1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
5 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
6 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * a) Redistributions of source code must retain the above copyright notice,
12 *    this list of conditions and the following disclaimer.
13 *
14 * b) Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in
16 *    the documentation and/or other materials provided with the distribution.
17 *
18 * c) Neither the name of Cisco Systems, Inc. nor the names of its
19 *    contributors may be used to endorse or promote products derived
20 *    from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
26 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
29 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
30 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
31 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
32 * THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35#include <netinet/sctp_os.h>
36#include <sys/proc.h>
37#include <netinet/sctp_var.h>
38#include <netinet/sctp_sysctl.h>
39#include <netinet/sctp_pcb.h>
40#include <netinet/sctputil.h>
41#include <netinet/sctp.h>
42#include <netinet/sctp_header.h>
43#include <netinet/sctp_asconf.h>
44#include <netinet/sctp_output.h>
45#include <netinet/sctp_timer.h>
46#include <netinet/sctp_bsd_addr.h>
47#if defined(INET) || defined(INET6)
48#include <netinet/udp.h>
49#endif
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/* FIX: we don't handle multiple link local scopes */
58/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
59#ifdef INET6
60int
61SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
62{
63	struct sockaddr_in6 tmp_a, tmp_b;
64
65	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
66	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
67		return (0);
68	}
69	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
70	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
71		return (0);
72	}
73	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
74}
75#endif
76
77void
78sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
79{
80	/*
81	 * We really don't need to lock this, but I will just because it
82	 * does not hurt.
83	 */
84	SCTP_INP_INFO_RLOCK();
85	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
86	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
87	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
88	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
89	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
90	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
91	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
92	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
93	SCTP_INP_INFO_RUNLOCK();
94}
95
96/*-
97 * Addresses are added to VRF's (Virtual Router's). For BSD we
98 * have only the default VRF 0. We maintain a hash list of
99 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
100 * these has a list of addresses. When we add a new address
101 * to a VRF we lookup the ifn/ifn_index, if the ifn does
102 * not exist we create it and add it to the list of IFN's
103 * within the VRF. Once we have the sctp_ifn, we add the
104 * address to the list. So we look something like:
105 *
106 * hash-vrf-table
107 *   vrf-> ifn-> ifn -> ifn
108 *   vrf    |
109 *    ...   +--ifa-> ifa -> ifa
110 *   vrf
111 *
112 * We keep these separate lists since the SCTP subsystem will
113 * point to these from its source address selection nets structure.
114 * When an address is deleted it does not happen right away on
115 * the SCTP side, it gets scheduled. What we do when a
116 * delete happens is immediately remove the address from
117 * the master list and decrement the refcount. As our
118 * addip iterator works through and frees the src address
119 * selection pointing to the sctp_ifa, eventually the refcount
120 * will reach 0 and we will delete it. Note that it is assumed
121 * that any locking on system level ifn/ifa is done at the
122 * caller of these functions and these routines will only
123 * lock the SCTP structures as they add or delete things.
124 *
125 * Other notes on VRF concepts.
126 *  - An endpoint can be in multiple VRF's
127 *  - An association lives within a VRF and only one VRF.
128 *  - Any incoming packet we can deduce the VRF for by
129 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
130 *  - Any downward send call or connect call must supply the
131 *    VRF via ancillary data or via some sort of set default
132 *    VRF socket option call (again for BSD no brainer since
133 *    the VRF is always 0).
134 *  - An endpoint may add multiple VRF's to it.
135 *  - Listening sockets can accept associations in any
136 *    of the VRF's they are in but the assoc will end up
137 *    in only one VRF (gotten from the packet or connect/send).
138 *
139 */
140
141struct sctp_vrf *
142sctp_allocate_vrf(int vrf_id)
143{
144	struct sctp_vrf *vrf = NULL;
145	struct sctp_vrflist *bucket;
146
147	/* First allocate the VRF structure */
148	vrf = sctp_find_vrf(vrf_id);
149	if (vrf) {
150		/* Already allocated */
151		return (vrf);
152	}
153	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
154	    SCTP_M_VRF);
155	if (vrf == NULL) {
156		/* No memory */
157#ifdef INVARIANTS
158		panic("No memory for VRF:%d", vrf_id);
159#endif
160		return (NULL);
161	}
162	/* setup the VRF */
163	memset(vrf, 0, sizeof(struct sctp_vrf));
164	vrf->vrf_id = vrf_id;
165	LIST_INIT(&vrf->ifnlist);
166	vrf->total_ifa_count = 0;
167	vrf->refcount = 0;
168	/* now also setup table ids */
169	SCTP_INIT_VRF_TABLEID(vrf);
170	/* Init the HASH of addresses */
171	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
172	    &vrf->vrf_addr_hashmark);
173	if (vrf->vrf_addr_hash == NULL) {
174		/* No memory */
175#ifdef INVARIANTS
176		panic("No memory for VRF:%d", vrf_id);
177#endif
178		SCTP_FREE(vrf, SCTP_M_VRF);
179		return (NULL);
180	}
181
182	/* Add it to the hash table */
183	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
184	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
185	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
186	return (vrf);
187}
188
189struct sctp_ifn *
190sctp_find_ifn(void *ifn, uint32_t ifn_index)
191{
192	struct sctp_ifn *sctp_ifnp;
193	struct sctp_ifnlist *hash_ifn_head;
194
195	/*
196	 * We assume the lock is held for the addresses if that's wrong
197	 * problems could occur :-)
198	 */
199	SCTP_IPI_ADDR_LOCK_ASSERT();
200	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
201	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
202		if (sctp_ifnp->ifn_index == ifn_index) {
203			return (sctp_ifnp);
204		}
205		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
206			return (sctp_ifnp);
207		}
208	}
209	return (NULL);
210}
211
212struct sctp_vrf *
213sctp_find_vrf(uint32_t vrf_id)
214{
215	struct sctp_vrflist *bucket;
216	struct sctp_vrf *liste;
217
218	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
219	LIST_FOREACH(liste, bucket, next_vrf) {
220		if (vrf_id == liste->vrf_id) {
221			return (liste);
222		}
223	}
224	return (NULL);
225}
226
227void
228sctp_free_vrf(struct sctp_vrf *vrf)
229{
230	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
231		if (vrf->vrf_addr_hash) {
232			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
233			vrf->vrf_addr_hash = NULL;
234		}
235		/* We zero'd the count */
236		LIST_REMOVE(vrf, next_vrf);
237		SCTP_FREE(vrf, SCTP_M_VRF);
238		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
239	}
240}
241
242void
243sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
244{
245	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
246		/* We zero'd the count */
247		if (sctp_ifnp->vrf) {
248			sctp_free_vrf(sctp_ifnp->vrf);
249		}
250		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
251		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
252	}
253}
254
255void
256sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
257{
258	struct sctp_ifn *sctp_ifnp;
259
260	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
261	if (sctp_ifnp != NULL) {
262		sctp_ifnp->ifn_mtu = mtu;
263	}
264}
265
266void
267sctp_free_ifa(struct sctp_ifa *sctp_ifap)
268{
269	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
270		/* We zero'd the count */
271		if (sctp_ifap->ifn_p) {
272			sctp_free_ifn(sctp_ifap->ifn_p);
273		}
274		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
275		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
276	}
277}
278
279static void
280sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
281{
282	struct sctp_ifn *found;
283
284	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
285	if (found == NULL) {
286		/* Not in the list.. sorry */
287		return;
288	}
289	if (hold_addr_lock == 0) {
290		SCTP_IPI_ADDR_WLOCK();
291	} else {
292		SCTP_IPI_ADDR_WLOCK_ASSERT();
293	}
294	LIST_REMOVE(sctp_ifnp, next_bucket);
295	LIST_REMOVE(sctp_ifnp, next_ifn);
296	if (hold_addr_lock == 0) {
297		SCTP_IPI_ADDR_WUNLOCK();
298	}
299	/* Take away the reference, and possibly free it */
300	sctp_free_ifn(sctp_ifnp);
301}
302
303void
304sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
305    const char *if_name, uint32_t ifn_index)
306{
307	struct sctp_vrf *vrf;
308	struct sctp_ifa *sctp_ifap;
309
310	SCTP_IPI_ADDR_RLOCK();
311	vrf = sctp_find_vrf(vrf_id);
312	if (vrf == NULL) {
313		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
314		goto out;
315	}
316	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
317	if (sctp_ifap == NULL) {
318		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
319		goto out;
320	}
321	if (sctp_ifap->ifn_p == NULL) {
322		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
323		goto out;
324	}
325	if (if_name) {
326		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
327			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
328			    sctp_ifap->ifn_p->ifn_name, if_name);
329			goto out;
330		}
331	} else {
332		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
333			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
334			    sctp_ifap->ifn_p->ifn_index, ifn_index);
335			goto out;
336		}
337	}
338
339	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
340	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
341out:
342	SCTP_IPI_ADDR_RUNLOCK();
343}
344
345void
346sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
347    const char *if_name, uint32_t ifn_index)
348{
349	struct sctp_vrf *vrf;
350	struct sctp_ifa *sctp_ifap;
351
352	SCTP_IPI_ADDR_RLOCK();
353	vrf = sctp_find_vrf(vrf_id);
354	if (vrf == NULL) {
355		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
356		goto out;
357	}
358	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
359	if (sctp_ifap == NULL) {
360		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
361		goto out;
362	}
363	if (sctp_ifap->ifn_p == NULL) {
364		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
365		goto out;
366	}
367	if (if_name) {
368		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
369			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
370			    sctp_ifap->ifn_p->ifn_name, if_name);
371			goto out;
372		}
373	} else {
374		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
375			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
376			    sctp_ifap->ifn_p->ifn_index, ifn_index);
377			goto out;
378		}
379	}
380
381	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
382	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
383out:
384	SCTP_IPI_ADDR_RUNLOCK();
385}
386
387/*-
388 * Add an ifa to an ifn.
389 * Register the interface as necessary.
390 * NOTE: ADDR write lock MUST be held.
391 */
392static void
393sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
394{
395	int ifa_af;
396
397	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
398	sctp_ifap->ifn_p = sctp_ifnp;
399	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
400	/* update address counts */
401	sctp_ifnp->ifa_count++;
402	ifa_af = sctp_ifap->address.sa.sa_family;
403	switch (ifa_af) {
404#ifdef INET
405	case AF_INET:
406		sctp_ifnp->num_v4++;
407		break;
408#endif
409#ifdef INET6
410	case AF_INET6:
411		sctp_ifnp->num_v6++;
412		break;
413#endif
414	default:
415		break;
416	}
417	if (sctp_ifnp->ifa_count == 1) {
418		/* register the new interface */
419		sctp_ifnp->registered_af = ifa_af;
420	}
421}
422
423/*-
424 * Remove an ifa from its ifn.
425 * If no more addresses exist, remove the ifn too. Otherwise, re-register
426 * the interface based on the remaining address families left.
427 * NOTE: ADDR write lock MUST be held.
428 */
429static void
430sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
431{
432	LIST_REMOVE(sctp_ifap, next_ifa);
433	if (sctp_ifap->ifn_p) {
434		/* update address counts */
435		sctp_ifap->ifn_p->ifa_count--;
436		switch (sctp_ifap->address.sa.sa_family) {
437#ifdef INET
438		case AF_INET:
439			sctp_ifap->ifn_p->num_v4--;
440			break;
441#endif
442#ifdef INET6
443		case AF_INET6:
444			sctp_ifap->ifn_p->num_v6--;
445			break;
446#endif
447		default:
448			break;
449		}
450
451		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
452			/* remove the ifn, possibly freeing it */
453			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
454		} else {
455			/* re-register address family type, if needed */
456			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
457			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
458				sctp_ifap->ifn_p->registered_af = AF_INET;
459			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
460			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
461				sctp_ifap->ifn_p->registered_af = AF_INET6;
462			}
463			/* free the ifn refcount */
464			sctp_free_ifn(sctp_ifap->ifn_p);
465		}
466		sctp_ifap->ifn_p = NULL;
467	}
468}
469
470struct sctp_ifa *
471sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
472    uint32_t ifn_type, const char *if_name, void *ifa,
473    struct sockaddr *addr, uint32_t ifa_flags,
474    int dynamic_add)
475{
476	struct sctp_vrf *vrf;
477	struct sctp_ifn *sctp_ifnp, *new_sctp_ifnp;
478	struct sctp_ifa *sctp_ifap, *new_sctp_ifap;
479	struct sctp_ifalist *hash_addr_head;
480	struct sctp_ifnlist *hash_ifn_head;
481	uint32_t hash_of_addr;
482	int new_ifn_af = 0;
483
484#ifdef SCTP_DEBUG
485	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
486	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
487#endif
488	SCTP_MALLOC(new_sctp_ifnp, struct sctp_ifn *,
489	    sizeof(struct sctp_ifn), SCTP_M_IFN);
490	if (new_sctp_ifnp == NULL) {
491#ifdef INVARIANTS
492		panic("No memory for IFN");
493#endif
494		return (NULL);
495	}
496	SCTP_MALLOC(new_sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
497	if (new_sctp_ifap == NULL) {
498#ifdef INVARIANTS
499		panic("No memory for IFA");
500#endif
501		SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
502		return (NULL);
503	}
504
505	SCTP_IPI_ADDR_WLOCK();
506	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
507	if (sctp_ifnp) {
508		vrf = sctp_ifnp->vrf;
509	} else {
510		vrf = sctp_find_vrf(vrf_id);
511		if (vrf == NULL) {
512			vrf = sctp_allocate_vrf(vrf_id);
513			if (vrf == NULL) {
514				SCTP_IPI_ADDR_WUNLOCK();
515				SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
516				SCTP_FREE(new_sctp_ifap, SCTP_M_IFA);
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_ifnp = new_sctp_ifnp;
527		new_sctp_ifnp = NULL;
528		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
529		sctp_ifnp->ifn_index = ifn_index;
530		sctp_ifnp->ifn_p = ifn;
531		sctp_ifnp->ifn_type = ifn_type;
532		sctp_ifnp->refcount = 0;
533		sctp_ifnp->vrf = vrf;
534		atomic_add_int(&vrf->refcount, 1);
535		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index);
536		if (if_name != NULL) {
537			SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
538		} else {
539			SCTP_SNPRINTF(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
540		}
541		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
542		LIST_INIT(&sctp_ifnp->ifalist);
543		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
544		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
545		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
546		new_ifn_af = 1;
547	}
548	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
549	if (sctp_ifap) {
550		/* Hmm, it already exists? */
551		if ((sctp_ifap->ifn_p) &&
552		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
553			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
554			    sctp_ifap->ifn_p->ifn_name, ifn_index,
555			    (void *)sctp_ifap);
556			if (new_ifn_af) {
557				/* Remove the created one that we don't want */
558				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
559			}
560			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
561				/* easy to solve, just switch back to active */
562				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
563				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
564				sctp_ifap->ifn_p = sctp_ifnp;
565				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
566			}
567	exit_stage_left:
568			SCTP_IPI_ADDR_WUNLOCK();
569			if (new_sctp_ifnp != NULL) {
570				SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
571			}
572			SCTP_FREE(new_sctp_ifap, SCTP_M_IFA);
573			return (sctp_ifap);
574		} else {
575			if (sctp_ifap->ifn_p) {
576				/*
577				 * The last IFN gets the address, remove the
578				 * old one
579				 */
580				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
581				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
582				    sctp_ifap->ifn_p->ifn_index, if_name,
583				    ifn_index);
584				/* remove the address from the old ifn */
585				sctp_remove_ifa_from_ifn(sctp_ifap);
586				/* move the address over to the new ifn */
587				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
588				goto exit_stage_left;
589			} else {
590				/* repair ifnp which was NULL ? */
591				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
592				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
593				    (void *)sctp_ifnp, (void *)sctp_ifap);
594				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
595			}
596			goto exit_stage_left;
597		}
598	}
599	sctp_ifap = new_sctp_ifap;
600	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
601	sctp_ifap->ifn_p = sctp_ifnp;
602	atomic_add_int(&sctp_ifnp->refcount, 1);
603	sctp_ifap->vrf_id = vrf_id;
604	sctp_ifap->ifa = ifa;
605	memcpy(&sctp_ifap->address, addr, addr->sa_len);
606	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
607	sctp_ifap->flags = ifa_flags;
608	/* Set scope */
609	switch (sctp_ifap->address.sa.sa_family) {
610#ifdef INET
611	case AF_INET:
612		{
613			struct sockaddr_in *sin;
614
615			sin = &sctp_ifap->address.sin;
616			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
617			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
618				sctp_ifap->src_is_loop = 1;
619			}
620			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
621				sctp_ifap->src_is_priv = 1;
622			}
623			sctp_ifnp->num_v4++;
624			if (new_ifn_af)
625				new_ifn_af = AF_INET;
626			break;
627		}
628#endif
629#ifdef INET6
630	case AF_INET6:
631		{
632			/* ok to use deprecated addresses? */
633			struct sockaddr_in6 *sin6;
634
635			sin6 = &sctp_ifap->address.sin6;
636			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
637			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
638				sctp_ifap->src_is_loop = 1;
639			}
640			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
641				sctp_ifap->src_is_priv = 1;
642			}
643			sctp_ifnp->num_v6++;
644			if (new_ifn_af)
645				new_ifn_af = AF_INET6;
646			break;
647		}
648#endif
649	default:
650		new_ifn_af = 0;
651		break;
652	}
653	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
654
655	if ((sctp_ifap->src_is_priv == 0) &&
656	    (sctp_ifap->src_is_loop == 0)) {
657		sctp_ifap->src_is_glob = 1;
658	}
659	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
660	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
661	sctp_ifap->refcount = 1;
662	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
663	sctp_ifnp->ifa_count++;
664	vrf->total_ifa_count++;
665	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
666	if (new_ifn_af) {
667		sctp_ifnp->registered_af = new_ifn_af;
668	}
669	SCTP_IPI_ADDR_WUNLOCK();
670	if (new_sctp_ifnp != NULL) {
671		SCTP_FREE(new_sctp_ifnp, SCTP_M_IFN);
672	}
673
674	if (dynamic_add) {
675		/*
676		 * Bump up the refcount so that when the timer completes it
677		 * will drop back down.
678		 */
679		struct sctp_laddr *wi;
680
681		atomic_add_int(&sctp_ifap->refcount, 1);
682		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
683		if (wi == NULL) {
684			/*
685			 * Gak, what can we do? We have lost an address
686			 * change can you say HOSED?
687			 */
688			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
689			/* Opps, must decrement the count */
690			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
691			    if_name);
692			return (NULL);
693		}
694		SCTP_INCR_LADDR_COUNT();
695		memset(wi, 0, sizeof(*wi));
696		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
697		wi->ifa = sctp_ifap;
698		wi->action = SCTP_ADD_IP_ADDRESS;
699
700		SCTP_WQ_ADDR_LOCK();
701		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
702		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
703		    (struct sctp_inpcb *)NULL,
704		    (struct sctp_tcb *)NULL,
705		    (struct sctp_nets *)NULL);
706		SCTP_WQ_ADDR_UNLOCK();
707	} else {
708		/* it's ready for use */
709		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
710	}
711	return (sctp_ifap);
712}
713
714void
715sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
716    uint32_t ifn_index, const char *if_name)
717{
718	struct sctp_vrf *vrf;
719	struct sctp_ifa *sctp_ifap = NULL;
720
721	SCTP_IPI_ADDR_WLOCK();
722	vrf = sctp_find_vrf(vrf_id);
723	if (vrf == NULL) {
724		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
725		goto out_now;
726	}
727
728#ifdef SCTP_DEBUG
729	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
730	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
731#endif
732	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
733	if (sctp_ifap) {
734		/* Validate the delete */
735		if (sctp_ifap->ifn_p) {
736			int valid = 0;
737
738			/*-
739			 * The name has priority over the ifn_index
740			 * if its given.
741			 */
742			if (if_name) {
743				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
744					/* They match its a correct delete */
745					valid = 1;
746				}
747			}
748			if (!valid) {
749				/* last ditch check ifn_index */
750				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
751					valid = 1;
752				}
753			}
754			if (!valid) {
755				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
756				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
757				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
758				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
759				SCTP_IPI_ADDR_WUNLOCK();
760				return;
761			}
762		}
763		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
764		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
765		/*
766		 * We don't set the flag. This means that the structure will
767		 * hang around in EP's that have bound specific to it until
768		 * they close. This gives us TCP like behavior if someone
769		 * removes an address (or for that matter adds it right
770		 * back).
771		 */
772		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
773		vrf->total_ifa_count--;
774		LIST_REMOVE(sctp_ifap, next_bucket);
775		sctp_remove_ifa_from_ifn(sctp_ifap);
776	}
777#ifdef SCTP_DEBUG
778	else {
779		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
780		    ifn_index);
781		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
782	}
783#endif
784
785out_now:
786	SCTP_IPI_ADDR_WUNLOCK();
787	if (sctp_ifap) {
788		struct sctp_laddr *wi;
789
790		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
791		if (wi == NULL) {
792			/*
793			 * Gak, what can we do? We have lost an address
794			 * change can you say HOSED?
795			 */
796			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
797
798			/* Oops, must decrement the count */
799			sctp_free_ifa(sctp_ifap);
800			return;
801		}
802		SCTP_INCR_LADDR_COUNT();
803		memset(wi, 0, sizeof(*wi));
804		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
805		wi->ifa = sctp_ifap;
806		wi->action = SCTP_DEL_IP_ADDRESS;
807		SCTP_WQ_ADDR_LOCK();
808		/*
809		 * Should this really be a tailq? As it is we will process
810		 * the newest first :-0
811		 */
812		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
813		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
814		    (struct sctp_inpcb *)NULL,
815		    (struct sctp_tcb *)NULL,
816		    (struct sctp_nets *)NULL);
817		SCTP_WQ_ADDR_UNLOCK();
818	}
819	return;
820}
821
822static int
823sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
824{
825	int loopback_scope;
826#if defined(INET)
827	int ipv4_local_scope, ipv4_addr_legal;
828#endif
829#if defined(INET6)
830	int local_scope, site_scope, ipv6_addr_legal;
831#endif
832	struct sctp_vrf *vrf;
833	struct sctp_ifn *sctp_ifn;
834	struct sctp_ifa *sctp_ifa;
835
836	loopback_scope = stcb->asoc.scope.loopback_scope;
837#if defined(INET)
838	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
839	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
840#endif
841#if defined(INET6)
842	local_scope = stcb->asoc.scope.local_scope;
843	site_scope = stcb->asoc.scope.site_scope;
844	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
845#endif
846
847	SCTP_IPI_ADDR_RLOCK();
848	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
849	if (vrf == NULL) {
850		/* no vrf, no addresses */
851		SCTP_IPI_ADDR_RUNLOCK();
852		return (0);
853	}
854
855	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
856		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
857			if ((loopback_scope == 0) &&
858			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
859				continue;
860			}
861			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
862				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
863				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
864					/*
865					 * We allow pending addresses, where
866					 * we have sent an asconf-add to be
867					 * considered valid.
868					 */
869					continue;
870				}
871				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
872					continue;
873				}
874				switch (sctp_ifa->address.sa.sa_family) {
875#ifdef INET
876				case AF_INET:
877					if (ipv4_addr_legal) {
878						struct sockaddr_in *sin,
879						           *rsin;
880
881						sin = &sctp_ifa->address.sin;
882						rsin = (struct sockaddr_in *)to;
883						if ((ipv4_local_scope == 0) &&
884						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
885							continue;
886						}
887						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
888						    &sin->sin_addr) != 0) {
889							continue;
890						}
891						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
892							SCTP_IPI_ADDR_RUNLOCK();
893							return (1);
894						}
895					}
896					break;
897#endif
898#ifdef INET6
899				case AF_INET6:
900					if (ipv6_addr_legal) {
901						struct sockaddr_in6 *sin6,
902						            *rsin6;
903
904						sin6 = &sctp_ifa->address.sin6;
905						rsin6 = (struct sockaddr_in6 *)to;
906						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
907						    &sin6->sin6_addr) != 0) {
908							continue;
909						}
910						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
911							if (local_scope == 0)
912								continue;
913							if (sin6->sin6_scope_id == 0) {
914								if (sa6_recoverscope(sin6) != 0)
915									continue;
916							}
917						}
918						if ((site_scope == 0) &&
919						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
920							continue;
921						}
922						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
923							SCTP_IPI_ADDR_RUNLOCK();
924							return (1);
925						}
926					}
927					break;
928#endif
929				default:
930					/* TSNH */
931					break;
932				}
933			}
934		}
935	} else {
936		struct sctp_laddr *laddr;
937
938		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
939			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
940				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
941				continue;
942			}
943			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
944			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
945				/*
946				 * We allow pending addresses, where we have
947				 * sent an asconf-add to be considered
948				 * valid.
949				 */
950				continue;
951			}
952			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
953				continue;
954			}
955			switch (to->sa_family) {
956#ifdef INET
957			case AF_INET:
958				{
959					struct sockaddr_in *sin, *rsin;
960
961					sin = &laddr->ifa->address.sin;
962					rsin = (struct sockaddr_in *)to;
963					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
964						SCTP_IPI_ADDR_RUNLOCK();
965						return (1);
966					}
967					break;
968				}
969#endif
970#ifdef INET6
971			case AF_INET6:
972				{
973					struct sockaddr_in6 *sin6, *rsin6;
974
975					sin6 = &laddr->ifa->address.sin6;
976					rsin6 = (struct sockaddr_in6 *)to;
977					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
978						SCTP_IPI_ADDR_RUNLOCK();
979						return (1);
980					}
981					break;
982				}
983
984#endif
985			default:
986				/* TSNH */
987				break;
988			}
989		}
990	}
991	SCTP_IPI_ADDR_RUNLOCK();
992	return (0);
993}
994
995static struct sctp_tcb *
996sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
997    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
998{
999	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1000	/*
1001	 * If we support the TCP model, then we must now dig through to see
1002	 * if we can find our endpoint in the list of tcp ep's.
