue;
2123			}
2124			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2125				SCTP_TCB_UNLOCK(stcb);
2126				continue;
2127			}
2128			/* RRS:Need toaddr check here */
2129			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2130				/* Endpoint does not own this address */
2131				SCTP_TCB_UNLOCK(stcb);
2132				continue;
2133			}
2134			if (remote_tag) {
2135				/*
2136				 * If we have both vtags that's all we match
2137				 * on
2138				 */
2139				if (stcb->asoc.peer_vtag == remote_tag) {
2140					/*
2141					 * If both tags match we consider it
2142					 * conclusive and check NO
2143					 * source/destination addresses
2144					 */
2145					goto conclusive;
2146				}
2147			}
2148			if (skip_src_check) {
2149		conclusive:
2150				if (from) {
2151					*netp = sctp_findnet(stcb, from);
2152				} else {
2153					*netp = NULL;	/* unknown */
2154				}
2155				if (inp_p)
2156					*inp_p = stcb->sctp_ep;
2157				SCTP_INP_INFO_RUNLOCK();
2158				return (stcb);
2159			}
2160			net = sctp_findnet(stcb, from);
2161			if (net) {
2162				/* yep its him. */
2163				*netp = net;
2164				SCTP_STAT_INCR(sctps_vtagexpress);
2165				*inp_p = stcb->sctp_ep;
2166				SCTP_INP_INFO_RUNLOCK();
2167				return (stcb);
2168			} else {
2169				/*
2170				 * not him, this should only happen in rare
2171				 * cases so I peg it.
2172				 */
2173				SCTP_STAT_INCR(sctps_vtagbogus);
2174			}
2175		}
2176		SCTP_TCB_UNLOCK(stcb);
2177	}
2178	SCTP_INP_INFO_RUNLOCK();
2179	return (NULL);
2180}
2181
2182
2183/*
2184 * Find an association with the pointer to the inbound IP packet. This can be
2185 * a IPv4 or IPv6 packet.
2186 */
2187struct sctp_tcb *
2188sctp_findassociation_addr(struct mbuf *m, int offset,
2189    struct sockaddr *src, struct sockaddr *dst,
2190    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2191    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2192{
2193	int find_tcp_pool;
2194	struct sctp_tcb *stcb;
2195	struct sctp_inpcb *inp;
2196
2197	if (sh->v_tag) {
2198		/* we only go down this path if vtag is non-zero */
2199		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2200		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2201		if (stcb) {
2202			return (stcb);
2203		}
2204	}
2205	find_tcp_pool = 0;
2206	if ((ch->chunk_type != SCTP_INITIATION) &&
2207	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2208	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2209	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2210		/* Other chunk types go to the tcp pool. */
2211		find_tcp_pool = 1;
2212	}
2213	if (inp_p) {
2214		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2215		    find_tcp_pool, vrf_id);
2216		inp = *inp_p;
2217	} else {
2218		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2219		    find_tcp_pool, vrf_id);
2220	}
2221	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2222	if (stcb == NULL && inp) {
2223		/* Found a EP but not this address */
2224		if ((ch->chunk_type == SCTP_INITIATION) ||
2225		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2226			/*-
2227			 * special hook, we do NOT return linp or an
2228			 * association that is linked to an existing
2229			 * association that is under the TCP pool (i.e. no
2230			 * listener exists). The endpoint finding routine
2231			 * will always find a listener before examining the
2232			 * TCP pool.
2233			 */
2234			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2235				if (inp_p) {
2236					*inp_p = NULL;
2237				}
2238				return (NULL);
2239			}
2240			stcb = sctp_findassociation_special_addr(m,
2241			    offset, sh, &inp, netp, dst);
2242			if (inp_p != NULL) {
2243				*inp_p = inp;
2244			}
2245		}
2246	}
2247	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2248	return (stcb);
2249}
2250
2251/*
2252 * lookup an association by an ASCONF lookup address.
2253 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2254 */
2255struct sctp_tcb *
2256sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2257    struct sockaddr *dst, struct sctphdr *sh,
2258    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2259{
2260	struct sctp_tcb *stcb;
2261	struct sockaddr_storage remote_store;
2262	struct sctp_paramhdr parm_buf, *phdr;
2263	int ptype;
2264	int zero_address = 0;
2265
2266#ifdef INET
2267	struct sockaddr_in *sin;
2268
2269#endif
2270#ifdef INET6
2271	struct sockaddr_in6 *sin6;
2272
2273#endif
2274
2275	memset(&remote_store, 0, sizeof(remote_store));
2276	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2277	    &parm_buf, sizeof(struct sctp_paramhdr));
2278	if (phdr == NULL) {
2279		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2280		    __FUNCTION__);
2281		return NULL;
2282	}
2283	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2284	/* get the correlation address */
2285	switch (ptype) {
2286#ifdef INET6
2287	case SCTP_IPV6_ADDRESS:
2288		{
2289			/* ipv6 address param */
2290			struct sctp_ipv6addr_param *p6, p6_buf;
2291
2292			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2293				return NULL;
2294			}
2295			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2296			    offset + sizeof(struct sctp_asconf_chunk),
2297			    &p6_buf.ph, sizeof(*p6));
2298			if (p6 == NULL) {
2299				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2300				    __FUNCTION__);
2301				return (NULL);
2302			}
2303			sin6 = (struct sockaddr_in6 *)&remote_store;
2304			sin6->sin6_family = AF_INET6;
2305			sin6->sin6_len = sizeof(*sin6);
2306			sin6->sin6_port = sh->src_port;
2307			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2308			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2309				zero_address = 1;
2310			break;
2311		}
2312#endif
2313#ifdef INET
2314	case SCTP_IPV4_ADDRESS:
2315		{
2316			/* ipv4 address param */
2317			struct sctp_ipv4addr_param *p4, p4_buf;
2318
2319			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2320				return NULL;
2321			}
2322			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2323			    offset + sizeof(struct sctp_asconf_chunk),
2324			    &p4_buf.ph, sizeof(*p4));
2325			if (p4 == NULL) {
2326				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2327				    __FUNCTION__);
2328				return (NULL);
2329			}
2330			sin = (struct sockaddr_in *)&remote_store;
2331			sin->sin_family = AF_INET;
2332			sin->sin_len = sizeof(*sin);
2333			sin->sin_port = sh->src_port;
2334			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2335			if (sin->sin_addr.s_addr == INADDR_ANY)
2336				zero_address = 1;
2337			break;
2338		}
2339#endif
2340	default:
2341		/* invalid address param type */
2342		return NULL;
2343	}
2344
2345	if (zero_address) {
2346		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2347		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2348		if (stcb != NULL) {
2349			SCTP_INP_DECR_REF(*inp_p);
2350		}
2351	} else {
2352		stcb = sctp_findassociation_ep_addr(inp_p,
2353		    (struct sockaddr *)&remote_store, netp,
2354		    dst, NULL);
2355	}
2356	return (stcb);
2357}
2358
2359
2360/*
2361 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2362 * addresses. This way if we don't get a bind we by default pick a ephemeral
2363 * port with all addresses bound.
2364 */
2365int
2366sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2367{
2368	/*
2369	 * we get called when a new endpoint starts up. We need to allocate
2370	 * the sctp_inpcb structure from the zone and init it. Mark it as
2371	 * unbound and find a port that we can use as an ephemeral with
2372	 * INADDR_ANY. If the user binds later no problem we can then add in
2373	 * the specific addresses. And setup the default parameters for the
2374	 * EP.
2375	 */
2376	int i, error;
2377	struct sctp_inpcb *inp;
2378	struct sctp_pcb *m;
2379	struct timeval time;
2380	sctp_sharedkey_t *null_key;
2381
2382	error = 0;
2383
2384	SCTP_INP_INFO_WLOCK();
2385	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2386	if (inp == NULL) {
2387		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2388		SCTP_INP_INFO_WUNLOCK();
2389		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2390		return (ENOBUFS);
2391	}
2392	/* zap it */
2393	bzero(inp, sizeof(*inp));
2394
2395	/* bump generations */
2396	/* setup socket pointers */
2397	inp->sctp_socket = so;
2398	inp->ip_inp.inp.inp_socket = so;
2399#ifdef INET6
2400	if (INP_SOCKAF(so) == AF_INET6) {
2401		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2402			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2403		}
2404		if (MODULE_GLOBAL(ip6_v6only)) {
2405			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2406		}
2407	}
2408#endif
2409	inp->sctp_associd_counter = 1;
2410	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2411	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2412	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2413	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2414	/* init the small hash table we use to track asocid <-> tcb */
2415	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2416	if (inp->sctp_asocidhash == NULL) {
2417		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2418		SCTP_INP_INFO_WUNLOCK();
2419		return (ENOBUFS);
2420	}
2421#ifdef IPSEC
2422	{
2423		struct inpcbpolicy *pcb_sp = NULL;
2424
2425		error = ipsec_init_policy(so, &pcb_sp);
2426		/* Arrange to share the policy */
2427		inp->ip_inp.inp.inp_sp = pcb_sp;
2428		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2429	}
2430	if (error != 0) {
2431		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2432		SCTP_INP_INFO_WUNLOCK();
2433		return error;
2434	}
2435#endif				/* IPSEC */
2436	SCTP_INCR_EP_COUNT();
2437	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2438	SCTP_INP_INFO_WUNLOCK();
2439
2440	so->so_pcb = (caddr_t)inp;
2441
2442	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2443		/* UDP style socket */
2444		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2445		    SCTP_PCB_FLAGS_UNBOUND);
2446		/* Be sure it is NON-BLOCKING IO for UDP */
2447		/* SCTP_SET_SO_NBIO(so); */
2448	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2449		/* TCP style socket */
2450		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2451		    SCTP_PCB_FLAGS_UNBOUND);
2452		/* Be sure we have blocking IO by default */
2453		SCTP_CLEAR_SO_NBIO(so);
2454	} else {
2455		/*
2456		 * unsupported socket type (RAW, etc)- in case we missed it
2457		 * in protosw
2458		 */
2459		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2460		so->so_pcb = NULL;
2461		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2462		return (EOPNOTSUPP);
2463	}
2464	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2465		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2466		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2467	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2468		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2469		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2470	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2471		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2472		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2473	}
2474	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2475	    &inp->sctp_hashmark);
2476	if (inp->sctp_tcbhash == NULL) {
2477		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2478		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2479		so->so_pcb = NULL;
2480		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2481		return (ENOBUFS);
2482	}
2483	inp->def_vrf_id = vrf_id;
2484
2485	SCTP_INP_INFO_WLOCK();
2486	SCTP_INP_LOCK_INIT(inp);
2487	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2488	SCTP_INP_READ_INIT(inp);
2489	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2490	/* lock the new ep */
2491	SCTP_INP_WLOCK(inp);
2492
2493	/* add it to the info area */
2494	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2495	SCTP_INP_INFO_WUNLOCK();
2496
2497	TAILQ_INIT(&inp->read_queue);
2498	LIST_INIT(&inp->sctp_addr_list);
2499
2500	LIST_INIT(&inp->sctp_asoc_list);
2501
2502#ifdef SCTP_TRACK_FREED_ASOCS
2503	/* TEMP CODE */
2504	LIST_INIT(&inp->sctp_asoc_free_list);
2505#endif
2506	/* Init the timer structure for signature change */
2507	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2508	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2509
2510	/* now init the actual endpoint default data */
2511	m = &inp->sctp_ep;
2512
2513	/* setup the base timeout information */
2514	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2515	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2516	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2517	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2518	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2519	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2520	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2521	/* all max/min max are in ms */
2522	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2523	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2524	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2525	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2526	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2527	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2528	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2529	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2530	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2531	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2532	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2533	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2534	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2535
2536	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2537	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2538	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2539	/* number of streams to pre-open on a association */
2540	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2541
2542	/* Add adaptation cookie */
2543	m->adaptation_layer_indicator = 0;
2544	m->adaptation_layer_indicator_provided = 0;
2545
2546	/* seed random number generator */
2547	m->random_counter = 1;
2548	m->store_at = SCTP_SIGNATURE_SIZE;
2549	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2550	sctp_fill_random_store(m);
2551
2552	/* Minimum cookie size */
2553	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2554	    sizeof(struct sctp_state_cookie);
2555	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2556
2557	/* Setup the initial secret */
2558	(void)SCTP_GETTIME_TIMEVAL(&time);
2559	m->time_of_secret_change = time.tv_sec;
2560
2561	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2562		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2563	}
2564	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2565
2566	/* How long is a cookie good for ? */
2567	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2568	/*
2569	 * Initialize authentication parameters
2570	 */
2571	m->local_hmacs = sctp_default_supported_hmaclist();
2572	m->local_auth_chunks = sctp_alloc_chunklist();
2573	m->default_dscp = 0;
2574#ifdef INET6
2575	m->default_flowlabel = 0;
2576#endif
2577	m->port = 0;		/* encapsulation disabled by default */
2578	sctp_auth_set_default_chunks(m->local_auth_chunks);
2579	LIST_INIT(&m->shared_keys);
2580	/* add default NULL key as key id 0 */
2581	null_key = sctp_alloc_sharedkey();
2582	sctp_insert_sharedkey(&m->shared_keys, null_key);
2583	SCTP_INP_WUNLOCK(inp);
2584#ifdef SCTP_LOG_CLOSING
2585	sctp_log_closing(inp, NULL, 12);
2586#endif
2587	return (error);
2588}
2589
2590
2591void
2592sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2593    struct sctp_tcb *stcb)
2594{
2595	struct sctp_nets *net;
2596	uint16_t lport, rport;
2597	struct sctppcbhead *head;
2598	struct sctp_laddr *laddr, *oladdr;
2599
2600	atomic_add_int(&stcb->asoc.refcnt, 1);
2601	SCTP_TCB_UNLOCK(stcb);
2602	SCTP_INP_INFO_WLOCK();
2603	SCTP_INP_WLOCK(old_inp);
2604	SCTP_INP_WLOCK(new_inp);
2605	SCTP_TCB_LOCK(stcb);
2606	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2607
2608	new_inp->sctp_ep.time_of_secret_change =
2609	    old_inp->sctp_ep.time_of_secret_change;
2610	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2611	    sizeof(old_inp->sctp_ep.secret_key));
2612	new_inp->sctp_ep.current_secret_number =
2613	    old_inp->sctp_ep.current_secret_number;
2614	new_inp->sctp_ep.last_secret_number =
2615	    old_inp->sctp_ep.last_secret_number;
2616	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2617
2618	/* make it so new data pours into the new socket */
2619	stcb->sctp_socket = new_inp->sctp_socket;
2620	stcb->sctp_ep = new_inp;
2621
2622	/* Copy the port across */
2623	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2624	rport = stcb->rport;
2625	/* Pull the tcb from the old association */
2626	LIST_REMOVE(stcb, sctp_tcbhash);
2627	LIST_REMOVE(stcb, sctp_tcblist);
2628	if (stcb->asoc.in_asocid_hash) {
2629		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2630	}
2631	/* Now insert the new_inp into the TCP connected hash */
2632	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2633
2634	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2635	/* Its safe to access */
2636	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2637
2638	/* Now move the tcb into the endpoint list */
2639	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2640	/*
2641	 * Question, do we even need to worry about the ep-hash since we
2642	 * only have one connection? Probably not :> so lets get rid of it
2643	 * and not suck up any kernel memory in that.
2644	 */
2645	if (stcb->asoc.in_asocid_hash) {
2646		struct sctpasochead *lhd;
2647
2648		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2649		    new_inp->hashasocidmark)];
2650		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2651	}
2652	/* Ok. Let's restart timer. */
2653	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2654		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2655		    stcb, net);
2656	}
2657
2658	SCTP_INP_INFO_WUNLOCK();
2659	if (new_inp->sctp_tcbhash != NULL) {
2660		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2661		new_inp->sctp_tcbhash = NULL;
2662	}
2663	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2664		/* Subset bound, so copy in the laddr list from the old_inp */
2665		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2666			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2667			if (laddr == NULL) {
2668				/*
2669				 * Gak, what can we do? This assoc is really
2670				 * HOSED. We probably should send an abort
2671				 * here.
2672				 */
2673				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2674				continue;
2675			}
2676			SCTP_INCR_LADDR_COUNT();
2677			bzero(laddr, sizeof(*laddr));
2678			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2679			laddr->ifa = oladdr->ifa;
2680			atomic_add_int(&laddr->ifa->refcount, 1);
2681			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2682			    sctp_nxt_addr);
2683			new_inp->laddr_count++;
2684			if (oladdr == stcb->asoc.last_used_address) {
2685				stcb->asoc.last_used_address = laddr;
2686			}
2687		}
2688	}
2689	/*
2690	 * Now any running timers need to be adjusted since we really don't
2691	 * care if they are running or not just blast in the new_inp into
2692	 * all of them.