1003	 */
1004	uint16_t lport, rport;
1005	struct sctppcbhead *ephead;
1006	struct sctp_inpcb *inp;
1007	struct sctp_laddr *laddr;
1008	struct sctp_tcb *stcb;
1009	struct sctp_nets *net;
1010
1011	if ((to == NULL) || (from == NULL)) {
1012		return (NULL);
1013	}
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		switch (to->sa_family) {
1057#ifdef INET
1058		case AF_INET:
1059			{
1060				struct sockaddr_in *sin;
1061
1062				sin = (struct sockaddr_in *)to;
1063				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1064				    &sin->sin_addr) != 0) {
1065					SCTP_INP_RUNLOCK(inp);
1066					continue;
1067				}
1068				break;
1069			}
1070#endif
1071#ifdef INET6
1072		case AF_INET6:
1073			{
1074				struct sockaddr_in6 *sin6;
1075
1076				sin6 = (struct sockaddr_in6 *)to;
1077				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1078				    &sin6->sin6_addr) != 0) {
1079					SCTP_INP_RUNLOCK(inp);
1080					continue;
1081				}
1082				break;
1083			}
1084#endif
1085		default:
1086			SCTP_INP_RUNLOCK(inp);
1087			continue;
1088		}
1089		if (inp->def_vrf_id != vrf_id) {
1090			SCTP_INP_RUNLOCK(inp);
1091			continue;
1092		}
1093		/* check to see if the ep has one of the addresses */
1094		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1095			/* We are NOT bound all, so look further */
1096			int match = 0;
1097
1098			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1099				if (laddr->ifa == NULL) {
1100					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__);
1101					continue;
1102				}
1103				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1104					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1105					continue;
1106				}
1107				if (laddr->ifa->address.sa.sa_family ==
1108				    to->sa_family) {
1109					/* see if it matches */
1110#ifdef INET
1111					if (from->sa_family == AF_INET) {
1112						struct sockaddr_in *intf_addr,
1113						           *sin;
1114
1115						intf_addr = &laddr->ifa->address.sin;
1116						sin = (struct sockaddr_in *)to;
1117						if (sin->sin_addr.s_addr ==
1118						    intf_addr->sin_addr.s_addr) {
1119							match = 1;
1120							break;
1121						}
1122					}
1123#endif
1124#ifdef INET6
1125					if (from->sa_family == AF_INET6) {
1126						struct sockaddr_in6 *intf_addr6;
1127						struct sockaddr_in6 *sin6;
1128
1129						sin6 = (struct sockaddr_in6 *)
1130						    to;
1131						intf_addr6 = &laddr->ifa->address.sin6;
1132
1133						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1134						    intf_addr6)) {
1135							match = 1;
1136							break;
1137						}
1138					}
1139#endif
1140				}
1141			}
1142			if (match == 0) {
1143				/* This endpoint does not have this address */
1144				SCTP_INP_RUNLOCK(inp);
1145				continue;
1146			}
1147		}
1148		/*
1149		 * Ok if we hit here the ep has the address, does it hold
1150		 * the tcb?
1151		 */
1152		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1153		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1154		if (stcb == NULL) {
1155			SCTP_INP_RUNLOCK(inp);
1156			continue;
1157		}
1158		SCTP_TCB_LOCK(stcb);
1159		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1160			SCTP_TCB_UNLOCK(stcb);
1161			SCTP_INP_RUNLOCK(inp);
1162			continue;
1163		}
1164		if (stcb->rport != rport) {
1165			/* remote port does not match. */
1166			SCTP_TCB_UNLOCK(stcb);
1167			SCTP_INP_RUNLOCK(inp);
1168			continue;
1169		}
1170		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1171			SCTP_TCB_UNLOCK(stcb);
1172			SCTP_INP_RUNLOCK(inp);
1173			continue;
1174		}
1175		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1176			SCTP_TCB_UNLOCK(stcb);
1177			SCTP_INP_RUNLOCK(inp);
1178			continue;
1179		}
1180		/* Does this TCB have a matching address? */
1181		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1182			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1183				/* not the same family, can't be a match */
1184				continue;
1185			}
1186			switch (from->sa_family) {
1187#ifdef INET
1188			case AF_INET:
1189				{
1190					struct sockaddr_in *sin, *rsin;
1191
1192					sin = (struct sockaddr_in *)&net->ro._l_addr;
1193					rsin = (struct sockaddr_in *)from;
1194					if (sin->sin_addr.s_addr ==
1195					    rsin->sin_addr.s_addr) {
1196						/* found it */
1197						if (netp != NULL) {
1198							*netp = net;
1199						}
1200						/*
1201						 * Update the endpoint
1202						 * pointer
1203						 */
1204						*inp_p = inp;
1205						SCTP_INP_RUNLOCK(inp);
1206						return (stcb);
1207					}
1208					break;
1209				}
1210#endif
1211#ifdef INET6
1212			case AF_INET6:
1213				{
1214					struct sockaddr_in6 *sin6, *rsin6;
1215
1216					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1217					rsin6 = (struct sockaddr_in6 *)from;
1218					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1219					    rsin6)) {
1220						/* found it */
1221						if (netp != NULL) {
1222							*netp = net;
1223						}
1224						/*
1225						 * Update the endpoint
1226						 * pointer
1227						 */
1228						*inp_p = inp;
1229						SCTP_INP_RUNLOCK(inp);
1230						return (stcb);
1231					}
1232					break;
1233				}
1234#endif
1235			default:
1236				/* TSNH */
1237				break;
1238			}
1239		}
1240		SCTP_TCB_UNLOCK(stcb);
1241		SCTP_INP_RUNLOCK(inp);
1242	}
1243	return (NULL);
1244}
1245
1246/*
1247 * rules for use
1248 *
1249 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1250 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1251 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1252 * NULL.
1253 */
1254
1255struct sctp_tcb *
1256sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1257    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1258{
1259	struct sctpasochead *head;
1260	struct sctp_inpcb *inp;
1261	struct sctp_tcb *stcb = NULL;
1262	struct sctp_nets *net;
1263	uint16_t rport;
1264
1265	inp = *inp_p;
1266	switch (remote->sa_family) {
1267#ifdef INET
1268	case AF_INET:
1269		rport = (((struct sockaddr_in *)remote)->sin_port);
1270		break;
1271#endif
1272#ifdef INET6
1273	case AF_INET6:
1274		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1275		break;
1276#endif
1277	default:
1278		return (NULL);
1279	}
1280	if (locked_tcb) {
1281		/*
1282		 * UN-lock so we can do proper locking here this occurs when
1283		 * called from load_addresses_from_init.
1284		 */
1285		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1286		SCTP_TCB_UNLOCK(locked_tcb);
1287	}
1288	SCTP_INP_INFO_RLOCK();
1289	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1290	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1291		/*-
1292		 * Now either this guy is our listener or it's the
1293		 * connector. If it is the one that issued the connect, then
1294		 * it's only chance is to be the first TCB in the list. If
1295		 * it is the acceptor, then do the special_lookup to hash
1296		 * and find the real inp.
1297		 */
1298		if ((inp->sctp_socket) && SCTP_IS_LISTENING(inp)) {
1299			/* to is peer addr, from is my addr */
1300			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1301			    netp, inp->def_vrf_id);
1302			if ((stcb != NULL) && (locked_tcb == NULL)) {
1303				/* we have a locked tcb, lower refcount */
1304				SCTP_INP_DECR_REF(inp);
1305			}
1306			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1307				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1308				SCTP_TCB_LOCK(locked_tcb);
1309				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1310				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1311			}
1312			SCTP_INP_INFO_RUNLOCK();
1313			return (stcb);
1314		} else {
1315			SCTP_INP_WLOCK(inp);
1316			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1317				goto null_return;
1318			}
1319			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1320			if (stcb == NULL) {
1321				goto null_return;
1322			}
1323			SCTP_TCB_LOCK(stcb);
1324
1325			if (stcb->rport != rport) {
1326				/* remote port does not match. */
1327				SCTP_TCB_UNLOCK(stcb);
1328				goto null_return;
1329			}
1330			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1331				SCTP_TCB_UNLOCK(stcb);
1332				goto null_return;
1333			}
1334			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1335				SCTP_TCB_UNLOCK(stcb);
1336				goto null_return;
1337			}
1338			/* now look at the list of remote addresses */
1339			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1340#ifdef INVARIANTS
1341				if (net == (TAILQ_NEXT(net, sctp_next))) {
1342					panic("Corrupt net list");
1343				}
1344#endif
1345				if (net->ro._l_addr.sa.sa_family !=
1346				    remote->sa_family) {
1347					/* not the same family */
1348					continue;
1349				}
1350				switch (remote->sa_family) {
1351#ifdef INET
1352				case AF_INET:
1353					{
1354						struct sockaddr_in *sin,
1355						           *rsin;
1356
1357						sin = (struct sockaddr_in *)
1358						    &net->ro._l_addr;
1359						rsin = (struct sockaddr_in *)remote;
1360						if (sin->sin_addr.s_addr ==
1361						    rsin->sin_addr.s_addr) {
1362							/* found it */
1363							if (netp != NULL) {
1364								*netp = net;
1365							}
1366							if (locked_tcb == NULL) {
1367								SCTP_INP_DECR_REF(inp);
1368							} else if (locked_tcb != stcb) {
1369								SCTP_TCB_LOCK(locked_tcb);
1370							}
1371							if (locked_tcb) {
1372								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1373							}
1374
1375							SCTP_INP_WUNLOCK(inp);
1376							SCTP_INP_INFO_RUNLOCK();
1377							return (stcb);
1378						}
1379						break;
1380					}
1381#endif
1382#ifdef INET6
1383				case AF_INET6:
1384					{
1385						struct sockaddr_in6 *sin6,
1386						            *rsin6;
1387
1388						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1389						rsin6 = (struct sockaddr_in6 *)remote;
1390						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1391						    rsin6)) {
1392							/* found it */
1393							if (netp != NULL) {
1394								*netp = net;
1395							}
1396							if (locked_tcb == NULL) {
1397								SCTP_INP_DECR_REF(inp);
1398							} else if (locked_tcb != stcb) {
1399								SCTP_TCB_LOCK(locked_tcb);
1400							}
1401							if (locked_tcb) {
1402								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1403							}
1404							SCTP_INP_WUNLOCK(inp);
1405							SCTP_INP_INFO_RUNLOCK();
1406							return (stcb);
1407						}
1408						break;
1409					}
1410#endif
1411				default:
1412					/* TSNH */
1413					break;
1414				}
1415			}
1416			SCTP_TCB_UNLOCK(stcb);
1417		}
1418	} else {
1419		SCTP_INP_WLOCK(inp);
1420		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1421			goto null_return;
1422		}
1423		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1424		    inp->sctp_hashmark)];
1425		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1426			if (stcb->rport != rport) {
1427				/* remote port does not match */
1428				continue;
1429			}
1430			SCTP_TCB_LOCK(stcb);
1431			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1432				SCTP_TCB_UNLOCK(stcb);
1433				continue;
1434			}
1435			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1436				SCTP_TCB_UNLOCK(stcb);
1437				continue;
1438			}
1439			/* now look at the list of remote addresses */
1440			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1441#ifdef INVARIANTS
1442				if (net == (TAILQ_NEXT(net, sctp_next))) {
1443					panic("Corrupt net list");
1444				}
1445#endif
1446				if (net->ro._l_addr.sa.sa_family !=
1447				    remote->sa_family) {
1448					/* not the same family */
1449					continue;
1450				}
1451				switch (remote->sa_family) {
1452#ifdef INET
1453				case AF_INET:
1454					{
1455						struct sockaddr_in *sin,
1456						           *rsin;
1457
1458						sin = (struct sockaddr_in *)
1459						    &net->ro._l_addr;
1460						rsin = (struct sockaddr_in *)remote;
1461						if (sin->sin_addr.s_addr ==
1462						    rsin->sin_addr.s_addr) {
1463							/* found it */
1464							if (netp != NULL) {
1465								*netp = net;
1466							}
1467							if (locked_tcb == NULL) {
1468								SCTP_INP_DECR_REF(inp);
1469							} else if (locked_tcb != stcb) {
1470								SCTP_TCB_LOCK(locked_tcb);
1471							}
1472							if (locked_tcb) {
1473								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1474							}
1475							SCTP_INP_WUNLOCK(inp);
1476							SCTP_INP_INFO_RUNLOCK();
1477							return (stcb);
1478						}
1479						break;
1480					}
1481#endif
1482#ifdef INET6
1483				case AF_INET6:
1484					{
1485						struct sockaddr_in6 *sin6,
1486						            *rsin6;
1487
1488						sin6 = (struct sockaddr_in6 *)
1489						    &net->ro._l_addr;
1490						rsin6 = (struct sockaddr_in6 *)remote;
1491						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1492						    rsin6)) {
1493							/* found it */
1494							if (netp != NULL) {
1495								*netp = net;
1496							}
1497							if (locked_tcb == NULL) {
1498								SCTP_INP_DECR_REF(inp);
1499							} else if (locked_tcb != stcb) {
1500								SCTP_TCB_LOCK(locked_tcb);
1501							}
1502							if (locked_tcb) {
1503								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1504							}
1505							SCTP_INP_WUNLOCK(inp);
1506							SCTP_INP_INFO_RUNLOCK();
1507							return (stcb);
1508						}
1509						break;
1510					}
1511#endif
1512				default:
1513					/* TSNH */
1514					break;
1515				}
1516			}
1517			SCTP_TCB_UNLOCK(stcb);
1518		}
1519	}
1520null_return:
1521	/* clean up for returning null */
1522	if (locked_tcb) {
1523		SCTP_TCB_LOCK(locked_tcb);
1524		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1525	}
1526	SCTP_INP_WUNLOCK(inp);
1527	SCTP_INP_INFO_RUNLOCK();
1528	/* not found */
1529	return (NULL);
1530}
1531
1532/*
1533 * Find an association for a specific endpoint using the association id given
1534 * out in the COMM_UP notification
1535 */
1536struct sctp_tcb *
1537sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1538{
1539	/*
1540	 * Use my the assoc_id to find a endpoint
1541	 */
1542	struct sctpasochead *head;
1543	struct sctp_tcb *stcb;
1544	uint32_t id;
1545
1546	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1547		SCTP_PRINTF("TSNH ep_associd0\n");
1548		return (NULL);
1549	}
1550	id = (uint32_t)asoc_id;
1551	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1552	if (head == NULL) {
1553		/* invalid id TSNH */
1554		SCTP_PRINTF("TSNH ep_associd1\n");
1555		return (NULL);
1556	}
1557	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1558		if (stcb->asoc.assoc_id == id) {
1559			if (inp != stcb->sctp_ep) {
1560				/*
1561				 * some other guy has the same id active (id
1562				 * collision ??).
1563				 */
1564				SCTP_PRINTF("TSNH ep_associd2\n");
1565				continue;
1566			}
1567			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1568				continue;
1569			}
1570			if (want_lock) {
1571				SCTP_TCB_LOCK(stcb);
1572			}
1573			return (stcb);
1574		}
1575	}
1576	return (NULL);
1577}
1578
1579struct sctp_tcb *
1580sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1581{
1582	struct sctp_tcb *stcb;
1583
1584	SCTP_INP_RLOCK(inp);
1585	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1586	SCTP_INP_RUNLOCK(inp);
1587	return (stcb);
1588}
1589
1590/*
1591 * Endpoint probe expects that the INP_INFO is locked.
1592 */
1593static struct sctp_inpcb *
1594sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1595    uint16_t lport, uint32_t vrf_id)
1596{
1597	struct sctp_inpcb *inp;
1598	struct sctp_laddr *laddr;
1599#ifdef INET
1600	struct sockaddr_in *sin;
1601#endif
1602#ifdef INET6
1603	struct sockaddr_in6 *sin6;
1604	struct sockaddr_in6 *intf_addr6;
1605#endif
1606	int fnd;
1607
1608#ifdef INET
1609	sin = NULL;
1610#endif
1611#ifdef INET6
1612	sin6 = NULL;
1613#endif
1614	switch (nam->sa_family) {
1615#ifdef INET
1616	case AF_INET:
1617		sin = (struct sockaddr_in *)nam;
1618		break;
1619#endif
1620#ifdef INET6
1621	case AF_INET6:
1622		sin6 = (struct sockaddr_in6 *)nam;
1623		break;
1624#endif
1625	default:
1626		/* unsupported family */
1627		return (NULL);
1628	}
1629
1630	if (head == NULL)
1631		return (NULL);
1632
1633	LIST_FOREACH(inp, head, sctp_hash) {
1634		SCTP_INP_RLOCK(inp);
1635		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1636			SCTP_INP_RUNLOCK(inp);
1637			continue;
1638		}
1639		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1640		    (inp->sctp_lport == lport)) {
1641			/* got it */
1642			switch (nam->sa_family) {
1643#ifdef INET
1644			case AF_INET:
1645				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1646				    SCTP_IPV6_V6ONLY(inp)) {
1647					/*
1648					 * IPv4 on a IPv6 socket with ONLY
1649					 * IPv6 set
1650					 */
1651					SCTP_INP_RUNLOCK(inp);
1652					continue;
1653				}
1654				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1655				    &sin->sin_addr) != 0) {
1656					SCTP_INP_RUNLOCK(inp);
1657					continue;
1658				}
1659				break;
1660#endif
1661#ifdef INET6
1662			case AF_INET6:
1663				/*
1664				 * A V6 address and the endpoint is NOT
1665				 * bound V6
1666				 */
1667				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1668					SCTP_INP_RUNLOCK(inp);
1669					continue;
1670				}
1671				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1672				    &sin6->sin6_addr) != 0) {
1673					SCTP_INP_RUNLOCK(inp);
1674					continue;
1675				}
1676				break;
1677#endif
1678			default:
1679				break;
1680			}
1681			/* does a VRF id match? */
1682			fnd = 0;
1683			if (inp->def_vrf_id == vrf_id)
1684				fnd = 1;
1685
1686			SCTP_INP_RUNLOCK(inp);
1687			if (!fnd)
1688				continue;
1689			return (inp);
1690		}
1691		SCTP_INP_RUNLOCK(inp);
1692	}
1693	switch (nam->sa_family) {
1694#ifdef INET
1695	case AF_INET:
1696		if (sin->sin_addr.s_addr == INADDR_ANY) {
1697			/* Can't hunt for one that has no address specified */
1698			return (NULL);
1699		}
1700		break;
1701#endif
1702#ifdef INET6
1703	case AF_INET6:
1704		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1705			/* Can't hunt for one that has no address specified */
1706			return (NULL);
1707		}
1708		break;
1709#endif
1710	default:
1711		break;
1712	}
1713	/*
1714	 * ok, not bound to all so see if we can find a EP bound to this
1715	 * address.
1716	 */
1717	LIST_FOREACH(inp, head, sctp_hash) {
1718		SCTP_INP_RLOCK(inp);
1719		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1720			SCTP_INP_RUNLOCK(inp);
1721			continue;
1722		}
1723		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1724			SCTP_INP_RUNLOCK(inp);
1725			continue;
1726		}
1727		/*
1728		 * Ok this could be a likely candidate, look at all of its
1729		 * addresses
1730		 */
1731		if (inp->sctp_lport != lport) {
1732			SCTP_INP_RUNLOCK(inp);
1733			continue;
1734		}
1735		/* does a VRF id match? */
1736		fnd = 0;
1737		if (inp->def_vrf_id == vrf_id)
1738			fnd = 1;
1739
1740		if (!fnd) {
1741			SCTP_INP_RUNLOCK(inp);
1742			continue;
1743		}
1744		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1745			if (laddr->ifa == NULL) {
1746				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1747				    __func__);
1748				continue;
1749			}
1750			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1751			    (void *)laddr->ifa);
1752			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1753				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1754				continue;
1755			}
1756			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1757				/* possible, see if it matches */
1758				switch (nam->sa_family) {
1759#ifdef INET
1760				case AF_INET:
1761					if (sin->sin_addr.s_addr ==
1762					    laddr->ifa->address.sin.sin_addr.s_addr) {
1763						SCTP_INP_RUNLOCK(inp);
1764						return (inp);
1765					}
1766					break;
1767#endif
1768#ifdef INET6
1769				case AF_INET6:
1770					intf_addr6 = &laddr->ifa->address.sin6;
1771					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1772					    intf_addr6)) {
1773						SCTP_INP_RUNLOCK(inp);
1774						return (inp);
1775					}
1776					break;
1777#endif
1778				}
1779			}
1780		}
1781		SCTP_INP_RUNLOCK(inp);
1782	}
1783	return (NULL);
1784}
1785
1786static struct sctp_inpcb *
1787sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1788{
1789	struct sctppcbhead *head;
1790	struct sctp_inpcb *t_inp;
1791	int fnd;
1792
1793	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1794	    SCTP_BASE_INFO(hashmark))];
1795	LIST_FOREACH(t_inp, head, sctp_hash) {
1796		if (t_inp->sctp_lport != lport) {
1797			continue;
1798		}
1799		/* is it in the VRF in question */
1800		fnd = 0;
1801		if (t_inp->def_vrf_id == vrf_id)
1802			fnd = 1;
1803		if (!fnd)
1804			continue;
1805
1806		/* This one is in use. */
1807		/* check the v6/v4 binding issue */
1808		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1809		    SCTP_IPV6_V6ONLY(t_inp)) {
1810			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1811				/* collision in V6 space */
1812				return (t_inp);
1813			} else {
1814				/* inp is BOUND_V4 no conflict */
1815				continue;
1816			}
1817		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1818			/* t_inp is bound v4 and v6, conflict always */
1819			return (t_inp);
1820		} else {
1821			/* t_inp is bound only V4 */
1822			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1823			    SCTP_IPV6_V6ONLY(inp)) {
1824				/* no conflict */
1825				continue;
1826			}
1827			/* else fall through to conflict */
1828		}
1829		return (t_inp);
1830	}
1831	return (NULL);
1832}
1833
1834int
1835sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1836{
1837	/* For 1-2-1 with port reuse */
1838	struct sctppcbhead *head;
1839	struct sctp_inpcb *tinp, *ninp;
1840
1841	SCTP_INP_INFO_WLOCK_ASSERT();
1842	SCTP_INP_WLOCK_ASSERT(inp);
1843
1844	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1845		/* only works with port reuse on */
1846		return (-1);
1847	}
1848	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1849		return (0);
1850	}
1851	SCTP_INP_WUNLOCK(inp);
1852	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1853	    SCTP_BASE_INFO(hashmark))];
1854	/* Kick out all non-listeners to the TCP hash */
1855	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1856		if (tinp->sctp_lport != inp->sctp_lport) {
1857			continue;
1858		}
1859		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1860			continue;
1861		}
1862		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1863			continue;
1864		}
1865		if (SCTP_IS_LISTENING(tinp)) {
1866			continue;
1867		}
1868		SCTP_INP_WLOCK(tinp);
1869		LIST_REMOVE(tinp, sctp_hash);
1870		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1871		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1872		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1873		SCTP_INP_WUNLOCK(tinp);
1874	}
1875	SCTP_INP_WLOCK(inp);
1876	/* Pull from where he was */
1877	LIST_REMOVE(inp, sctp_hash);
1878	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1879	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1880	LIST_INSERT_HEAD(head, inp, sctp_hash);
1881	return (0);
1882}
1883
1884struct sctp_inpcb *
1885sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1886    uint32_t vrf_id)
1887{
1888	/*
1889	 * First we check the hash table to see if someone has this port
1890	 * bound with just the port.
1891	 */
1892	struct sctp_inpcb *inp;
1893	struct sctppcbhead *head;
1894	int lport;
1895	unsigned int i;
1896#ifdef INET
1897	struct sockaddr_in *sin;
1898#endif
1899#ifdef INET6
1900	struct sockaddr_in6 *sin6;
1901#endif
1902
1903	switch (nam->sa_family) {
1904#ifdef INET
1905	case AF_INET:
1906		sin = (struct sockaddr_in *)nam;
1907		lport = sin->sin_port;
1908		break;
1909#endif
1910#ifdef INET6
1911	case AF_INET6:
1912		sin6 = (struct sockaddr_in6 *)nam;
1913		lport = sin6->sin6_port;
1914		break;
1915#endif
1916	default:
1917		return (NULL);
1918	}
1919	/*
1920	 * I could cheat here and just cast to one of the types but we will
1921	 * do it right. It also provides the check against an Unsupported
1922	 * type too.
1923	 */
1924	/* Find the head of the ALLADDR chain */
1925	if (have_lock == 0) {
1926		SCTP_INP_INFO_RLOCK();
1927	}
1928	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1929	    SCTP_BASE_INFO(hashmark))];
1930	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1931
1932	/*
1933	 * If the TCP model exists it could be that the main listening
1934	 * endpoint is gone but there still exists a connected socket for
1935	 * this guy. If so we can return the first one that we find. This
1936	 * may NOT be the correct one so the caller should be wary on the
1937	 * returned INP. Currently the only caller that sets find_tcp_pool
1938	 * is in bindx where we are verifying that a user CAN bind the
1939	 * address. He either has bound it already, or someone else has, or
1940	 * its open to bind, so this is good enough.
1941	 */
1942	if (inp == NULL && find_tcp_pool) {
1943		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1944			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1945			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1946			if (inp) {
1947				break;
1948			}
1949		}
1950	}
1951	if (inp) {
1952		SCTP_INP_INCR_REF(inp);
1953	}
1954	if (have_lock == 0) {
1955		SCTP_INP_INFO_RUNLOCK();
1956	}
1957	return (inp);
1958}
1959
1960/*
1961 * Find an association for an endpoint with the pointer to whom you want to
1962 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1963 * need to change the *to to some other struct like a mbuf...
1964 */
1965struct sctp_tcb *
1966sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1967    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1968    uint32_t vrf_id)
1969{
1970	struct sctp_inpcb *inp = NULL;
1971	struct sctp_tcb *stcb;
1972
1973	SCTP_INP_INFO_RLOCK();
1974	if (find_tcp_pool) {
1975		if (inp_p != NULL) {
1976			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1977			    vrf_id);
1978		} else {
1979			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1980			    vrf_id);
1981		}
1982		if (stcb != NULL) {
1983			SCTP_INP_INFO_RUNLOCK();
1984			return (stcb);
1985		}
1986	}
1987	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1988	if (inp_p != NULL) {
1989		*inp_p = inp;
1990	}
1991	SCTP_INP_INFO_RUNLOCK();
1992	if (inp == NULL) {
1993		return (NULL);
1994	}
1995	/*
1996	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1997	 * we now place the source address or from in the to of the find
1998	 * endpoint call. Since in reality this chain is used from the
1999	 * inbound packet side.