2693	 */
2694
2695	stcb->asoc.dack_timer.ep = (void *)new_inp;
2696	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2697	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2698	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2699	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2700	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2701	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2702	/* now what about the nets? */
2703	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2704		net->pmtu_timer.ep = (void *)new_inp;
2705		net->hb_timer.ep = (void *)new_inp;
2706		net->rxt_timer.ep = (void *)new_inp;
2707	}
2708	SCTP_INP_WUNLOCK(new_inp);
2709	SCTP_INP_WUNLOCK(old_inp);
2710}
2711
2712
2713
2714
2715/* sctp_ifap is used to bypass normal local address validation checks */
2716int
2717sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2718    struct sctp_ifa *sctp_ifap, struct thread *p)
2719{
2720	/* bind a ep to a socket address */
2721	struct sctppcbhead *head;
2722	struct sctp_inpcb *inp, *inp_tmp;
2723	struct inpcb *ip_inp;
2724	int port_reuse_active = 0;
2725	int bindall;
2726	uint16_t lport;
2727	int error;
2728	uint32_t vrf_id;
2729
2730	lport = 0;
2731	error = 0;
2732	bindall = 1;
2733	inp = (struct sctp_inpcb *)so->so_pcb;
2734	ip_inp = (struct inpcb *)so->so_pcb;
2735#ifdef SCTP_DEBUG
2736	if (addr) {
2737		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2738		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2739		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2740		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2741	}
2742#endif
2743	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2744		/* already did a bind, subsequent binds NOT allowed ! */
2745		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2746		return (EINVAL);
2747	}
2748#ifdef INVARIANTS
2749	if (p == NULL)
2750		panic("null proc/thread");
2751#endif
2752	if (addr != NULL) {
2753		switch (addr->sa_family) {
2754#ifdef INET
2755		case AF_INET:
2756			{
2757				struct sockaddr_in *sin;
2758
2759				/* IPV6_V6ONLY socket? */
2760				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2761					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2762					return (EINVAL);
2763				}
2764				if (addr->sa_len != sizeof(*sin)) {
2765					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2766					return (EINVAL);
2767				}
2768				sin = (struct sockaddr_in *)addr;
2769				lport = sin->sin_port;
2770				/*
2771				 * For LOOPBACK the prison_local_ip4() call
2772				 * will transmute the ip address to the
2773				 * proper value.
2774				 */
2775				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2776					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2777					return (error);
2778				}
2779				if (sin->sin_addr.s_addr != INADDR_ANY) {
2780					bindall = 0;
2781				}
2782				break;
2783			}
2784#endif
2785#ifdef INET6
2786		case AF_INET6:
2787			{
2788				/*
2789				 * Only for pure IPv6 Address. (No IPv4
2790				 * Mapped!)
2791				 */
2792				struct sockaddr_in6 *sin6;
2793
2794				sin6 = (struct sockaddr_in6 *)addr;
2795
2796				if (addr->sa_len != sizeof(*sin6)) {
2797					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2798					return (EINVAL);
2799				}
2800				lport = sin6->sin6_port;
2801				/*
2802				 * For LOOPBACK the prison_local_ip6() call
2803				 * will transmute the ipv6 address to the
2804				 * proper value.
2805				 */
2806				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2807				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2808					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2809					return (error);
2810				}
2811				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2812					bindall = 0;
2813					/* KAME hack: embed scopeid */
2814					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2815						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2816						return (EINVAL);
2817					}
2818				}
2819				/* this must be cleared for ifa_ifwithaddr() */
2820				sin6->sin6_scope_id = 0;
2821				break;
2822			}
2823#endif
2824		default:
2825			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2826			return (EAFNOSUPPORT);
2827		}
2828	}
2829	SCTP_INP_INFO_WLOCK();
2830	SCTP_INP_WLOCK(inp);
2831	/* Setup a vrf_id to be the default for the non-bind-all case. */
2832	vrf_id = inp->def_vrf_id;
2833
2834	/* increase our count due to the unlock we do */
2835	SCTP_INP_INCR_REF(inp);
2836	if (lport) {
2837		/*
2838		 * Did the caller specify a port? if so we must see if an ep
2839		 * already has this one bound.
2840		 */
2841		/* got to be root to get at low ports */
2842		if (ntohs(lport) < IPPORT_RESERVED) {
2843			if (p && (error =
2844			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2845			    )) {
2846				SCTP_INP_DECR_REF(inp);
2847				SCTP_INP_WUNLOCK(inp);
2848				SCTP_INP_INFO_WUNLOCK();
2849				return (error);
2850			}
2851		}
2852		if (p == NULL) {
2853			SCTP_INP_DECR_REF(inp);
2854			SCTP_INP_WUNLOCK(inp);
2855			SCTP_INP_INFO_WUNLOCK();
2856			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2857			return (error);
2858		}
2859		SCTP_INP_WUNLOCK(inp);
2860		if (bindall) {
2861			vrf_id = inp->def_vrf_id;
2862			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2863			if (inp_tmp != NULL) {
2864				/*
2865				 * lock guy returned and lower count note
2866				 * that we are not bound so inp_tmp should
2867				 * NEVER be inp. And it is this inp
2868				 * (inp_tmp) that gets the reference bump,
2869				 * so we must lower it.
2870				 */
2871				SCTP_INP_DECR_REF(inp_tmp);
2872				/* unlock info */
2873				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2874				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2875					/*
2876					 * Ok, must be one-2-one and
2877					 * allowing port re-use
2878					 */
2879					port_reuse_active = 1;
2880					goto continue_anyway;
2881				}
2882				SCTP_INP_DECR_REF(inp);
2883				SCTP_INP_INFO_WUNLOCK();
2884				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2885				return (EADDRINUSE);
2886			}
2887		} else {
2888			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2889			if (inp_tmp != NULL) {
2890				/*
2891				 * lock guy returned and lower count note
2892				 * that we are not bound so inp_tmp should
2893				 * NEVER be inp. And it is this inp
2894				 * (inp_tmp) that gets the reference bump,
2895				 * so we must lower it.
2896				 */
2897				SCTP_INP_DECR_REF(inp_tmp);
2898				/* unlock info */
2899				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2900				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2901					/*
2902					 * Ok, must be one-2-one and
2903					 * allowing port re-use
2904					 */
2905					port_reuse_active = 1;
2906					goto continue_anyway;
2907				}
2908				SCTP_INP_DECR_REF(inp);
2909				SCTP_INP_INFO_WUNLOCK();
2910				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2911				return (EADDRINUSE);
2912			}
2913		}
2914continue_anyway:
2915		SCTP_INP_WLOCK(inp);
2916		if (bindall) {
2917			/* verify that no lport is not used by a singleton */
2918			if ((port_reuse_active == 0) &&
2919			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2920				/* Sorry someone already has this one bound */
2921				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2922				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2923					port_reuse_active = 1;
2924				} else {
2925					SCTP_INP_DECR_REF(inp);
2926					SCTP_INP_WUNLOCK(inp);
2927					SCTP_INP_INFO_WUNLOCK();
2928					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2929					return (EADDRINUSE);
2930				}
2931			}
2932		}
2933	} else {
2934		uint16_t first, last, candidate;
2935		uint16_t count;
2936		int done;
2937
2938		if (ip_inp->inp_flags & INP_HIGHPORT) {
2939			first = MODULE_GLOBAL(ipport_hifirstauto);
2940			last = MODULE_GLOBAL(ipport_hilastauto);
2941		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2942			if (p && (error =
2943			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2944			    )) {
2945				SCTP_INP_DECR_REF(inp);
2946				SCTP_INP_WUNLOCK(inp);
2947				SCTP_INP_INFO_WUNLOCK();
2948				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2949				return (error);
2950			}
2951			first = MODULE_GLOBAL(ipport_lowfirstauto);
2952			last = MODULE_GLOBAL(ipport_lowlastauto);
2953		} else {
2954			first = MODULE_GLOBAL(ipport_firstauto);
2955			last = MODULE_GLOBAL(ipport_lastauto);
2956		}
2957		if (first > last) {
2958			uint16_t temp;
2959
2960			temp = first;
2961			first = last;
2962			last = temp;
2963		}
2964		count = last - first + 1;	/* number of candidates */
2965		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2966
2967		done = 0;
2968		while (!done) {
2969			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2970				done = 1;
2971			}
2972			if (!done) {
2973				if (--count == 0) {
2974					SCTP_INP_DECR_REF(inp);
2975					SCTP_INP_WUNLOCK(inp);
2976					SCTP_INP_INFO_WUNLOCK();
2977					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2978					return (EADDRINUSE);
2979				}
2980				if (candidate == last)
2981					candidate = first;
2982				else
2983					candidate = candidate + 1;
2984			}
2985		}
2986		lport = htons(candidate);
2987	}
2988	SCTP_INP_DECR_REF(inp);
2989	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2990	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2991		/*
2992		 * this really should not happen. The guy did a non-blocking
2993		 * bind and then did a close at the same time.
2994		 */
2995		SCTP_INP_WUNLOCK(inp);
2996		SCTP_INP_INFO_WUNLOCK();
2997		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2998		return (EINVAL);
2999	}
3000	/* ok we look clear to give out this port, so lets setup the binding */
3001	if (bindall) {
3002		/* binding to all addresses, so just set in the proper flags */
3003		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3004		/* set the automatic addr changes from kernel flag */
3005		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3006			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3007			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3008		} else {
3009			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3010			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3011		}
3012		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3013			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3014		} else {
3015			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3016		}
3017		/*
3018		 * set the automatic mobility_base from kernel flag (by
3019		 * micchie)
3020		 */
3021		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3022			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3023			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3024		} else {
3025			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3026			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3027		}
3028		/*
3029		 * set the automatic mobility_fasthandoff from kernel flag
3030		 * (by micchie)
3031		 */
3032		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3033			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3034			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3035		} else {
3036			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3037			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3038		}
3039	} else {
3040		/*
3041		 * bind specific, make sure flags is off and add a new
3042		 * address structure to the sctp_addr_list inside the ep
3043		 * structure.
3044		 *
3045		 * We will need to allocate one and insert it at the head. The
3046		 * socketopt call can just insert new addresses in there as
3047		 * well. It will also have to do the embed scope kame hack
3048		 * too (before adding).
3049		 */
3050		struct sctp_ifa *ifa;
3051		struct sockaddr_storage store_sa;
3052
3053		memset(&store_sa, 0, sizeof(store_sa));
3054		switch (addr->sa_family) {
3055#ifdef INET
3056		case AF_INET:
3057			{
3058				struct sockaddr_in *sin;
3059
3060				sin = (struct sockaddr_in *)&store_sa;
3061				memcpy(sin, addr, sizeof(struct sockaddr_in));
3062				sin->sin_port = 0;
3063				break;
3064			}
3065#endif
3066#ifdef INET6
3067		case AF_INET6:
3068			{
3069				struct sockaddr_in6 *sin6;
3070
3071				sin6 = (struct sockaddr_in6 *)&store_sa;
3072				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3073				sin6->sin6_port = 0;
3074				break;
3075			}
3076#endif
3077		default:
3078			break;
3079		}
3080		/*
3081		 * first find the interface with the bound address need to
3082		 * zero out the port to find the address! yuck! can't do
3083		 * this earlier since need port for sctp_pcb_findep()
3084		 */
3085		if (sctp_ifap != NULL) {
3086			ifa = sctp_ifap;
3087		} else {
3088			/*
3089			 * Note for BSD we hit here always other O/S's will
3090			 * pass things in via the sctp_ifap argument
3091			 * (Panda).
3092			 */
3093			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3094			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3095		}
3096		if (ifa == NULL) {
3097			/* Can't find an interface with that address */
3098			SCTP_INP_WUNLOCK(inp);
3099			SCTP_INP_INFO_WUNLOCK();
3100			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3101			return (EADDRNOTAVAIL);
3102		}
3103#ifdef INET6
3104		if (addr->sa_family == AF_INET6) {
3105			/* GAK, more FIXME IFA lock? */
3106			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3107				/* Can't bind a non-existent addr. */
3108				SCTP_INP_WUNLOCK(inp);
3109				SCTP_INP_INFO_WUNLOCK();
3110				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3111				return (EINVAL);
3112			}
3113		}
3114#endif
3115		/* we're not bound all */
3116		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3117		/* allow bindx() to send ASCONF's for binding changes */
3118		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3119		/* clear automatic addr changes from kernel flag */
3120		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3121
3122		/* add this address to the endpoint list */
3123		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3124		if (error != 0) {
3125			SCTP_INP_WUNLOCK(inp);
3126			SCTP_INP_INFO_WUNLOCK();
3127			return (error);
3128		}
3129		inp->laddr_count++;
3130	}
3131	/* find the bucket */
3132	if (port_reuse_active) {
3133		/* Put it into tcp 1-2-1 hash */
3134		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3135		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3136	} else {
3137		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3138	}
3139	/* put it in the bucket */
3140	LIST_INSERT_HEAD(head, inp, sctp_hash);
3141	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3142	    (void *)head, ntohs(lport), port_reuse_active);
3143	/* set in the port */
3144	inp->sctp_lport = lport;
3145
3146	/* turn off just the unbound flag */
3147	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3148	SCTP_INP_WUNLOCK(inp);
3149	SCTP_INP_INFO_WUNLOCK();
3150	return (0);
3151}
3152
3153
3154static void
3155sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3156{
3157	struct sctp_iterator *it, *nit;
3158
3159	/*
3160	 * We enter with the only the ITERATOR_LOCK in place and a write
3161	 * lock on the inp_info stuff.
3162	 */
3163	it = sctp_it_ctl.cur_it;
3164	if (it && (it->vn != curvnet)) {
3165		/* Its not looking at our VNET */
3166		return;
3167	}
3168	if (it && (it->inp == inp)) {
3169		/*
3170		 * This is tricky and we hold the iterator lock, but when it
3171		 * returns and gets the lock (when we release it) the
3172		 * iterator will try to operate on inp. We need to stop that
3173		 * from happening. But of course the iterator has a
3174		 * reference on the stcb and inp. We can mark it and it will
3175		 * stop.
3176		 *
3177		 * If its a single iterator situation, we set the end iterator
3178		 * flag. Otherwise we set the iterator to go to the next
3179		 * inp.
3180		 *
3181		 */
3182		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3183			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3184		} else {
3185			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3186		}
3187	}
3188	/*
3189	 * Now go through and remove any single reference to our inp that
3190	 * may be still pending on the list
3191	 */
3192	SCTP_IPI_ITERATOR_WQ_LOCK();
3193	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3194		if (it->vn != curvnet) {
3195			continue;
3196		}
3197		if (it->inp == inp) {
3198			/* This one points to me is it inp specific? */
3199			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3200				/* Remove and free this one */
3201				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3202				    it, sctp_nxt_itr);
3203				if (it->function_atend != NULL) {
3204					(*it->function_atend) (it->pointer, it->val);
3205				}
3206				SCTP_FREE(it, SCTP_M_ITER);
3207			} else {
3208				it->inp = LIST_NEXT(it->inp, sctp_list);
3209				if (it->inp) {
3210					SCTP_INP_INCR_REF(it->inp);
3211				}
3212			}
3213			/*
3214			 * When its put in the refcnt is incremented so decr
3215			 * it
3216			 */
3217			SCTP_INP_DECR_REF(inp);
3218		}
3219	}
3220	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3221}
3222
3223/* release sctp_inpcb unbind the port */
3224void
3225sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3226{
3227	/*
3228	 * Here we free a endpoint. We must find it (if it is in the Hash
3229	 * table) and remove it from there. Then we must also find it in the
3230	 * overall list and remove it from there. After all removals are
3231	 * complete then any timer has to be stopped. Then start the actual
3232	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3233	 * all associations. d) finally the ep itself.
3234	 */
3235	struct sctp_tcb *asoc, *nasoc;
3236	struct sctp_laddr *laddr, *nladdr;
3237	struct inpcb *ip_pcb;
3238	struct socket *so;
3239	int being_refed = 0;
3240	struct sctp_queued_to_read *sq, *nsq;
3241	int cnt;
3242	sctp_sharedkey_t *shared_key, *nshared_key;
3243
3244
3245#ifdef SCTP_LOG_CLOSING
3246	sctp_log_closing(inp, NULL, 0);
3247#endif
3248	SCTP_ITERATOR_LOCK();
3249	/* mark any iterators on the list or being processed */
3250	sctp_iterator_inp_being_freed(inp);
3251	SCTP_ITERATOR_UNLOCK();
3252	so = inp->sctp_socket;
3253	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3254		/* been here before.. eeks.. get out of here */
3255		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3256#ifdef SCTP_LOG_CLOSING
3257		sctp_log_closing(inp, NULL, 1);
3258#endif
3259		return;
3260	}
3261	SCTP_ASOC_CREATE_LOCK(inp);
3262	SCTP_INP_INFO_WLOCK();
3263
3264	SCTP_INP_WLOCK(inp);
3265	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3266		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3267		/* socket is gone, so no more wakeups allowed */
3268		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3269		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3270		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3271
3272	}
3273	/* First time through we have the socket lock, after that no more. */
3274	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3275	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3276
3277	if (inp->control) {
3278		sctp_m_freem(inp->control);
3279		inp->control = NULL;
3280	}
3281	if (inp->pkt) {
3282		sctp_m_freem(inp->pkt);
3283		inp->pkt = NULL;
3284	}
3285	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3286					 * here but I will be nice :> (i.e.
3287					 * ip_pcb = ep;) */
3288	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3289		int cnt_in_sd;
3290
3291		cnt_in_sd = 0;
3292		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3293			SCTP_TCB_LOCK(asoc);
3294			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3295				/* Skip guys being freed */
3296				cnt_in_sd++;
3297				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3298					/*
3299					 * Special case - we did not start a
3300					 * kill timer on the asoc due to it
3301					 * was not closed. So go ahead and
3302					 * start it now.
3303					 */
3304					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3305					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3306				}
3307				SCTP_TCB_UNLOCK(asoc);
3308				continue;
3309			}
3310			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3311			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3312			    (asoc->asoc.total_output_queue_size == 0)) {
3313				/*
3314				 * If we have data in queue, we don't want
3315				 * to just free since the app may have done,
3316				 * send()/close or connect/send/close. And
3317				 * it wants the data to get across first.