2000	 */
2001	if (inp_p != NULL) {
2002		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2003		    NULL);
2004	} else {
2005		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2006		    NULL);
2007	}
2008	return (stcb);
2009}
2010
2011/*
2012 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2013 * find all addresses that the sender has specified in any address list. Each
2014 * address will be used to lookup the TCB and see if one exits.
2015 */
2016static struct sctp_tcb *
2017sctp_findassociation_special_addr(struct mbuf *m, int offset,
2018    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2019    struct sockaddr *dst)
2020{
2021	struct sctp_paramhdr *phdr, param_buf;
2022#if defined(INET) || defined(INET6)
2023	struct sctp_tcb *stcb;
2024	uint16_t ptype;
2025#endif
2026	uint16_t plen;
2027#ifdef INET
2028	struct sockaddr_in sin4;
2029#endif
2030#ifdef INET6
2031	struct sockaddr_in6 sin6;
2032#endif
2033
2034#ifdef INET
2035	memset(&sin4, 0, sizeof(sin4));
2036	sin4.sin_len = sizeof(sin4);
2037	sin4.sin_family = AF_INET;
2038	sin4.sin_port = sh->src_port;
2039#endif
2040#ifdef INET6
2041	memset(&sin6, 0, sizeof(sin6));
2042	sin6.sin6_len = sizeof(sin6);
2043	sin6.sin6_family = AF_INET6;
2044	sin6.sin6_port = sh->src_port;
2045#endif
2046
2047	offset += sizeof(struct sctp_init_chunk);
2048
2049	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
2050	while (phdr != NULL) {
2051		/* now we must see if we want the parameter */
2052#if defined(INET) || defined(INET6)
2053		ptype = ntohs(phdr->param_type);
2054#endif
2055		plen = ntohs(phdr->param_length);
2056		if (plen == 0) {
2057			break;
2058		}
2059#ifdef INET
2060		if (ptype == SCTP_IPV4_ADDRESS &&
2061		    plen == sizeof(struct sctp_ipv4addr_param)) {
2062			/* Get the rest of the address */
2063			struct sctp_ipv4addr_param ip4_param, *p4;
2064
2065			phdr = sctp_get_next_param(m, offset,
2066			    (struct sctp_paramhdr *)&ip4_param, sizeof(ip4_param));
2067			if (phdr == NULL) {
2068				return (NULL);
2069			}
2070			p4 = (struct sctp_ipv4addr_param *)phdr;
2071			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2072			/* look it up */
2073			stcb = sctp_findassociation_ep_addr(inp_p,
2074			    (struct sockaddr *)&sin4, netp, dst, NULL);
2075			if (stcb != NULL) {
2076				return (stcb);
2077			}
2078		}
2079#endif
2080#ifdef INET6
2081		if (ptype == SCTP_IPV6_ADDRESS &&
2082		    plen == sizeof(struct sctp_ipv6addr_param)) {
2083			/* Get the rest of the address */
2084			struct sctp_ipv6addr_param ip6_param, *p6;
2085
2086			phdr = sctp_get_next_param(m, offset,
2087			    (struct sctp_paramhdr *)&ip6_param, sizeof(ip6_param));
2088			if (phdr == NULL) {
2089				return (NULL);
2090			}
2091			p6 = (struct sctp_ipv6addr_param *)phdr;
2092			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2093			/* look it up */
2094			stcb = sctp_findassociation_ep_addr(inp_p,
2095			    (struct sockaddr *)&sin6, netp, dst, NULL);
2096			if (stcb != NULL) {
2097				return (stcb);
2098			}
2099		}
2100#endif
2101		offset += SCTP_SIZE32(plen);
2102		phdr = sctp_get_next_param(m, offset, &param_buf,
2103		    sizeof(param_buf));
2104	}
2105	return (NULL);
2106}
2107
2108static struct sctp_tcb *
2109sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2110    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2111    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2112{
2113	/*
2114	 * Use my vtag to hash. If we find it we then verify the source addr
2115	 * is in the assoc. If all goes well we save a bit on rec of a
2116	 * packet.
2117	 */
2118	struct sctpasochead *head;
2119	struct sctp_nets *net;
2120	struct sctp_tcb *stcb;
2121
2122	SCTP_INP_INFO_RLOCK();
2123	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2124	    SCTP_BASE_INFO(hashasocmark))];
2125	LIST_FOREACH(stcb, head, sctp_asocs) {
2126		SCTP_INP_RLOCK(stcb->sctp_ep);
2127		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2128			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2129			continue;
2130		}
2131		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2132			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2133			continue;
2134		}
2135		SCTP_TCB_LOCK(stcb);
2136		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2137		if (stcb->asoc.my_vtag == vtag) {
2138			/* candidate */
2139			if (stcb->rport != rport) {
2140				SCTP_TCB_UNLOCK(stcb);
2141				continue;
2142			}
2143			if (stcb->sctp_ep->sctp_lport != lport) {
2144				SCTP_TCB_UNLOCK(stcb);
2145				continue;
2146			}
2147			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2148				SCTP_TCB_UNLOCK(stcb);
2149				continue;
2150			}
2151			/* RRS:Need toaddr check here */
2152			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2153				/* Endpoint does not own this address */
2154				SCTP_TCB_UNLOCK(stcb);
2155				continue;
2156			}
2157			if (remote_tag) {
2158				/*
2159				 * If we have both vtags that's all we match
2160				 * on
2161				 */
2162				if (stcb->asoc.peer_vtag == remote_tag) {
2163					/*
2164					 * If both tags match we consider it
2165					 * conclusive and check NO
2166					 * source/destination addresses
2167					 */
2168					goto conclusive;
2169				}
2170			}
2171			if (skip_src_check) {
2172		conclusive:
2173				if (from) {
2174					*netp = sctp_findnet(stcb, from);
2175				} else {
2176					*netp = NULL;	/* unknown */
2177				}
2178				if (inp_p)
2179					*inp_p = stcb->sctp_ep;
2180				SCTP_INP_INFO_RUNLOCK();
2181				return (stcb);
2182			}
2183			net = sctp_findnet(stcb, from);
2184			if (net) {
2185				/* yep its him. */
2186				*netp = net;
2187				SCTP_STAT_INCR(sctps_vtagexpress);
2188				*inp_p = stcb->sctp_ep;
2189				SCTP_INP_INFO_RUNLOCK();
2190				return (stcb);
2191			} else {
2192				/*
2193				 * not him, this should only happen in rare
2194				 * cases so I peg it.
2195				 */
2196				SCTP_STAT_INCR(sctps_vtagbogus);
2197			}
2198		}
2199		SCTP_TCB_UNLOCK(stcb);
2200	}
2201	SCTP_INP_INFO_RUNLOCK();
2202	return (NULL);
2203}
2204
2205/*
2206 * Find an association with the pointer to the inbound IP packet. This can be
2207 * a IPv4 or IPv6 packet.
2208 */
2209struct sctp_tcb *
2210sctp_findassociation_addr(struct mbuf *m, int offset,
2211    struct sockaddr *src, struct sockaddr *dst,
2212    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2213    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2214{
2215	struct sctp_tcb *stcb;
2216	struct sctp_inpcb *inp;
2217
2218	if (sh->v_tag) {
2219		/* we only go down this path if vtag is non-zero */
2220		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2221		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2222		if (stcb) {
2223			return (stcb);
2224		}
2225	}
2226
2227	if (inp_p) {
2228		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2229		    1, vrf_id);
2230		inp = *inp_p;
2231	} else {
2232		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2233		    1, vrf_id);
2234	}
2235	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2236	if (stcb == NULL && inp) {
2237		/* Found a EP but not this address */
2238		if ((ch->chunk_type == SCTP_INITIATION) ||
2239		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2240			/*-
2241			 * special hook, we do NOT return linp or an
2242			 * association that is linked to an existing
2243			 * association that is under the TCP pool (i.e. no
2244			 * listener exists). The endpoint finding routine
2245			 * will always find a listener before examining the
2246			 * TCP pool.
2247			 */
2248			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2249				if (inp_p) {
2250					*inp_p = NULL;
2251				}
2252				return (NULL);
2253			}
2254			stcb = sctp_findassociation_special_addr(m,
2255			    offset, sh, &inp, netp, dst);
2256			if (inp_p != NULL) {
2257				*inp_p = inp;
2258			}
2259		}
2260	}
2261	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2262	return (stcb);
2263}
2264
2265/*
2266 * lookup an association by an ASCONF lookup address.
2267 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2268 */
2269struct sctp_tcb *
2270sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2271    struct sockaddr *dst, struct sctphdr *sh,
2272    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2273{
2274	struct sctp_tcb *stcb;
2275	union sctp_sockstore remote_store;
2276	struct sctp_paramhdr param_buf, *phdr;
2277	int ptype;
2278	int zero_address = 0;
2279#ifdef INET
2280	struct sockaddr_in *sin;
2281#endif
2282#ifdef INET6
2283	struct sockaddr_in6 *sin6;
2284#endif
2285
2286	memset(&remote_store, 0, sizeof(remote_store));
2287	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2288	    &param_buf, sizeof(struct sctp_paramhdr));
2289	if (phdr == NULL) {
2290		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2291		    __func__);
2292		return NULL;
2293	}
2294	ptype = (int)((uint32_t)ntohs(phdr->param_type));
2295	/* get the correlation address */
2296	switch (ptype) {
2297#ifdef INET6
2298	case SCTP_IPV6_ADDRESS:
2299		{
2300			/* ipv6 address param */
2301			struct sctp_ipv6addr_param *p6, p6_buf;
2302
2303			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2304				return NULL;
2305			}
2306			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2307			    offset + sizeof(struct sctp_asconf_chunk),
2308			    &p6_buf.ph, sizeof(p6_buf));
2309			if (p6 == NULL) {
2310				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2311				    __func__);
2312				return (NULL);
2313			}
2314			sin6 = &remote_store.sin6;
2315			sin6->sin6_family = AF_INET6;
2316			sin6->sin6_len = sizeof(*sin6);
2317			sin6->sin6_port = sh->src_port;
2318			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2319			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2320				zero_address = 1;
2321			break;
2322		}
2323#endif
2324#ifdef INET
2325	case SCTP_IPV4_ADDRESS:
2326		{
2327			/* ipv4 address param */
2328			struct sctp_ipv4addr_param *p4, p4_buf;
2329
2330			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2331				return NULL;
2332			}
2333			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2334			    offset + sizeof(struct sctp_asconf_chunk),
2335			    &p4_buf.ph, sizeof(p4_buf));
2336			if (p4 == NULL) {
2337				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2338				    __func__);
2339				return (NULL);
2340			}
2341			sin = &remote_store.sin;
2342			sin->sin_family = AF_INET;
2343			sin->sin_len = sizeof(*sin);
2344			sin->sin_port = sh->src_port;
2345			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2346			if (sin->sin_addr.s_addr == INADDR_ANY)
2347				zero_address = 1;
2348			break;
2349		}
2350#endif
2351	default:
2352		/* invalid address param type */
2353		return NULL;
2354	}
2355
2356	if (zero_address) {
2357		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2358		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2359		if (stcb != NULL) {
2360			SCTP_INP_DECR_REF(*inp_p);
2361		}
2362	} else {
2363		stcb = sctp_findassociation_ep_addr(inp_p,
2364		    &remote_store.sa, netp,
2365		    dst, NULL);
2366	}
2367	return (stcb);
2368}
2369
2370/*
2371 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2372 * addresses. This way if we don't get a bind we by default pick a ephemeral
2373 * port with all addresses bound.
2374 */
2375int
2376sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2377{
2378	/*
2379	 * we get called when a new endpoint starts up. We need to allocate
2380	 * the sctp_inpcb structure from the zone and init it. Mark it as
2381	 * unbound and find a port that we can use as an ephemeral with
2382	 * INADDR_ANY. If the user binds later no problem we can then add in
2383	 * the specific addresses. And setup the default parameters for the
2384	 * EP.
2385	 */
2386	int i, error;
2387	struct sctp_inpcb *inp;
2388	struct sctp_pcb *m;
2389	struct timeval time;
2390	sctp_sharedkey_t *null_key;
2391
2392	error = 0;
2393
2394	SCTP_INP_INFO_WLOCK();
2395	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2396	if (inp == NULL) {
2397		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2398		SCTP_INP_INFO_WUNLOCK();
2399		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2400		return (ENOBUFS);
2401	}
2402	/* zap it */
2403	memset(inp, 0, sizeof(*inp));
2404
2405	/* bump generations */
2406	/* setup socket pointers */
2407	inp->sctp_socket = so;
2408	inp->ip_inp.inp.inp_socket = so;
2409	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2410#ifdef INET6
2411	if (INP_SOCKAF(so) == AF_INET6) {
2412		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2413			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2414		}
2415		if (MODULE_GLOBAL(ip6_v6only)) {
2416			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2417		}
2418	}
2419#endif
2420	inp->sctp_associd_counter = 1;
2421	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2422	inp->sctp_frag_point = 0;
2423	inp->max_cwnd = 0;
2424	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2425	inp->ecn_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_ecn_enable);
2426	inp->prsctp_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pr_enable);
2427	inp->auth_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_auth_enable);
2428	inp->asconf_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_asconf_enable);
2429	inp->reconfig_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2430	inp->nrsack_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2431	inp->pktdrop_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2432	inp->idata_supported = 0;
2433	inp->rcv_edmid = SCTP_EDMID_NONE;
2434
2435	inp->fibnum = so->so_fibnum;
2436	/* init the small hash table we use to track asocid <-> tcb */
2437	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2438	if (inp->sctp_asocidhash == NULL) {
2439		crfree(inp->ip_inp.inp.inp_cred);
2440		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2441		SCTP_INP_INFO_WUNLOCK();
2442		return (ENOBUFS);
2443	}
2444	SCTP_INCR_EP_COUNT();
2445	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2446	SCTP_INP_INFO_WUNLOCK();
2447
2448	so->so_pcb = (caddr_t)inp;
2449
2450	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2451		/* UDP style socket */
2452		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2453		    SCTP_PCB_FLAGS_UNBOUND);
2454		/* Be sure it is NON-BLOCKING IO for UDP */
2455		/* SCTP_SET_SO_NBIO(so); */
2456	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2457		/* TCP style socket */
2458		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2459		    SCTP_PCB_FLAGS_UNBOUND);
2460		/* Be sure we have blocking IO by default */
2461		SOCK_LOCK(so);
2462		SCTP_CLEAR_SO_NBIO(so);
2463		SOCK_UNLOCK(so);
2464	} else {
2465		/*
2466		 * unsupported socket type (RAW, etc)- in case we missed it
2467		 * in protosw
2468		 */
2469		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2470		so->so_pcb = NULL;
2471		crfree(inp->ip_inp.inp.inp_cred);
2472		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2473		return (EOPNOTSUPP);
2474	}
2475	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2476		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2477		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2478	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2479		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2480		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2481	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2482		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2483		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2484	}
2485	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2486	    &inp->sctp_hashmark);
2487	if (inp->sctp_tcbhash == NULL) {
2488		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2489		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2490		so->so_pcb = NULL;
2491		crfree(inp->ip_inp.inp.inp_cred);
2492		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2493		return (ENOBUFS);
2494	}
2495	inp->def_vrf_id = vrf_id;
2496
2497	SCTP_INP_INFO_WLOCK();
2498	SCTP_INP_LOCK_INIT(inp);
2499	rw_init_flags(&inp->ip_inp.inp.inp_lock, "sctpinp",
2500	    RW_RECURSE | RW_DUPOK);
2501	SCTP_INP_READ_LOCK_INIT(inp);
2502	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2503	/* lock the new ep */
2504	SCTP_INP_WLOCK(inp);
2505
2506	/* add it to the info area */
2507	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2508	SCTP_INP_INFO_WUNLOCK();
2509
2510	TAILQ_INIT(&inp->read_queue);
2511	LIST_INIT(&inp->sctp_addr_list);
2512
2513	LIST_INIT(&inp->sctp_asoc_list);
2514
2515#ifdef SCTP_TRACK_FREED_ASOCS
2516	/* TEMP CODE */
2517	LIST_INIT(&inp->sctp_asoc_free_list);
2518#endif
2519	/* Init the timer structure for signature change */
2520	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2521	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2522
2523	/* now init the actual endpoint default data */
2524	m = &inp->sctp_ep;
2525
2526	/* setup the base timeout information */
2527	m->sctp_timeoutticks[SCTP_TIMER_SEND] = sctp_secs_to_ticks(SCTP_SEND_SEC);	/* needed ? */
2528	m->sctp_timeoutticks[SCTP_TIMER_INIT] = sctp_secs_to_ticks(SCTP_INIT_SEC);	/* needed ? */
2529	m->sctp_timeoutticks[SCTP_TIMER_RECV] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2530	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2531	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2532	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2533	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = sctp_secs_to_ticks(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2534	/* all max/min max are in ms */
2535	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2536	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2537	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2538	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2539	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2540	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2541	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2542	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2543	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2544	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2545	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2546	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2547	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2548
2549	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2550	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2551	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2552	/* number of streams to pre-open on a association */
2553	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2554
2555	m->default_mtu = 0;
2556	/* Add adaptation cookie */
2557	m->adaptation_layer_indicator = 0;
2558	m->adaptation_layer_indicator_provided = 0;
2559
2560	/* seed random number generator */
2561	m->random_counter = 1;
2562	m->store_at = SCTP_SIGNATURE_SIZE;
2563	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2564	sctp_fill_random_store(m);
2565
2566	/* Minimum cookie size */
2567	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2568	    sizeof(struct sctp_state_cookie);
2569	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2570
2571	/* Setup the initial secret */
2572	(void)SCTP_GETTIME_TIMEVAL(&time);
2573	m->time_of_secret_change = time.tv_sec;
2574
2575	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2576		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2577	}
2578	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2579
2580	/* How long is a cookie good for ? */
2581	m->def_cookie_life = sctp_msecs_to_ticks(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2582	/*
2583	 * Initialize authentication parameters
2584	 */
2585	m->local_hmacs = sctp_default_supported_hmaclist();
2586	m->local_auth_chunks = sctp_alloc_chunklist();
2587	if (inp->asconf_supported) {
2588		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2589		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2590	}
2591	m->default_dscp = 0;
2592#ifdef INET6
2593	m->default_flowlabel = 0;
2594#endif
2595	m->port = 0;		/* encapsulation disabled by default */
2596	LIST_INIT(&m->shared_keys);
2597	/* add default NULL key as key id 0 */
2598	null_key = sctp_alloc_sharedkey();
2599	sctp_insert_sharedkey(&m->shared_keys, null_key);
2600	SCTP_INP_WUNLOCK(inp);
2601#ifdef SCTP_LOG_CLOSING
2602	sctp_log_closing(inp, NULL, 12);
2603#endif
2604	return (error);
2605}
2606
2607void
2608sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2609    struct sctp_tcb *stcb)
2610{
2611	struct sctp_nets *net;
2612	uint16_t lport, rport;
2613	struct sctppcbhead *head;
2614	struct sctp_laddr *laddr, *oladdr;
2615
2616	atomic_add_int(&stcb->asoc.refcnt, 1);
2617	SCTP_TCB_UNLOCK(stcb);
2618	SCTP_INP_INFO_WLOCK();
2619	SCTP_INP_WLOCK(old_inp);
2620	SCTP_INP_WLOCK(new_inp);
2621	SCTP_TCB_LOCK(stcb);
2622	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2623
2624#ifdef INET6
2625	if (old_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2626		new_inp->ip_inp.inp.inp_flags |= old_inp->ip_inp.inp.inp_flags & INP_CONTROLOPTS;
2627		if (old_inp->ip_inp.inp.in6p_outputopts) {
2628			new_inp->ip_inp.inp.in6p_outputopts = ip6_copypktopts(old_inp->ip_inp.inp.in6p_outputopts, M_NOWAIT);
2629		}
2630	}
2631#endif
2632#if defined(INET) && defined(INET6)
2633	else
2634#endif
2635#ifdef INET
2636	{
2637		new_inp->ip_inp.inp.inp_ip_tos = old_inp->ip_inp.inp.inp_ip_tos;
2638		new_inp->ip_inp.inp.inp_ip_ttl = old_inp->ip_inp.inp.inp_ip_ttl;
2639	}
2640#endif
2641	new_inp->sctp_ep.time_of_secret_change =
2642	    old_inp->sctp_ep.time_of_secret_change;
2643	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2644	    sizeof(old_inp->sctp_ep.secret_key));
2645	new_inp->sctp_ep.current_secret_number =
2646	    old_inp->sctp_ep.current_secret_number;
2647	new_inp->sctp_ep.last_secret_number =
2648	    old_inp->sctp_ep.last_secret_number;
2649	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2650
2651	/* make it so new data pours into the new socket */
2652	stcb->sctp_socket = new_inp->sctp_socket;
2653	stcb->sctp_ep = new_inp;
2654
2655	/* Copy the port across */
2656	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2657	rport = stcb->rport;
2658	/* Pull the tcb from the old association */
2659	LIST_REMOVE(stcb, sctp_tcbhash);
2660	LIST_REMOVE(stcb, sctp_tcblist);
2661	if (stcb->asoc.in_asocid_hash) {
2662		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2663	}
2664	/* Now insert the new_inp into the TCP connected hash */
2665	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2666
2667	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2668	/* Its safe to access */
2669	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2670
2671	/* Now move the tcb into the endpoint list */
2672	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2673	/*
2674	 * Question, do we even need to worry about the ep-hash since we
2675	 * only have one connection? Probably not :> so lets get rid of it
2676	 * and not suck up any kernel memory in that.
2677	 */
2678	if (stcb->asoc.in_asocid_hash) {
2679		struct sctpasochead *lhd;
2680
2681		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2682		    new_inp->hashasocidmark)];
2683		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2684	}
2685	/* Ok. Let's restart timer. */
2686	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2687		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2688		    stcb, net);
2689	}
2690
2691	SCTP_INP_INFO_WUNLOCK();
2692	if (new_inp->sctp_tcbhash != NULL) {
2693		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2694		new_inp->sctp_tcbhash = NULL;
2695	}
2696	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2697		/* Subset bound, so copy in the laddr list from the old_inp */
2698		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2699			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2700			if (laddr == NULL) {
2701				/*
2702				 * Gak, what can we do? This assoc is really
2703				 * HOSED. We probably should send an abort
2704				 * here.
2705				 */
2706				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2707				continue;
2708			}
2709			SCTP_INCR_LADDR_COUNT();
2710			memset(laddr, 0, sizeof(*laddr));
2711			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2712			laddr->ifa = oladdr->ifa;
2713			atomic_add_int(&laddr->ifa->refcount, 1);
2714			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2715			    sctp_nxt_addr);
2716			new_inp->laddr_count++;
2717			if (oladdr == stcb->asoc.last_used_address) {
2718				stcb->asoc.last_used_address = laddr;
2719			}
2720		}
2721	}
2722	/* Now any running timers need to be adjusted. */
2723	if (stcb->asoc.dack_timer.ep == old_inp) {
2724		SCTP_INP_DECR_REF(old_inp);
2725		stcb->asoc.dack_timer.ep = new_inp;
2726		SCTP_INP_INCR_REF(new_inp);
2727	}
2728	if (stcb->asoc.asconf_timer.ep == old_inp) {
2729		SCTP_INP_DECR_REF(old_inp);
2730		stcb->asoc.asconf_timer.ep = new_inp;
2731		SCTP_INP_INCR_REF(new_inp);
2732	}
2733	if (stcb->asoc.strreset_timer.ep == old_inp) {
2734		SCTP_INP_DECR_REF(old_inp);
2735		stcb->asoc.strreset_timer.ep = new_inp;
2736		SCTP_INP_INCR_REF(new_inp);
2737	}
2738	if (stcb->asoc.shut_guard_timer.ep == old_inp) {
2739		SCTP_INP_DECR_REF(old_inp);
2740		stcb->asoc.shut_guard_timer.ep = new_inp;
2741		SCTP_INP_INCR_REF(new_inp);
2742	}
2743	if (stcb->asoc.autoclose_timer.ep == old_inp) {
2744		SCTP_INP_DECR_REF(old_inp);
2745		stcb->asoc.autoclose_timer.ep = new_inp;
2746		SCTP_INP_INCR_REF(new_inp);
2747	}
2748	if (stcb->asoc.delete_prim_timer.ep == old_inp) {
2749		SCTP_INP_DECR_REF(old_inp);
2750		stcb->asoc.delete_prim_timer.ep = new_inp;
2751		SCTP_INP_INCR_REF(new_inp);
2752	}
2753	/* now what about the nets? */
2754	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2755		if (net->pmtu_timer.ep == old_inp) {
2756			SCTP_INP_DECR_REF(old_inp);
2757			net->pmtu_timer.ep = new_inp;
2758			SCTP_INP_INCR_REF(new_inp);
2759		}
2760		if (net->hb_timer.ep == old_inp) {
2761			SCTP_INP_DECR_REF(old_inp);
2762			net->hb_timer.ep = new_inp;
2763			SCTP_INP_INCR_REF(new_inp);
2764		}
2765		if (net->rxt_timer.ep == old_inp) {
2766			SCTP_INP_DECR_REF(old_inp);
2767			net->rxt_timer.ep = new_inp;
2768			SCTP_INP_INCR_REF(new_inp);
2769		}
2770	}
2771	SCTP_INP_WUNLOCK(new_inp);
2772	SCTP_INP_WUNLOCK(old_inp);
2773}
2774
2775/*
2776 * insert an laddr entry with the given ifa for the desired list
2777 */
2778static int
2779sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
2780{
2781	struct sctp_laddr *laddr;
2782
2783	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2784	if (laddr == NULL) {
2785		/* out of memory? */
2786		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2787		return (EINVAL);
2788	}
2789	SCTP_INCR_LADDR_COUNT();
2790	memset(laddr, 0, sizeof(*laddr));
2791	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2792	laddr->ifa = ifa;
2793	laddr->action = act;
2794	atomic_add_int(&ifa->refcount, 1);
2795	/* insert it */
2796	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
2797
2798	return (0);
2799}
2800
2801/*
2802 * Remove an laddr entry from the local address list (on an assoc)
2803 */
2804static void
2805sctp_remove_laddr(struct sctp_laddr *laddr)
2806{
2807
2808	/* remove from the list */
2809	LIST_REMOVE(laddr, sctp_nxt_addr);
2810	sctp_free_ifa(laddr->ifa);
2811	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
2812	SCTP_DECR_LADDR_COUNT();
2813}
2814
2815/*
2816 * Bind the socket, with the PCB and global info locks held.  Note, if a
2817 * socket address is specified, the PCB lock may be dropped and re-acquired.
2818 *
2819 * sctp_ifap is used to bypass normal local address validation checks.
2820 */
2821int
2822sctp_inpcb_bind_locked(struct sctp_inpcb *inp, struct sockaddr *addr,
2823    struct sctp_ifa *sctp_ifap, struct thread *td)
2824{
2825	/* bind a ep to a socket address */
2826	struct sctppcbhead *head;
2827	struct sctp_inpcb *inp_tmp;
2828	struct inpcb *ip_inp;
2829	int port_reuse_active = 0;
2830	int bindall;
2831	uint16_t lport;
2832	int error;
2833	uint32_t vrf_id;
2834
2835	KASSERT(td != NULL, ("%s: null thread", __func__));
2836
2837	error = 0;
2838	lport = 0;
2839	bindall = 1;
2840	ip_inp = &inp->ip_inp.inp;
2841
2842	SCTP_INP_INFO_WLOCK_ASSERT();
2843	SCTP_INP_WLOCK_ASSERT(inp);
2844
2845#ifdef SCTP_DEBUG
2846	if (addr) {
2847		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2848		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2849		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2850		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2851	}
2852#endif
2853	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2854		error = EINVAL;
2855		/* already did a bind, subsequent binds NOT allowed ! */
2856		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2857		goto out;
2858	}
2859	if (addr != NULL) {
2860		switch (addr->sa_family) {
2861#ifdef INET
2862		case AF_INET:
2863			{
2864				struct sockaddr_in *sin;
2865
2866				/* IPV6_V6ONLY socket? */
2867				if (SCTP_IPV6_V6ONLY(inp)) {
2868					error = EINVAL;
2869					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2870					goto out;
2871				}
2872				if (addr->sa_len != sizeof(*sin)) {
2873					error = EINVAL;
2874					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2875					goto out;
2876				}
2877
2878				sin = (struct sockaddr_in *)addr;
2879				lport = sin->sin_port;
2880				/*
2881				 * For LOOPBACK the prison_local_ip4() call
2882				 * will transmute the ip address to the
2883				 * proper value.
2884				 */
2885				if ((error = prison_local_ip4(td->td_ucred, &sin->sin_addr)) != 0) {
2886					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2887					goto out;
2888				}
2889				if (sin->sin_addr.s_addr != INADDR_ANY) {
2890					bindall = 0;
2891				}
2892				break;
2893			}
2894#endif
2895#ifdef INET6
2896		case AF_INET6:
2897			{
2898				/*
2899				 * Only for pure IPv6 Address. (No IPv4
2900				 * Mapped!)
2901				 */
2902				struct sockaddr_in6 *sin6;
2903
2904				sin6 = (struct sockaddr_in6 *)addr;
2905				if (addr->sa_len != sizeof(*sin6)) {
2906					error = EINVAL;
2907					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2908					goto out;
2909				}
2910				lport = sin6->sin6_port;
2911				/*
2912				 * For LOOPBACK the prison_local_ip6() call
2913				 * will transmute the ipv6 address to the
2914				 * proper value.
2915				 */
2916				if ((error = prison_local_ip6(td->td_ucred, &sin6->sin6_addr,
2917				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2918					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2919					goto out;
2920				}
2921				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2922					bindall = 0;
2923					/* KAME hack: embed scopeid */
2924					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2925						error = EINVAL;
2926						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2927						goto out;
2928					}
2929				}
2930				/* this must be cleared for ifa_ifwithaddr() */
2931				sin6->sin6_scope_id = 0;
2932				break;
2933			}
2934#endif
2935		default:
2936			error = EAFNOSUPPORT;
2937			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2938			goto out;
2939		}
2940	}
2941	/* Setup a vrf_id to be the default for the non-bind-all case. */
2942	vrf_id = inp->def_vrf_id;
2943
2944	if (lport) {
2945		/*
2946		 * Did the caller specify a port? if so we must see if an ep
2947		 * already has this one bound.
2948		 */
2949		/* got to be root to get at low ports */
2950		if (ntohs(lport) < IPPORT_RESERVED &&
2951		    (error = priv_check(td, PRIV_NETINET_RESERVEDPORT)) != 0) {
2952			goto out;
2953		}
2954		SCTP_INP_INCR_REF(inp);
2955		SCTP_INP_WUNLOCK(inp);
2956		if (bindall) {
2957			vrf_id = inp->def_vrf_id;
2958			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2959			if (inp_tmp != NULL) {
2960				/*
2961				 * lock guy returned and lower count note
2962				 * that we are not bound so inp_tmp should
2963				 * NEVER be inp. And it is this inp
2964				 * (inp_tmp) that gets the reference bump,
2965				 * so we must lower it.
2966				 */
2967				SCTP_INP_DECR_REF(inp_tmp);
2968				/* unlock info */
2969				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2970				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2971					/*
2972					 * Ok, must be one-2-one and
2973					 * allowing port re-use
2974					 */
2975					port_reuse_active = 1;
2976					goto continue_anyway;
2977				}
2978				SCTP_INP_WLOCK(inp);
2979				SCTP_INP_DECR_REF(inp);
2980				error = EADDRINUSE;
2981				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2982				goto out;
2983			}
2984		} else {
2985			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2986			if (inp_tmp != NULL) {
2987				/*
2988				 * lock guy returned and lower count note
2989				 * that we are not bound so inp_tmp should
2990				 * NEVER be inp. And it is this inp
2991				 * (inp_tmp) that gets the reference bump,
2992				 * so we must lower it.
2993				 */
2994				SCTP_INP_DECR_REF(inp_tmp);
2995				/* unlock info */
2996				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2997				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2998					/*
2999					 * Ok, must be one-2-one and
3000					 * allowing port re-use
3001					 */
3002					port_reuse_active = 1;
3003					goto continue_anyway;
3004				}
3005				SCTP_INP_WLOCK(inp);
3006				SCTP_INP_DECR_REF(inp);
3007				error = EADDRINUSE;
3008				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3009				goto out;
3010			}
3011		}
3012continue_anyway:
3013		SCTP_INP_WLOCK(inp);
3014		SCTP_INP_DECR_REF(inp);
3015		if (bindall) {
3016			/* verify that no lport is not used by a singleton */
3017			if ((port_reuse_active == 0) &&
3018			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
3019				/* Sorry someone already has this one bound */
3020				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
3021				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
3022					port_reuse_active = 1;
3023				} else {
3024					error = EADDRINUSE;
3025					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3026					goto out;
3027				}
3028			}
3029		}
3030	} else {
3031		uint16_t first, last, candidate;
3032		uint16_t count;
3033
3034		if (ip_inp->inp_flags & INP_HIGHPORT) {
3035			first = MODULE_GLOBAL(ipport_hifirstauto);
3036			last = MODULE_GLOBAL(ipport_hilastauto);
3037		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3038			if ((error = priv_check(td, PRIV_NETINET_RESERVEDPORT)) != 0) {
3039				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3040				goto out;
3041			}
3042			first = MODULE_GLOBAL(ipport_lowfirstauto);
3043			last = MODULE_GLOBAL(ipport_lowlastauto);
3044		} else {
3045			first = MODULE_GLOBAL(ipport_firstauto);
3046			last = MODULE_GLOBAL(ipport_lastauto);
3047		}
3048		if (first > last) {
3049			uint16_t temp;
3050
3051			temp = first;
3052			first = last;
3053			last = temp;
3054		}
3055		count = last - first + 1;	/* number of candidates */
3056		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3057
3058		for (;;) {
3059			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3060				lport = htons(candidate);
3061				break;
3062			}
3063			if (--count == 0) {
3064				error = EADDRINUSE;
3065				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3066				goto out;
3067			}
3068			if (candidate == last)
3069				candidate = first;
3070			else
3071				candidate = candidate + 1;
3072		}
3073	}
3074	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3075	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3076		/*
3077		 * this really should not happen. The guy did a non-blocking
3078		 * bind and then did a close at the same time.
3079		 */
3080		error = EINVAL;
3081		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3082		goto out;
3083	}
3084	/* ok we look clear to give out this port, so lets setup the binding */
3085	if (bindall) {
3086		/* binding to all addresses, so just set in the proper flags */
3087		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3088		/* set the automatic addr changes from kernel flag */
3089		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3090			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3091			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3092		} else {
3093			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3094			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3095		}
3096		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3097			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3098		} else {
3099			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3100		}
3101		/*
3102		 * set the automatic mobility_base from kernel flag (by
3103		 * micchie)
3104		 */
3105		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3106			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3107			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3108		} else {
3109			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3110			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3111		}
3112		/*
3113		 * set the automatic mobility_fasthandoff from kernel flag
3114		 * (by micchie)
3115		 */
3116		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3117			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3118			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3119		} else {
3120			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3121			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3122		}
3123	} else {
3124		/*
3125		 * bind specific, make sure flags is off and add a new
3126		 * address structure to the sctp_addr_list inside the ep
3127		 * structure.
3128		 *
3129		 * We will need to allocate one and insert it at the head.
3130		 * The socketopt call can just insert new addresses in there
3131		 * as well. It will also have to do the embed scope kame
3132		 * hack too (before adding).
3133		 */
3134		struct sctp_ifa *ifa;
3135		union sctp_sockstore store;
3136
3137		memset(&store, 0, sizeof(store));
3138		switch (addr->sa_family) {
3139#ifdef INET
3140		case AF_INET:
3141			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3142			store.sin.sin_port = 0;
3143			break;
3144#endif
3145#ifdef INET6
3146		case AF_INET6:
3147			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3148			store.sin6.sin6_port = 0;
3149			break;
3150#endif
3151		default:
3152			break;
3153		}
3154		/*
3155		 * first find the interface with the bound address need to
3156		 * zero out the port to find the address! yuck! can't do
3157		 * this earlier since need port for sctp_pcb_findep()
3158		 */
3159		if (sctp_ifap != NULL) {
3160			ifa = sctp_ifap;
3161		} else {
3162			/*
3163			 * Note for BSD we hit here always other O/S's will
3164			 * pass things in via the sctp_ifap argument.
3165			 */
3166			ifa = sctp_find_ifa_by_addr(&store.sa,
3167			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3168		}
3169		if (ifa == NULL) {
3170			error = EADDRNOTAVAIL;
3171			/* Can't find an interface with that address */
3172			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3173			goto out;
3174		}
3175#ifdef INET6
3176		if (addr->sa_family == AF_INET6) {
3177			/* GAK, more FIXME IFA lock? */
3178			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3179				/* Can't bind a non-existent addr. */
3180				error = EINVAL;
3181				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3182				goto out;
3183			}
3184		}
3185#endif
3186		/* we're not bound all */
3187		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3188		/* allow bindx() to send ASCONF's for binding changes */
3189		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3190		/* clear automatic addr changes from kernel flag */
3191		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3192
3193		/* add this address to the endpoint list */
3194		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3195		if (error != 0)
3196			goto out;
3197		inp->laddr_count++;
3198	}
3199	/* find the bucket */
3200	if (port_reuse_active) {
3201		/* Put it into tcp 1-2-1 hash */
3202		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3203		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3204	} else {
3205		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3206	}
3207	/* put it in the bucket */
3208	LIST_INSERT_HEAD(head, inp, sctp_hash);
3209	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3210	    (void *)head, ntohs(lport), port_reuse_active);
3211	/* set in the port */
3212	inp->sctp_lport = lport;
3213
3214	/* turn off just the unbound flag */
3215	KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != 0,
3216	    ("%s: inp %p is already bound", __func__, inp));
3217	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3218out:
3219	return (error);
3220}
3221
3222int
3223sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
3224    struct sctp_ifa *sctp_ifap, struct thread *td)
3225{
3226	struct sctp_inpcb *inp;
3227	int error;
3228
3229	inp = so->so_pcb;
3230	SCTP_INP_INFO_WLOCK();
3231	SCTP_INP_WLOCK(inp);
3232	error = sctp_inpcb_bind_locked(inp, addr, sctp_ifap, td);
3233	SCTP_INP_WUNLOCK(inp);
3234	SCTP_INP_INFO_WUNLOCK();
3235	return (error);
3236}
3237
3238static void
3239sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3240{
3241	struct sctp_iterator *it, *nit;
3242
3243	/*
3244	 * We enter with the only the ITERATOR_LOCK in place and a write
3245	 * lock on the inp_info stuff.
3246	 */
3247	it = sctp_it_ctl.cur_it;
3248	if (it && (it->vn != curvnet)) {
3249		/* Its not looking at our VNET */
3250		return;
3251	}
3252	if (it && (it->inp == inp)) {
3253		/*
3254		 * This is tricky and we hold the iterator lock, but when it
3255		 * returns and gets the lock (when we release it) the
3256		 * iterator will try to operate on inp. We need to stop that
3257		 * from happening. But of course the iterator has a
3258		 * reference on the stcb and inp. We can mark it and it will
3259		 * stop.
3260		 *
3261		 * If its a single iterator situation, we set the end
3262		 * iterator flag. Otherwise we set the iterator to go to the
3263		 * next inp.
3264		 *
3265		 */
3266		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3267			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3268		} else {
3269			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3270		}
3271	}
3272	/*
3273	 * Now go through and remove any single reference to our inp that
3274	 * may be still pending on the list
3275	 */
3276	SCTP_IPI_ITERATOR_WQ_LOCK();
3277	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3278		if (it->vn != curvnet) {
3279			continue;
3280		}
3281		if (it->inp == inp) {
3282			/* This one points to me is it inp specific? */
3283			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3284				/* Remove and free this one */
3285				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3286				    it, sctp_nxt_itr);
3287				if (it->function_atend != NULL) {
3288					(*it->function_atend) (it->pointer, it->val);
3289				}
3290				SCTP_FREE(it, SCTP_M_ITER);
3291			} else {
3292				it->inp = LIST_NEXT(it->inp, sctp_list);
3293				if (it->inp) {
3294					SCTP_INP_INCR_REF(it->inp);
3295				}
3296			}
3297			/*
3298			 * When its put in the refcnt is incremented so decr
3299			 * it
3300			 */
3301			SCTP_INP_DECR_REF(inp);
3302		}
3303	}
3304	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3305}
3306
3307/* release sctp_inpcb unbind the port */
3308void
3309sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3310{
3311	/*
3312	 * Here we free a endpoint. We must find it (if it is in the Hash
3313	 * table) and remove it from there. Then we must also find it in the
3314	 * overall list and remove it from there. After all removals are
3315	 * complete then any timer has to be stopped. Then start the actual
3316	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3317	 * all associations. d) finally the ep itself.
3318	 */
3319	struct sctp_tcb *stcb, *nstcb;
3320	struct sctp_laddr *laddr, *nladdr;
3321	struct inpcb *ip_pcb;
3322	struct socket *so;
3323	int being_refed = 0;
3324	struct sctp_queued_to_read *sq, *nsq;
3325	int cnt;
3326	sctp_sharedkey_t *shared_key, *nshared_key;
3327
3328#ifdef SCTP_LOG_CLOSING
3329	sctp_log_closing(inp, NULL, 0);
3330#endif
3331	SCTP_ITERATOR_LOCK();
3332	/* mark any iterators on the list or being processed */
3333	sctp_iterator_inp_being_freed(inp);
3334	SCTP_ITERATOR_UNLOCK();
3335
3336	SCTP_ASOC_CREATE_LOCK(inp);
3337	SCTP_INP_INFO_WLOCK();
3338	SCTP_INP_WLOCK(inp);
3339	so = inp->sctp_socket;
3340	KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) != 0,
3341	    ("%s: inp %p still has socket", __func__, inp));
3342	KASSERT((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0,
3343	    ("%s: double free of inp %p", __func__, inp));
3344	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3345		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3346		/* socket is gone, so no more wakeups allowed */
3347		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3348		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3349		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3350	}
3351	/* First time through we have the socket lock, after that no more. */
3352	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3353	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3354
3355	if (inp->control) {
3356		sctp_m_freem(inp->control);
3357		inp->control = NULL;
3358	}
3359	if (inp->pkt) {
3360		sctp_m_freem(inp->pkt);
3361		inp->pkt = NULL;
3362	}
3363	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3364					 * here but I will be nice :> (i.e.
3365					 * ip_pcb = ep;) */
3366	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3367		int cnt_in_sd;
3368
3369		cnt_in_sd = 0;
3370		LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_list, sctp_tcblist, nstcb) {
3371			SCTP_TCB_LOCK(stcb);
3372			/* Disconnect the socket please. */
3373			stcb->sctp_socket = NULL;
3374			SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_CLOSED_SOCKET);
3375			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3376				/* Skip guys being freed */
3377				cnt_in_sd++;
3378				if (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3379					/*
3380					 * Special case - we did not start a
3381					 * kill timer on the asoc due to it
3382					 * was not closed. So go ahead and
3383					 * start it now.
3384					 */
3385					SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
3386					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3387				}
3388				SCTP_TCB_UNLOCK(stcb);
3389				continue;
3390			}
3391			if (((SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_WAIT) ||
3392			    (SCTP_GET_STATE(stcb) == SCTP_STATE_COOKIE_ECHOED)) &&
3393			    (stcb->asoc.total_output_queue_size == 0)) {
3394				/*
3395				 * If we have data in queue, we don't want
3396				 * to just free since the app may have done,
3397				 * send()/close or connect/send/close. And
3398				 * it wants the data to get across first.
3399				 */
3400				/* Just abandon things in the front states */
3401				if (sctp_free_assoc(inp, stcb, SCTP_PCBFREE_NOFORCE,
3402				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3403					cnt_in_sd++;
3404				}
3405				continue;
3406			}
3407			if ((stcb->asoc.size_on_reasm_queue > 0) ||
3408			    (stcb->asoc.size_on_all_streams > 0) ||
3409			    ((so != NULL) && (SCTP_SBAVAIL(&so->so_rcv) > 0))) {
3410				/* Left with Data unread */
3411				struct mbuf *op_err;
3412
3413				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3414				stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3415				sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED);
3416				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3417				if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
3418				    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3419					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3420				}
3421				if (sctp_free_assoc(inp, stcb,
3422				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3423					cnt_in_sd++;
3424				}
3425				continue;
3426			} else if (TAILQ_EMPTY(&stcb->asoc.send_queue) &&
3427				    TAILQ_EMPTY(&stcb->asoc.sent_queue) &&
3428			    (stcb->asoc.stream_queue_cnt == 0)) {
3429				if ((*stcb->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (stcb, &stcb->asoc)) {
3430					goto abort_anyway;
3431				}
3432				if ((SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_SENT) &&
3433				    (SCTP_GET_STATE(stcb) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3434					struct sctp_nets *netp;
3435
3436					/*
3437					 * there is nothing queued to send,
3438					 * so I send shutdown
3439					 */
3440					if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
3441					    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3442						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3443					}
3444					SCTP_SET_STATE(stcb, SCTP_STATE_SHUTDOWN_SENT);
3445					sctp_stop_timers_for_shutdown(stcb);
3446					if (stcb->asoc.alternate) {
3447						netp = stcb->asoc.alternate;
3448					} else {
3449						netp = stcb->asoc.primary_destination;
3450					}
3451					sctp_send_shutdown(stcb, netp);
3452					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb,
3453					    netp);
3454					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, NULL);
3455					sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3456				}
3457			} else {
3458				/* mark into shutdown pending */
3459				SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_SHUTDOWN_PENDING);
3460				if ((*stcb->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (stcb, &stcb->asoc)) {
3461					SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_PARTIAL_MSG_LEFT);
3462				}
3463				if (TAILQ_EMPTY(&stcb->asoc.send_queue) &&
3464				    TAILQ_EMPTY(&stcb->asoc.sent_queue) &&
3465				    (stcb->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3466					struct mbuf *op_err;
3467
3468			abort_anyway:
3469					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3470					stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3471					sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED);
3472					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3473					if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
3474					    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3475						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3476					}
3477					if (sctp_free_assoc(inp, stcb,
3478					    SCTP_PCBFREE_NOFORCE,
3479					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3480						cnt_in_sd++;
3481					}
3482					continue;
3483				} else {
3484					sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3485				}
3486			}
3487			cnt_in_sd++;
3488			SCTP_TCB_UNLOCK(stcb);
3489		}
3490		/* now is there some left in our SHUTDOWN state? */
3491		if (cnt_in_sd) {
3492#ifdef SCTP_LOG_CLOSING
3493			sctp_log_closing(inp, NULL, 2);
3494#endif
3495			inp->sctp_socket = NULL;
3496			SCTP_INP_WUNLOCK(inp);
3497			SCTP_ASOC_CREATE_UNLOCK(inp);
3498			SCTP_INP_INFO_WUNLOCK();
3499			return;
3500		}
3501	}
3502	inp->sctp_socket = NULL;
3503	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
3504		/*
3505		 * ok, this guy has been bound. It's port is somewhere in
3506		 * the SCTP_BASE_INFO(hash table). Remove it!
3507		 */
3508		LIST_REMOVE(inp, sctp_hash);
3509		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3510	}
3511
3512	/*
3513	 * If there is a timer running to kill us, forget it, since it may
3514	 * have a contest on the INP lock.. which would cause us to die ...
3515	 */
3516	cnt = 0;
3517	LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_list, sctp_tcblist, nstcb) {
3518		SCTP_TCB_LOCK(stcb);
3519		if (immediate != SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3520			/* Disconnect the socket please */
3521			stcb->sctp_socket = NULL;
3522			SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_CLOSED_SOCKET);
3523		}
3524		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3525			if (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3526				SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
3527				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3528			}
3529			cnt++;
3530			SCTP_TCB_UNLOCK(stcb);
3531			continue;
3532		}
3533		/* Free associations that are NOT killing us */
3534		if ((SCTP_GET_STATE(stcb) != SCTP_STATE_COOKIE_WAIT) &&
3535		    ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3536			struct mbuf *op_err;
3537
3538			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3539			stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3540			sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED);
3541			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3542		} else if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3543			cnt++;
3544			SCTP_TCB_UNLOCK(stcb);
3545			continue;
3546		}
3547		if ((SCTP_GET_STATE(stcb) == SCTP_STATE_OPEN) ||
3548		    (SCTP_GET_STATE(stcb) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3549			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3550		}
3551		if (sctp_free_assoc(inp, stcb, SCTP_PCBFREE_FORCE,
3552		    SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3553			cnt++;
3554		}
3555	}
3556	if (cnt) {
3557		/* Ok we have someone out there that will kill us */
3558#ifdef SCTP_LOG_CLOSING
3559		sctp_log_closing(inp, NULL, 3);
3560#endif
3561		SCTP_INP_WUNLOCK(inp);
3562		SCTP_ASOC_CREATE_UNLOCK(inp);
3563		SCTP_INP_INFO_WUNLOCK();
3564		return;
3565	}
3566	if (SCTP_INP_LOCK_CONTENDED(inp))
3567		being_refed++;
3568	if (SCTP_INP_READ_CONTENDED(inp))
3569		being_refed++;
3570	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3571		being_refed++;
3572	/* NOTE: 0 refcount also means no timers are referencing us. */
3573	if ((inp->refcount) ||
3574	    (being_refed) ||
3575	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3576#ifdef SCTP_LOG_CLOSING
3577		sctp_log_closing(inp, NULL, 4);
3578#endif
3579		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3580		SCTP_INP_WUNLOCK(inp);
3581		SCTP_ASOC_CREATE_UNLOCK(inp);
3582		SCTP_INP_INFO_WUNLOCK();
3583		return;
3584	}
3585	inp->sctp_ep.signature_change.type = 0;
3586	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3587	/*
3588	 * Remove it from the list .. last thing we need a lock for.
3589	 */
3590	LIST_REMOVE(inp, sctp_list);
3591	SCTP_INP_WUNLOCK(inp);
3592	SCTP_ASOC_CREATE_UNLOCK(inp);
3593	SCTP_INP_INFO_WUNLOCK();
3594
3595#ifdef SCTP_LOG_CLOSING
3596	sctp_log_closing(inp, NULL, 5);
3597#endif
3598	if ((inp->sctp_asocidhash) != NULL) {
3599		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3600		inp->sctp_asocidhash = NULL;
3601	}
3602	/* sa_ignore FREED_MEMORY */
3603	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3604		/* Its only abandoned if it had data left */
3605		if (sq->length)
3606			SCTP_STAT_INCR(sctps_left_abandon);
3607
3608		TAILQ_REMOVE(&inp->read_queue, sq, next);
3609		sctp_free_remote_addr(sq->whoFrom);
3610		if (so)
3611			SCTP_SB_DECR(&so->so_rcv, sq->length);
3612		if (sq->data) {
3613			sctp_m_freem(sq->data);
3614			sq->data = NULL;
3615		}
3616		/*
3617		 * no need to free the net count, since at this point all
3618		 * assoc's are gone.
3619		 */
3620		sctp_free_a_readq(NULL, sq);
3621	}
3622	/* Now the sctp_pcb things */
3623	/*
3624	 * free each asoc if it is not already closed/free. we can't use the
3625	 * macro here since le_next will get freed as part of the
3626	 * sctp_free_assoc() call.
3627	 */
3628	if (ip_pcb->inp_options) {
3629		(void)sctp_m_free(ip_pcb->inp_options);
3630		ip_pcb->inp_options = 0;
3631	}
3632#ifdef INET6
3633	if (ip_pcb->inp_vflag & INP_IPV6) {
3634		ip6_freepcbopts(ip_pcb->in6p_outputopts);
3635	}
3636#endif				/* INET6 */
3637	ip_pcb->inp_vflag = 0;
3638	/* free up authentication fields */
3639	if (inp->sctp_ep.local_auth_chunks != NULL)
3640		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3641	if (inp->sctp_ep.local_hmacs != NULL)
3642		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3643
3644	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3645		LIST_REMOVE(shared_key, next);
3646		sctp_free_sharedkey(shared_key);
3647		/* sa_ignore FREED_MEMORY */
3648	}
3649
3650	/*
3651	 * if we have an address list the following will free the list of
3652	 * ifaddr's that are set into this ep. Again macro limitations here,
3653	 * since the LIST_FOREACH could be a bad idea.