3318				 */
3319				/* Just abandon things in the front states */
3320				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3321				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3322					cnt_in_sd++;
3323				}
3324				continue;
3325			}
3326			/* Disconnect the socket please */
3327			asoc->sctp_socket = NULL;
3328			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3329			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3330			    (asoc->asoc.control_pdapi) ||
3331			    (asoc->asoc.size_on_all_streams > 0) ||
3332			    (so && (so->so_rcv.sb_cc > 0))) {
3333				/* Left with Data unread */
3334				struct mbuf *op_err;
3335
3336				op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3337				    0, M_NOWAIT, 1, MT_DATA);
3338				if (op_err) {
3339					/* Fill in the user initiated abort */
3340					struct sctp_paramhdr *ph;
3341
3342					SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3343					ph = mtod(op_err, struct sctp_paramhdr *);
3344					ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3345					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3346				}
3347				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3348				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3349				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3350				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3351				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3352					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3353				}
3354				if (sctp_free_assoc(inp, asoc,
3355				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3356					cnt_in_sd++;
3357				}
3358				continue;
3359			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3360				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3361			    (asoc->asoc.stream_queue_cnt == 0)) {
3362				if (asoc->asoc.locked_on_sending) {
3363					goto abort_anyway;
3364				}
3365				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3366				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3367					struct sctp_nets *netp;
3368
3369					/*
3370					 * there is nothing queued to send,
3371					 * so I send shutdown
3372					 */
3373					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3374					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3375						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3376					}
3377					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3378					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3379					sctp_stop_timers_for_shutdown(asoc);
3380					if (asoc->asoc.alternate) {
3381						netp = asoc->asoc.alternate;
3382					} else {
3383						netp = asoc->asoc.primary_destination;
3384					}
3385					sctp_send_shutdown(asoc, netp);
3386					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3387					    netp);
3388					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3389					    asoc->asoc.primary_destination);
3390					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3391				}
3392			} else {
3393				/* mark into shutdown pending */
3394				struct sctp_stream_queue_pending *sp;
3395
3396				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3397				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3398				    asoc->asoc.primary_destination);
3399				if (asoc->asoc.locked_on_sending) {
3400					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3401					    sctp_streamhead);
3402					if (sp == NULL) {
3403						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3404						    (void *)asoc->asoc.locked_on_sending,
3405						    asoc->asoc.locked_on_sending->stream_no);
3406					} else {
3407						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3408							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3409					}
3410				}
3411				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3412				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3413				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3414					struct mbuf *op_err;
3415
3416			abort_anyway:
3417					op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3418					    0, M_NOWAIT, 1, MT_DATA);
3419					if (op_err) {
3420						/*
3421						 * Fill in the user
3422						 * initiated abort
3423						 */
3424						struct sctp_paramhdr *ph;
3425
3426						SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3427						ph = mtod(op_err, struct sctp_paramhdr *);
3428						ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3429						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3430					}
3431					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3432					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3433					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3434					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3435					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3436						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3437					}
3438					if (sctp_free_assoc(inp, asoc,
3439					    SCTP_PCBFREE_NOFORCE,
3440					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3441						cnt_in_sd++;
3442					}
3443					continue;
3444				} else {
3445					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3446				}
3447			}
3448			cnt_in_sd++;
3449			SCTP_TCB_UNLOCK(asoc);
3450		}
3451		/* now is there some left in our SHUTDOWN state? */
3452		if (cnt_in_sd) {
3453#ifdef SCTP_LOG_CLOSING
3454			sctp_log_closing(inp, NULL, 2);
3455#endif
3456			inp->sctp_socket = NULL;
3457			SCTP_INP_WUNLOCK(inp);
3458			SCTP_ASOC_CREATE_UNLOCK(inp);
3459			SCTP_INP_INFO_WUNLOCK();
3460			return;
3461		}
3462	}
3463	inp->sctp_socket = NULL;
3464	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3465	    SCTP_PCB_FLAGS_UNBOUND) {
3466		/*
3467		 * ok, this guy has been bound. It's port is somewhere in
3468		 * the SCTP_BASE_INFO(hash table). Remove it!
3469		 */
3470		LIST_REMOVE(inp, sctp_hash);
3471		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3472	}
3473	/*
3474	 * If there is a timer running to kill us, forget it, since it may
3475	 * have a contest on the INP lock.. which would cause us to die ...
3476	 */
3477	cnt = 0;
3478	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3479		SCTP_TCB_LOCK(asoc);
3480		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3481			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3482				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3483				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3484			}
3485			cnt++;
3486			SCTP_TCB_UNLOCK(asoc);
3487			continue;
3488		}
3489		/* Free associations that are NOT killing us */
3490		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3491		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3492			struct mbuf *op_err;
3493
3494			op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3495			    0, M_NOWAIT, 1, MT_DATA);
3496			if (op_err) {
3497				/* Fill in the user initiated abort */
3498				struct sctp_paramhdr *ph;
3499
3500				SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3501				ph = mtod(op_err, struct sctp_paramhdr *);
3502				ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3503				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3504			}
3505			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3506			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3507			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3508		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3509			cnt++;
3510			SCTP_TCB_UNLOCK(asoc);
3511			continue;
3512		}
3513		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3514		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3515			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3516		}
3517		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3518			cnt++;
3519		}
3520	}
3521	if (cnt) {
3522		/* Ok we have someone out there that will kill us */
3523		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3524#ifdef SCTP_LOG_CLOSING
3525		sctp_log_closing(inp, NULL, 3);
3526#endif
3527		SCTP_INP_WUNLOCK(inp);
3528		SCTP_ASOC_CREATE_UNLOCK(inp);
3529		SCTP_INP_INFO_WUNLOCK();
3530		return;
3531	}
3532	if (SCTP_INP_LOCK_CONTENDED(inp))
3533		being_refed++;
3534	if (SCTP_INP_READ_CONTENDED(inp))
3535		being_refed++;
3536	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3537		being_refed++;
3538
3539	if ((inp->refcount) ||
3540	    (being_refed) ||
3541	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3542		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3543#ifdef SCTP_LOG_CLOSING
3544		sctp_log_closing(inp, NULL, 4);
3545#endif
3546		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3547		SCTP_INP_WUNLOCK(inp);
3548		SCTP_ASOC_CREATE_UNLOCK(inp);
3549		SCTP_INP_INFO_WUNLOCK();
3550		return;
3551	}
3552	inp->sctp_ep.signature_change.type = 0;
3553	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3554	/*
3555	 * Remove it from the list .. last thing we need a lock for.
3556	 */
3557	LIST_REMOVE(inp, sctp_list);
3558	SCTP_INP_WUNLOCK(inp);
3559	SCTP_ASOC_CREATE_UNLOCK(inp);
3560	SCTP_INP_INFO_WUNLOCK();
3561	/*
3562	 * Now we release all locks. Since this INP cannot be found anymore
3563	 * except possibly by the kill timer that might be running. We call
3564	 * the drain function here. It should hit the case were it sees the
3565	 * ACTIVE flag cleared and exit out freeing us to proceed and
3566	 * destroy everything.
3567	 */
3568	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3569		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3570	} else {
3571		/* Probably un-needed */
3572		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3573	}
3574
3575#ifdef SCTP_LOG_CLOSING
3576	sctp_log_closing(inp, NULL, 5);
3577#endif
3578
3579
3580	if ((inp->sctp_asocidhash) != NULL) {
3581		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3582		inp->sctp_asocidhash = NULL;
3583	}
3584	/* sa_ignore FREED_MEMORY */
3585	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3586		/* Its only abandoned if it had data left */
3587		if (sq->length)
3588			SCTP_STAT_INCR(sctps_left_abandon);
3589
3590		TAILQ_REMOVE(&inp->read_queue, sq, next);
3591		sctp_free_remote_addr(sq->whoFrom);
3592		if (so)
3593			so->so_rcv.sb_cc -= sq->length;
3594		if (sq->data) {
3595			sctp_m_freem(sq->data);
3596			sq->data = NULL;
3597		}
3598		/*
3599		 * no need to free the net count, since at this point all
3600		 * assoc's are gone.
3601		 */
3602		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3603		SCTP_DECR_READQ_COUNT();
3604	}
3605	/* Now the sctp_pcb things */
3606	/*
3607	 * free each asoc if it is not already closed/free. we can't use the
3608	 * macro here since le_next will get freed as part of the
3609	 * sctp_free_assoc() call.
3610	 */
3611	if (so) {
3612#ifdef IPSEC
3613		ipsec_delete_pcbpolicy(ip_pcb);
3614#endif				/* IPSEC */
3615
3616		/* Unlocks not needed since the socket is gone now */
3617	}
3618	if (ip_pcb->inp_options) {
3619		(void)sctp_m_free(ip_pcb->inp_options);
3620		ip_pcb->inp_options = 0;
3621	}
3622#ifdef INET6
3623	if (ip_pcb->inp_vflag & INP_IPV6) {
3624		struct in6pcb *in6p;
3625
3626		in6p = (struct in6pcb *)inp;
3627		ip6_freepcbopts(in6p->in6p_outputopts);
3628	}
3629#endif				/* INET6 */
3630	ip_pcb->inp_vflag = 0;
3631	/* free up authentication fields */
3632	if (inp->sctp_ep.local_auth_chunks != NULL)
3633		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3634	if (inp->sctp_ep.local_hmacs != NULL)
3635		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3636
3637	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3638		LIST_REMOVE(shared_key, next);
3639		sctp_free_sharedkey(shared_key);
3640		/* sa_ignore FREED_MEMORY */
3641	}
3642
3643	/*
3644	 * if we have an address list the following will free the list of
3645	 * ifaddr's that are set into this ep. Again macro limitations here,
3646	 * since the LIST_FOREACH could be a bad idea.
3647	 */
3648	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3649		sctp_remove_laddr(laddr);
3650	}
3651
3652#ifdef SCTP_TRACK_FREED_ASOCS
3653	/* TEMP CODE */
3654	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3655		LIST_REMOVE(asoc, sctp_tcblist);
3656		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3657		SCTP_DECR_ASOC_COUNT();
3658	}
3659	/* *** END TEMP CODE *** */
3660#endif
3661	/* Now lets see about freeing the EP hash table. */
3662	if (inp->sctp_tcbhash != NULL) {
3663		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3664		inp->sctp_tcbhash = NULL;
3665	}
3666	/* Now we must put the ep memory back into the zone pool */
3667	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3668	SCTP_INP_LOCK_DESTROY(inp);
3669	SCTP_INP_READ_DESTROY(inp);
3670	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3671	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3672	SCTP_DECR_EP_COUNT();
3673}
3674
3675
3676struct sctp_nets *
3677sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3678{
3679	struct sctp_nets *net;
3680
3681	/* locate the address */
3682	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3683		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3684			return (net);
3685	}
3686	return (NULL);
3687}
3688
3689
3690int
3691sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3692{
3693	struct sctp_ifa *sctp_ifa;
3694
3695	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3696	if (sctp_ifa) {
3697		return (1);
3698	} else {
3699		return (0);
3700	}
3701}
3702
3703/*
3704 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3705 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3706 * stats of stuff.
3707 */
3708int
3709sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3710    struct sctp_nets **netp, int set_scope, int from)
3711{
3712	/*
3713	 * The following is redundant to the same lines in the
3714	 * sctp_aloc_assoc() but is needed since others call the add address
3715	 * function
3716	 */
3717	struct sctp_nets *net, *netfirst;
3718	int addr_inscope;
3719
3720	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3721	    from);
3722	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3723
3724	netfirst = sctp_findnet(stcb, newaddr);
3725	if (netfirst) {
3726		/*
3727		 * Lie and return ok, we don't want to make the association
3728		 * go away for this behavior. It will happen in the TCP
3729		 * model in a connected socket. It does not reach the hash
3730		 * table until after the association is built so it can't be
3731		 * found. Mark as reachable, since the initial creation will
3732		 * have been cleared and the NOT_IN_ASSOC flag will have
3733		 * been added... and we don't want to end up removing it
3734		 * back out.
3735		 */
3736		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3737			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3738			    SCTP_ADDR_UNCONFIRMED);
3739		} else {
3740			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3741		}
3742
3743		return (0);
3744	}
3745	addr_inscope = 1;
3746	switch (newaddr->sa_family) {
3747#ifdef INET
3748	case AF_INET:
3749		{
3750			struct sockaddr_in *sin;
3751
3752			sin = (struct sockaddr_in *)newaddr;
3753			if (sin->sin_addr.s_addr == 0) {
3754				/* Invalid address */
3755				return (-1);
3756			}
3757			/* zero out the bzero area */
3758			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3759
3760			/* assure len is set */
3761			sin->sin_len = sizeof(struct sockaddr_in);
3762			if (set_scope) {
3763#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3764				stcb->asoc.scope.ipv4_local_scope = 1;
3765#else
3766				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3767					stcb->asoc.scope.ipv4_local_scope = 1;
3768				}
3769#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3770			} else {
3771				/* Validate the address is in scope */
3772				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3773				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3774					addr_inscope = 0;
3775				}
3776			}
3777			break;
3778		}
3779#endif
3780#ifdef INET6
3781	case AF_INET6:
3782		{
3783			struct sockaddr_in6 *sin6;
3784
3785			sin6 = (struct sockaddr_in6 *)newaddr;
3786			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3787				/* Invalid address */
3788				return (-1);
3789			}
3790			/* assure len is set */
3791			sin6->sin6_len = sizeof(struct sockaddr_in6);
3792			if (set_scope) {
3793				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3794					stcb->asoc.scope.loopback_scope = 1;
3795					stcb->asoc.scope.local_scope = 0;
3796					stcb->asoc.scope.ipv4_local_scope = 1;
3797					stcb->asoc.scope.site_scope = 1;
3798				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3799					/*
3800					 * If the new destination is a
3801					 * LINK_LOCAL we must have common
3802					 * site scope. Don't set the local
3803					 * scope since we may not share all
3804					 * links, only loopback can do this.
3805					 * Links on the local network would
3806					 * also be on our private network
3807					 * for v4 too.
3808					 */
3809					stcb->asoc.scope.ipv4_local_scope = 1;
3810					stcb->asoc.scope.site_scope = 1;
3811				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3812					/*
3813					 * If the new destination is
3814					 * SITE_LOCAL then we must have site
3815					 * scope in common.
3816					 */
3817					stcb->asoc.scope.site_scope = 1;
3818				}
3819			} else {
3820				/* Validate the address is in scope */
3821				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3822				    (stcb->asoc.scope.loopback_scope == 0)) {
3823					addr_inscope = 0;
3824				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3825				    (stcb->asoc.scope.local_scope == 0)) {
3826					addr_inscope = 0;
3827				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3828				    (stcb->asoc.scope.site_scope == 0)) {
3829					addr_inscope = 0;
3830				}
3831			}
3832			break;
3833		}
3834#endif
3835	default:
3836		/* not supported family type */
3837		return (-1);
3838	}
3839	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3840	if (net == NULL) {
3841		return (-1);
3842	}
3843	SCTP_INCR_RADDR_COUNT();
3844	bzero(net, sizeof(struct sctp_nets));
3845	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3846	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3847	switch (newaddr->sa_family) {
3848#ifdef INET
3849	case AF_INET:
3850		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3851		break;
3852#endif
3853#ifdef INET6
3854	case AF_INET6:
3855		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3856		break;
3857#endif
3858	default:
3859		break;
3860	}
3861	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3862	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3863		stcb->asoc.scope.loopback_scope = 1;
3864		stcb->asoc.scope.ipv4_local_scope = 1;
3865		stcb->asoc.scope.local_scope = 0;
3866		stcb->asoc.scope.site_scope = 1;
3867		addr_inscope = 1;
3868	}
3869	net->failure_threshold = stcb->asoc.def_net_failure;
3870	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3871	if (addr_inscope == 0) {
3872		net->dest_state = (SCTP_ADDR_REACHABLE |
3873		    SCTP_ADDR_OUT_OF_SCOPE);
3874	} else {
3875		if (from == SCTP_ADDR_IS_CONFIRMED)
3876			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3877			net->dest_state = SCTP_ADDR_REACHABLE;
3878		else
3879			net->dest_state = SCTP_ADDR_REACHABLE |
3880			    SCTP_ADDR_UNCONFIRMED;
3881	}
3882	/*
3883	 * We set this to 0, the timer code knows that this means its an
3884	 * initial value
3885	 */
3886	net->rto_needed = 1;
3887	net->RTO = 0;
3888	net->RTO_measured = 0;
3889	stcb->asoc.numnets++;
3890	net->ref_count = 1;
3891	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3892	net->port = stcb->asoc.port;
3893	net->dscp = stcb->asoc.default_dscp;
3894#ifdef INET6
3895	net->flowlabel = stcb->asoc.default_flowlabel;
3896#endif
3897	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3898		net->dest_state |= SCTP_ADDR_NOHB;
3899	} else {
3900		net->dest_state &= ~SCTP_ADDR_NOHB;
3901	}
3902	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3903		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3904	} else {
3905		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3906	}
3907	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3908	/* Init the timer structure */
3909	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3910	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3911	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3912
3913	/* Now generate a route for this guy */
3914#ifdef INET6
3915	/* KAME hack: embed scopeid */
3916	if (newaddr->sa_family == AF_INET6) {
3917		struct sockaddr_in6 *sin6;
3918
3919		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3920		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3921		sin6->sin6_scope_id = 0;
3922	}
3923#endif
3924	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3925
3926	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3927		/* Get source address */
3928		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3929		    stcb,
3930		    (sctp_route_t *) & net->ro,
3931		    net,
3932		    0,
3933		    stcb->asoc.vrf_id);
3934		/* Now get the interface MTU */
3935		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3936			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3937		}
3938		if (net->mtu > 0) {
3939			uint32_t rmtu;
3940
3941			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3942			if (rmtu == 0) {
3943				/*
3944				 * Start things off to match mtu of
3945				 * interface please.
3946				 */
3947				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3948				    net->ro.ro_rt, net->mtu);
3949			} else {
3950				/*
3951				 * we take the route mtu over the interface,
3952				 * since the route may be leading out the
3953				 * loopback, or a different interface.
3954				 */
3955				net->mtu = rmtu;
3956			}
3957		}
3958	}
3959	if (net->mtu == 0) {
3960		switch (newaddr->sa_family) {
3961#ifdef INET
3962		case AF_INET:
3963			net->mtu = SCTP_DEFAULT_MTU;
3964			break;
3965#endif
3966#ifdef INET6
3967		case AF_INET6:
3968			net->mtu = 1280;
3969			break;
3970#endif
3971		default:
3972			break;
3973		}
3974	}
3975	if (net->port) {
3976		net->mtu -= (uint32_t) sizeof(struct udphdr);
3977	}
3978	if (from == SCTP_ALLOC_ASOC) {
3979		stcb->asoc.smallest_mtu = net->mtu;
3980	}
3981	if (stcb->asoc.smallest_mtu > net->mtu) {
3982		stcb->asoc.smallest_mtu = net->mtu;
3983	}
3984#ifdef INET6
3985	if (newaddr->sa_family == AF_INET6) {
3986		struct sockaddr_in6 *sin6;
3987
3988		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3989		(void)sa6_recoverscope(sin6);
3990	}
3991#endif
3992
3993	/* JRS - Use the congestion control given in the CC module */
3994	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
3995		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
3996
3997	/*
3998	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3999	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4000	 */
4001	net->find_pseudo_cumack = 1;
4002	net->find_rtx_pseudo_cumack = 1;
4003	net->src_addr_selected = 0;
4004	/* Choose an initial flowid. */
4005	net->flowid = stcb->asoc.my_vtag ^
4006	    ntohs(stcb->rport) ^
4007	    ntohs(stcb->sctp_ep->sctp_lport);
4008#ifdef INVARIANTS
4009	net->flowidset = 1;
4010#endif
4011	if (netp) {
4012		*netp = net;
4013	}
4014	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4015	if (net->ro.ro_rt == NULL) {
4016		/* Since we have no route put it at the back */
4017		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4018	} else if (netfirst == NULL) {
4019		/* We are the first one in the pool. */
4020		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4021	} else if (netfirst->ro.ro_rt == NULL) {
4022		/*
4023		 * First one has NO route. Place this one ahead of the first
4024		 * one.
4025		 */
4026		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4027	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4028		/*
4029		 * This one has a different interface than the one at the
4030		 * top of the list. Place it ahead.
4031		 */
4032		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4033	} else {
4034		/*
4035		 * Ok we have the same interface as the first one. Move
4036		 * forward until we find either a) one with a NULL route...
4037		 * insert ahead of that b) one with a different ifp.. insert
4038		 * after that. c) end of the list.. insert at the tail.
4039		 */
4040		struct sctp_nets *netlook;
4041
4042		do {
4043			netlook = TAILQ_NEXT(netfirst, sctp_next);
4044			if (netlook == NULL) {
4045				/* End of the list */
4046				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4047				break;
4048			} else if (netlook->ro.ro_rt == NULL) {
4049				/* next one has NO route */
4050				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4051				break;
4052			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4053				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4054				    net, sctp_next);
4055				break;
4056			}
4057			/* Shift forward */
4058			netfirst = netlook;
4059		} while (netlook != NULL);
4060	}
4061
4062	/* got to have a primary set */
4063	if (stcb->asoc.primary_destination == 0) {
4064		stcb->asoc.primary_destination = net;
4065	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4066		    (net->ro.ro_rt) &&
4067	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4068		/* No route to current primary adopt new primary */
4069		stcb->asoc.primary_destination = net;
4070	}
4071	/* Validate primary is first */
4072	net = TAILQ_FIRST(&stcb->asoc.nets);
4073	if ((net != stcb->asoc.primary_destination) &&
4074	    (stcb->asoc.primary_destination)) {
4075		/*
4076		 * first one on the list is NOT the primary sctp_cmpaddr()
4077		 * is much more efficient if the primary is the first on the
4078		 * list, make it so.
4079		 */
4080		TAILQ_REMOVE(&stcb->asoc.nets,
4081		    stcb->asoc.primary_destination, sctp_next);
4082		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4083		    stcb->asoc.primary_destination, sctp_next);
4084	}
4085	return (0);
4086}
4087
4088
4089static uint32_t
4090sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4091{
4092	uint32_t id;
4093	struct sctpasochead *head;
4094	struct sctp_tcb *lstcb;
4095
4096	SCTP_INP_WLOCK(inp);
4097try_again:
4098	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4099		/* TSNH */
4100		SCTP_INP_WUNLOCK(inp);
4101		return (0);
4102	}
4103	/*
4104	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4105	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4106	 */
4107	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4108		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4109	}
4110	id = inp->sctp_associd_counter;
4111	inp->sctp_associd_counter++;
4112	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4113	if (lstcb) {
4114		goto try_again;
4115	}
4116	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4117	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4118	stcb->asoc.in_asocid_hash = 1;
4119	SCTP_INP_WUNLOCK(inp);
4120	return id;
4121}
4122
4123/*
4124 * allocate an association and add it to the endpoint. The caller must be
4125 * careful to add all additional addresses once they are know right away or
4126 * else the assoc will be may experience a blackout scenario.
4127 */
4128struct sctp_tcb *
4129sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4130    int *error, uint32_t override_tag, uint32_t vrf_id,
4131    struct thread *p
4132)
4133{
4134	/* note the p argument is only valid in unbound sockets */
4135
4136	struct sctp_tcb *stcb;
4137	struct sctp_association *asoc;
4138	struct sctpasochead *head;
4139	uint16_t rport;
4140	int err;
4141
4142	/*
4143	 * Assumption made here: Caller has done a
4144	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4145	 * address does not exist already.
4146	 */
4147	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4148		/* Hit max assoc, sorry no more */
4149		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4150		*error = ENOBUFS;
4151		return (NULL);
4152	}
4153	if (firstaddr == NULL) {
4154		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4155		*error = EINVAL;
4156		return (NULL);
4157	}
4158	SCTP_INP_RLOCK(inp);
4159	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4160	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4161	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4162		/*
4163		 * If its in the TCP pool, its NOT allowed to create an
4164		 * association. The parent listener needs to call
4165		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4166		 * off, or connected one does this.. its an error.
4167		 */
4168		SCTP_INP_RUNLOCK(inp);
4169		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4170		*error = EINVAL;
4171		return (NULL);
4172	}
4173	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4174	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4175		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4176		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4177			SCTP_INP_RUNLOCK(inp);
4178			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4179			*error = EINVAL;
4180			return (NULL);
4181		}
4182	}
4183	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4184#ifdef SCTP_DEBUG
4185	if (firstaddr) {
4186		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4187		switch (firstaddr->sa_family) {
4188#ifdef INET
4189		case AF_INET:
4190			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4191			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4192			break;
4193#endif
4194#ifdef INET6
4195		case AF_INET6:
4196			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4197			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4198			break;
4199#endif
4200		default:
4201			break;
4202		}
4203	} else {
4204		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4205	}
4206#endif				/* SCTP_DEBUG */
4207	switch (firstaddr->sa_family) {
4208#ifdef INET
4209	case AF_INET:
4210		{
4211			struct sockaddr_in *sin;
4212
4213			sin = (struct sockaddr_in *)firstaddr;
4214			if ((ntohs(sin->sin_port) == 0) ||
4215			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4216			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4217			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4218				/* Invalid address */
4219				SCTP_INP_RUNLOCK(inp);
4220				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4221				*error = EINVAL;
4222				return (NULL);
4223			}
4224			rport = sin->sin_port;
4225			break;
4226		}
4227#endif
4228#ifdef INET6
4229	case AF_INET6:
4230		{
4231			struct sockaddr_in6 *sin6;
4232
4233			sin6 = (struct sockaddr_in6 *)firstaddr;
4234			if ((ntohs(sin6->sin6_port) == 0) ||
4235			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4236			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4237				/* Invalid address */
4238				SCTP_INP_RUNLOCK(inp);
4239				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4240				*error = EINVAL;
4241				return (NULL);
4242			}
4243			rport = sin6->sin6_port;
4244			break;
4245		}
4246#endif
4247	default:
4248		/* not supported family type */
4249		SCTP_INP_RUNLOCK(inp);
4250		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4251		*error = EINVAL;
4252		return (NULL);
4253	}
4254	SCTP_INP_RUNLOCK(inp);
4255	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4256		/*
4257		 * If you have not performed a bind, then we need to do the
4258		 * ephemeral bind for you.
4259		 */
4260		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4261		    (struct sockaddr *)NULL,
4262		    (struct sctp_ifa *)NULL,
4263		    p
4264		    ))) {
4265			/* bind error, probably perm */
4266			*error = err;
4267			return (NULL);
4268		}
4269	}
4270	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4271	if (stcb == NULL) {
4272		/* out of memory? */
4273		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4274		*error = ENOMEM;
4275		return (NULL);
4276	}
4277	SCTP_INCR_ASOC_COUNT();
4278
4279	bzero(stcb, sizeof(*stcb));
4280	asoc = &stcb->asoc;
4281
4282	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4283	SCTP_TCB_LOCK_INIT(stcb);
4284	SCTP_TCB_SEND_LOCK_INIT(stcb);
4285	stcb->rport = rport;
4286	/* setup back pointer's */
4287	stcb->sctp_ep = inp;
4288	stcb->sctp_socket = inp->sctp_socket;
4289	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4290		/* failed */
4291		SCTP_TCB_LOCK_DESTROY(stcb);
4292		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4293		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4294		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4295		SCTP_DECR_ASOC_COUNT();
4296		*error = err;
4297		return (NULL);
4298	}
4299	/* and the port */
4300	SCTP_INP_INFO_WLOCK();
4301	SCTP_INP_WLOCK(inp);
4302	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4303		/* inpcb freed while alloc going on */
4304		SCTP_TCB_LOCK_DESTROY(stcb);
4305		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4306		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4307		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4308		SCTP_INP_WUNLOCK(inp);
4309		SCTP_INP_INFO_WUNLOCK();
4310		SCTP_DECR_ASOC_COUNT();
4311		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4312		*error = EINVAL;
4313		return (NULL);
4314	}
4315	SCTP_TCB_LOCK(stcb);
4316
4317	/* now that my_vtag is set, add it to the hash */
4318	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4319	/* put it in the bucket in the vtag hash of assoc's for the system */
4320	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4321	SCTP_INP_INFO_WUNLOCK();
4322
4323	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4324		/* failure.. memory error? */
4325		if (asoc->strmout) {
4326			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4327			asoc->strmout = NULL;
4328		}
4329		if (asoc->mapping_array) {
4330			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4331			asoc->mapping_array = NULL;
4332		}
4333		if (asoc->nr_mapping_array) {
4334			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4335			asoc->nr_mapping_array = NULL;
4336		}
4337		SCTP_DECR_ASOC_COUNT();
4338		SCTP_TCB_UNLOCK(stcb);
4339		SCTP_TCB_LOCK_DESTROY(stcb);
4340		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4341		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4342		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4343		SCTP_INP_WUNLOCK(inp);
4344		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4345		*error = ENOBUFS;
4346		return (NULL);
4347	}
4348	/* Init all the timers */
4349	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4350	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4351	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4352	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4353	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4354	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4355	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4356
4357	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4358	/* now file the port under the hash as well */
4359	if (inp->sctp_tcbhash != NULL) {
4360		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4361		    inp->sctp_hashmark)];
4362		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4363	}
4364	SCTP_INP_WUNLOCK(inp);
4365	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4366	return (stcb);
4367}
4368
4369
4370void
4371sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4372{
4373	struct sctp_association *asoc;
4374
4375	asoc = &stcb->asoc;
4376	asoc->numnets--;
4377	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4378	if (net == asoc->primary_destination) {
4379		/* Reset primary */
4380		struct sctp_nets *lnet;
4381
4382		lnet = TAILQ_FIRST(&asoc->nets);
4383		/*
4384		 * Mobility adaptation Ideally, if deleted destination is
4385		 * the primary, it becomes a fast retransmission trigger by
4386		 * the subsequent SET PRIMARY. (by micchie)
4387		 */
4388		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4389		    SCTP_MOBILITY_BASE) ||
4390		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4391		    SCTP_MOBILITY_FASTHANDOFF)) {
4392			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4393			if (asoc->deleted_primary != NULL) {
4394				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4395				goto out;
4396			}
4397			asoc->deleted_primary = net;
4398			atomic_add_int(&net->ref_count, 1);
4399			memset(&net->lastsa, 0, sizeof(net->lastsa));
4400			memset(&net->lastsv, 0, sizeof(net->lastsv));
4401			sctp_mobility_feature_on(stcb->sctp_ep,
4402			    SCTP_MOBILITY_PRIM_DELETED);
4403			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4404			    stcb->sctp_ep, stcb, NULL);
4405		}
4406out:
4407		/* Try to find a confirmed primary */
4408		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4409	}
4410	if (net == asoc->last_data_chunk_from) {
4411		/* Reset primary */
4412		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4413	}
4414	if (net == asoc->last_control_chunk_from) {
4415		/* Clear net */
4416		asoc->last_control_chunk_from = NULL;
4417	}
4418	if (net == stcb->asoc.alternate) {
4419		sctp_free_remote_addr(stcb->asoc.alternate);
4420		stcb->asoc.alternate = NULL;
4421	}
4422	sctp_free_remote_addr(net);
4423}
4424
4425/*
4426 * remove a remote endpoint address from an association, it will fail if the
4427 * address does not exist.
4428 */
4429int
4430sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4431{
4432	/*
4433	 * Here we need to remove a remote address. This is quite simple, we
4434	 * first find it in the list of address for the association
4435	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4436	 * on that item. Note we do not allow it to be removed if there are
4437	 * no other addresses.
4438	 */
4439	struct sctp_association *asoc;
4440	struct sctp_nets *net, *nnet;
4441
4442	asoc = &stcb->asoc;
4443
4444	/* locate the address */
4445	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4446		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4447			continue;
4448		}
4449		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4450		    remaddr)) {
4451			/* we found the guy */
4452			if (asoc->numnets < 2) {
4453				/* Must have at LEAST two remote addresses */
4454				return (-1);
4455			} else {
4456				sctp_remove_net(stcb, net);
4457				return (0);
4458			}
4459		}
4460	}
4461	/* not found. */
4462	return (-2);
4463}
4464
4465void
4466sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4467{
4468	struct sctpvtaghead *chain;
4469	struct sctp_tagblock *twait_block;
4470	int found = 0;
4471	int i;
4472
4473	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4474	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4475		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4476			if ((twait_block->vtag_block[i].v_tag == tag) &&
4477			    (twait_block->vtag_block[i].lport == lport) &&
4478			    (twait_block->vtag_block[i].rport == rport)) {
4479				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4480				twait_block->vtag_block[i].v_tag = 0;
4481				twait_block->vtag_block[i].lport = 0;
4482				twait_block->vtag_block[i].rport = 0;
4483				found = 1;
4484				break;
4485			}
4486		}
4487		if (found)
4488			break;
4489	}
4490}
4491
4492int
4493sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4494{
4495	struct sctpvtaghead *chain;
4496	struct sctp_tagblock *twait_block;
4497	int found = 0;
4498	int i;
4499
4500	SCTP_INP_INFO_WLOCK();
4501	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4502	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4503		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4504			if ((twait_block->vtag_block[i].v_tag == tag) &&
4505			    (twait_block->vtag_block[i].lport == lport) &&
4506			    (twait_block->vtag_block[i].rport == rport)) {
4507				found = 1;
4508				break;
4509			}
4510		}
4511		if (found)
4512			break;
4513	}
4514	SCTP_INP_INFO_WUNLOCK();
4515	return (found);
4516}
4517
4518
4519void
4520sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4521{
4522	struct sctpvtaghead *chain;
4523	struct sctp_tagblock *twait_block;
4524	struct timeval now;
4525	int set, i;
4526
4527	if (time == 0) {
4528		/* Its disabled */
4529		return;
4530	}
4531	(void)SCTP_GETTIME_TIMEVAL(&now);
4532	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4533	set = 0;
4534	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4535		/* Block(s) present, lets find space, and expire on the fly */
4536		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4537			if ((twait_block->vtag_block[i].v_tag == 0) &&
4538			    !set) {
4539				twait_block->vtag_block[i].tv_sec_at_expire =
4540				    now.tv_sec + time;
4541				twait_block->vtag_block[i].v_tag = tag;
4542				twait_block->vtag_block[i].lport = lport;
4543				twait_block->vtag_block[i].rport = rport;
4544				set = 1;
4545			} else if ((twait_block->vtag_block[i].v_tag) &&
4546			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4547				/* Audit expires this guy */
4548				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4549				twait_block->vtag_block[i].v_tag = 0;
4550				twait_block->vtag_block[i].lport = 0;
4551				twait_block->vtag_block[i].rport = 0;
4552				if (set == 0) {
4553					/* Reuse it for my new tag */
4554					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4555					twait_block->vtag_block[i].v_tag = tag;
4556					twait_block->vtag_block[i].lport = lport;
4557					twait_block->vtag_block[i].rport = rport;
4558					set = 1;
4559				}
4560			}
4561		}
4562		if (set) {
4563			/*
4564			 * We only do up to the block where we can place our
4565			 * tag for audits
4566			 */
4567			break;
4568		}
4569	}
4570	/* Need to add a new block to chain */
4571	if (!set) {
4572		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4573		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4574		if (twait_block == NULL) {
4575#ifdef INVARIANTS
4576			panic("Can not alloc tagblock");
4577#endif
4578			return;
4579		}
4580		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4581		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4582		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4583		twait_block->vtag_block[0].v_tag = tag;
4584		twait_block->vtag_block[0].lport = lport;
4585		twait_block->vtag_block[0].rport = rport;
4586	}
4587}
4588
4589
4590
4591/*-
4592 * Free the association after un-hashing the remote port. This
4593 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4594 * expect that the input to this function IS a locked TCB.
4595 * It will return 0, if it did NOT destroy the association (instead
4596 * it unlocks it. It will return NON-zero if it either destroyed the
4597 * association OR the association is already destroyed.