3654	 */
3655	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3656		sctp_remove_laddr(laddr);
3657	}
3658
3659#ifdef SCTP_TRACK_FREED_ASOCS
3660	/* TEMP CODE */
3661	LIST_FOREACH_SAFE(stcb, &inp->sctp_asoc_free_list, sctp_tcblist, nstcb) {
3662		LIST_REMOVE(stcb, sctp_tcblist);
3663		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
3664		SCTP_DECR_ASOC_COUNT();
3665	}
3666	/* *** END TEMP CODE *** */
3667#endif
3668	/* Now lets see about freeing the EP hash table. */
3669	if (inp->sctp_tcbhash != NULL) {
3670		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3671		inp->sctp_tcbhash = NULL;
3672	}
3673	/* Now we must put the ep memory back into the zone pool */
3674	crfree(inp->ip_inp.inp.inp_cred);
3675	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3676	SCTP_INP_LOCK_DESTROY(inp);
3677	SCTP_INP_READ_LOCK_DESTROY(inp);
3678	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3679	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3680	SCTP_DECR_EP_COUNT();
3681}
3682
3683struct sctp_nets *
3684sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3685{
3686	struct sctp_nets *net;
3687
3688	/* locate the address */
3689	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3690		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3691			return (net);
3692	}
3693	return (NULL);
3694}
3695
3696int
3697sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3698{
3699	struct sctp_ifa *sctp_ifa;
3700
3701	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3702	if (sctp_ifa) {
3703		return (1);
3704	} else {
3705		return (0);
3706	}
3707}
3708
3709/*
3710 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3711 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3712 * stats of stuff.
3713 */
3714int
3715sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3716    struct sctp_nets **netp, uint16_t port, int set_scope, int from)
3717{
3718	/*
3719	 * The following is redundant to the same lines in the
3720	 * sctp_aloc_assoc() but is needed since others call the add address
3721	 * function
3722	 */
3723	struct sctp_nets *net, *netfirst;
3724	int addr_inscope;
3725
3726	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3727	    from);
3728	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3729
3730	netfirst = sctp_findnet(stcb, newaddr);
3731	if (netfirst) {
3732		/*
3733		 * Lie and return ok, we don't want to make the association
3734		 * go away for this behavior. It will happen in the TCP
3735		 * model in a connected socket. It does not reach the hash
3736		 * table until after the association is built so it can't be
3737		 * found. Mark as reachable, since the initial creation will
3738		 * have been cleared and the NOT_IN_ASSOC flag will have
3739		 * been added... and we don't want to end up removing it
3740		 * back out.
3741		 */
3742		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3743			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3744			    SCTP_ADDR_UNCONFIRMED);
3745		} else {
3746			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3747		}
3748
3749		return (0);
3750	}
3751	addr_inscope = 1;
3752	switch (newaddr->sa_family) {
3753#ifdef INET
3754	case AF_INET:
3755		{
3756			struct sockaddr_in *sin;
3757
3758			sin = (struct sockaddr_in *)newaddr;
3759			if (sin->sin_addr.s_addr == 0) {
3760				/* Invalid address */
3761				return (-1);
3762			}
3763			/* zero out the zero area */
3764			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3765
3766			/* assure len is set */
3767			sin->sin_len = sizeof(struct sockaddr_in);
3768			if (set_scope) {
3769				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3770					stcb->asoc.scope.ipv4_local_scope = 1;
3771				}
3772			} else {
3773				/* Validate the address is in scope */
3774				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3775				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3776					addr_inscope = 0;
3777				}
3778			}
3779			break;
3780		}
3781#endif
3782#ifdef INET6
3783	case AF_INET6:
3784		{
3785			struct sockaddr_in6 *sin6;
3786
3787			sin6 = (struct sockaddr_in6 *)newaddr;
3788			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3789				/* Invalid address */
3790				return (-1);
3791			}
3792			/* assure len is set */
3793			sin6->sin6_len = sizeof(struct sockaddr_in6);
3794			if (set_scope) {
3795				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3796					stcb->asoc.scope.loopback_scope = 1;
3797					stcb->asoc.scope.local_scope = 0;
3798					stcb->asoc.scope.ipv4_local_scope = 1;
3799					stcb->asoc.scope.site_scope = 1;
3800				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3801					/*
3802					 * If the new destination is a
3803					 * LINK_LOCAL we must have common
3804					 * site scope. Don't set the local
3805					 * scope since we may not share all
3806					 * links, only loopback can do this.
3807					 * Links on the local network would
3808					 * also be on our private network
3809					 * for v4 too.
3810					 */
3811					stcb->asoc.scope.ipv4_local_scope = 1;
3812					stcb->asoc.scope.site_scope = 1;
3813				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3814					/*
3815					 * If the new destination is
3816					 * SITE_LOCAL then we must have site
3817					 * scope in common.
3818					 */
3819					stcb->asoc.scope.site_scope = 1;
3820				}
3821			} else {
3822				/* Validate the address is in scope */
3823				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3824				    (stcb->asoc.scope.loopback_scope == 0)) {
3825					addr_inscope = 0;
3826				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3827				    (stcb->asoc.scope.local_scope == 0)) {
3828					addr_inscope = 0;
3829				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3830				    (stcb->asoc.scope.site_scope == 0)) {
3831					addr_inscope = 0;
3832				}
3833			}
3834			break;
3835		}
3836#endif
3837	default:
3838		/* not supported family type */
3839		return (-1);
3840	}
3841	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3842	if (net == NULL) {
3843		return (-1);
3844	}
3845	SCTP_INCR_RADDR_COUNT();
3846	memset(net, 0, sizeof(struct sctp_nets));
3847	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3848	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3849	switch (newaddr->sa_family) {
3850#ifdef INET
3851	case AF_INET:
3852		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3853		break;
3854#endif
3855#ifdef INET6
3856	case AF_INET6:
3857		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3858		break;
3859#endif
3860	default:
3861		break;
3862	}
3863	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3864	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3865		stcb->asoc.scope.loopback_scope = 1;
3866		stcb->asoc.scope.ipv4_local_scope = 1;
3867		stcb->asoc.scope.local_scope = 0;
3868		stcb->asoc.scope.site_scope = 1;
3869		addr_inscope = 1;
3870	}
3871	net->failure_threshold = stcb->asoc.def_net_failure;
3872	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3873	if (addr_inscope == 0) {
3874		net->dest_state = (SCTP_ADDR_REACHABLE |
3875		    SCTP_ADDR_OUT_OF_SCOPE);
3876	} else {
3877		if (from == SCTP_ADDR_IS_CONFIRMED)
3878			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3879			net->dest_state = SCTP_ADDR_REACHABLE;
3880		else
3881			net->dest_state = SCTP_ADDR_REACHABLE |
3882			    SCTP_ADDR_UNCONFIRMED;
3883	}
3884	/*
3885	 * We set this to 0, the timer code knows that this means its an
3886	 * initial value
3887	 */
3888	net->rto_needed = 1;
3889	net->RTO = 0;
3890	net->RTO_measured = 0;
3891	stcb->asoc.numnets++;
3892	net->ref_count = 1;
3893	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3894	net->port = port;
3895	net->dscp = stcb->asoc.default_dscp;
3896#ifdef INET6
3897	net->flowlabel = stcb->asoc.default_flowlabel;
3898#endif
3899	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3900		net->dest_state |= SCTP_ADDR_NOHB;
3901	} else {
3902		net->dest_state &= ~SCTP_ADDR_NOHB;
3903	}
3904	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3905		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3906	} else {
3907		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3908	}
3909	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3910	/* Init the timer structure */
3911	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3912	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3913	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3914
3915	/* Now generate a route for this guy */
3916#ifdef INET6
3917	/* KAME hack: embed scopeid */
3918	if (newaddr->sa_family == AF_INET6) {
3919		struct sockaddr_in6 *sin6;
3920
3921		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3922		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3923		sin6->sin6_scope_id = 0;
3924	}
3925#endif
3926	SCTP_RTALLOC((sctp_route_t *)&net->ro,
3927	    stcb->asoc.vrf_id,
3928	    stcb->sctp_ep->fibnum);
3929
3930	net->src_addr_selected = 0;
3931	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3932		/* Get source address */
3933		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3934		    stcb,
3935		    (sctp_route_t *)&net->ro,
3936		    net,
3937		    0,
3938		    stcb->asoc.vrf_id);
3939		if (stcb->asoc.default_mtu > 0) {
3940			net->mtu = stcb->asoc.default_mtu;
3941			switch (net->ro._l_addr.sa.sa_family) {
3942#ifdef INET
3943			case AF_INET:
3944				net->mtu += SCTP_MIN_V4_OVERHEAD;
3945				break;
3946#endif
3947#ifdef INET6
3948			case AF_INET6:
3949				net->mtu += SCTP_MIN_OVERHEAD;
3950				break;
3951#endif
3952			default:
3953				break;
3954			}
3955#if defined(INET) || defined(INET6)
3956			if (net->port) {
3957				net->mtu += (uint32_t)sizeof(struct udphdr);
3958			}
3959#endif
3960		} else if (net->ro._s_addr != NULL) {
3961			uint32_t imtu, rmtu, hcmtu;
3962
3963			net->src_addr_selected = 1;
3964			/* Now get the interface MTU */
3965			if (net->ro._s_addr->ifn_p != NULL) {
3966				/*
3967				 * XXX: Should we here just use
3968				 * net->ro._s_addr->ifn_p->ifn_mtu
3969				 */
3970				imtu = SCTP_GATHER_MTU_FROM_IFN_INFO(net->ro._s_addr->ifn_p->ifn_p,
3971				    net->ro._s_addr->ifn_p->ifn_index);
3972			} else {
3973				imtu = 0;
3974			}
3975			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_nh);
3976			hcmtu = sctp_hc_get_mtu(&net->ro._l_addr, stcb->sctp_ep->fibnum);
3977			net->mtu = sctp_min_mtu(hcmtu, rmtu, imtu);
3978		}
3979	}
3980	if (net->mtu == 0) {
3981		if (stcb->asoc.default_mtu > 0) {
3982			net->mtu = stcb->asoc.default_mtu;
3983			switch (net->ro._l_addr.sa.sa_family) {
3984#ifdef INET
3985			case AF_INET:
3986				net->mtu += SCTP_MIN_V4_OVERHEAD;
3987				break;
3988#endif
3989#ifdef INET6
3990			case AF_INET6:
3991				net->mtu += SCTP_MIN_OVERHEAD;
3992				break;
3993#endif
3994			default:
3995				break;
3996			}
3997#if defined(INET) || defined(INET6)
3998			if (net->port) {
3999				net->mtu += (uint32_t)sizeof(struct udphdr);
4000			}
4001#endif
4002		} else {
4003			switch (newaddr->sa_family) {
4004#ifdef INET
4005			case AF_INET:
4006				net->mtu = SCTP_DEFAULT_MTU;
4007				break;
4008#endif
4009#ifdef INET6
4010			case AF_INET6:
4011				net->mtu = 1280;
4012				break;
4013#endif
4014			default:
4015				break;
4016			}
4017		}
4018	}
4019#if defined(INET) || defined(INET6)
4020	if (net->port) {
4021		net->mtu -= (uint32_t)sizeof(struct udphdr);
4022	}
4023#endif
4024	if (from == SCTP_ALLOC_ASOC) {
4025		stcb->asoc.smallest_mtu = net->mtu;
4026	}
4027	if (stcb->asoc.smallest_mtu > net->mtu) {
4028		sctp_pathmtu_adjustment(stcb, net->mtu, true);
4029	}
4030#ifdef INET6
4031	if (newaddr->sa_family == AF_INET6) {
4032		struct sockaddr_in6 *sin6;
4033
4034		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4035		(void)sa6_recoverscope(sin6);
4036	}
4037#endif
4038
4039	/* JRS - Use the congestion control given in the CC module */
4040	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4041		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4042
4043	/*
4044	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4045	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4046	 */
4047	net->find_pseudo_cumack = 1;
4048	net->find_rtx_pseudo_cumack = 1;
4049	/* Choose an initial flowid. */
4050	net->flowid = stcb->asoc.my_vtag ^
4051	    ntohs(stcb->rport) ^
4052	    ntohs(stcb->sctp_ep->sctp_lport);
4053	net->flowtype = M_HASHTYPE_OPAQUE_HASH;
4054	if (netp) {
4055		*netp = net;
4056	}
4057	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4058	if (net->ro.ro_nh == NULL) {
4059		/* Since we have no route put it at the back */
4060		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4061	} else if (netfirst == NULL) {
4062		/* We are the first one in the pool. */
4063		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4064	} else if (netfirst->ro.ro_nh == NULL) {
4065		/*
4066		 * First one has NO route. Place this one ahead of the first
4067		 * one.
4068		 */
4069		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4070	} else if (net->ro.ro_nh->nh_ifp != netfirst->ro.ro_nh->nh_ifp) {
4071		/*
4072		 * This one has a different interface than the one at the
4073		 * top of the list. Place it ahead.
4074		 */
4075		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4076	} else {
4077		/*
4078		 * Ok we have the same interface as the first one. Move
4079		 * forward until we find either a) one with a NULL route...
4080		 * insert ahead of that b) one with a different ifp.. insert
4081		 * after that. c) end of the list.. insert at the tail.
4082		 */
4083		struct sctp_nets *netlook;
4084
4085		do {
4086			netlook = TAILQ_NEXT(netfirst, sctp_next);
4087			if (netlook == NULL) {
4088				/* End of the list */
4089				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4090				break;
4091			} else if (netlook->ro.ro_nh == NULL) {
4092				/* next one has NO route */
4093				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4094				break;
4095			} else if (netlook->ro.ro_nh->nh_ifp != net->ro.ro_nh->nh_ifp) {
4096				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4097				    net, sctp_next);
4098				break;
4099			}
4100			/* Shift forward */
4101			netfirst = netlook;
4102		} while (netlook != NULL);
4103	}
4104
4105	/* got to have a primary set */
4106	if (stcb->asoc.primary_destination == 0) {
4107		stcb->asoc.primary_destination = net;
4108	} else if ((stcb->asoc.primary_destination->ro.ro_nh == NULL) &&
4109		    (net->ro.ro_nh) &&
4110	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4111		/* No route to current primary adopt new primary */
4112		stcb->asoc.primary_destination = net;
4113	}
4114	/* Validate primary is first */
4115	net = TAILQ_FIRST(&stcb->asoc.nets);
4116	if ((net != stcb->asoc.primary_destination) &&
4117	    (stcb->asoc.primary_destination)) {
4118		/*
4119		 * first one on the list is NOT the primary sctp_cmpaddr()
4120		 * is much more efficient if the primary is the first on the
4121		 * list, make it so.
4122		 */
4123		TAILQ_REMOVE(&stcb->asoc.nets,
4124		    stcb->asoc.primary_destination, sctp_next);
4125		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4126		    stcb->asoc.primary_destination, sctp_next);
4127	}
4128	return (0);
4129}
4130
4131static uint32_t
4132sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4133{
4134	uint32_t id;
4135	struct sctpasochead *head;
4136	struct sctp_tcb *lstcb;
4137
4138try_again:
4139	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4140		/* TSNH */
4141		return (0);
4142	}
4143	/*
4144	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4145	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4146	 */
4147	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4148		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4149	}
4150	id = inp->sctp_associd_counter;
4151	inp->sctp_associd_counter++;
4152	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t)id, 0);
4153	if (lstcb) {
4154		goto try_again;
4155	}
4156	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4157	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4158	stcb->asoc.in_asocid_hash = 1;
4159	return (id);
4160}
4161
4162/*
4163 * allocate an association and add it to the endpoint. The caller must be
4164 * careful to add all additional addresses once they are know right away or
4165 * else the assoc will be may experience a blackout scenario.
4166 */
4167static struct sctp_tcb *
4168sctp_aloc_assoc_locked(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4169    int *error, uint32_t override_tag, uint32_t initial_tsn,
4170    uint32_t vrf_id, uint16_t o_streams, uint16_t port,
4171    struct thread *p,
4172    int initialize_auth_params)
4173{
4174	/* note the p argument is only valid in unbound sockets */
4175
4176	struct sctp_tcb *stcb;
4177	struct sctp_association *asoc;
4178	struct sctpasochead *head;
4179	uint16_t rport;
4180	int err;
4181
4182	SCTP_INP_INFO_WLOCK_ASSERT();
4183	SCTP_INP_WLOCK_ASSERT(inp);
4184
4185	/*
4186	 * Assumption made here: Caller has done a
4187	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4188	 * address does not exist already.
4189	 */
4190	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4191		/* Hit max assoc, sorry no more */
4192		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4193		*error = ENOBUFS;
4194		return (NULL);
4195	}
4196	if (firstaddr == NULL) {
4197		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4198		*error = EINVAL;
4199		return (NULL);
4200	}
4201	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4202		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4203		*error = EINVAL;
4204		return (NULL);
4205	}
4206	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4207	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4208	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4209		/*
4210		 * If its in the TCP pool, its NOT allowed to create an
4211		 * association. The parent listener needs to call
4212		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4213		 * off, or connected one does this.. its an error.
4214		 */
4215		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4216		*error = EINVAL;
4217		return (NULL);
4218	}
4219	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4220	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4221		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4222		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4223			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4224			*error = EINVAL;
4225			return (NULL);
4226		}
4227	}
4228	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4229#ifdef SCTP_DEBUG
4230	if (firstaddr) {
4231		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4232		switch (firstaddr->sa_family) {
4233#ifdef INET
4234		case AF_INET:
4235			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4236			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4237			break;
4238#endif
4239#ifdef INET6
4240		case AF_INET6:
4241			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4242			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4243			break;
4244#endif
4245		default:
4246			break;
4247		}
4248	} else {
4249		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4250	}
4251#endif				/* SCTP_DEBUG */
4252	switch (firstaddr->sa_family) {
4253#ifdef INET
4254	case AF_INET:
4255		{
4256			struct sockaddr_in *sin;
4257
4258			sin = (struct sockaddr_in *)firstaddr;
4259			if ((ntohs(sin->sin_port) == 0) ||
4260			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4261			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4262			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) ||
4263			    ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
4264			    (SCTP_IPV6_V6ONLY(inp) != 0))) {
4265				/* Invalid address */
4266				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4267				*error = EINVAL;
4268				return (NULL);
4269			}
4270			rport = sin->sin_port;
4271			break;
4272		}
4273#endif
4274#ifdef INET6
4275	case AF_INET6:
4276		{
4277			struct sockaddr_in6 *sin6;
4278
4279			sin6 = (struct sockaddr_in6 *)firstaddr;
4280			if ((ntohs(sin6->sin6_port) == 0) ||
4281			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4282			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) ||
4283			    ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0)) {
4284				/* Invalid address */
4285				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4286				*error = EINVAL;
4287				return (NULL);
4288			}
4289			rport = sin6->sin6_port;
4290			break;
4291		}
4292#endif
4293	default:
4294		/* not supported family type */
4295		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4296		*error = EINVAL;
4297		return (NULL);
4298	}
4299	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4300		/*
4301		 * If you have not performed a bind, then we need to do the
4302		 * ephemeral bind for you.
4303		 */
4304		if ((err = sctp_inpcb_bind_locked(inp, NULL, NULL, p))) {
4305			/* bind error, probably perm */
4306			*error = err;
4307			return (NULL);
4308		}
4309	}
4310	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4311	if (stcb == NULL) {
4312		/* out of memory? */
4313		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4314		*error = ENOMEM;
4315		return (NULL);
4316	}
4317	SCTP_INCR_ASOC_COUNT();
4318
4319	memset(stcb, 0, sizeof(*stcb));
4320	asoc = &stcb->asoc;
4321
4322	SCTP_TCB_LOCK_INIT(stcb);
4323	stcb->rport = rport;
4324	/* setup back pointer's */
4325	stcb->sctp_ep = inp;
4326	stcb->sctp_socket = inp->sctp_socket;
4327	if ((err = sctp_init_asoc(inp, stcb, override_tag, initial_tsn, vrf_id, o_streams))) {
4328		/* failed */
4329		SCTP_TCB_LOCK_DESTROY(stcb);
4330		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4331		SCTP_DECR_ASOC_COUNT();
4332		*error = err;
4333		return (NULL);
4334	}
4335	SCTP_TCB_LOCK(stcb);
4336
4337	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4338	/* now that my_vtag is set, add it to the hash */
4339	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4340	/* put it in the bucket in the vtag hash of assoc's for the system */
4341	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4342
4343	if (sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC)) {
4344		/* failure.. memory error? */
4345		if (asoc->strmout) {
4346			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4347			asoc->strmout = NULL;
4348		}
4349		if (asoc->mapping_array) {
4350			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4351			asoc->mapping_array = NULL;
4352		}
4353		if (asoc->nr_mapping_array) {
4354			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4355			asoc->nr_mapping_array = NULL;
4356		}
4357		SCTP_DECR_ASOC_COUNT();
4358		SCTP_TCB_UNLOCK(stcb);
4359		SCTP_TCB_LOCK_DESTROY(stcb);
4360		LIST_REMOVE(stcb, sctp_asocs);
4361		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4362		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4363		SCTP_INP_WUNLOCK(inp);
4364		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4365		*error = ENOBUFS;
4366		return (NULL);
4367	}
4368	/* Init all the timers */
4369	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4370	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4371	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4372	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4373	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4374	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4375
4376	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4377	/* now file the port under the hash as well */
4378	if (inp->sctp_tcbhash != NULL) {
4379		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4380		    inp->sctp_hashmark)];
4381		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4382	}
4383	if (initialize_auth_params == SCTP_INITIALIZE_AUTH_PARAMS) {
4384		sctp_initialize_auth_params(inp, stcb);
4385	}
4386	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4387	return (stcb);
4388}
4389
4390struct sctp_tcb *
4391sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4392    int *error, uint32_t override_tag, uint32_t initial_tsn,
4393    uint32_t vrf_id, uint16_t o_streams, uint16_t port,
4394    struct thread *p,
4395    int initialize_auth_params)
4396{
4397	struct sctp_tcb *stcb;
4398
4399	SCTP_INP_INFO_WLOCK();
4400	SCTP_INP_WLOCK(inp);
4401	stcb = sctp_aloc_assoc_locked(inp, firstaddr, error, override_tag,
4402	    initial_tsn, vrf_id, o_streams, port, p, initialize_auth_params);
4403	SCTP_INP_INFO_WUNLOCK();
4404	SCTP_INP_WUNLOCK(inp);
4405	return (stcb);
4406}
4407
4408struct sctp_tcb *
4409sctp_aloc_assoc_connected(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4410    int *error, uint32_t override_tag, uint32_t initial_tsn,
4411    uint32_t vrf_id, uint16_t o_streams, uint16_t port,
4412    struct thread *p,
4413    int initialize_auth_params)
4414{
4415	struct sctp_tcb *stcb;
4416
4417	SCTP_INP_INFO_WLOCK();
4418	SCTP_INP_WLOCK(inp);
4419	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) &&
4420	    SCTP_IS_LISTENING(inp)) {
4421		SCTP_INP_INFO_WUNLOCK();
4422		SCTP_INP_WUNLOCK(inp);
4423		*error = EINVAL;
4424		return (NULL);
4425	}
4426	stcb = sctp_aloc_assoc_locked(inp, firstaddr, error, override_tag,
4427	    initial_tsn, vrf_id, o_streams, port, p, initialize_auth_params);
4428	SCTP_INP_INFO_WUNLOCK();
4429	if (stcb != NULL && (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4430		inp->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED;
4431		soisconnecting(inp->sctp_socket);
4432	}
4433	SCTP_INP_WUNLOCK(inp);
4434	return (stcb);
4435}
4436
4437void
4438sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4439{
4440	struct sctp_inpcb *inp;
4441	struct sctp_association *asoc;
4442
4443	inp = stcb->sctp_ep;
4444	asoc = &stcb->asoc;
4445	asoc->numnets--;
4446	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4447	if (net == asoc->primary_destination) {
4448		/* Reset primary */
4449		struct sctp_nets *lnet;
4450
4451		lnet = TAILQ_FIRST(&asoc->nets);
4452		/*
4453		 * Mobility adaptation Ideally, if deleted destination is
4454		 * the primary, it becomes a fast retransmission trigger by
4455		 * the subsequent SET PRIMARY. (by micchie)
4456		 */
4457		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4458		    SCTP_MOBILITY_BASE) ||
4459		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4460		    SCTP_MOBILITY_FASTHANDOFF)) {
4461			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4462			if (asoc->deleted_primary != NULL) {
4463				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4464				goto out;
4465			}
4466			asoc->deleted_primary = net;
4467			atomic_add_int(&net->ref_count, 1);
4468			memset(&net->lastsa, 0, sizeof(net->lastsa));
4469			memset(&net->lastsv, 0, sizeof(net->lastsv));
4470			sctp_mobility_feature_on(stcb->sctp_ep,
4471			    SCTP_MOBILITY_PRIM_DELETED);
4472			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4473			    stcb->sctp_ep, stcb, NULL);
4474		}
4475out:
4476		/* Try to find a confirmed primary */
4477		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4478	}
4479	if (net == asoc->last_data_chunk_from) {
4480		/* Reset primary */
4481		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4482	}
4483	if (net == asoc->last_control_chunk_from) {
4484		/* Clear net */
4485		asoc->last_control_chunk_from = NULL;
4486	}
4487	if (net == asoc->last_net_cmt_send_started) {
4488		/* Clear net */
4489		asoc->last_net_cmt_send_started = NULL;
4490	}
4491	if (net == stcb->asoc.alternate) {
4492		sctp_free_remote_addr(stcb->asoc.alternate);
4493		stcb->asoc.alternate = NULL;
4494	}
4495	sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net,
4496	    SCTP_FROM_SCTP_PCB + SCTP_LOC_9);
4497	sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net,
4498	    SCTP_FROM_SCTP_PCB + SCTP_LOC_10);
4499	net->dest_state |= SCTP_ADDR_BEING_DELETED;
4500	sctp_free_remote_addr(net);
4501}
4502
4503/*
4504 * remove a remote endpoint address from an association, it will fail if the
4505 * address does not exist.