4598 */
4599int
4600sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4601{
4602	int i;
4603	struct sctp_association *asoc;
4604	struct sctp_nets *net, *nnet;
4605	struct sctp_laddr *laddr, *naddr;
4606	struct sctp_tmit_chunk *chk, *nchk;
4607	struct sctp_asconf_addr *aparam, *naparam;
4608	struct sctp_asconf_ack *aack, *naack;
4609	struct sctp_stream_reset_list *strrst, *nstrrst;
4610	struct sctp_queued_to_read *sq, *nsq;
4611	struct sctp_stream_queue_pending *sp, *nsp;
4612	sctp_sharedkey_t *shared_key, *nshared_key;
4613	struct socket *so;
4614
4615	/* first, lets purge the entry from the hash table. */
4616
4617#ifdef SCTP_LOG_CLOSING
4618	sctp_log_closing(inp, stcb, 6);
4619#endif
4620	if (stcb->asoc.state == 0) {
4621#ifdef SCTP_LOG_CLOSING
4622		sctp_log_closing(inp, NULL, 7);
4623#endif
4624		/* there is no asoc, really TSNH :-0 */
4625		return (1);
4626	}
4627	if (stcb->asoc.alternate) {
4628		sctp_free_remote_addr(stcb->asoc.alternate);
4629		stcb->asoc.alternate = NULL;
4630	}
4631	/* TEMP CODE */
4632	if (stcb->freed_from_where == 0) {
4633		/* Only record the first place free happened from */
4634		stcb->freed_from_where = from_location;
4635	}
4636	/* TEMP CODE */
4637
4638	asoc = &stcb->asoc;
4639	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4640	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4641		/* nothing around */
4642		so = NULL;
4643	else
4644		so = inp->sctp_socket;
4645
4646	/*
4647	 * We used timer based freeing if a reader or writer is in the way.
4648	 * So we first check if we are actually being called from a timer,
4649	 * if so we abort early if a reader or writer is still in the way.
4650	 */
4651	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4652	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4653		/*
4654		 * is it the timer driving us? if so are the reader/writers
4655		 * gone?
4656		 */
4657		if (stcb->asoc.refcnt) {
4658			/* nope, reader or writer in the way */
4659			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4660			/* no asoc destroyed */
4661			SCTP_TCB_UNLOCK(stcb);
4662#ifdef SCTP_LOG_CLOSING
4663			sctp_log_closing(inp, stcb, 8);
4664#endif
4665			return (0);
4666		}
4667	}
4668	/* now clean up any other timers */
4669	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4670	asoc->dack_timer.self = NULL;
4671	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4672	/*-
4673	 * For stream reset we don't blast this unless
4674	 * it is a str-reset timer, it might be the
4675	 * free-asoc timer which we DON'T want to
4676	 * disturb.
4677	 */
4678	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4679		asoc->strreset_timer.self = NULL;
4680	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4681	asoc->asconf_timer.self = NULL;
4682	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4683	asoc->autoclose_timer.self = NULL;
4684	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4685	asoc->shut_guard_timer.self = NULL;
4686	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4687	asoc->delayed_event_timer.self = NULL;
4688	/* Mobility adaptation */
4689	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4690	asoc->delete_prim_timer.self = NULL;
4691	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4692		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4693		net->rxt_timer.self = NULL;
4694		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4695		net->pmtu_timer.self = NULL;
4696		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4697		net->hb_timer.self = NULL;
4698	}
4699	/* Now the read queue needs to be cleaned up (only once) */
4700	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4701		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4702		SCTP_INP_READ_LOCK(inp);
4703		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4704			if (sq->stcb == stcb) {
4705				sq->do_not_ref_stcb = 1;
4706				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4707				/*
4708				 * If there is no end, there never will be
4709				 * now.
4710				 */
4711				if (sq->end_added == 0) {
4712					/* Held for PD-API clear that. */
4713					sq->pdapi_aborted = 1;
4714					sq->held_length = 0;
4715					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4716						/*
4717						 * Need to add a PD-API
4718						 * aborted indication.
4719						 * Setting the control_pdapi
4720						 * assures that it will be
4721						 * added right after this
4722						 * msg.
4723						 */
4724						uint32_t strseq;
4725
4726						stcb->asoc.control_pdapi = sq;
4727						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4728						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4729						    stcb,
4730						    SCTP_PARTIAL_DELIVERY_ABORTED,
4731						    (void *)&strseq,
4732						    SCTP_SO_LOCKED);
4733						stcb->asoc.control_pdapi = NULL;
4734					}
4735				}
4736				/* Add an end to wake them */
4737				sq->end_added = 1;
4738			}
4739		}
4740		SCTP_INP_READ_UNLOCK(inp);
4741		if (stcb->block_entry) {
4742			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4743			stcb->block_entry->error = ECONNRESET;
4744			stcb->block_entry = NULL;
4745		}
4746	}
4747	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4748		/*
4749		 * Someone holds a reference OR the socket is unaccepted
4750		 * yet.
4751		 */
4752		if ((stcb->asoc.refcnt) ||
4753		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4754		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4755			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4756			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4757		}
4758		SCTP_TCB_UNLOCK(stcb);
4759		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4760		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4761			/* nothing around */
4762			so = NULL;
4763		if (so) {
4764			/* Wake any reader/writers */
4765			sctp_sorwakeup(inp, so);
4766			sctp_sowwakeup(inp, so);
4767		}
4768#ifdef SCTP_LOG_CLOSING
4769		sctp_log_closing(inp, stcb, 9);
4770#endif
4771		/* no asoc destroyed */
4772		return (0);
4773	}
4774#ifdef SCTP_LOG_CLOSING
4775	sctp_log_closing(inp, stcb, 10);
4776#endif
4777	/*
4778	 * When I reach here, no others want to kill the assoc yet.. and I
4779	 * own the lock. Now its possible an abort comes in when I do the
4780	 * lock exchange below to grab all the locks to do the final take
4781	 * out. to prevent this we increment the count, which will start a
4782	 * timer and blow out above thus assuring us that we hold exclusive
4783	 * killing of the asoc. Note that after getting back the TCB lock we
4784	 * will go ahead and increment the counter back up and stop any
4785	 * timer a passing stranger may have started :-S
4786	 */
4787	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4788		atomic_add_int(&stcb->asoc.refcnt, 1);
4789
4790		SCTP_TCB_UNLOCK(stcb);
4791		SCTP_INP_INFO_WLOCK();
4792		SCTP_INP_WLOCK(inp);
4793		SCTP_TCB_LOCK(stcb);
4794	}
4795	/* Double check the GONE flag */
4796	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4797	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4798		/* nothing around */
4799		so = NULL;
4800
4801	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4802	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4803		/*
4804		 * For TCP type we need special handling when we are
4805		 * connected. We also include the peel'ed off ones to.
4806		 */
4807		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4808			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4809			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4810			if (so) {
4811				SOCK_LOCK(so);
4812				if (so->so_rcv.sb_cc == 0) {
4813					so->so_state &= ~(SS_ISCONNECTING |
4814					    SS_ISDISCONNECTING |
4815					    SS_ISCONFIRMING |
4816					    SS_ISCONNECTED);
4817				}
4818				socantrcvmore_locked(so);
4819				sctp_sowwakeup(inp, so);
4820				sctp_sorwakeup(inp, so);
4821				SCTP_SOWAKEUP(so);
4822			}
4823		}
4824	}
4825	/*
4826	 * Make it invalid too, that way if its about to run it will abort
4827	 * and return.
4828	 */
4829	/* re-increment the lock */
4830	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4831		atomic_add_int(&stcb->asoc.refcnt, -1);
4832	}
4833	if (stcb->asoc.refcnt) {
4834		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4835		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4836		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4837			SCTP_INP_INFO_WUNLOCK();
4838			SCTP_INP_WUNLOCK(inp);
4839		}
4840		SCTP_TCB_UNLOCK(stcb);
4841		return (0);
4842	}
4843	asoc->state = 0;
4844	if (inp->sctp_tcbhash) {
4845		LIST_REMOVE(stcb, sctp_tcbhash);
4846	}
4847	if (stcb->asoc.in_asocid_hash) {
4848		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4849	}
4850	/* Now lets remove it from the list of ALL associations in the EP */
4851	LIST_REMOVE(stcb, sctp_tcblist);
4852	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4853		SCTP_INP_INCR_REF(inp);
4854		SCTP_INP_WUNLOCK(inp);
4855	}
4856	/* pull from vtag hash */
4857	LIST_REMOVE(stcb, sctp_asocs);
4858	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4859	    inp->sctp_lport, stcb->rport);
4860
4861	/*
4862	 * Now restop the timers to be sure this is paranoia at is finest!
4863	 */
4864	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4865	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4866	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4867	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4868	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4869	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4870	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4871	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4872		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4873		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4874		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4875	}
4876
4877	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4878	/*
4879	 * The chunk lists and such SHOULD be empty but we check them just
4880	 * in case.
4881	 */
4882	/* anything on the wheel needs to be removed */
4883	for (i = 0; i < asoc->streamoutcnt; i++) {
4884		struct sctp_stream_out *outs;
4885
4886		outs = &asoc->strmout[i];
4887		/* now clean up any chunks here */
4888		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4889			TAILQ_REMOVE(&outs->outqueue, sp, next);
4890			sctp_free_spbufspace(stcb, asoc, sp);
4891			if (sp->data) {
4892				if (so) {
4893					/* Still an open socket - report */
4894					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4895					    0, (void *)sp, SCTP_SO_LOCKED);
4896				}
4897				if (sp->data) {
4898					sctp_m_freem(sp->data);
4899					sp->data = NULL;
4900					sp->tail_mbuf = NULL;
4901					sp->length = 0;
4902				}
4903			}
4904			if (sp->net) {
4905				sctp_free_remote_addr(sp->net);
4906				sp->net = NULL;
4907			}
4908			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4909		}
4910	}
4911	/* sa_ignore FREED_MEMORY */
4912	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4913		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4914		SCTP_FREE(strrst, SCTP_M_STRESET);
4915	}
4916	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4917		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4918		if (sq->data) {
4919			sctp_m_freem(sq->data);
4920			sq->data = NULL;
4921		}
4922		sctp_free_remote_addr(sq->whoFrom);
4923		sq->whoFrom = NULL;
4924		sq->stcb = NULL;
4925		/* Free the ctl entry */
4926		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4927		SCTP_DECR_READQ_COUNT();
4928		/* sa_ignore FREED_MEMORY */
4929	}
4930	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4931		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4932		if (chk->data) {
4933			sctp_m_freem(chk->data);
4934			chk->data = NULL;
4935		}
4936		if (chk->holds_key_ref)
4937			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4938		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4939		SCTP_DECR_CHK_COUNT();
4940		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4941		asoc->free_chunk_cnt--;
4942		/* sa_ignore FREED_MEMORY */
4943	}
4944	/* pending send queue SHOULD be empty */
4945	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4946		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4947			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4948#ifdef INVARIANTS
4949		} else {
4950			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4951#endif
4952		}
4953		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4954		if (chk->data) {
4955			if (so) {
4956				/* Still a socket? */
4957				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4958				    0, chk, SCTP_SO_LOCKED);
4959			}
4960			if (chk->data) {
4961				sctp_m_freem(chk->data);
4962				chk->data = NULL;
4963			}
4964		}
4965		if (chk->holds_key_ref)
4966			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4967		if (chk->whoTo) {
4968			sctp_free_remote_addr(chk->whoTo);
4969			chk->whoTo = NULL;
4970		}
4971		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4972		SCTP_DECR_CHK_COUNT();
4973		/* sa_ignore FREED_MEMORY */
4974	}
4975	/* sent queue SHOULD be empty */
4976	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4977		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
4978			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4979				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4980#ifdef INVARIANTS
4981			} else {
4982				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4983#endif
4984			}
4985		}
4986		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4987		if (chk->data) {
4988			if (so) {
4989				/* Still a socket? */
4990				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
4991				    0, chk, SCTP_SO_LOCKED);
4992			}
4993			if (chk->data) {
4994				sctp_m_freem(chk->data);
4995				chk->data = NULL;
4996			}
4997		}
4998		if (chk->holds_key_ref)
4999			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5000		sctp_free_remote_addr(chk->whoTo);
5001		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5002		SCTP_DECR_CHK_COUNT();
5003		/* sa_ignore FREED_MEMORY */
5004	}
5005#ifdef INVARIANTS
5006	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5007		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5008			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5009		}
5010	}
5011#endif
5012	/* control queue MAY not be empty */
5013	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5014		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5015		if (chk->data) {
5016			sctp_m_freem(chk->data);
5017			chk->data = NULL;
5018		}
5019		if (chk->holds_key_ref)
5020			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5021		sctp_free_remote_addr(chk->whoTo);
5022		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5023		SCTP_DECR_CHK_COUNT();
5024		/* sa_ignore FREED_MEMORY */
5025	}
5026	/* ASCONF queue MAY not be empty */
5027	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5028		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5029		if (chk->data) {
5030			sctp_m_freem(chk->data);
5031			chk->data = NULL;
5032		}
5033		if (chk->holds_key_ref)
5034			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5035		sctp_free_remote_addr(chk->whoTo);
5036		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5037		SCTP_DECR_CHK_COUNT();
5038		/* sa_ignore FREED_MEMORY */
5039	}
5040	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5041		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5042		if (chk->data) {
5043			sctp_m_freem(chk->data);
5044			chk->data = NULL;
5045		}
5046		if (chk->holds_key_ref)
5047			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5048		sctp_free_remote_addr(chk->whoTo);
5049		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5050		SCTP_DECR_CHK_COUNT();
5051		/* sa_ignore FREED_MEMORY */
5052	}
5053
5054	if (asoc->mapping_array) {
5055		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5056		asoc->mapping_array = NULL;
5057	}
5058	if (asoc->nr_mapping_array) {
5059		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5060		asoc->nr_mapping_array = NULL;
5061	}
5062	/* the stream outs */
5063	if (asoc->strmout) {
5064		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5065		asoc->strmout = NULL;
5066	}
5067	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5068	if (asoc->strmin) {
5069		struct sctp_queued_to_read *ctl, *nctl;
5070
5071		for (i = 0; i < asoc->streamincnt; i++) {
5072			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5073				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5074				sctp_free_remote_addr(ctl->whoFrom);
5075				if (ctl->data) {
5076					sctp_m_freem(ctl->data);
5077					ctl->data = NULL;
5078				}
5079				/*
5080				 * We don't free the address here since all
5081				 * the net's were freed above.
5082				 */
5083				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5084				SCTP_DECR_READQ_COUNT();
5085			}
5086		}
5087		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5088		asoc->strmin = NULL;
5089	}
5090	asoc->streamincnt = 0;
5091	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5092#ifdef INVARIANTS
5093		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5094			panic("no net's left alloc'ed, or list points to itself");
5095		}
5096#endif
5097		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5098		sctp_free_remote_addr(net);
5099	}
5100	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5101		/* sa_ignore FREED_MEMORY */
5102		sctp_remove_laddr(laddr);
5103	}
5104
5105	/* pending asconf (address) parameters */
5106	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5107		/* sa_ignore FREED_MEMORY */
5108		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5109		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5110	}
5111	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5112		/* sa_ignore FREED_MEMORY */
5113		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5114		if (aack->data != NULL) {
5115			sctp_m_freem(aack->data);
5116		}
5117		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5118	}
5119	/* clean up auth stuff */
5120	if (asoc->local_hmacs)
5121		sctp_free_hmaclist(asoc->local_hmacs);
5122	if (asoc->peer_hmacs)
5123		sctp_free_hmaclist(asoc->peer_hmacs);
5124
5125	if (asoc->local_auth_chunks)
5126		sctp_free_chunklist(asoc->local_auth_chunks);
5127	if (asoc->peer_auth_chunks)
5128		sctp_free_chunklist(asoc->peer_auth_chunks);
5129
5130	sctp_free_authinfo(&asoc->authinfo);
5131
5132	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5133		LIST_REMOVE(shared_key, next);
5134		sctp_free_sharedkey(shared_key);
5135		/* sa_ignore FREED_MEMORY */
5136	}
5137
5138	/* Insert new items here :> */
5139
5140	/* Get rid of LOCK */
5141	SCTP_TCB_UNLOCK(stcb);
5142	SCTP_TCB_LOCK_DESTROY(stcb);
5143	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5144	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5145		SCTP_INP_INFO_WUNLOCK();
5146		SCTP_INP_RLOCK(inp);
5147	}
5148#ifdef SCTP_TRACK_FREED_ASOCS
5149	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5150		/* now clean up the tasoc itself */
5151		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5152		SCTP_DECR_ASOC_COUNT();
5153	} else {
5154		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5155	}
5156#else
5157	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5158	SCTP_DECR_ASOC_COUNT();
5159#endif
5160	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5161		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5162			/*
5163			 * If its NOT the inp_free calling us AND sctp_close
5164			 * as been called, we call back...
5165			 */
5166			SCTP_INP_RUNLOCK(inp);
5167			/*
5168			 * This will start the kill timer (if we are the
5169			 * last one) since we hold an increment yet. But
5170			 * this is the only safe way to do this since
5171			 * otherwise if the socket closes at the same time
5172			 * we are here we might collide in the cleanup.