4506 */
4507int
4508sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4509{
4510	/*
4511	 * Here we need to remove a remote address. This is quite simple, we
4512	 * first find it in the list of address for the association
4513	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4514	 * on that item. Note we do not allow it to be removed if there are
4515	 * no other addresses.
4516	 */
4517	struct sctp_association *asoc;
4518	struct sctp_nets *net, *nnet;
4519
4520	asoc = &stcb->asoc;
4521
4522	/* locate the address */
4523	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4524		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4525			continue;
4526		}
4527		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4528		    remaddr)) {
4529			/* we found the guy */
4530			if (asoc->numnets < 2) {
4531				/* Must have at LEAST two remote addresses */
4532				return (-1);
4533			} else {
4534				sctp_remove_net(stcb, net);
4535				return (0);
4536			}
4537		}
4538	}
4539	/* not found. */
4540	return (-2);
4541}
4542
4543static bool
4544sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport, time_t now)
4545{
4546	struct sctpvtaghead *chain;
4547	struct sctp_tagblock *twait_block;
4548	int i;
4549
4550	SCTP_INP_INFO_LOCK_ASSERT();
4551	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4552	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4553		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4554			if ((twait_block->vtag_block[i].tv_sec_at_expire >= now) &&
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				return (true);
4559			}
4560		}
4561	}
4562	return (false);
4563}
4564
4565static void
4566sctp_set_vtag_block(struct sctp_timewait *vtag_block, time_t time,
4567    uint32_t tag, uint16_t lport, uint16_t rport)
4568{
4569	vtag_block->tv_sec_at_expire = time;
4570	vtag_block->v_tag = tag;
4571	vtag_block->lport = lport;
4572	vtag_block->rport = rport;
4573}
4574
4575static void
4576sctp_add_vtag_to_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4577{
4578	struct sctpvtaghead *chain;
4579	struct sctp_tagblock *twait_block;
4580	struct timeval now;
4581	time_t time;
4582	int i;
4583	bool set;
4584
4585	SCTP_INP_INFO_WLOCK_ASSERT();
4586	(void)SCTP_GETTIME_TIMEVAL(&now);
4587	time = now.tv_sec + SCTP_BASE_SYSCTL(sctp_vtag_time_wait);
4588	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4589	set = false;
4590	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4591		/* Block(s) present, lets find space, and expire on the fly */
4592		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4593			if ((twait_block->vtag_block[i].v_tag == 0) && !set) {
4594				sctp_set_vtag_block(twait_block->vtag_block + i, time, tag, lport, rport);
4595				set = true;
4596				continue;
4597			}
4598			if ((twait_block->vtag_block[i].v_tag != 0) &&
4599			    (twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4600				if (set) {
4601					/* Audit expires this guy */
4602					sctp_set_vtag_block(twait_block->vtag_block + i, 0, 0, 0, 0);
4603				} else {
4604					/* Reuse it for the new tag */
4605					sctp_set_vtag_block(twait_block->vtag_block + i, time, tag, lport, rport);
4606					set = true;
4607				}
4608			}
4609		}
4610		if (set) {
4611			/*
4612			 * We only do up to the block where we can place our
4613			 * tag for audits
4614			 */
4615			break;
4616		}
4617	}
4618	/* Need to add a new block to chain */
4619	if (!set) {
4620		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4621		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4622		if (twait_block == NULL) {
4623			return;
4624		}
4625		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4626		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4627		sctp_set_vtag_block(twait_block->vtag_block, time, tag, lport, rport);
4628	}
4629}
4630
4631void
4632sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh)
4633{
4634	struct sctp_tmit_chunk *chk, *nchk;
4635	struct sctp_queued_to_read *control, *ncontrol;
4636
4637	TAILQ_FOREACH_SAFE(control, rh, next_instrm, ncontrol) {
4638		TAILQ_REMOVE(rh, control, next_instrm);
4639		control->on_strm_q = 0;
4640		if (control->on_read_q == 0) {
4641			sctp_free_remote_addr(control->whoFrom);
4642			if (control->data) {
4643				sctp_m_freem(control->data);
4644				control->data = NULL;
4645			}
4646		}
4647		/* Reassembly free? */
4648		TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
4649			TAILQ_REMOVE(&control->reasm, chk, sctp_next);
4650			if (chk->data) {
4651				sctp_m_freem(chk->data);
4652				chk->data = NULL;
4653			}
4654			if (chk->holds_key_ref)
4655				sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4656			sctp_free_remote_addr(chk->whoTo);
4657			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4658			SCTP_DECR_CHK_COUNT();
4659			/* sa_ignore FREED_MEMORY */
4660		}
4661		/*
4662		 * We don't free the address here since all the net's were
4663		 * freed above.
4664		 */
4665		if (control->on_read_q == 0) {
4666			sctp_free_a_readq(stcb, control);
4667		}
4668	}
4669}
4670
4671/*-
4672 * Free the association after un-hashing the remote port. This
4673 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4674 * expect that the input to this function IS a locked TCB.
4675 * It will return 0, if it did NOT destroy the association (instead
4676 * it unlocks it. It will return NON-zero if it either destroyed the
4677 * association OR the association is already destroyed.
4678 */
4679int
4680sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4681{
4682	int i;
4683	struct sctp_association *asoc;
4684	struct sctp_nets *net, *nnet;
4685	struct sctp_laddr *laddr, *naddr;
4686	struct sctp_tmit_chunk *chk, *nchk;
4687	struct sctp_asconf_addr *aparam, *naparam;
4688	struct sctp_asconf_ack *aack, *naack;
4689	struct sctp_stream_reset_list *strrst, *nstrrst;
4690	struct sctp_queued_to_read *sq, *nsq;
4691	struct sctp_stream_queue_pending *sp, *nsp;
4692	sctp_sharedkey_t *shared_key, *nshared_key;
4693	struct socket *so;
4694
4695	/* first, lets purge the entry from the hash table. */
4696	SCTP_TCB_LOCK_ASSERT(stcb);
4697
4698#ifdef SCTP_LOG_CLOSING
4699	sctp_log_closing(inp, stcb, 6);
4700#endif
4701	if (stcb->asoc.state == 0) {
4702#ifdef SCTP_LOG_CLOSING
4703		sctp_log_closing(inp, NULL, 7);
4704#endif
4705		/* there is no asoc, really TSNH :-0 */
4706		return (1);
4707	}
4708	if (stcb->asoc.alternate) {
4709		sctp_free_remote_addr(stcb->asoc.alternate);
4710		stcb->asoc.alternate = NULL;
4711	}
4712	/* TEMP CODE */
4713	if (stcb->freed_from_where == 0) {
4714		/* Only record the first place free happened from */
4715		stcb->freed_from_where = from_location;
4716	}
4717	/* TEMP CODE */
4718
4719	asoc = &stcb->asoc;
4720	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4721	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4722		/* nothing around */
4723		so = NULL;
4724	else
4725		so = inp->sctp_socket;
4726
4727	/*
4728	 * We used timer based freeing if a reader or writer is in the way.
4729	 * So we first check if we are actually being called from a timer,
4730	 * if so we abort early if a reader or writer is still in the way.
4731	 */
4732	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4733	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4734		/*
4735		 * is it the timer driving us? if so are the reader/writers
4736		 * gone?
4737		 */
4738		if (stcb->asoc.refcnt) {
4739			/* nope, reader or writer in the way */
4740			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4741			/* no asoc destroyed */
4742			SCTP_TCB_UNLOCK(stcb);
4743#ifdef SCTP_LOG_CLOSING
4744			sctp_log_closing(inp, stcb, 8);
4745#endif
4746			return (0);
4747		}
4748	}
4749	/* Now clean up any other timers */
4750	sctp_stop_association_timers(stcb, false);
4751	/* Now the read queue needs to be cleaned up (only once) */
4752	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4753		SCTP_ADD_SUBSTATE(stcb, SCTP_STATE_ABOUT_TO_BE_FREED);
4754		SCTP_INP_READ_LOCK(inp);
4755		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4756			if (sq->stcb == stcb) {
4757				sq->do_not_ref_stcb = 1;
4758				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4759				/*
4760				 * If there is no end, there never will be
4761				 * now.
4762				 */
4763				if (sq->end_added == 0) {
4764					/* Held for PD-API, clear that. */
4765					sq->pdapi_aborted = 1;
4766					sq->held_length = 0;
4767					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4768						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4769						    stcb,
4770						    SCTP_PARTIAL_DELIVERY_ABORTED,
4771						    (void *)sq,
4772						    SCTP_SO_LOCKED);
4773					}
4774					/* Add an end to wake them */
4775					sq->end_added = 1;
4776				}
4777			}
4778		}
4779		SCTP_INP_READ_UNLOCK(inp);
4780		if (stcb->block_entry) {
4781			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4782			stcb->block_entry->error = ECONNRESET;
4783			stcb->block_entry = NULL;
4784		}
4785	}
4786	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4787		/*
4788		 * Someone holds a reference OR the socket is unaccepted
4789		 * yet.
4790		 */
4791		if ((stcb->asoc.refcnt) ||
4792		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4793		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4794			SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
4795			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4796		}
4797		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4798		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4799			/* nothing around */
4800			so = NULL;
4801		if (so) {
4802			/* Wake any reader/writers */
4803			sctp_sorwakeup(inp, so);
4804			sctp_sowwakeup(inp, so);
4805		}
4806		SCTP_TCB_UNLOCK(stcb);
4807
4808#ifdef SCTP_LOG_CLOSING
4809		sctp_log_closing(inp, stcb, 9);
4810#endif
4811		/* no asoc destroyed */
4812		return (0);
4813	}
4814#ifdef SCTP_LOG_CLOSING
4815	sctp_log_closing(inp, stcb, 10);
4816#endif
4817	/*
4818	 * When I reach here, no others want to kill the assoc yet.. and I
4819	 * own the lock. Now its possible an abort comes in when I do the
4820	 * lock exchange below to grab all the locks to do the final take
4821	 * out. to prevent this we increment the count, which will start a
4822	 * timer and blow out above thus assuring us that we hold exclusive
4823	 * killing of the asoc. Note that after getting back the TCB lock we
4824	 * will go ahead and increment the counter back up and stop any
4825	 * timer a passing stranger may have started :-S
4826	 */
4827	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4828		atomic_add_int(&stcb->asoc.refcnt, 1);
4829
4830		SCTP_TCB_UNLOCK(stcb);
4831		SCTP_INP_INFO_WLOCK();
4832		SCTP_INP_WLOCK(inp);
4833		SCTP_TCB_LOCK(stcb);
4834	}
4835	/* Double check the GONE flag */
4836	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4837	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4838		/* nothing around */
4839		so = NULL;
4840
4841	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4842	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4843		/*
4844		 * For TCP type we need special handling when we are
4845		 * connected. We also include the peel'ed off ones to.
4846		 */
4847		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4848			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4849			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4850			if (so) {
4851				SOCKBUF_LOCK(&so->so_rcv);
4852				so->so_state &= ~(SS_ISCONNECTING |
4853				    SS_ISDISCONNECTING |
4854				    SS_ISCONNECTED);
4855				so->so_state |= SS_ISDISCONNECTED;
4856				socantrcvmore_locked(so);
4857				socantsendmore(so);
4858				sctp_sowwakeup(inp, so);
4859				sctp_sorwakeup(inp, so);
4860				SCTP_SOWAKEUP(so);
4861			}
4862		}
4863	}
4864
4865	/*
4866	 * Make it invalid too, that way if its about to run it will abort
4867	 * and return.
4868	 */
4869	/* re-increment the lock */
4870	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4871		atomic_subtract_int(&stcb->asoc.refcnt, 1);
4872	}
4873	if (stcb->asoc.refcnt) {
4874		SCTP_CLEAR_SUBSTATE(stcb, SCTP_STATE_IN_ACCEPT_QUEUE);
4875		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4876		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4877			SCTP_INP_INFO_WUNLOCK();
4878			SCTP_INP_WUNLOCK(inp);
4879		}
4880		SCTP_TCB_UNLOCK(stcb);
4881		return (0);
4882	}
4883	asoc->state = 0;
4884	if (inp->sctp_tcbhash) {
4885		LIST_REMOVE(stcb, sctp_tcbhash);
4886	}
4887	if (stcb->asoc.in_asocid_hash) {
4888		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4889	}
4890	if (inp->sctp_socket == NULL) {
4891		stcb->sctp_socket = NULL;
4892	}
4893	/* Now lets remove it from the list of ALL associations in the EP */
4894	LIST_REMOVE(stcb, sctp_tcblist);
4895	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4896		SCTP_INP_INCR_REF(inp);
4897		SCTP_INP_WUNLOCK(inp);
4898	}
4899	/* pull from vtag hash */
4900	LIST_REMOVE(stcb, sctp_asocs);
4901	sctp_add_vtag_to_timewait(asoc->my_vtag, inp->sctp_lport, stcb->rport);
4902
4903	/*
4904	 * Now restop the timers to be sure this is paranoia at is finest!
4905	 */
4906	sctp_stop_association_timers(stcb, true);
4907
4908	/*
4909	 * The chunk lists and such SHOULD be empty but we check them just
4910	 * in case.
4911	 */
4912	/* anything on the wheel needs to be removed */
4913	for (i = 0; i < asoc->streamoutcnt; i++) {
4914		struct sctp_stream_out *outs;
4915
4916		outs = &asoc->strmout[i];
4917		/* now clean up any chunks here */
4918		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4919			atomic_subtract_int(&asoc->stream_queue_cnt, 1);
4920			TAILQ_REMOVE(&outs->outqueue, sp, next);
4921			stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp);
4922			sctp_free_spbufspace(stcb, asoc, sp);
4923			if (sp->data) {
4924				if (so) {
4925					/* Still an open socket - report */
4926					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4927					    0, (void *)sp, SCTP_SO_LOCKED);
4928				}
4929				if (sp->data) {
4930					sctp_m_freem(sp->data);
4931					sp->data = NULL;
4932					sp->tail_mbuf = NULL;
4933					sp->length = 0;
4934				}
4935			}
4936			if (sp->net) {
4937				sctp_free_remote_addr(sp->net);
4938				sp->net = NULL;
4939			}
4940			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4941		}
4942	}
4943	/* sa_ignore FREED_MEMORY */
4944	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4945		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4946		SCTP_FREE(strrst, SCTP_M_STRESET);
4947	}
4948	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4949		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4950		if (sq->data) {
4951			sctp_m_freem(sq->data);
4952			sq->data = NULL;
4953		}
4954		sctp_free_remote_addr(sq->whoFrom);
4955		sq->whoFrom = NULL;
4956		sq->stcb = NULL;
4957		/* Free the ctl entry */
4958		sctp_free_a_readq(stcb, sq);
4959		/* sa_ignore FREED_MEMORY */
4960	}
4961	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4962		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4963		if (chk->data) {
4964			sctp_m_freem(chk->data);
4965			chk->data = NULL;
4966		}
4967		if (chk->holds_key_ref)
4968			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4969		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4970		SCTP_DECR_CHK_COUNT();
4971		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4972		asoc->free_chunk_cnt--;
4973		/* sa_ignore FREED_MEMORY */
4974	}
4975	/* pending send queue SHOULD be empty */
4976	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4977		if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
4978			asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
4979#ifdef INVARIANTS
4980		} else {
4981			panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
4982#endif
4983		}
4984		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4985		if (chk->data) {
4986			if (so) {
4987				/* Still a socket? */
4988				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4989				    0, chk, SCTP_SO_LOCKED);
4990			}
4991			if (chk->data) {
4992				sctp_m_freem(chk->data);
4993				chk->data = NULL;
4994			}
4995		}
4996		if (chk->holds_key_ref)
4997			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4998		if (chk->whoTo) {
4999			sctp_free_remote_addr(chk->whoTo);
5000			chk->whoTo = NULL;
5001		}
5002		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5003		SCTP_DECR_CHK_COUNT();
5004		/* sa_ignore FREED_MEMORY */
5005	}
5006	/* sent queue SHOULD be empty */
5007	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5008		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5009			if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
5010				asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
5011#ifdef INVARIANTS
5012			} else {
5013				panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
5014#endif
5015			}
5016		}
5017		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5018		if (chk->data) {
5019			if (so) {
5020				/* Still a socket? */
5021				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5022				    0, chk, SCTP_SO_LOCKED);
5023			}
5024			if (chk->data) {
5025				sctp_m_freem(chk->data);
5026				chk->data = NULL;
5027			}
5028		}
5029		if (chk->holds_key_ref)
5030			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5031		sctp_free_remote_addr(chk->whoTo);
5032		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5033		SCTP_DECR_CHK_COUNT();
5034		/* sa_ignore FREED_MEMORY */
5035	}
5036#ifdef INVARIANTS
5037	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5038		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5039			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5040		}
5041	}
5042#endif
5043	/* control queue MAY not be empty */
5044	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5045		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5046		if (chk->data) {
5047			sctp_m_freem(chk->data);
5048			chk->data = NULL;
5049		}
5050		if (chk->holds_key_ref)
5051			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5052		sctp_free_remote_addr(chk->whoTo);
5053		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5054		SCTP_DECR_CHK_COUNT();
5055		/* sa_ignore FREED_MEMORY */
5056	}
5057	/* ASCONF queue MAY not be empty */
5058	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5059		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5060		if (chk->data) {
5061			sctp_m_freem(chk->data);
5062			chk->data = NULL;
5063		}
5064		if (chk->holds_key_ref)
5065			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5066		sctp_free_remote_addr(chk->whoTo);
5067		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5068		SCTP_DECR_CHK_COUNT();
5069		/* sa_ignore FREED_MEMORY */
5070	}
5071	if (asoc->mapping_array) {
5072		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5073		asoc->mapping_array = NULL;
5074	}
5075	if (asoc->nr_mapping_array) {
5076		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5077		asoc->nr_mapping_array = NULL;
5078	}
5079	/* the stream outs */
5080	if (asoc->strmout) {
5081		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5082		asoc->strmout = NULL;
5083	}
5084	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5085	if (asoc->strmin) {
5086		for (i = 0; i < asoc->streamincnt; i++) {
5087			sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue);
5088			sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue);
5089		}
5090		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5091		asoc->strmin = NULL;
5092	}
5093	asoc->streamincnt = 0;
5094	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5095#ifdef INVARIANTS
5096		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5097			panic("no net's left alloc'ed, or list points to itself");
5098		}
5099#endif
5100		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5101		sctp_free_remote_addr(net);
5102	}
5103	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5104		/* sa_ignore FREED_MEMORY */
5105		sctp_remove_laddr(laddr);
5106	}
5107
5108	/* pending asconf (address) parameters */
5109	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5110		/* sa_ignore FREED_MEMORY */
5111		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5112		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5113	}
5114	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5115		/* sa_ignore FREED_MEMORY */
5116		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5117		if (aack->data != NULL) {
5118			sctp_m_freem(aack->data);
5119		}
5120		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5121	}
5122	/* clean up auth stuff */
5123	if (asoc->local_hmacs)
5124		sctp_free_hmaclist(asoc->local_hmacs);
5125	if (asoc->peer_hmacs)
5126		sctp_free_hmaclist(asoc->peer_hmacs);
5127
5128	if (asoc->local_auth_chunks)
5129		sctp_free_chunklist(asoc->local_auth_chunks);
5130	if (asoc->peer_auth_chunks)
5131		sctp_free_chunklist(asoc->peer_auth_chunks);
5132
5133	sctp_free_authinfo(&asoc->authinfo);
5134
5135	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5136		LIST_REMOVE(shared_key, next);
5137		sctp_free_sharedkey(shared_key);
5138		/* sa_ignore FREED_MEMORY */
5139	}
5140
5141	/* Insert new items here :> */
5142
5143	/* Get rid of LOCK */
5144	SCTP_TCB_UNLOCK(stcb);
5145	SCTP_TCB_LOCK_DESTROY(stcb);
5146	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5147		SCTP_INP_INFO_WUNLOCK();
5148		SCTP_INP_RLOCK(inp);
5149	}
5150#ifdef SCTP_TRACK_FREED_ASOCS
5151	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5152		/* now clean up the tasoc itself */
5153		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5154		SCTP_DECR_ASOC_COUNT();
5155	} else {
5156		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5157	}
5158#else
5159	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5160	SCTP_DECR_ASOC_COUNT();
5161#endif
5162	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5163		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5164			/*
5165			 * If its NOT the inp_free calling us AND sctp_close
5166			 * as been called, we call back...
5167			 */
5168			SCTP_INP_RUNLOCK(inp);
5169			/*
5170			 * This will start the kill timer (if we are the
5171			 * last one) since we hold an increment yet. But
5172			 * this is the only safe way to do this since
5173			 * otherwise if the socket closes at the same time
5174			 * we are here we might collide in the cleanup.
5175			 */
5176			sctp_inpcb_free(inp,
5177			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5178			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5179			SCTP_INP_DECR_REF(inp);
5180		} else {
5181			/* The socket is still open. */
5182			SCTP_INP_DECR_REF(inp);
5183			SCTP_INP_RUNLOCK(inp);
5184		}
5185	}
5186	/* destroyed the asoc */
5187#ifdef SCTP_LOG_CLOSING
5188	sctp_log_closing(inp, NULL, 11);
5189#endif
5190	return (1);
5191}
5192
5193/*
5194 * determine if a destination is "reachable" based upon the addresses bound
5195 * to the current endpoint (e.g. only v4 or v6 currently bound)
5196 */
5197/*
5198 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5199 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5200 * types bound as its endpoint
5201 */
5202int
5203sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5204{
5205	struct sctp_inpcb *inp;
5206	int answer;
5207
5208	/*
5209	 * No locks here, the TCB, in all cases is already locked and an
5210	 * assoc is up. There is either a INP lock by the caller applied (in
5211	 * asconf case when deleting an address) or NOT in the HB case,
5212	 * however if HB then the INP increment is up and the INP will not
5213	 * be removed (on top of the fact that we have a TCB lock). So we
5214	 * only want to read the sctp_flags, which is either bound-all or
5215	 * not.. no protection needed since once an assoc is up you can't be
5216	 * changing your binding.
5217	 */
5218	inp = stcb->sctp_ep;
5219	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5220		/* if bound all, destination is not restricted */
5221		/*
5222		 * RRS: Question during lock work: Is this correct? If you
5223		 * are bound-all you still might need to obey the V4--V6
5224		 * flags??? IMO this bound-all stuff needs to be removed!
5225		 */
5226		return (1);
5227	}
5228	/* NOTE: all "scope" checks are done when local addresses are added */
5229	switch (destaddr->sa_family) {
5230#ifdef INET6
5231	case AF_INET6:
5232		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5233		break;
5234#endif
5235#ifdef INET
5236	case AF_INET:
5237		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5238		break;
5239#endif
5240	default:
5241		/* invalid family, so it's unreachable */
5242		answer = 0;
5243		break;
5244	}
5245	return (answer);
5246}
5247
5248/*
5249 * update the inp_vflags on an endpoint
5250 */
5251static void
5252sctp_update_ep_vflag(struct sctp_inpcb *inp)
5253{
5254	struct sctp_laddr *laddr;
5255
5256	/* first clear the flag */
5257	inp->ip_inp.inp.inp_vflag = 0;
5258	/* set the flag based on addresses on the ep list */
5259	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5260		if (laddr->ifa == NULL) {
5261			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5262			    __func__);
5263			continue;
5264		}
5265
5266		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5267			continue;
5268		}
5269		switch (laddr->ifa->address.sa.sa_family) {
5270#ifdef INET6
5271		case AF_INET6:
5272			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5273			break;
5274#endif
5275#ifdef INET
5276		case AF_INET:
5277			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5278			break;
5279#endif
5280		default:
5281			break;
5282		}
5283	}
5284}
5285
5286/*
5287 * Add the address to the endpoint local address list There is nothing to be
5288 * done if we are bound to all addresses
5289 */
5290void
5291sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5292{
5293	struct sctp_laddr *laddr;
5294	struct sctp_tcb *stcb;
5295	int fnd, error = 0;
5296
5297	fnd = 0;
5298
5299	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5300		/* You are already bound to all. You have it already */
5301		return;
5302	}
5303#ifdef INET6
5304	if (ifa->address.sa.sa_family == AF_INET6) {
5305		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5306			/* Can't bind a non-useable addr. */
5307			return;
5308		}
5309	}
5310#endif
5311	/* first, is it already present? */
5312	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5313		if (laddr->ifa == ifa) {
5314			fnd = 1;
5315			break;
5316		}
5317	}
5318
5319	if (fnd == 0) {
5320		/* Not in the ep list */
5321		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5322		if (error != 0)
5323			return;
5324		inp->laddr_count++;
5325		/* update inp_vflag flags */
5326		switch (ifa->address.sa.sa_family) {
5327#ifdef INET6
5328		case AF_INET6:
5329			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5330			break;
5331#endif
5332#ifdef INET
5333		case AF_INET:
5334			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5335			break;
5336#endif
5337		default:
5338			break;
5339		}
5340		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5341			sctp_add_local_addr_restricted(stcb, ifa);
5342		}
5343	}
5344	return;
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 * Delete the address from the endpoint local address list. There is nothing
5372 * to be done if we are bound to all addresses
5373 */
5374void
5375sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5376{
5377	struct sctp_laddr *laddr;
5378	int fnd;
5379
5380	fnd = 0;
5381	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5382		/* You are already bound to all. You have it already */
5383		return;
5384	}
5385	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5386		if (laddr->ifa == ifa) {
5387			fnd = 1;
5388			break;
5389		}
5390	}
5391	if (fnd && (inp->laddr_count < 2)) {
5392		/* can't delete unless there are at LEAST 2 addresses */
5393		return;
5394	}
5395	if (fnd) {
5396		/*
5397		 * clean up any use of this address go through our
5398		 * associations and clear any last_used_address that match
5399		 * this one for each assoc, see if a new primary_destination
5400		 * is needed
5401		 */
5402		struct sctp_tcb *stcb;
5403
5404		/* clean up "next_addr_touse" */
5405		if (inp->next_addr_touse == laddr)
5406			/* delete this address */
5407			inp->next_addr_touse = NULL;
5408
5409		/* clean up "last_used_address" */
5410		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5411			struct sctp_nets *net;
5412
5413			SCTP_TCB_LOCK(stcb);
5414			if (stcb->asoc.last_used_address == laddr)
5415				/* delete this address */
5416				stcb->asoc.last_used_address = NULL;
5417			/*
5418			 * Now spin through all the nets and purge any ref
5419			 * to laddr
5420			 */
5421			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5422				if (net->ro._s_addr == laddr->ifa) {
5423					/* Yep, purge src address selected */
5424					RO_NHFREE(&net->ro);
5425					sctp_free_ifa(net->ro._s_addr);
5426					net->ro._s_addr = NULL;
5427					net->src_addr_selected = 0;
5428				}
5429			}
5430			SCTP_TCB_UNLOCK(stcb);
5431		}		/* for each tcb */
5432		/* remove it from the ep list */
5433		sctp_remove_laddr(laddr);
5434		inp->laddr_count--;
5435		/* update inp_vflag flags */
5436		sctp_update_ep_vflag(inp);
5437	}
5438	return;
5439}
5440
5441/*
5442 * Add the address to the TCB local address restricted list.
5443 * This is a "pending" address list (eg. addresses waiting for an
5444 * ASCONF-ACK response) and cannot be used as a valid source address.
5445 */
5446void
5447sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5448{
5449	struct sctp_laddr *laddr;
5450	struct sctpladdr *list;
5451
5452	/*
5453	 * Assumes TCB is locked.. and possibly the INP. May need to
5454	 * confirm/fix that if we need it and is not the case.
5455	 */
5456	list = &stcb->asoc.sctp_restricted_addrs;
5457
5458#ifdef INET6
5459	if (ifa->address.sa.sa_family == AF_INET6) {
5460		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5461			/* Can't bind a non-existent addr. */
5462			return;
5463		}
5464	}
5465#endif
5466	/* does the address already exist? */
5467	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5468		if (laddr->ifa == ifa) {
5469			return;
5470		}
5471	}
5472
5473	/* add to the list */
5474	(void)sctp_insert_laddr(list, ifa, 0);
5475	return;
5476}
5477
5478/*
5479 * Remove a local address from the TCB local address restricted list
5480 */
5481void
5482sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5483{
5484	struct sctp_inpcb *inp;
5485	struct sctp_laddr *laddr;
5486
5487	/*
5488	 * This is called by asconf work. It is assumed that a) The TCB is
5489	 * locked and b) The INP is locked. This is true in as much as I can
5490	 * trace through the entry asconf code where I did these locks.
5491	 * Again, the ASCONF code is a bit different in that it does lock
5492	 * the INP during its work often times. This must be since we don't
5493	 * want other proc's looking up things while what they are looking
5494	 * up is changing :-D
5495	 */
5496
5497	inp = stcb->sctp_ep;
5498	/* if subset bound and don't allow ASCONF's, can't delete last */
5499	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5500	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5501		if (stcb->sctp_ep->laddr_count < 2) {
5502			/* can't delete last address */
5503			return;
5504		}
5505	}
5506	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5507		/* remove the address if it exists */
5508		if (laddr->ifa == NULL)
5509			continue;
5510		if (laddr->ifa == ifa) {
5511			sctp_remove_laddr(laddr);
5512			return;
5513		}
5514	}
5515
5516	/* address not found! */
5517	return;
5518}
5519
5520/* sysctl */
5521static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5522static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5523
5524#if defined(SCTP_MCORE_INPUT) && defined(SMP)
5525struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5526int *sctp_cpuarry = NULL;
5527
5528void
5529sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5530{
5531	/* Queue a packet to a processor for the specified core */
5532	struct sctp_mcore_queue *qent;
5533	struct sctp_mcore_ctrl *wkq;
5534	int need_wake = 0;
5535
5536	if (sctp_mcore_workers == NULL) {
5537		/* Something went way bad during setup */
5538		sctp_input_with_port(m, off, 0);
5539		return;
5540	}
5541	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5542	    (sizeof(struct sctp_mcore_queue)),
5543	    SCTP_M_MCORE);
5544	if (qent == NULL) {
5545		/* This is trouble  */
5546		sctp_input_with_port(m, off, 0);
5547		return;
5548	}
5549	qent->vn = curvnet;
5550	qent->m = m;
5551	qent->off = off;
5552	qent->v6 = 0;
5553	wkq = &sctp_mcore_workers[cpu_to_use];
5554	SCTP_MCORE_QLOCK(wkq);
5555
5556	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5557	if (wkq->running == 0) {
5558		need_wake = 1;
5559	}
5560	SCTP_MCORE_QUNLOCK(wkq);
5561	if (need_wake) {
5562		wakeup(&wkq->running);
5563	}
5564}
5565
5566static void
5567sctp_mcore_thread(void *arg)
5568{
5569
5570	struct sctp_mcore_ctrl *wkq;
5571	struct sctp_mcore_queue *qent;
5572
5573	wkq = (struct sctp_mcore_ctrl *)arg;
5574	struct mbuf *m;
5575	int off, v6;
5576
5577	/* Wait for first tickle */
5578	SCTP_MCORE_LOCK(wkq);
5579	wkq->running = 0;
5580	msleep(&wkq->running,
5581	    &wkq->core_mtx,
5582	    0, "wait for pkt", 0);
5583	SCTP_MCORE_UNLOCK(wkq);
5584
5585	/* Bind to our cpu */
5586	thread_lock(curthread);
5587	sched_bind(curthread, wkq->cpuid);
5588	thread_unlock(curthread);
5589
5590	/* Now lets start working */
5591	SCTP_MCORE_LOCK(wkq);
5592	/* Now grab lock and go */
5593	for (;;) {
5594		SCTP_MCORE_QLOCK(wkq);
5595skip_sleep:
5596		wkq->running = 1;
5597		qent = TAILQ_FIRST(&wkq->que);
5598		if (qent) {
5599			TAILQ_REMOVE(&wkq->que, qent, next);
5600			SCTP_MCORE_QUNLOCK(wkq);
5601			CURVNET_SET(qent->vn);
5602			m = qent->m;
5603			off = qent->off;
5604			v6 = qent->v6;
5605			SCTP_FREE(qent, SCTP_M_MCORE);
5606			if (v6 == 0) {
5607				sctp_input_with_port(m, off, 0);
5608			} else {
5609				SCTP_PRINTF("V6 not yet supported\n");
5610				sctp_m_freem(m);
5611			}
5612			CURVNET_RESTORE();
5613			SCTP_MCORE_QLOCK(wkq);
5614		}
5615		wkq->running = 0;
5616		if (!TAILQ_EMPTY(&wkq->que)) {
5617			goto skip_sleep;
5618		}
5619		SCTP_MCORE_QUNLOCK(wkq);
5620		msleep(&wkq->running,
5621		    &wkq->core_mtx,
5622		    0, "wait for pkt", 0);
5623	}
5624}
5625
5626static void
5627sctp_startup_mcore_threads(void)
5628{
5629	int i, cpu;
5630
5631	if (mp_ncpus == 1)
5632		return;
5633
5634	if (sctp_mcore_workers != NULL) {
5635		/*
5636		 * Already been here in some previous vnet?
5637		 */
5638		return;
5639	}
5640	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5641	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5642	    SCTP_M_MCORE);
5643	if (sctp_mcore_workers == NULL) {
5644		/* TSNH I hope */
5645		return;
5646	}
5647	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5648	    sizeof(struct sctp_mcore_ctrl)));
5649	/* Init the structures */
5650	for (i = 0; i <= mp_maxid; i++) {
5651		TAILQ_INIT(&sctp_mcore_workers[i].que);
5652		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5653		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5654		sctp_mcore_workers[i].cpuid = i;
5655	}
5656	if (sctp_cpuarry == NULL) {
5657		SCTP_MALLOC(sctp_cpuarry, int *,
5658		    (mp_ncpus * sizeof(int)),
5659		    SCTP_M_MCORE);
5660		i = 0;
5661		CPU_FOREACH(cpu) {
5662			sctp_cpuarry[i] = cpu;
5663			i++;
5664		}
5665	}
5666	/* Now start them all */
5667	CPU_FOREACH(cpu) {
5668		(void)kproc_create(sctp_mcore_thread,
5669		    (void *)&sctp_mcore_workers[cpu],
5670		    &sctp_mcore_workers[cpu].thread_proc,
5671		    0,
5672		    SCTP_KTHREAD_PAGES,
5673		    SCTP_MCORE_NAME);
5674	}
5675}
5676#endif
5677
5678#define VALIDATE_LOADER_TUNABLE(var_name, prefix)		\
5679	if (SCTP_BASE_SYSCTL(var_name) < prefix##_MIN ||	\
5680	    SCTP_BASE_SYSCTL(var_name) > prefix##_MAX)		\
5681		SCTP_BASE_SYSCTL(var_name) = prefix##_DEFAULT
5682
5683void
5684sctp_pcb_init(void)
5685{
5686	/*
5687	 * SCTP initialization for the PCB structures should be called by
5688	 * the sctp_init() function.
5689	 */
5690	int i;
5691	struct timeval tv;
5692
5693	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5694		/* error I was called twice */
5695		return;
5696	}
5697	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5698
5699#if defined(SCTP_LOCAL_TRACE_BUF)
5700	memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log));
5701#endif
5702#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5703	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5704	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5705	    SCTP_M_MCORE);
5706#endif
5707	(void)SCTP_GETTIME_TIMEVAL(&tv);
5708#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5709	memset(SCTP_BASE_STATS, 0, sizeof(struct sctpstat) * (mp_maxid + 1));
5710	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t)tv.tv_sec;
5711	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t)tv.tv_usec;
5712#else
5713	memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat));
5714	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec;
5715	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec;
5716#endif
5717	/* init the empty list of (All) Endpoints */
5718	LIST_INIT(&SCTP_BASE_INFO(listhead));
5719
5720	/* init the hash table of endpoints */
5721	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5722	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5723	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5724	VALIDATE_LOADER_TUNABLE(sctp_hashtblsize, SCTPCTL_TCBHASHSIZE);
5725	VALIDATE_LOADER_TUNABLE(sctp_pcbtblsize, SCTPCTL_PCBHASHSIZE);
5726	VALIDATE_LOADER_TUNABLE(sctp_chunkscale, SCTPCTL_CHUNKSCALE);
5727	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5728	    &SCTP_BASE_INFO(hashasocmark));
5729	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5730	    &SCTP_BASE_INFO(hashmark));
5731	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5732	    &SCTP_BASE_INFO(hashtcpmark));
5733	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5734	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5735	    &SCTP_BASE_INFO(hashvrfmark));
5736
5737	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5738	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5739	/* init the zones */
5740	/*
5741	 * FIX ME: Should check for NULL returns, but if it does fail we are
5742	 * doomed to panic anyways... add later maybe.
5743	 */
5744	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5745	    sizeof(struct sctp_inpcb), maxsockets);
5746
5747	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5748	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5749
5750	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5751	    sizeof(struct sctp_laddr),
5752	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5753
5754	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5755	    sizeof(struct sctp_nets),
5756	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5757
5758	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5759	    sizeof(struct sctp_tmit_chunk),
5760	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5761
5762	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5763	    sizeof(struct sctp_queued_to_read),
5764	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5765
5766	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5767	    sizeof(struct sctp_stream_queue_pending),
5768	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5769
5770	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5771	    sizeof(struct sctp_asconf),
5772	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5773
5774	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5775	    sizeof(struct sctp_asconf_ack),
5776	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5777
5778	/* Master Lock INIT for info structure */
5779	SCTP_INP_INFO_LOCK_INIT();
5780	SCTP_STATLOG_INIT_LOCK();
5781
5782	SCTP_IPI_COUNT_INIT();
5783	SCTP_IPI_ADDR_INIT();
5784#ifdef SCTP_PACKET_LOGGING
5785	SCTP_IP_PKTLOG_INIT();
5786#endif
5787	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5788
5789	SCTP_WQ_ADDR_INIT();
5790	/* not sure if we need all the counts */
5791	SCTP_BASE_INFO(ipi_count_ep) = 0;
5792	/* assoc/tcb zone info */
5793	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5794	/* local addrlist zone info */
5795	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5796	/* remote addrlist zone info */
5797	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5798	/* chunk info */
5799	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5800
5801	/* socket queue zone info */
5802	SCTP_BASE_INFO(ipi_count_readq) = 0;
5803
5804	/* stream out queue cont */
5805	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5806
5807	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5808	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5809
5810	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5811
5812	/* Init the TIMEWAIT list */
5813	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5814		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5815	}
5816	sctp_startup_iterator();
5817
5818#if defined(SCTP_MCORE_INPUT) && defined(SMP)
5819	sctp_startup_mcore_threads();
5820#endif
5821
5822	/*
5823	 * INIT the default VRF which for BSD is the only one, other O/S's
5824	 * may have more. But initially they must start with one and then
5825	 * add the VRF's as addresses are added.
5826	 */
5827	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5828}
5829
5830/*
5831 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5832 */
5833void
5834sctp_pcb_finish(void)
5835{
5836	struct sctp_vrflist *vrf_bucket;
5837	struct sctp_vrf *vrf, *nvrf;
5838	struct sctp_ifn *ifn, *nifn;
5839	struct sctp_ifa *ifa, *nifa;
5840	struct sctpvtaghead *chain;
5841	struct sctp_tagblock *twait_block, *prev_twait_block;
5842	struct sctp_laddr *wi, *nwi;
5843	int i;
5844	struct sctp_iterator *it, *nit;
5845
5846	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
5847		SCTP_PRINTF("%s: race condition on teardown.\n", __func__);
5848		return;
5849	}
5850	SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
5851	/*
5852	 * In FreeBSD the iterator thread never exits but we do clean up.
5853	 * The only way FreeBSD reaches here is if we have VRF's but we
5854	 * still add the ifdef to make it compile on old versions.
5855	 */
5856retry:
5857	SCTP_IPI_ITERATOR_WQ_LOCK();
5858	/*
5859	 * sctp_iterator_worker() might be working on an it entry without
5860	 * holding the lock.  We won't find it on the list either and
5861	 * continue and free/destroy it.  While holding the lock, spin, to
5862	 * avoid the race condition as sctp_iterator_worker() will have to
5863	 * wait to re-acquire the lock.
5864	 */
5865	if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) {
5866		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5867		SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. "
5868		    "cur_it=%p\n", __func__, sctp_it_ctl.cur_it);
5869		DELAY(10);
5870		goto retry;
5871	}
5872	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5873		if (it->vn != curvnet) {
5874			continue;
5875		}
5876		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5877		if (it->function_atend != NULL) {
5878			(*it->function_atend) (it->pointer, it->val);
5879		}
5880		SCTP_FREE(it, SCTP_M_ITER);
5881	}
5882	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5883	SCTP_ITERATOR_LOCK();
5884	if ((sctp_it_ctl.cur_it) &&
5885	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5886		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5887	}
5888	SCTP_ITERATOR_UNLOCK();
5889	SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer));
5890	SCTP_WQ_ADDR_LOCK();
5891	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5892		LIST_REMOVE(wi, sctp_nxt_addr);
5893		SCTP_DECR_LADDR_COUNT();
5894		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5895			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5896		}
5897		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5898	}
5899	SCTP_WQ_ADDR_UNLOCK();
5900
5901	/*
5902	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5903	 * destroyed first).
5904	 */
5905	SCTP_IPI_ADDR_WLOCK();
5906	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5907	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5908		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5909			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5910				/* free the ifa */
5911				LIST_REMOVE(ifa, next_bucket);
5912				LIST_REMOVE(ifa, next_ifa);
5913				SCTP_FREE(ifa, SCTP_M_IFA);
5914			}
5915			/* free the ifn */
5916			LIST_REMOVE(ifn, next_bucket);
5917			LIST_REMOVE(ifn, next_ifn);
5918			SCTP_FREE(ifn, SCTP_M_IFN);
5919		}
5920		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5921		/* free the vrf */
5922		LIST_REMOVE(vrf, next_vrf);
5923		SCTP_FREE(vrf, SCTP_M_VRF);
5924	}
5925	SCTP_IPI_ADDR_WUNLOCK();
5926	/* free the vrf hashes */
5927	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5928	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5929
5930	/*
5931	 * free the TIMEWAIT list elements malloc'd in the function
5932	 * sctp_add_vtag_to_timewait()...
5933	 */
5934	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5935		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5936		if (!LIST_EMPTY(chain)) {
5937			prev_twait_block = NULL;
5938			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5939				if (prev_twait_block) {
5940					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5941				}
5942				prev_twait_block = twait_block;
5943			}
5944			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5945		}
5946	}
5947
5948	/* free the locks and mutexes */
5949#ifdef SCTP_PACKET_LOGGING
5950	SCTP_IP_PKTLOG_DESTROY();
5951#endif
5952	SCTP_IPI_ADDR_DESTROY();
5953	SCTP_STATLOG_DESTROY();
5954	SCTP_INP_INFO_LOCK_DESTROY();
5955
5956	SCTP_WQ_ADDR_DESTROY();
5957
5958	/* Get rid of other stuff too. */
5959	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5960		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5961	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5962		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5963	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5964		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5965
5966	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5967	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5968	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5969	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5970	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5971	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5972	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5973	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5974	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5975#if defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5976	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5977#endif
5978}
5979
5980int
5981sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5982    int offset, int limit,
5983    struct sockaddr *src, struct sockaddr *dst,
5984    struct sockaddr *altsa, uint16_t port)
5985{
5986	/*
5987	 * grub through the INIT pulling addresses and loading them to the
5988	 * nets structure in the asoc. The from address in the mbuf should
5989	 * also be loaded (if it is not already). This routine can be called
5990	 * with either INIT or INIT-ACK's as long as the m points to the IP
5991	 * packet and the offset points to the beginning of the parameters.
5992	 */
5993	struct sctp_inpcb *inp;
5994	struct sctp_nets *net, *nnet, *net_tmp;
5995	struct sctp_paramhdr *phdr, param_buf;
5996	struct sctp_tcb *stcb_tmp;
5997	uint16_t ptype, plen;
5998	struct sockaddr *sa;
5999	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6000	struct sctp_auth_random *p_random = NULL;
6001	uint16_t random_len = 0;
6002	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6003	struct sctp_auth_hmac_algo *hmacs = NULL;
6004	uint16_t hmacs_len = 0;
6005	uint8_t saw_asconf = 0;
6006	uint8_t saw_asconf_ack = 0;
6007	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6008	struct sctp_auth_chunk_list *chunks = NULL;
6009	uint16_t num_chunks = 0;
6010	sctp_key_t *new_key;
6011	uint32_t keylen;
6012	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6013	uint8_t peer_supports_ecn;
6014	uint8_t peer_supports_prsctp;
6015	uint8_t peer_supports_auth;
6016	uint8_t peer_supports_asconf;
6017	uint8_t peer_supports_asconf_ack;
6018	uint8_t peer_supports_reconfig;
6019	uint8_t peer_supports_nrsack;
6020	uint8_t peer_supports_pktdrop;
6021	uint8_t peer_supports_idata;
6022#ifdef INET
6023	struct sockaddr_in sin;
6024#endif
6025#ifdef INET6
6026	struct sockaddr_in6 sin6;
6027#endif
6028
6029	/* First get the destination address setup too. */
6030#ifdef INET
6031	memset(&sin, 0, sizeof(sin));
6032	sin.sin_family = AF_INET;
6033	sin.sin_len = sizeof(sin);
6034	sin.sin_port = stcb->rport;
6035#endif
6036#ifdef INET6
6037	memset(&sin6, 0, sizeof(sin6));
6038	sin6.sin6_family = AF_INET6;
6039	sin6.sin6_len = sizeof(struct sockaddr_in6);
6040	sin6.sin6_port = stcb->rport;
6041#endif
6042	if (altsa) {
6043		sa = altsa;
6044	} else {
6045		sa = src;
6046	}
6047	peer_supports_idata = 0;
6048	peer_supports_ecn = 0;
6049	peer_supports_prsctp = 0;
6050	peer_supports_auth = 0;
6051	peer_supports_asconf = 0;
6052	peer_supports_asconf_ack = 0;
6053	peer_supports_reconfig = 0;
6054	peer_supports_nrsack = 0;
6055	peer_supports_pktdrop = 0;
6056	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6057		/* mark all addresses that we have currently on the list */
6058		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6059	}
6060	/* does the source address already exist? if so skip it */
6061	inp = stcb->sctp_ep;
6062	atomic_add_int(&stcb->asoc.refcnt, 1);
6063	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6064	atomic_subtract_int(&stcb->asoc.refcnt, 1);
6065
6066	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6067		/* we must add the source address */
6068		/* no scope set here since we have a tcb already. */
6069		switch (sa->sa_family) {
6070#ifdef INET
6071		case AF_INET:
6072			if (stcb->asoc.scope.ipv4_addr_legal) {
6073				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6074					return (-1);
6075				}
6076			}
6077			break;
6078#endif
6079#ifdef INET6
6080		case AF_INET6:
6081			if (stcb->asoc.scope.ipv6_addr_legal) {
6082				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6083					return (-2);
6084				}
6085			}
6086			break;
6087#endif
6088		default:
6089			break;
6090		}
6091	} else {
6092		if (net_tmp != NULL && stcb_tmp == stcb) {
6093			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6094		} else if (stcb_tmp != stcb) {
6095			/* It belongs to another association? */
6096			if (stcb_tmp)
6097				SCTP_TCB_UNLOCK(stcb_tmp);
6098			return (-3);
6099		}
6100	}
6101	if (stcb->asoc.state == 0) {
6102		/* the assoc was freed? */
6103		return (-4);
6104	}
6105	/* now we must go through each of the params. */
6106	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
6107	while (phdr) {
6108		ptype = ntohs(phdr->param_type);
6109		plen = ntohs(phdr->param_length);
6110		/*
6111		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6112		 * (uint32_t)ptype, (int)plen);
6113		 */
6114		if (offset + plen > limit) {
6115			break;
6116		}
6117		if (plen < sizeof(struct sctp_paramhdr)) {
6118			break;
6119		}
6120#ifdef INET
6121		if (ptype == SCTP_IPV4_ADDRESS) {
6122			if (stcb->asoc.scope.ipv4_addr_legal) {
6123				struct sctp_ipv4addr_param *p4, p4_buf;
6124
6125				/* ok get the v4 address and check/add */
6126				phdr = sctp_get_next_param(m, offset,
6127				    (struct sctp_paramhdr *)&p4_buf,
6128				    sizeof(p4_buf));
6129				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6130				    phdr == NULL) {
6131					return (-5);
6132				}
6133				p4 = (struct sctp_ipv4addr_param *)phdr;
6134				sin.sin_addr.s_addr = p4->addr;
6135				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6136					/* Skip multi-cast addresses */
6137					goto next_param;
6138				}
6139				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6140				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6141					goto next_param;
6142				}
6143				sa = (struct sockaddr *)&sin;
6144				inp = stcb->sctp_ep;
6145				atomic_add_int(&stcb->asoc.refcnt, 1);
6146				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6147				    dst, stcb);
6148				atomic_subtract_int(&stcb->asoc.refcnt, 1);
6149
6150				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6151				    inp == NULL) {
6152					/* we must add the source address */
6153					/*
6154					 * no scope set since we have a tcb
6155					 * already
6156					 */
6157
6158					/*
6159					 * we must validate the state again
6160					 * here
6161					 */
6162			add_it_now:
6163					if (stcb->asoc.state == 0) {
6164						/* the assoc was freed? */
6165						return (-7);
6166					}
6167					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6168						return (-8);
6169					}
6170				} else if (stcb_tmp == stcb) {
6171					if (stcb->asoc.state == 0) {
6172						/* the assoc was freed? */
6173						return (-10);
6174					}
6175					if (net != NULL) {
6176						/* clear flag */
6177						net->dest_state &=
6178						    ~SCTP_ADDR_NOT_IN_ASSOC;
6179					}
6180				} else {
6181					/*
6182					 * strange, address is in another
6183					 * assoc? straighten out locks.
6184					 */
6185					if (stcb_tmp) {
6186						if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) {
6187							struct mbuf *op_err;
6188							char msg[SCTP_DIAG_INFO_LEN];
6189
6190							/*
6191							 * in setup state we
6192							 * abort this guy
6193							 */
6194							SCTP_SNPRINTF(msg, sizeof(msg),
6195							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6196							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6197							    msg);
6198							sctp_abort_an_association(stcb_tmp->sctp_ep,
6199							    stcb_tmp, op_err, false,
6200							    SCTP_SO_NOT_LOCKED);
6201							goto add_it_now;
6202						}
6203						SCTP_TCB_UNLOCK(stcb_tmp);
6204					}
6205
6206					if (stcb->asoc.state == 0) {
6207						/* the assoc was freed? */
6208						return (-12);
6209					}
6210					return (-13);
6211				}
6212			}
6213		} else
6214#endif
6215#ifdef INET6
6216		if (ptype == SCTP_IPV6_ADDRESS) {
6217			if (stcb->asoc.scope.ipv6_addr_legal) {
6218				/* ok get the v6 address and check/add */
6219				struct sctp_ipv6addr_param *p6, p6_buf;
6220
6221				phdr = sctp_get_next_param(m, offset,
6222				    (struct sctp_paramhdr *)&p6_buf,
6223				    sizeof(p6_buf));
6224				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6225				    phdr == NULL) {
6226					return (-14);
6227				}
6228				p6 = (struct sctp_ipv6addr_param *)phdr;
6229				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6230				    sizeof(p6->addr));
6231				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6232					/* Skip multi-cast addresses */
6233					goto next_param;
6234				}
6235				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6236					/*
6237					 * Link local make no sense without
6238					 * scope
6239					 */
6240					goto next_param;
6241				}
6242				sa = (struct sockaddr *)&sin6;
6243				inp = stcb->sctp_ep;
6244				atomic_add_int(&stcb->asoc.refcnt, 1);
6245				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6246				    dst, stcb);
6247				atomic_subtract_int(&stcb->asoc.refcnt, 1);
6248				if (stcb_tmp == NULL &&
6249				    (inp == stcb->sctp_ep || inp == NULL)) {
6250					/*
6251					 * we must validate the state again
6252					 * here
6253					 */
6254			add_it_now6:
6255					if (stcb->asoc.state == 0) {
6256						/* the assoc was freed? */
6257						return (-16);
6258					}
6259					/*
6260					 * we must add the address, no scope
6261					 * set
6262					 */
6263					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6264						return (-17);
6265					}
6266				} else if (stcb_tmp == stcb) {
6267					/*
6268					 * we must validate the state again
6269					 * here
6270					 */
6271					if (stcb->asoc.state == 0) {
6272						/* the assoc was freed? */
6273						return (-19);
6274					}
6275					if (net != NULL) {
6276						/* clear flag */
6277						net->dest_state &=
6278						    ~SCTP_ADDR_NOT_IN_ASSOC;
6279					}
6280				} else {
6281					/*
6282					 * strange, address is in another
6283					 * assoc? straighten out locks.