5173			 */
5174			sctp_inpcb_free(inp,
5175			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5176			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5177			SCTP_INP_DECR_REF(inp);
5178			goto out_of;
5179		} else {
5180			/* The socket is still open. */
5181			SCTP_INP_DECR_REF(inp);
5182		}
5183	}
5184	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5185		SCTP_INP_RUNLOCK(inp);
5186	}
5187out_of:
5188	/* destroyed the asoc */
5189#ifdef SCTP_LOG_CLOSING
5190	sctp_log_closing(inp, NULL, 11);
5191#endif
5192	return (1);
5193}
5194
5195
5196
5197/*
5198 * determine if a destination is "reachable" based upon the addresses bound
5199 * to the current endpoint (e.g. only v4 or v6 currently bound)
5200 */
5201/*
5202 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5203 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5204 * types bound as its endpoint
5205 */
5206int
5207sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5208{
5209	struct sctp_inpcb *inp;
5210	int answer;
5211
5212	/*
5213	 * No locks here, the TCB, in all cases is already locked and an
5214	 * assoc is up. There is either a INP lock by the caller applied (in
5215	 * asconf case when deleting an address) or NOT in the HB case,
5216	 * however if HB then the INP increment is up and the INP will not
5217	 * be removed (on top of the fact that we have a TCB lock). So we
5218	 * only want to read the sctp_flags, which is either bound-all or
5219	 * not.. no protection needed since once an assoc is up you can't be
5220	 * changing your binding.
5221	 */
5222	inp = stcb->sctp_ep;
5223	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5224		/* if bound all, destination is not restricted */
5225		/*
5226		 * RRS: Question during lock work: Is this correct? If you
5227		 * are bound-all you still might need to obey the V4--V6
5228		 * flags??? IMO this bound-all stuff needs to be removed!
5229		 */
5230		return (1);
5231	}
5232	/* NOTE: all "scope" checks are done when local addresses are added */
5233	switch (destaddr->sa_family) {
5234#ifdef INET6
5235	case AF_INET6:
5236		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5237		break;
5238#endif
5239#ifdef INET
5240	case AF_INET:
5241		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5242		break;
5243#endif
5244	default:
5245		/* invalid family, so it's unreachable */
5246		answer = 0;
5247		break;
5248	}
5249	return (answer);
5250}
5251
5252/*
5253 * update the inp_vflags on an endpoint
5254 */
5255static void
5256sctp_update_ep_vflag(struct sctp_inpcb *inp)
5257{
5258	struct sctp_laddr *laddr;
5259
5260	/* first clear the flag */
5261	inp->ip_inp.inp.inp_vflag = 0;
5262	/* set the flag based on addresses on the ep list */
5263	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5264		if (laddr->ifa == NULL) {
5265			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5266			    __FUNCTION__);
5267			continue;
5268		}
5269		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5270			continue;
5271		}
5272		switch (laddr->ifa->address.sa.sa_family) {
5273#ifdef INET6
5274		case AF_INET6:
5275			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5276			break;
5277#endif
5278#ifdef INET
5279		case AF_INET:
5280			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5281			break;
5282#endif
5283		default:
5284			break;
5285		}
5286	}
5287}
5288
5289/*
5290 * Add the address to the endpoint local address list There is nothing to be
5291 * done if we are bound to all addresses
5292 */
5293void
5294sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5295{
5296	struct sctp_laddr *laddr;
5297	int fnd, error = 0;
5298
5299	fnd = 0;
5300
5301	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5302		/* You are already bound to all. You have it already */
5303		return;
5304	}
5305#ifdef INET6
5306	if (ifa->address.sa.sa_family == AF_INET6) {
5307		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5308			/* Can't bind a non-useable addr. */
5309			return;
5310		}
5311	}
5312#endif
5313	/* first, is it already present? */
5314	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5315		if (laddr->ifa == ifa) {
5316			fnd = 1;
5317			break;
5318		}
5319	}
5320
5321	if (fnd == 0) {
5322		/* Not in the ep list */
5323		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5324		if (error != 0)
5325			return;
5326		inp->laddr_count++;
5327		/* update inp_vflag flags */
5328		switch (ifa->address.sa.sa_family) {
5329#ifdef INET6
5330		case AF_INET6:
5331			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5332			break;
5333#endif
5334#ifdef INET
5335		case AF_INET:
5336			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5337			break;
5338#endif
5339		default:
5340			break;
5341		}
5342	}
5343	return;
5344}
5345
5346
5347/*
5348 * select a new (hopefully reachable) destination net (should only be used
5349 * when we deleted an ep addr that is the only usable source address to reach
5350 * the destination net)
5351 */
5352static void
5353sctp_select_primary_destination(struct sctp_tcb *stcb)
5354{
5355	struct sctp_nets *net;
5356
5357	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5358		/* for now, we'll just pick the first reachable one we find */
5359		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5360			continue;
5361		if (sctp_destination_is_reachable(stcb,
5362		    (struct sockaddr *)&net->ro._l_addr)) {
5363			/* found a reachable destination */
5364			stcb->asoc.primary_destination = net;
5365		}
5366	}
5367	/* I can't there from here! ...we're gonna die shortly... */
5368}
5369
5370
5371/*
5372 * Delete the address from the endpoint local address list There is nothing
5373 * to be done if we are bound to all addresses
5374 */
5375void
5376sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5377{
5378	struct sctp_laddr *laddr;
5379	int fnd;
5380
5381	fnd = 0;
5382	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5383		/* You are already bound to all. You have it already */
5384		return;
5385	}
5386	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5387		if (laddr->ifa == ifa) {
5388			fnd = 1;
5389			break;
5390		}
5391	}
5392	if (fnd && (inp->laddr_count < 2)) {
5393		/* can't delete unless there are at LEAST 2 addresses */
5394		return;
5395	}
5396	if (fnd) {
5397		/*
5398		 * clean up any use of this address go through our
5399		 * associations and clear any last_used_address that match
5400		 * this one for each assoc, see if a new primary_destination
5401		 * is needed
5402		 */
5403		struct sctp_tcb *stcb;
5404
5405		/* clean up "next_addr_touse" */
5406		if (inp->next_addr_touse == laddr)
5407			/* delete this address */
5408			inp->next_addr_touse = NULL;
5409
5410		/* clean up "last_used_address" */
5411		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5412			struct sctp_nets *net;
5413
5414			SCTP_TCB_LOCK(stcb);
5415			if (stcb->asoc.last_used_address == laddr)
5416				/* delete this address */
5417				stcb->asoc.last_used_address = NULL;
5418			/*
5419			 * Now spin through all the nets and purge any ref
5420			 * to laddr
5421			 */
5422			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5423				if (net->ro._s_addr &&
5424				    (net->ro._s_addr->ifa == laddr->ifa)) {
5425					/* Yep, purge src address selected */
5426					sctp_rtentry_t *rt;
5427
5428					/* delete this address if cached */
5429					rt = net->ro.ro_rt;
5430					if (rt != NULL) {
5431						RTFREE(rt);
5432						net->ro.ro_rt = NULL;
5433					}
5434					sctp_free_ifa(net->ro._s_addr);
5435					net->ro._s_addr = NULL;
5436					net->src_addr_selected = 0;
5437				}
5438			}
5439			SCTP_TCB_UNLOCK(stcb);
5440		}		/* for each tcb */
5441		/* remove it from the ep list */
5442		sctp_remove_laddr(laddr);
5443		inp->laddr_count--;
5444		/* update inp_vflag flags */
5445		sctp_update_ep_vflag(inp);
5446	}
5447	return;
5448}
5449
5450/*
5451 * Add the address to the TCB local address restricted list.
5452 * This is a "pending" address list (eg. addresses waiting for an
5453 * ASCONF-ACK response) and cannot be used as a valid source address.
5454 */
5455void
5456sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5457{
5458	struct sctp_laddr *laddr;
5459	struct sctpladdr *list;
5460
5461	/*
5462	 * Assumes TCB is locked.. and possibly the INP. May need to
5463	 * confirm/fix that if we need it and is not the case.
5464	 */
5465	list = &stcb->asoc.sctp_restricted_addrs;
5466
5467#ifdef INET6
5468	if (ifa->address.sa.sa_family == AF_INET6) {
5469		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5470			/* Can't bind a non-existent addr. */
5471			return;
5472		}
5473	}
5474#endif
5475	/* does the address already exist? */
5476	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5477		if (laddr->ifa == ifa) {
5478			return;
5479		}
5480	}
5481
5482	/* add to the list */
5483	(void)sctp_insert_laddr(list, ifa, 0);
5484	return;
5485}
5486
5487/*
5488 * insert an laddr entry with the given ifa for the desired list
5489 */
5490int
5491sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5492{
5493	struct sctp_laddr *laddr;
5494
5495	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5496	if (laddr == NULL) {
5497		/* out of memory? */
5498		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5499		return (EINVAL);
5500	}
5501	SCTP_INCR_LADDR_COUNT();
5502	bzero(laddr, sizeof(*laddr));
5503	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5504	laddr->ifa = ifa;
5505	laddr->action = act;
5506	atomic_add_int(&ifa->refcount, 1);
5507	/* insert it */
5508	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5509
5510	return (0);
5511}
5512
5513/*
5514 * Remove an laddr entry from the local address list (on an assoc)
5515 */
5516void
5517sctp_remove_laddr(struct sctp_laddr *laddr)
5518{
5519
5520	/* remove from the list */
5521	LIST_REMOVE(laddr, sctp_nxt_addr);
5522	sctp_free_ifa(laddr->ifa);
5523	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5524	SCTP_DECR_LADDR_COUNT();
5525}
5526
5527/*
5528 * Remove a local address from the TCB local address restricted list
5529 */
5530void
5531sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5532{
5533	struct sctp_inpcb *inp;
5534	struct sctp_laddr *laddr;
5535
5536	/*
5537	 * This is called by asconf work. It is assumed that a) The TCB is
5538	 * locked and b) The INP is locked. This is true in as much as I can
5539	 * trace through the entry asconf code where I did these locks.
5540	 * Again, the ASCONF code is a bit different in that it does lock
5541	 * the INP during its work often times. This must be since we don't
5542	 * want other proc's looking up things while what they are looking
5543	 * up is changing :-D
5544	 */
5545
5546	inp = stcb->sctp_ep;
5547	/* if subset bound and don't allow ASCONF's, can't delete last */
5548	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5549	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5550		if (stcb->sctp_ep->laddr_count < 2) {
5551			/* can't delete last address */
5552			return;
5553		}
5554	}
5555	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5556		/* remove the address if it exists */
5557		if (laddr->ifa == NULL)
5558			continue;
5559		if (laddr->ifa == ifa) {
5560			sctp_remove_laddr(laddr);
5561			return;
5562		}
5563	}
5564
5565	/* address not found! */
5566	return;
5567}
5568
5569/*
5570 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5571 */
5572/* sysctl */
5573static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5574static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5575
5576
5577
5578#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5579struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5580int *sctp_cpuarry = NULL;
5581void
5582sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5583{
5584	/* Queue a packet to a processor for the specified core */
5585	struct sctp_mcore_queue *qent;
5586	struct sctp_mcore_ctrl *wkq;
5587	int need_wake = 0;
5588
5589	if (sctp_mcore_workers == NULL) {
5590		/* Something went way bad during setup */
5591		sctp_input_with_port(m, off, 0);
5592		return;
5593	}
5594	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5595	    (sizeof(struct sctp_mcore_queue)),
5596	    SCTP_M_MCORE);
5597	if (qent == NULL) {
5598		/* This is trouble  */
5599		sctp_input_with_port(m, off, 0);
5600		return;
5601	}
5602	qent->vn = curvnet;
5603	qent->m = m;
5604	qent->off = off;
5605	qent->v6 = 0;
5606	wkq = &sctp_mcore_workers[cpu_to_use];
5607	SCTP_MCORE_QLOCK(wkq);
5608
5609	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5610	if (wkq->running == 0) {
5611		need_wake = 1;
5612	}
5613	SCTP_MCORE_QUNLOCK(wkq);
5614	if (need_wake) {
5615		wakeup(&wkq->running);
5616	}
5617}
5618
5619static void
5620sctp_mcore_thread(void *arg)
5621{
5622
5623	struct sctp_mcore_ctrl *wkq;
5624	struct sctp_mcore_queue *qent;
5625
5626	wkq = (struct sctp_mcore_ctrl *)arg;
5627	struct mbuf *m;
5628	int off, v6;
5629
5630	/* Wait for first tickle */
5631	SCTP_MCORE_LOCK(wkq);
5632	wkq->running = 0;
5633	msleep(&wkq->running,
5634	    &wkq->core_mtx,
5635	    0, "wait for pkt", 0);
5636	SCTP_MCORE_UNLOCK(wkq);
5637
5638	/* Bind to our cpu */
5639	thread_lock(curthread);
5640	sched_bind(curthread, wkq->cpuid);
5641	thread_unlock(curthread);
5642
5643	/* Now lets start working */
5644	SCTP_MCORE_LOCK(wkq);
5645	/* Now grab lock and go */
5646	for (;;) {
5647		SCTP_MCORE_QLOCK(wkq);
5648skip_sleep:
5649		wkq->running = 1;
5650		qent = TAILQ_FIRST(&wkq->que);
5651		if (qent) {
5652			TAILQ_REMOVE(&wkq->que, qent, next);
5653			SCTP_MCORE_QUNLOCK(wkq);
5654			CURVNET_SET(qent->vn);
5655			m = qent->m;
5656			off = qent->off;
5657			v6 = qent->v6;
5658			SCTP_FREE(qent, SCTP_M_MCORE);
5659			if (v6 == 0) {
5660				sctp_input_with_port(m, off, 0);
5661			} else {
5662				SCTP_PRINTF("V6 not yet supported\n");
5663				sctp_m_freem(m);
5664			}
5665			CURVNET_RESTORE();
5666			SCTP_MCORE_QLOCK(wkq);
5667		}
5668		wkq->running = 0;
5669		if (!TAILQ_EMPTY(&wkq->que)) {
5670			goto skip_sleep;
5671		}
5672		SCTP_MCORE_QUNLOCK(wkq);
5673		msleep(&wkq->running,
5674		    &wkq->core_mtx,
5675		    0, "wait for pkt", 0);
5676	}
5677}
5678
5679static void
5680sctp_startup_mcore_threads(void)
5681{
5682	int i, cpu;
5683
5684	if (mp_ncpus == 1)
5685		return;
5686
5687	if (sctp_mcore_workers != NULL) {
5688		/*
5689		 * Already been here in some previous vnet?
5690		 */
5691		return;
5692	}
5693	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5694	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5695	    SCTP_M_MCORE);
5696	if (sctp_mcore_workers == NULL) {
5697		/* TSNH I hope */
5698		return;
5699	}
5700	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5701	    sizeof(struct sctp_mcore_ctrl)));
5702	/* Init the structures */
5703	for (i = 0; i <= mp_maxid; i++) {
5704		TAILQ_INIT(&sctp_mcore_workers[i].que);
5705		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5706		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5707		sctp_mcore_workers[i].cpuid = i;
5708	}
5709	if (sctp_cpuarry == NULL) {
5710		SCTP_MALLOC(sctp_cpuarry, int *,
5711		    (mp_ncpus * sizeof(int)),
5712		    SCTP_M_MCORE);
5713		i = 0;
5714		CPU_FOREACH(cpu) {
5715			sctp_cpuarry[i] = cpu;
5716			i++;
5717		}
5718	}
5719	/* Now start them all */
5720	CPU_FOREACH(cpu) {
5721		(void)kproc_create(sctp_mcore_thread,
5722		    (void *)&sctp_mcore_workers[cpu],
5723		    &sctp_mcore_workers[cpu].thread_proc,
5724		    RFPROC,
5725		    SCTP_KTHREAD_PAGES,
5726		    SCTP_MCORE_NAME);
5727
5728	}
5729}
5730
5731#endif
5732
5733void
5734sctp_pcb_init()
5735{
5736	/*
5737	 * SCTP initialization for the PCB structures should be called by
5738	 * the sctp_init() funciton.
5739	 */
5740	int i;
5741	struct timeval tv;
5742
5743	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5744		/* error I was called twice */
5745		return;
5746	}
5747	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5748
5749#if defined(SCTP_LOCAL_TRACE_BUF)
5750	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5751#endif
5752#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5753	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5754	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5755	    SCTP_M_MCORE);
5756#endif
5757	(void)SCTP_GETTIME_TIMEVAL(&tv);
5758#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5759	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5760	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5761	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5762#else
5763	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5764	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5765	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5766#endif
5767	/* init the empty list of (All) Endpoints */
5768	LIST_INIT(&SCTP_BASE_INFO(listhead));
5769
5770
5771	/* init the hash table of endpoints */
5772	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5773	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5774	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5775	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5776	    &SCTP_BASE_INFO(hashasocmark));
5777	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5778	    &SCTP_BASE_INFO(hashmark));
5779	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5780	    &SCTP_BASE_INFO(hashtcpmark));
5781	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5782
5783
5784	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5785	    &SCTP_BASE_INFO(hashvrfmark));
5786
5787	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5788	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5789	/* init the zones */
5790	/*
5791	 * FIX ME: Should check for NULL returns, but if it does fail we are
5792	 * doomed to panic anyways... add later maybe.
5793	 */
5794	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5795	    sizeof(struct sctp_inpcb), maxsockets);
5796
5797	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5798	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5799
5800	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5801	    sizeof(struct sctp_laddr),
5802	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5803
5804	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5805	    sizeof(struct sctp_nets),
5806	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5807
5808	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5809	    sizeof(struct sctp_tmit_chunk),
5810	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5811
5812	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5813	    sizeof(struct sctp_queued_to_read),
5814	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5815
5816	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5817	    sizeof(struct sctp_stream_queue_pending),
5818	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5819
5820	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5821	    sizeof(struct sctp_asconf),
5822	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5823
5824	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5825	    sizeof(struct sctp_asconf_ack),
5826	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5827
5828
5829	/* Master Lock INIT for info structure */
5830	SCTP_INP_INFO_LOCK_INIT();
5831	SCTP_STATLOG_INIT_LOCK();
5832
5833	SCTP_IPI_COUNT_INIT();
5834	SCTP_IPI_ADDR_INIT();
5835#ifdef SCTP_PACKET_LOGGING
5836	SCTP_IP_PKTLOG_INIT();
5837#endif
5838	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5839
5840	SCTP_WQ_ADDR_INIT();
5841	/* not sure if we need all the counts */
5842	SCTP_BASE_INFO(ipi_count_ep) = 0;
5843	/* assoc/tcb zone info */
5844	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5845	/* local addrlist zone info */
5846	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5847	/* remote addrlist zone info */
5848	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5849	/* chunk info */
5850	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5851
5852	/* socket queue zone info */
5853	SCTP_BASE_INFO(ipi_count_readq) = 0;
5854
5855	/* stream out queue cont */
5856	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5857
5858	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5859	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5860
5861	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5862
5863	/* Init the TIMEWAIT list */
5864	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5865		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5866	}
5867	SCTP_ITERATOR_LOCK_INIT();
5868	SCTP_IPI_ITERATOR_WQ_INIT();
5869	sctp_startup_iterator();
5870
5871#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5872	sctp_startup_mcore_threads();
5873#endif
5874
5875	/*
5876	 * INIT the default VRF which for BSD is the only one, other O/S's
5877	 * may have more. But initially they must start with one and then
5878	 * add the VRF's as addresses are added.
5879	 */
5880	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5881}
5882
5883/*
5884 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5885 */
5886void
5887sctp_pcb_finish(void)
5888{
5889	struct sctp_vrflist *vrf_bucket;
5890	struct sctp_vrf *vrf, *nvrf;
5891	struct sctp_ifn *ifn, *nifn;
5892	struct sctp_ifa *ifa, *nifa;
5893	struct sctpvtaghead *chain;
5894	struct sctp_tagblock *twait_block, *prev_twait_block;
5895	struct sctp_laddr *wi, *nwi;
5896	int i;
5897	struct sctp_iterator *it, *nit;
5898
5899	/*
5900	 * In FreeBSD the iterator thread never exits but we do clean up.
5901	 * The only way FreeBSD reaches here is if we have VRF's but we
5902	 * still add the ifdef to make it compile on old versions.
5903	 */
5904	SCTP_IPI_ITERATOR_WQ_LOCK();
5905	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5906		if (it->vn != curvnet) {
5907			continue;
5908		}
5909		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5910		if (it->function_atend != NULL) {
5911			(*it->function_atend) (it->pointer, it->val);
5912		}
5913		SCTP_FREE(it, SCTP_M_ITER);
5914	}
5915	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5916	SCTP_ITERATOR_LOCK();
5917	if ((sctp_it_ctl.cur_it) &&
5918	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5919		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5920	}
5921	SCTP_ITERATOR_UNLOCK();
5922	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5923	SCTP_WQ_ADDR_LOCK();
5924	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5925		LIST_REMOVE(wi, sctp_nxt_addr);
5926		SCTP_DECR_LADDR_COUNT();
5927		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5928	}
5929	SCTP_WQ_ADDR_UNLOCK();
5930
5931	/*
5932	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5933	 * destroyed first).
5934	 */
5935	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5936	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5937		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5938			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5939				/* free the ifa */
5940				LIST_REMOVE(ifa, next_bucket);
5941				LIST_REMOVE(ifa, next_ifa);
5942				SCTP_FREE(ifa, SCTP_M_IFA);
5943			}
5944			/* free the ifn */
5945			LIST_REMOVE(ifn, next_bucket);
5946			LIST_REMOVE(ifn, next_ifn);
5947			SCTP_FREE(ifn, SCTP_M_IFN);
5948		}
5949		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5950		/* free the vrf */
5951		LIST_REMOVE(vrf, next_vrf);
5952		SCTP_FREE(vrf, SCTP_M_VRF);
5953	}
5954	/* free the vrf hashes */
5955	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5956	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5957
5958	/*
5959	 * free the TIMEWAIT list elements malloc'd in the function
5960	 * sctp_add_vtag_to_timewait()...
5961	 */
5962	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5963		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5964		if (!LIST_EMPTY(chain)) {
5965			prev_twait_block = NULL;
5966			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5967				if (prev_twait_block) {
5968					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5969				}
5970				prev_twait_block = twait_block;
5971			}
5972			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5973		}
5974	}
5975
5976	/* free the locks and mutexes */
5977#ifdef SCTP_PACKET_LOGGING
5978	SCTP_IP_PKTLOG_DESTROY();
5979#endif
5980	SCTP_IPI_ADDR_DESTROY();
5981	SCTP_STATLOG_DESTROY();
5982	SCTP_INP_INFO_LOCK_DESTROY();
5983
5984	SCTP_WQ_ADDR_DESTROY();
5985
5986	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5987	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5988	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5989	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5990	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5991	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5992	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5993	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5994	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5995	/* Get rid of other stuff to */
5996	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5997		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5998	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5999		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
6000	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
6001		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
6002#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6003	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6004#endif
6005}
6006
6007
6008int
6009sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6010    int offset, int limit,
6011    struct sockaddr *src, struct sockaddr *dst,
6012    struct sockaddr *altsa)
6013{
6014	/*
6015	 * grub through the INIT pulling addresses and loading them to the
6016	 * nets structure in the asoc. The from address in the mbuf should
6017	 * also be loaded (if it is not already). This routine can be called
6018	 * with either INIT or INIT-ACK's as long as the m points to the IP
6019	 * packet and the offset points to the beginning of the parameters.
6020	 */
6021	struct sctp_inpcb *inp;
6022	struct sctp_nets *net, *nnet, *net_tmp;
6023	struct sctp_paramhdr *phdr, parm_buf;
6024	struct sctp_tcb *stcb_tmp;
6025	uint16_t ptype, plen;
6026	struct sockaddr *sa;
6027	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6028	struct sctp_auth_random *p_random = NULL;
6029	uint16_t random_len = 0;
6030	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6031	struct sctp_auth_hmac_algo *hmacs = NULL;
6032	uint16_t hmacs_len = 0;
6033	uint8_t saw_asconf = 0;
6034	uint8_t saw_asconf_ack = 0;
6035	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6036	struct sctp_auth_chunk_list *chunks = NULL;
6037	uint16_t num_chunks = 0;
6038	sctp_key_t *new_key;
6039	uint32_t keylen;
6040	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6041	uint8_t ecn_allowed;
6042
6043#ifdef INET
6044	struct sockaddr_in sin;
6045
6046#endif
6047#ifdef INET6
6048	struct sockaddr_in6 sin6;
6049
6050#endif
6051
6052	/* First get the destination address setup too. */
6053#ifdef INET
6054	memset(&sin, 0, sizeof(sin));
6055	sin.sin_family = AF_INET;
6056	sin.sin_len = sizeof(sin);
6057	sin.sin_port = stcb->rport;
6058#endif
6059#ifdef INET6
6060	memset(&sin6, 0, sizeof(sin6));
6061	sin6.sin6_family = AF_INET6;
6062	sin6.sin6_len = sizeof(struct sockaddr_in6);
6063	sin6.sin6_port = stcb->rport;
6064#endif
6065	if (altsa) {
6066		sa = altsa;
6067	} else {
6068		sa = src;
6069	}
6070	/* Turn off ECN until we get through all params */
6071	ecn_allowed = 0;
6072	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6073		/* mark all addresses that we have currently on the list */
6074		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6075	}
6076	/* does the source address already exist? if so skip it */
6077	inp = stcb->sctp_ep;
6078	atomic_add_int(&stcb->asoc.refcnt, 1);
6079	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6080	atomic_add_int(&stcb->asoc.refcnt, -1);
6081
6082	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6083		/* we must add the source address */
6084		/* no scope set here since we have a tcb already. */
6085		switch (sa->sa_family) {
6086#ifdef INET
6087		case AF_INET:
6088			if (stcb->asoc.scope.ipv4_addr_legal) {
6089				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6090					return (-1);
6091				}
6092			}
6093			break;
6094#endif
6095#ifdef INET6
6096		case AF_INET6:
6097			if (stcb->asoc.scope.ipv6_addr_legal) {
6098				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6099					return (-2);
6100				}
6101			}
6102			break;
6103#endif
6104		default:
6105			break;
6106		}
6107	} else {
6108		if (net_tmp != NULL && stcb_tmp == stcb) {
6109			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6110		} else if (stcb_tmp != stcb) {
6111			/* It belongs to another association? */
6112			if (stcb_tmp)
6113				SCTP_TCB_UNLOCK(stcb_tmp);
6114			return (-3);
6115		}
6116	}
6117	if (stcb->asoc.state == 0) {
6118		/* the assoc was freed? */
6119		return (-4);
6120	}
6121	/*
6122	 * peer must explicitly turn this on. This may have been initialized
6123	 * to be "on" in order to allow local addr changes while INIT's are
6124	 * in flight.
6125	 */
6126	stcb->asoc.peer_supports_asconf = 0;
6127	/* now we must go through each of the params. */
6128	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6129	while (phdr) {
6130		ptype = ntohs(phdr->param_type);
6131		plen = ntohs(phdr->param_length);
6132		/*
6133		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6134		 * (uint32_t)ptype, (int)plen);
6135		 */
6136		if (offset + plen > limit) {
6137			break;
6138		}
6139		if (plen == 0) {
6140			break;
6141		}
6142#ifdef INET
6143		if (ptype == SCTP_IPV4_ADDRESS) {
6144			if (stcb->asoc.scope.ipv4_addr_legal) {
6145				struct sctp_ipv4addr_param *p4, p4_buf;
6146
6147				/* ok get the v4 address and check/add */
6148				phdr = sctp_get_next_param(m, offset,
6149				    (struct sctp_paramhdr *)&p4_buf,
6150				    sizeof(p4_buf));
6151				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6152				    phdr == NULL) {
6153					return (-5);
6154				}
6155				p4 = (struct sctp_ipv4addr_param *)phdr;
6156				sin.sin_addr.s_addr = p4->addr;
6157				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6158					/* Skip multi-cast addresses */
6159					goto next_param;
6160				}
6161				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6162				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6163					goto next_param;
6164				}
6165				sa = (struct sockaddr *)&sin;
6166				inp = stcb->sctp_ep;
6167				atomic_add_int(&stcb->asoc.refcnt, 1);
6168				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6169				    dst, stcb);
6170				atomic_add_int(&stcb->asoc.refcnt, -1);
6171
6172				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6173				    inp == NULL) {
6174					/* we must add the source address */
6175					/*
6176					 * no scope set since we have a tcb
6177					 * already
6178					 */
6179
6180					/*
6181					 * we must validate the state again
6182					 * here
6183					 */
6184			add_it_now:
6185					if (stcb->asoc.state == 0) {
6186						/* the assoc was freed? */
6187						return (-7);
6188					}
6189					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6190						return (-8);
6191					}
6192				} else if (stcb_tmp == stcb) {
6193					if (stcb->asoc.state == 0) {
6194						/* the assoc was freed? */
6195						return (-10);
6196					}
6197					if (net != NULL) {
6198						/* clear flag */
6199						net->dest_state &=
6200						    ~SCTP_ADDR_NOT_IN_ASSOC;
6201					}
6202				} else {
6203					/*
6204					 * strange, address is in another
6205					 * assoc? straighten out locks.
6206					 */
6207					if (stcb_tmp) {
6208						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6209							/*
6210							 * in setup state we
6211							 * abort this guy
6212							 */
6213							sctp_abort_an_association(stcb_tmp->sctp_ep,
6214							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6215							goto add_it_now;
6216						}
6217						SCTP_TCB_UNLOCK(stcb_tmp);
6218					}
6219					if (stcb->asoc.state == 0) {
6220						/* the assoc was freed? */
6221						return (-12);
6222					}
6223					return (-13);
6224				}
6225			}
6226		} else
6227#endif
6228#ifdef INET6
6229		if (ptype == SCTP_IPV6_ADDRESS) {
6230			if (stcb->asoc.scope.ipv6_addr_legal) {
6231				/* ok get the v6 address and check/add */
6232				struct sctp_ipv6addr_param *p6, p6_buf;
6233
6234				phdr = sctp_get_next_param(m, offset,
6235				    (struct sctp_paramhdr *)&p6_buf,
6236				    sizeof(p6_buf));
6237				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6238				    phdr == NULL) {
6239					return (-14);
6240				}
6241				p6 = (struct sctp_ipv6addr_param *)phdr;
6242				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6243				    sizeof(p6->addr));
6244				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6245					/* Skip multi-cast addresses */
6246					goto next_param;
6247				}
6248				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6249					/*
6250					 * Link local make no sense without
6251					 * scope
6252					 */
6253					goto next_param;
6254				}
6255				sa = (struct sockaddr *)&sin6;
6256				inp = stcb->sctp_ep;
6257				atomic_add_int(&stcb->asoc.refcnt, 1);
6258				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6259				    dst, stcb);
6260				atomic_add_int(&stcb->asoc.refcnt, -1);
6261				if (stcb_tmp == NULL &&
6262				    (inp == stcb->sctp_ep || inp == NULL)) {
6263					/*
6264					 * we must validate the state again
6265					 * here
6266					 */
6267			add_it_now6:
6268					if (stcb->asoc.state == 0) {
6269						/* the assoc was freed? */
6270						return (-16);
6271					}
6272					/*
6273					 * we must add the address, no scope
6274					 * set
6275					 */
6276					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6277						return (-17);
6278					}
6279				} else if (stcb_tmp == stcb) {
6280					/*
6281					 * we must validate the state again
6282					 * here
6283					 */
6284					if (stcb->asoc.state == 0) {
6285						/* the assoc was freed? */
6286						return (-19);
6287					}
6288					if (net != NULL) {
6289						/* clear flag */
6290						net->dest_state &=
6291						    ~SCTP_ADDR_NOT_IN_ASSOC;
6292					}
6293				} else {
6294					/*
6295					 * strange, address is in another
6296					 * assoc? straighten out locks.
6297					 */
6298					if (stcb_tmp)
6299						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6300							/*
6301							 * in setup state we
6302							 * abort this guy
6303							 */
6304							sctp_abort_an_association(stcb_tmp->sctp_ep,
6305							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6306							goto add_it_now6;
6307						}
6308					SCTP_TCB_UNLOCK(stcb_tmp);
6309
6310					if (stcb->asoc.state == 0) {
6311						/* the assoc was freed? */
6312						return (-21);
6313					}
6314					return (-22);
6315				}
6316			}
6317		} else
6318#endif
6319		if (ptype == SCTP_ECN_CAPABLE) {
6320			ecn_allowed = 1;
6321		} else if (ptype == SCTP_ULP_ADAPTATION) {
6322			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6323				struct sctp_adaptation_layer_indication ai,
6324				                                *aip;
6325
6326				phdr = sctp_get_next_param(m, offset,
6327				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6328				aip = (struct sctp_adaptation_layer_indication *)phdr;
6329				if (aip) {
6330					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6331					stcb->asoc.adaptation_needed = 1;
6332				}
6333			}
6334		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6335			struct sctp_asconf_addr_param lstore, *fee;
6336			int lptype;
6337			struct sockaddr *lsa = NULL;
6338
6339#ifdef INET
6340			struct sctp_asconf_addrv4_param *fii;
6341
6342#endif
6343
6344			stcb->asoc.peer_supports_asconf = 1;
6345			if (plen > sizeof(lstore)) {
6346				return (-23);
6347			}
6348			phdr = sctp_get_next_param(m, offset,
6349			    (struct sctp_paramhdr *)&lstore,
6350			    min(plen, sizeof(lstore)));
6351			if (phdr == NULL) {
6352				return (-24);
6353			}
6354			fee = (struct sctp_asconf_addr_param *)phdr;
6355			lptype = ntohs(fee->addrp.ph.param_type);
6356			switch (lptype) {
6357#ifdef INET
6358			case SCTP_IPV4_ADDRESS:
6359				if (plen !=
6360				    sizeof(struct sctp_asconf_addrv4_param)) {
6361					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6362					    (int)sizeof(struct sctp_asconf_addrv4_param),
6363					    plen);
6364				} else {
6365					fii = (struct sctp_asconf_addrv4_param *)fee;
6366					sin.sin_addr.s_addr = fii->addrp.addr;
6367					lsa = (struct sockaddr *)&sin;
6368				}
6369				break;
6370#endif
6371#ifdef INET6
6372			case SCTP_IPV6_ADDRESS:
6373				if (plen !=
6374				    sizeof(struct sctp_asconf_addr_param)) {
6375					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6376					    (int)sizeof(struct sctp_asconf_addr_param),
6377					    plen);
6378				} else {
6379					memcpy(sin6.sin6_addr.s6_addr,
6380					    fee->addrp.addr,
6381					    sizeof(fee->addrp.addr));
6382					lsa = (struct sockaddr *)&sin6;
6383				}
6384				break;
6385#endif
6386			default:
6387				break;
6388			}
6389			if (lsa) {
6390				(void)sctp_set_primary_addr(stcb, sa, NULL);
6391			}
6392		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6393			stcb->asoc.peer_supports_nat = 1;
6394		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6395			/* Peer supports pr-sctp */
6396			stcb->asoc.peer_supports_prsctp = 1;
6397		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6398			/* A supported extension chunk */
6399			struct sctp_supported_chunk_types_param *pr_supported;
6400			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6401			int num_ent, i;
6402
6403			phdr = sctp_get_next_param(m, offset,
6404			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6405			if (phdr == NULL) {
6406				return (-25);
6407			}
6408			stcb->asoc.peer_supports_asconf = 0;
6409			stcb->asoc.peer_supports_prsctp = 0;
6410			stcb->asoc.peer_supports_pktdrop = 0;
6411			stcb->asoc.peer_supports_strreset = 0;
6412			stcb->asoc.peer_supports_nr_sack = 0;
6413			stcb->asoc.peer_supports_auth = 0;
6414			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6415			num_ent = plen - sizeof(struct sctp_paramhdr);
6416			for (i = 0; i < num_ent; i++) {
6417				switch (pr_supported->chunk_types[i]) {
6418				case SCTP_ASCONF:
6419				case SCTP_ASCONF_ACK:
6420					stcb->asoc.peer_supports_asconf = 1;
6421					break;
6422				case SCTP_FORWARD_CUM_TSN:
6423					stcb->asoc.peer_supports_prsctp = 1;
6424					break;
6425				case SCTP_PACKET_DROPPED:
6426					stcb->asoc.peer_supports_pktdrop = 1;
6427					break;
6428				case SCTP_NR_SELECTIVE_ACK:
6429					stcb->asoc.peer_supports_nr_sack = 1;
6430					break;
6431				case SCTP_STREAM_RESET:
6432					stcb->asoc.peer_supports_strreset = 1;
6433					break;
6434				case SCTP_AUTHENTICATION:
6435					stcb->asoc.peer_supports_auth = 1;
6436					break;
6437				default:
6438					/* one I have not learned yet */
6439					break;
6440
6441				}
6442			}
6443		} else if (ptype == SCTP_RANDOM) {
6444			if (plen > sizeof(random_store))
6445				break;
6446			if (got_random) {
6447				/* already processed a RANDOM */
6448				goto next_param;
6449			}
6450			phdr = sctp_get_next_param(m, offset,
6451			    (struct sctp_paramhdr *)random_store,
6452			    min(sizeof(random_store), plen));
6453			if (phdr == NULL)
6454				return (-26);
6455			p_random = (struct sctp_auth_random *)phdr;
6456			random_len = plen - sizeof(*p_random);
6457			/* enforce the random length */
6458			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6459				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6460				return (-27);
6461			}
6462			got_random = 1;
6463		} else if (ptype == SCTP_HMAC_LIST) {
6464			int num_hmacs;
6465			int i;
6466
6467			if (plen > sizeof(hmacs_store))
6468				break;
6469			if (got_hmacs) {
6470				/* already processed a HMAC list */
6471				goto next_param;
6472			}
6473			phdr = sctp_get_next_param(m, offset,
6474			    (struct sctp_paramhdr *)hmacs_store,
6475			    min(plen, sizeof(hmacs_store)));
6476			if (phdr == NULL)
6477				return (-28);
6478			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6479			hmacs_len = plen - sizeof(*hmacs);
6480			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6481			/* validate the hmac list */
6482			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6483				return (-29);
6484			}
6485			if (stcb->asoc.peer_hmacs != NULL)
6486				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6487			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6488			if (stcb->asoc.peer_hmacs != NULL) {
6489				for (i = 0; i < num_hmacs; i++) {
6490					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6491					    ntohs(hmacs->hmac_ids[i]));
6492				}
6493			}
6494			got_hmacs = 1;
6495		} else if (ptype == SCTP_CHUNK_LIST) {
6496			int i;
6497
6498			if (plen > sizeof(chunks_store))
6499				break;
6500			if (got_chklist) {
6501				/* already processed a Chunks list */
6502				goto next_param;
6503			}
6504			phdr = sctp_get_next_param(m, offset,
6505			    (struct sctp_paramhdr *)chunks_store,
6506			    min(plen, sizeof(chunks_store)));
6507			if (phdr == NULL)
6508				return (-30);
6509			chunks = (struct sctp_auth_chunk_list *)phdr;
6510			num_chunks = plen - sizeof(*chunks);
6511			if (stcb->asoc.peer_auth_chunks != NULL)
6512				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6513			else
6514				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6515			for (i = 0; i < num_chunks; i++) {
6516				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6517				    stcb->asoc.peer_auth_chunks);
6518				/* record asconf/asconf-ack if listed */
6519				if (chunks->chunk_types[i] == SCTP_ASCONF)
6520					saw_asconf = 1;
6521				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6522					saw_asconf_ack = 1;
6523
6524			}
6525			got_chklist = 1;
6526		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6527			    (ptype == SCTP_STATE_COOKIE) ||
6528			    (ptype == SCTP_UNRECOG_PARAM) ||
6529			    (ptype == SCTP_COOKIE_PRESERVE) ||
6530			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6531			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6532			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6533			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6534		    (ptype == SCTP_SUCCESS_REPORT)) {
6535			 /* don't care */ ;
6536		} else {
6537			if ((ptype & 0x8000) == 0x0000) {
6538				/*
6539				 * must stop processing the rest of the
6540				 * param's. Any report bits were handled
6541				 * with the call to
6542				 * sctp_arethere_unrecognized_parameters()
6543				 * when the INIT or INIT-ACK was first seen.
6544				 */
6545				break;
6546			}
6547		}
6548
6549next_param:
6550		offset += SCTP_SIZE32(plen);
6551		if (offset >= limit) {
6552			break;
6553		}
6554		phdr = sctp_get_next_param(m, offset, &parm_buf,
6555		    sizeof(parm_buf));
6556	}
6557	/* Now check to see if we need to purge any addresses */
6558	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6559		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6560		    SCTP_ADDR_NOT_IN_ASSOC) {
6561			/* This address has been removed from the asoc */
6562			/* remove and free it */
6563			stcb->asoc.numnets--;
6564			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6565			sctp_free_remote_addr(net);
6566			if (net == stcb->asoc.primary_destination) {
6567				stcb->asoc.primary_destination = NULL;
6568				sctp_select_primary_destination(stcb);
6569			}
6570		}
6571	}
6572	if (ecn_allowed == 0) {
6573		stcb->asoc.ecn_allowed = 0;
6574	}
6575	/* validate authentication required parameters */
6576	if (got_random && got_hmacs) {
6577		stcb->asoc.peer_supports_auth = 1;
6578	} else {
6579		stcb->asoc.peer_supports_auth = 0;
6580	}
6581	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6582		/* peer does not support auth but sent a chunks list? */
6583		return (-31);
6584	}
6585	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6586	    !stcb->asoc.peer_supports_auth) {
6587		/* peer supports asconf but not auth? */
6588		return (-32);
6589	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6590	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6591		return (-33);
6592	}
6593	/* concatenate the full random key */
6594	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6595	if (chunks != NULL) {
6596		keylen += sizeof(*chunks) + num_chunks;
6597	}
6598	new_key = sctp_alloc_key(keylen);
6599	if (new_key != NULL) {
6600		/* copy in the RANDOM */
6601		if (p_random != NULL) {
6602			keylen = sizeof(*p_random) + random_len;
6603			bcopy(p_random, new_key->key, keylen);
6604		}
6605		/* append in the AUTH chunks */
6606		if (chunks != NULL) {
6607			bcopy(chunks, new_key->key + keylen,
6608			    sizeof(*chunks) + num_chunks);
6609			keylen += sizeof(*chunks) + num_chunks;
6610		}
6611		/* append in the HMACs */
6612		if (hmacs != NULL) {
6613			bcopy(hmacs, new_key->key + keylen,
6614			    sizeof(*hmacs) + hmacs_len);
6615		}
6616	} else {
6617		/* failed to get memory for the key */
6618		return (-34);
6619	}
6620	if (stcb->asoc.authinfo.peer_random != NULL)
6621		sctp_free_key(stcb->asoc.authinfo.peer_random);
6622	stcb->asoc.authinfo.peer_random = new_key;
6623	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6624	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6625
6626	return (0);
6627}
6628
6629int
6630sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6631    struct sctp_nets *net)
6632{
6633	/* make sure the requested primary address exists in the assoc */
6634	if (net == NULL && sa)
6635		net = sctp_findnet(stcb, sa);
6636
6637	if (net == NULL) {
6638		/* didn't find the requested primary address! */
6639		return (-1);
6640	} else {
6641		/* set the primary address */
6642		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6643			/* Must be confirmed, so queue to set */
6644			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6645			return (0);
6646		}
6647		stcb->asoc.primary_destination = net;
6648		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6649			sctp_free_remote_addr(stcb->asoc.alternate);
6650			stcb->asoc.alternate = NULL;
6651		}
6652		net = TAILQ_FIRST(&stcb->asoc.nets);
6653		if (net != stcb->asoc.primary_destination) {
6654			/*
6655			 * first one on the list is NOT the primary
6656			 * sctp_cmpaddr() is much more efficient if the
6657			 * primary is the first on the list, make it so.
6658			 */
6659			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6660			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6661		}
6662		return (0);
6663	}
6664}
6665
6666int
6667sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6668{
6669	/*
6670	 * This function serves two purposes. It will see if a TAG can be
6671	 * re-used and return 1 for yes it is ok and 0 for don't use that
6672	 * tag. A secondary function it will do is purge out old tags that
6673	 * can be removed.
6674	 */
6675	struct sctpvtaghead *chain;
6676	struct sctp_tagblock *twait_block;
6677	struct sctpasochead *head;
6678	struct sctp_tcb *stcb;
6679	int i;
6680
6681	SCTP_INP_INFO_RLOCK();
6682	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6683	    SCTP_BASE_INFO(hashasocmark))];
6684	if (head == NULL) {
6685		/* invalid vtag */
6686		goto skip_vtag_check;
6687	}
6688	LIST_FOREACH(stcb, head, sctp_asocs) {
6689		/*
6690		 * We choose not to lock anything here. TCB's can't be
6691		 * removed since we have the read lock, so they can't be
6692		 * freed on us, same thing for the INP. I may be wrong with
6693		 * this assumption, but we will go with it for now :-)
6694		 */
6695		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6696			continue;
6697		}
6698		if (stcb->asoc.my_vtag == tag) {
6699			/* candidate */
6700			if (stcb->rport != rport) {
6701				continue;
6702			}
6703			if (stcb->sctp_ep->sctp_lport != lport) {
6704				continue;
6705			}
6706			/* Its a used tag set */
6707			SCTP_INP_INFO_RUNLOCK();
6708			return (0);
6709		}
6710	}
6711skip_vtag_check:
6712
6713	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6714	/* Now what about timed wait ? */
6715	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6716		/*
6717		 * Block(s) are present, lets see if we have this tag in the
6718		 * list
6719		 */
6720		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6721			if (twait_block->vtag_block[i].v_tag == 0) {
6722				/* not used */
6723				continue;
6724			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6725			    now->tv_sec) {
6726				/* Audit expires this guy */
6727				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6728				twait_block->vtag_block[i].v_tag = 0;
6729				twait_block->vtag_block[i].lport = 0;
6730				twait_block->vtag_block[i].rport = 0;
6731			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6732				    (twait_block->vtag_block[i].lport == lport) &&
6733			    (twait_block->vtag_block[i].rport == rport)) {
6734				/* Bad tag, sorry :< */
6735				SCTP_INP_INFO_RUNLOCK();
6736				return (0);
6737			}
6738		}
6739	}
6740	SCTP_INP_INFO_RUNLOCK();
6741	return (1);
6742}
6743
6744static void
6745sctp_drain_mbufs(struct sctp_tcb *stcb)
6746{
6747	/*
6748	 * We must hunt this association for MBUF's past the cumack (i.e.
6749	 * out of order data that we can renege on).
6750	 */
6751	struct sctp_association *asoc;
6752	struct sctp_tmit_chunk *chk, *nchk;
6753	uint32_t cumulative_tsn_p1;
6754	struct sctp_queued_to_read *ctl, *nctl;
6755	int cnt, strmat;
6756	uint32_t gap, i;
6757	int fnd = 0;
6758
6759	/* We look for anything larger than the cum-ack + 1 */
6760
6761	asoc = &stcb->asoc;
6762	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6763		/* none we can reneg on. */
6764		return;
6765	}
6766	SCTP_STAT_INCR(sctps_protocol_drains_done);
6767	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6768	cnt = 0;
6769	/* First look in the re-assembly queue */
6770	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6771		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6772			/* Yep it is above cum-ack */
6773			cnt++;
6774			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6775			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6776			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6777			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6778			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6779			if (chk->data) {
6780				sctp_m_freem(chk->data);
6781				chk->data = NULL;
6782			}
6783			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6784		}
6785	}
6786	/* Ok that was fun, now we will drain all the inbound streams? */
6787	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6788		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6789			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6790				/* Yep it is above cum-ack */
6791				cnt++;
6792				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6793				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6794				sctp_ucount_decr(asoc->cnt_on_all_streams);
6795				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6796				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6797				if (ctl->data) {
6798					sctp_m_freem(ctl->data);
6799					ctl->data = NULL;
6800				}
6801				sctp_free_remote_addr(ctl->whoFrom);
6802				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6803				SCTP_DECR_READQ_COUNT();
6804			}
6805		}
6806	}
6807	if (cnt) {
6808		/* We must back down to see what the new highest is */
6809		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6810			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6811			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6812				asoc->highest_tsn_inside_map = i;
6813				fnd = 1;
6814				break;
6815			}
6816		}
6817		if (!fnd) {
6818			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6819		}
6820		/*
6821		 * Question, should we go through the delivery queue? The
6822		 * only reason things are on here is the app not reading OR
6823		 * a p-d-api up. An attacker COULD send enough in to
6824		 * initiate the PD-API and then send a bunch of stuff to
6825		 * other streams... these would wind up on the delivery
6826		 * queue.. and then we would not get to them. But in order
6827		 * to do this I then have to back-track and un-deliver
6828		 * sequence numbers in streams.. el-yucko. I think for now
6829		 * we will NOT look at the delivery queue and leave it to be
6830		 * something to consider later. An alternative would be to
6831		 * abort the P-D-API with a notification and then deliver
6832		 * the data.... Or another method might be to keep track of
6833		 * how many times the situation occurs and if we see a
6834		 * possible attack underway just abort the association.
6835		 */
6836#ifdef SCTP_DEBUG
6837		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6838#endif
6839		/*
6840		 * Now do we need to find a new
6841		 * asoc->highest_tsn_inside_map?
6842		 */
6843		asoc->last_revoke_count = cnt;
6844		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6845		/* sa_ignore NO_NULL_CHK */
6846		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6847		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6848	}
6849	/*
6850	 * Another issue, in un-setting the TSN's in the mapping array we
6851	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6852	 * two things to occur. It may cause us to do extra work in checking
6853	 * for our mapping array movement. More importantly it may cause us
6854	 * to SACK every datagram. This may not be a bad thing though since
6855	 * we will recover once we get our cum-ack above and all this stuff
6856	 * we dumped recovered.
6857	 */
6858}
6859
6860void
6861sctp_drain()
6862{
6863	/*
6864	 * We must walk the PCB lists for ALL associations here. The system
6865	 * is LOW on MBUF's and needs help. This is where reneging will
6866	 * occur. We really hope this does NOT happen!
6867	 */
6868	VNET_ITERATOR_DECL(vnet_iter);
6869	VNET_LIST_RLOCK_NOSLEEP();
6870	VNET_FOREACH(vnet_iter) {
6871		CURVNET_SET(vnet_iter);
6872		struct sctp_inpcb *inp;
6873		struct sctp_tcb *stcb;
6874
6875		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6876		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6877#ifdef VIMAGE
6878			continue;
6879#else
6880			return;
6881#endif
6882		}
6883		SCTP_INP_INFO_RLOCK();
6884		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6885			/* For each endpoint */
6886			SCTP_INP_RLOCK(inp);
6887			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6888				/* For each association */
6889				SCTP_TCB_LOCK(stcb);
6890				sctp_drain_mbufs(stcb);
6891				SCTP_TCB_UNLOCK(stcb);
6892			}
6893			SCTP_INP_RUNLOCK(inp);
6894		}
6895		SCTP_INP_INFO_RUNLOCK();
6896		CURVNET_RESTORE();
6897	}
6898	VNET_LIST_RUNLOCK_NOSLEEP();
6899}
6900
6901/*
6902 * start a new iterator
6903 * iterates through all endpoints and associations based on the pcb_state
6904 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6905 * assocs and "ef" (optional) is executed when the iterator completes.
6906 * "inpf" (optional) is executed for each new endpoint as it is being
6907 * iterated through. inpe (optional) is called when the inp completes
6908 * its way through all the stcbs.
6909 */
6910int
6911sctp_initiate_iterator(inp_func inpf,
6912    asoc_func af,
6913    inp_func inpe,
6914    uint32_t pcb_state,
6915    uint32_t pcb_features,
6916    uint32_t asoc_state,
6917    void *argp,
6918    uint32_t argi,
6919    end_func ef,
6920    struct sctp_inpcb *s_inp,
6921    uint8_t chunk_output_off)
6922{
6923	struct sctp_iterator *it = NULL;
6924
6925	if (af == NULL) {
6926		return (-1);
6927	}
6928	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6929	    SCTP_M_ITER);
6930	if (it == NULL) {
6931		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6932		return (ENOMEM);
6933	}
6934	memset(it, 0, sizeof(*it));
6935	it->function_assoc = af;
6936	it->function_inp = inpf;
6937	if (inpf)
6938		it->done_current_ep = 0;
6939	else
6940		it->done_current_ep = 1;
6941	it->function_atend = ef;
6942	it->pointer = argp;
6943	it->val = argi;
6944	it->pcb_flags = pcb_state;
6945	it->pcb_features = pcb_features;
6946	it->asoc_state = asoc_state;
6947	it->function_inp_end = inpe;
6948	it->no_chunk_output = chunk_output_off;
6949	it->vn = curvnet;
6950	if (s_inp) {
6951		/* Assume lock is held here */
6952		it->inp = s_inp;
6953		SCTP_INP_INCR_REF(it->inp);
6954		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6955	} else {
6956		SCTP_INP_INFO_RLOCK();
6957		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6958		if (it->inp) {
6959			SCTP_INP_INCR_REF(it->inp);
6960		}
6961		SCTP_INP_INFO_RUNLOCK();
6962		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6963
6964	}
6965	SCTP_IPI_ITERATOR_WQ_LOCK();
6966
6967	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6968	if (sctp_it_ctl.iterator_running == 0) {
6969		sctp_wakeup_iterator();
6970	}
6971	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6972	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6973	return (0);
6974}
6975