6284					 */
6285					if (stcb_tmp) {
6286						if (SCTP_GET_STATE(stcb_tmp) == SCTP_STATE_COOKIE_WAIT) {
6287							struct mbuf *op_err;
6288							char msg[SCTP_DIAG_INFO_LEN];
6289
6290							/*
6291							 * in setup state we
6292							 * abort this guy
6293							 */
6294							SCTP_SNPRINTF(msg, sizeof(msg),
6295							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6296							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6297							    msg);
6298							sctp_abort_an_association(stcb_tmp->sctp_ep,
6299							    stcb_tmp, op_err, false,
6300							    SCTP_SO_NOT_LOCKED);
6301							goto add_it_now6;
6302						}
6303						SCTP_TCB_UNLOCK(stcb_tmp);
6304					}
6305					if (stcb->asoc.state == 0) {
6306						/* the assoc was freed? */
6307						return (-21);
6308					}
6309					return (-22);
6310				}
6311			}
6312		} else
6313#endif
6314		if (ptype == SCTP_ECN_CAPABLE) {
6315			peer_supports_ecn = 1;
6316		} else if (ptype == SCTP_ULP_ADAPTATION) {
6317			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6318				struct sctp_adaptation_layer_indication ai,
6319				                                *aip;
6320
6321				phdr = sctp_get_next_param(m, offset,
6322				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6323				aip = (struct sctp_adaptation_layer_indication *)phdr;
6324				if (aip) {
6325					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6326					stcb->asoc.adaptation_needed = 1;
6327				}
6328			}
6329		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6330			struct sctp_asconf_addr_param lstore, *fee;
6331			int lptype;
6332			struct sockaddr *lsa = NULL;
6333#ifdef INET
6334			struct sctp_asconf_addrv4_param *fii;
6335#endif
6336
6337			if (stcb->asoc.asconf_supported == 0) {
6338				return (-100);
6339			}
6340			if (plen > sizeof(lstore)) {
6341				return (-23);
6342			}
6343			if (plen < sizeof(struct sctp_asconf_addrv4_param)) {
6344				return (-101);
6345			}
6346			phdr = sctp_get_next_param(m, offset,
6347			    (struct sctp_paramhdr *)&lstore,
6348			    plen);
6349			if (phdr == NULL) {
6350				return (-24);
6351			}
6352			fee = (struct sctp_asconf_addr_param *)phdr;
6353			lptype = ntohs(fee->addrp.ph.param_type);
6354			switch (lptype) {
6355#ifdef INET
6356			case SCTP_IPV4_ADDRESS:
6357				if (plen !=
6358				    sizeof(struct sctp_asconf_addrv4_param)) {
6359					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6360					    (int)sizeof(struct sctp_asconf_addrv4_param),
6361					    plen);
6362				} else {
6363					fii = (struct sctp_asconf_addrv4_param *)fee;
6364					sin.sin_addr.s_addr = fii->addrp.addr;
6365					lsa = (struct sockaddr *)&sin;
6366				}
6367				break;
6368#endif
6369#ifdef INET6
6370			case SCTP_IPV6_ADDRESS:
6371				if (plen !=
6372				    sizeof(struct sctp_asconf_addr_param)) {
6373					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6374					    (int)sizeof(struct sctp_asconf_addr_param),
6375					    plen);
6376				} else {
6377					memcpy(sin6.sin6_addr.s6_addr,
6378					    fee->addrp.addr,
6379					    sizeof(fee->addrp.addr));
6380					lsa = (struct sockaddr *)&sin6;
6381				}
6382				break;
6383#endif
6384			default:
6385				break;
6386			}
6387			if (lsa) {
6388				(void)sctp_set_primary_addr(stcb, sa, NULL);
6389			}
6390		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6391			stcb->asoc.peer_supports_nat = 1;
6392		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6393			/* Peer supports pr-sctp */
6394			peer_supports_prsctp = 1;
6395		} else if (ptype == SCTP_ZERO_CHECKSUM_ACCEPTABLE) {
6396			struct sctp_zero_checksum_acceptable zero_chksum,
6397			                             *zero_chksum_p;
6398
6399			phdr = sctp_get_next_param(m, offset,
6400			    (struct sctp_paramhdr *)&zero_chksum,
6401			    sizeof(struct sctp_zero_checksum_acceptable));
6402			if (phdr != NULL) {
6403				/*
6404				 * Only send zero checksums if the upper
6405				 * layer has enabled the support for the
6406				 * same method as allowed by the peer.
6407				 */
6408				zero_chksum_p = (struct sctp_zero_checksum_acceptable *)phdr;
6409				if ((ntohl(zero_chksum_p->edmid) != SCTP_EDMID_NONE) &&
6410				    (ntohl(zero_chksum_p->edmid) == stcb->asoc.rcv_edmid)) {
6411					stcb->asoc.snd_edmid = stcb->asoc.rcv_edmid;
6412				}
6413			}
6414		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6415			/* A supported extension chunk */
6416			struct sctp_supported_chunk_types_param *pr_supported;
6417			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6418			int num_ent, i;
6419
6420			if (plen > sizeof(local_store)) {
6421				return (-35);
6422			}
6423			phdr = sctp_get_next_param(m, offset,
6424			    (struct sctp_paramhdr *)&local_store, plen);
6425			if (phdr == NULL) {
6426				return (-25);
6427			}
6428			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6429			num_ent = plen - sizeof(struct sctp_paramhdr);
6430			for (i = 0; i < num_ent; i++) {
6431				switch (pr_supported->chunk_types[i]) {
6432				case SCTP_ASCONF:
6433					peer_supports_asconf = 1;
6434					break;
6435				case SCTP_ASCONF_ACK:
6436					peer_supports_asconf_ack = 1;
6437					break;
6438				case SCTP_FORWARD_CUM_TSN:
6439					peer_supports_prsctp = 1;
6440					break;
6441				case SCTP_PACKET_DROPPED:
6442					peer_supports_pktdrop = 1;
6443					break;
6444				case SCTP_NR_SELECTIVE_ACK:
6445					peer_supports_nrsack = 1;
6446					break;
6447				case SCTP_STREAM_RESET:
6448					peer_supports_reconfig = 1;
6449					break;
6450				case SCTP_AUTHENTICATION:
6451					peer_supports_auth = 1;
6452					break;
6453				case SCTP_IDATA:
6454					peer_supports_idata = 1;
6455					break;
6456				default:
6457					/* one I have not learned yet */
6458					break;
6459				}
6460			}
6461		} else if (ptype == SCTP_RANDOM) {
6462			if (plen > sizeof(random_store))
6463				break;
6464			if (got_random) {
6465				/* already processed a RANDOM */
6466				goto next_param;
6467			}
6468			phdr = sctp_get_next_param(m, offset,
6469			    (struct sctp_paramhdr *)random_store,
6470			    plen);
6471			if (phdr == NULL)
6472				return (-26);
6473			p_random = (struct sctp_auth_random *)phdr;
6474			random_len = plen - sizeof(*p_random);
6475			/* enforce the random length */
6476			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6477				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6478				return (-27);
6479			}
6480			got_random = 1;
6481		} else if (ptype == SCTP_HMAC_LIST) {
6482			uint16_t num_hmacs;
6483			uint16_t i;
6484
6485			if (plen > sizeof(hmacs_store))
6486				break;
6487			if (got_hmacs) {
6488				/* already processed a HMAC list */
6489				goto next_param;
6490			}
6491			phdr = sctp_get_next_param(m, offset,
6492			    (struct sctp_paramhdr *)hmacs_store,
6493			    plen);
6494			if (phdr == NULL)
6495				return (-28);
6496			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6497			hmacs_len = plen - sizeof(*hmacs);
6498			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6499			/* validate the hmac list */
6500			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6501				return (-29);
6502			}
6503			if (stcb->asoc.peer_hmacs != NULL)
6504				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6505			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6506			if (stcb->asoc.peer_hmacs != NULL) {
6507				for (i = 0; i < num_hmacs; i++) {
6508					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6509					    ntohs(hmacs->hmac_ids[i]));
6510				}
6511			}
6512			got_hmacs = 1;
6513		} else if (ptype == SCTP_CHUNK_LIST) {
6514			int i;
6515
6516			if (plen > sizeof(chunks_store))
6517				break;
6518			if (got_chklist) {
6519				/* already processed a Chunks list */
6520				goto next_param;
6521			}
6522			phdr = sctp_get_next_param(m, offset,
6523			    (struct sctp_paramhdr *)chunks_store,
6524			    plen);
6525			if (phdr == NULL)
6526				return (-30);
6527			chunks = (struct sctp_auth_chunk_list *)phdr;
6528			num_chunks = plen - sizeof(*chunks);
6529			if (stcb->asoc.peer_auth_chunks != NULL)
6530				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6531			else
6532				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6533			for (i = 0; i < num_chunks; i++) {
6534				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6535				    stcb->asoc.peer_auth_chunks);
6536				/* record asconf/asconf-ack if listed */
6537				if (chunks->chunk_types[i] == SCTP_ASCONF)
6538					saw_asconf = 1;
6539				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6540					saw_asconf_ack = 1;
6541			}
6542			got_chklist = 1;
6543		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6544			    (ptype == SCTP_STATE_COOKIE) ||
6545			    (ptype == SCTP_UNRECOG_PARAM) ||
6546			    (ptype == SCTP_COOKIE_PRESERVE) ||
6547			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6548			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6549			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6550			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6551		    (ptype == SCTP_SUCCESS_REPORT)) {
6552			/* don't care */
6553		} else {
6554			if ((ptype & 0x8000) == 0x0000) {
6555				/*
6556				 * must stop processing the rest of the
6557				 * param's. Any report bits were handled
6558				 * with the call to
6559				 * sctp_arethere_unrecognized_parameters()
6560				 * when the INIT or INIT-ACK was first seen.
6561				 */
6562				break;
6563			}
6564		}
6565
6566next_param:
6567		offset += SCTP_SIZE32(plen);
6568		if (offset >= limit) {
6569			break;
6570		}
6571		phdr = sctp_get_next_param(m, offset, &param_buf,
6572		    sizeof(param_buf));
6573	}
6574	/* Now check to see if we need to purge any addresses */
6575	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6576		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6577		    SCTP_ADDR_NOT_IN_ASSOC) {
6578			/* This address has been removed from the asoc */
6579			/* remove and free it */
6580			stcb->asoc.numnets--;
6581			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6582			if (net == stcb->asoc.alternate) {
6583				sctp_free_remote_addr(stcb->asoc.alternate);
6584				stcb->asoc.alternate = NULL;
6585			}
6586			if (net == stcb->asoc.primary_destination) {
6587				stcb->asoc.primary_destination = NULL;
6588				sctp_select_primary_destination(stcb);
6589			}
6590			sctp_free_remote_addr(net);
6591		}
6592	}
6593	if ((stcb->asoc.ecn_supported == 1) &&
6594	    (peer_supports_ecn == 0)) {
6595		stcb->asoc.ecn_supported = 0;
6596	}
6597	if ((stcb->asoc.prsctp_supported == 1) &&
6598	    (peer_supports_prsctp == 0)) {
6599		stcb->asoc.prsctp_supported = 0;
6600	}
6601	if ((stcb->asoc.auth_supported == 1) &&
6602	    ((peer_supports_auth == 0) ||
6603	    (got_random == 0) || (got_hmacs == 0))) {
6604		stcb->asoc.auth_supported = 0;
6605	}
6606	if ((stcb->asoc.asconf_supported == 1) &&
6607	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6608	    (stcb->asoc.auth_supported == 0) ||
6609	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6610		stcb->asoc.asconf_supported = 0;
6611	}
6612	if ((stcb->asoc.reconfig_supported == 1) &&
6613	    (peer_supports_reconfig == 0)) {
6614		stcb->asoc.reconfig_supported = 0;
6615	}
6616	if ((stcb->asoc.idata_supported == 1) &&
6617	    (peer_supports_idata == 0)) {
6618		stcb->asoc.idata_supported = 0;
6619	}
6620	if ((stcb->asoc.nrsack_supported == 1) &&
6621	    (peer_supports_nrsack == 0)) {
6622		stcb->asoc.nrsack_supported = 0;
6623	}
6624	if ((stcb->asoc.pktdrop_supported == 1) &&
6625	    (peer_supports_pktdrop == 0)) {
6626		stcb->asoc.pktdrop_supported = 0;
6627	}
6628	/* validate authentication required parameters */
6629	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6630		/* peer does not support auth but sent a chunks list? */
6631		return (-31);
6632	}
6633	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6634		/* peer supports asconf but not auth? */
6635		return (-32);
6636	} else if ((peer_supports_asconf == 1) &&
6637		    (peer_supports_auth == 1) &&
6638	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6639		return (-33);
6640	}
6641	/* concatenate the full random key */
6642	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6643	if (chunks != NULL) {
6644		keylen += sizeof(*chunks) + num_chunks;
6645	}
6646	new_key = sctp_alloc_key(keylen);
6647	if (new_key != NULL) {
6648		/* copy in the RANDOM */
6649		if (p_random != NULL) {
6650			keylen = sizeof(*p_random) + random_len;
6651			memcpy(new_key->key, p_random, keylen);
6652		} else {
6653			keylen = 0;
6654		}
6655		/* append in the AUTH chunks */
6656		if (chunks != NULL) {
6657			memcpy(new_key->key + keylen, chunks,
6658			    sizeof(*chunks) + num_chunks);
6659			keylen += sizeof(*chunks) + num_chunks;
6660		}
6661		/* append in the HMACs */
6662		if (hmacs != NULL) {
6663			memcpy(new_key->key + keylen, hmacs,
6664			    sizeof(*hmacs) + hmacs_len);
6665		}
6666	} else {
6667		/* failed to get memory for the key */
6668		return (-34);
6669	}
6670	if (stcb->asoc.authinfo.peer_random != NULL)
6671		sctp_free_key(stcb->asoc.authinfo.peer_random);
6672	stcb->asoc.authinfo.peer_random = new_key;
6673	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6674	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6675
6676	return (0);
6677}
6678
6679int
6680sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6681    struct sctp_nets *net)
6682{
6683	/* make sure the requested primary address exists in the assoc */
6684	if (net == NULL && sa)
6685		net = sctp_findnet(stcb, sa);
6686
6687	if (net == NULL) {
6688		/* didn't find the requested primary address! */
6689		return (-1);
6690	} else {
6691		/* set the primary address */
6692		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6693			/* Must be confirmed, so queue to set */
6694			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6695			return (0);
6696		}
6697		stcb->asoc.primary_destination = net;
6698		if (((net->dest_state & SCTP_ADDR_PF) == 0) &&
6699		    (stcb->asoc.alternate != NULL)) {
6700			sctp_free_remote_addr(stcb->asoc.alternate);
6701			stcb->asoc.alternate = NULL;
6702		}
6703		net = TAILQ_FIRST(&stcb->asoc.nets);
6704		if (net != stcb->asoc.primary_destination) {
6705			/*
6706			 * first one on the list is NOT the primary
6707			 * sctp_cmpaddr() is much more efficient if the
6708			 * primary is the first on the list, make it so.
6709			 */
6710			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6711			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6712		}
6713		return (0);
6714	}
6715}
6716
6717bool
6718sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6719{
6720	struct sctpasochead *head;
6721	struct sctp_tcb *stcb;
6722
6723	SCTP_INP_INFO_LOCK_ASSERT();
6724
6725	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))];
6726	LIST_FOREACH(stcb, head, sctp_asocs) {
6727		/*
6728		 * We choose not to lock anything here. TCB's can't be
6729		 * removed since we have the read lock, so they can't be
6730		 * freed on us, same thing for the INP. I may be wrong with
6731		 * this assumption, but we will go with it for now :-)
6732		 */
6733		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6734			continue;
6735		}
6736		if (stcb->asoc.my_vtag == tag) {
6737			/* candidate */
6738			if (stcb->rport != rport) {
6739				continue;
6740			}
6741			if (stcb->sctp_ep->sctp_lport != lport) {
6742				continue;
6743			}
6744			/* The tag is currently used, so don't use it. */
6745			return (false);
6746		}
6747	}
6748	return (!sctp_is_in_timewait(tag, lport, rport, now->tv_sec));
6749}
6750
6751static void
6752sctp_drain_mbufs(struct sctp_tcb *stcb)
6753{
6754	/*
6755	 * We must hunt this association for MBUF's past the cumack (i.e.
6756	 * out of order data that we can renege on).
6757	 */
6758	struct sctp_association *asoc;
6759	struct sctp_tmit_chunk *chk, *nchk;
6760	uint32_t cumulative_tsn_p1;
6761	struct sctp_queued_to_read *control, *ncontrol;
6762	int cnt, strmat;
6763	uint32_t gap, i;
6764	int fnd = 0;
6765
6766	/* We look for anything larger than the cum-ack + 1 */
6767
6768	asoc = &stcb->asoc;
6769	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6770		/* none we can reneg on. */
6771		return;
6772	}
6773	SCTP_STAT_INCR(sctps_protocol_drains_done);
6774	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6775	cnt = 0;
6776	/* Ok that was fun, now we will drain all the inbound streams? */
6777	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6778		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].inqueue, next_instrm, ncontrol) {
6779#ifdef INVARIANTS
6780			if (control->on_strm_q != SCTP_ON_ORDERED) {
6781				panic("Huh control: %p on_q: %d -- not ordered?",
6782				    control, control->on_strm_q);
6783			}
6784#endif
6785			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6786				/* Yep it is above cum-ack */
6787				cnt++;
6788				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6789				KASSERT(control->length > 0, ("control has zero length"));
6790				if (asoc->size_on_all_streams >= control->length) {
6791					asoc->size_on_all_streams -= control->length;
6792				} else {
6793#ifdef INVARIANTS
6794					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6795#else
6796					asoc->size_on_all_streams = 0;
6797#endif
6798				}
6799				sctp_ucount_decr(asoc->cnt_on_all_streams);
6800				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6801				if (control->on_read_q) {
6802					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6803					control->on_read_q = 0;
6804				}
6805				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, control, next_instrm);
6806				control->on_strm_q = 0;
6807				if (control->data) {
6808					sctp_m_freem(control->data);
6809					control->data = NULL;
6810				}
6811				sctp_free_remote_addr(control->whoFrom);
6812				/* Now its reasm? */
6813				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6814					cnt++;
6815					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6816					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6817					if (asoc->size_on_reasm_queue >= chk->send_size) {
6818						asoc->size_on_reasm_queue -= chk->send_size;
6819					} else {
6820#ifdef INVARIANTS
6821						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6822#else
6823						asoc->size_on_reasm_queue = 0;
6824#endif
6825					}
6826					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6827					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6828					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6829					if (chk->data) {
6830						sctp_m_freem(chk->data);
6831						chk->data = NULL;
6832					}
6833					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6834				}
6835				sctp_free_a_readq(stcb, control);
6836			}
6837		}
6838		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].uno_inqueue, next_instrm, ncontrol) {
6839#ifdef INVARIANTS
6840			if (control->on_strm_q != SCTP_ON_UNORDERED) {
6841				panic("Huh control: %p on_q: %d -- not unordered?",
6842				    control, control->on_strm_q);
6843			}
6844#endif
6845			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6846				/* Yep it is above cum-ack */
6847				cnt++;
6848				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6849				KASSERT(control->length > 0, ("control has zero length"));
6850				if (asoc->size_on_all_streams >= control->length) {
6851					asoc->size_on_all_streams -= control->length;
6852				} else {
6853#ifdef INVARIANTS
6854					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6855#else
6856					asoc->size_on_all_streams = 0;
6857#endif
6858				}
6859				sctp_ucount_decr(asoc->cnt_on_all_streams);
6860				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6861				if (control->on_read_q) {
6862					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6863					control->on_read_q = 0;
6864				}
6865				TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, control, next_instrm);
6866				control->on_strm_q = 0;
6867				if (control->data) {
6868					sctp_m_freem(control->data);
6869					control->data = NULL;
6870				}
6871				sctp_free_remote_addr(control->whoFrom);
6872				/* Now its reasm? */
6873				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6874					cnt++;
6875					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6876					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6877					if (asoc->size_on_reasm_queue >= chk->send_size) {
6878						asoc->size_on_reasm_queue -= chk->send_size;
6879					} else {
6880#ifdef INVARIANTS
6881						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6882#else
6883						asoc->size_on_reasm_queue = 0;
6884#endif
6885					}
6886					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6887					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6888					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6889					if (chk->data) {
6890						sctp_m_freem(chk->data);
6891						chk->data = NULL;
6892					}
6893					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6894				}
6895				sctp_free_a_readq(stcb, control);
6896			}
6897		}
6898	}
6899	if (cnt) {
6900		/* We must back down to see what the new highest is */
6901		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6902			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6903			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6904				asoc->highest_tsn_inside_map = i;
6905				fnd = 1;
6906				break;
6907			}
6908		}
6909		if (!fnd) {
6910			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6911		}
6912
6913		/*
6914		 * Question, should we go through the delivery queue? The
6915		 * only reason things are on here is the app not reading OR
6916		 * a p-d-api up. An attacker COULD send enough in to
6917		 * initiate the PD-API and then send a bunch of stuff to
6918		 * other streams... these would wind up on the delivery
6919		 * queue.. and then we would not get to them. But in order
6920		 * to do this I then have to back-track and un-deliver
6921		 * sequence numbers in streams.. el-yucko. I think for now
6922		 * we will NOT look at the delivery queue and leave it to be
6923		 * something to consider later. An alternative would be to
6924		 * abort the P-D-API with a notification and then deliver
6925		 * the data.... Or another method might be to keep track of
6926		 * how many times the situation occurs and if we see a
6927		 * possible attack underway just abort the association.
6928		 */
6929#ifdef SCTP_DEBUG
6930		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6931#endif
6932		/*
6933		 * Now do we need to find a new
6934		 * asoc->highest_tsn_inside_map?
6935		 */
6936		asoc->last_revoke_count = cnt;
6937		sctp_timer_stop(SCTP_TIMER_TYPE_RECV, stcb->sctp_ep, stcb, NULL,
6938		    SCTP_FROM_SCTP_PCB + SCTP_LOC_11);
6939		/* sa_ignore NO_NULL_CHK */
6940		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6941		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6942	}
6943	/*
6944	 * Another issue, in un-setting the TSN's in the mapping array we
6945	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6946	 * two things to occur. It may cause us to do extra work in checking
6947	 * for our mapping array movement. More importantly it may cause us
6948	 * to SACK every datagram. This may not be a bad thing though since
6949	 * we will recover once we get our cum-ack above and all this stuff
6950	 * we dumped recovered.
6951	 */
6952}
6953
6954static void
6955sctp_drain(void)
6956{
6957	struct epoch_tracker et;
6958
6959	VNET_ITERATOR_DECL(vnet_iter);
6960
6961	NET_EPOCH_ENTER(et);
6962	/*
6963	 * We must walk the PCB lists for ALL associations here. The system
6964	 * is LOW on MBUF's and needs help. This is where reneging will
6965	 * occur. We really hope this does NOT happen!
6966	 */
6967	VNET_LIST_RLOCK_NOSLEEP();
6968	VNET_FOREACH(vnet_iter) {
6969		CURVNET_SET(vnet_iter);
6970		struct sctp_inpcb *inp;
6971		struct sctp_tcb *stcb;
6972
6973		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6974		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6975#ifdef VIMAGE
6976			continue;
6977#else
6978			NET_EPOCH_EXIT(et);
6979			return;
6980#endif
6981		}
6982		SCTP_INP_INFO_RLOCK();
6983		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6984			/* For each endpoint */
6985			SCTP_INP_RLOCK(inp);
6986			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6987				/* For each association */
6988				SCTP_TCB_LOCK(stcb);
6989				sctp_drain_mbufs(stcb);
6990				SCTP_TCB_UNLOCK(stcb);
6991			}
6992			SCTP_INP_RUNLOCK(inp);
6993		}
6994		SCTP_INP_INFO_RUNLOCK();
6995		CURVNET_RESTORE();
6996	}
6997	VNET_LIST_RUNLOCK_NOSLEEP();
6998	NET_EPOCH_EXIT(et);
6999}
7000
7001EVENTHANDLER_DEFINE(vm_lowmem, sctp_drain, NULL, LOWMEM_PRI_DEFAULT);
7002EVENTHANDLER_DEFINE(mbuf_lowmem, sctp_drain, NULL, LOWMEM_PRI_DEFAULT);
7003
7004/*
7005 * start a new iterator
7006 * iterates through all endpoints and associations based on the pcb_state
7007 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7008 * assocs and "ef" (optional) is executed when the iterator completes.
7009 * "inpf" (optional) is executed for each new endpoint as it is being
7010 * iterated through. inpe (optional) is called when the inp completes
7011 * its way through all the stcbs.
7012 */
7013int
7014sctp_initiate_iterator(inp_func inpf,
7015    asoc_func af,
7016    inp_func inpe,
7017    uint32_t pcb_state,
7018    uint32_t pcb_features,
7019    uint32_t asoc_state,
7020    void *argp,
7021    uint32_t argi,
7022    end_func ef,
7023    struct sctp_inpcb *s_inp,
7024    uint8_t chunk_output_off)
7025{
7026	struct sctp_iterator *it = NULL;
7027
7028	if (af == NULL) {
7029		return (-1);
7030	}
7031	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7032		SCTP_PRINTF("%s: abort on initialize being %d\n", __func__,
7033		    SCTP_BASE_VAR(sctp_pcb_initialized));
7034		return (-1);
7035	}
7036	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7037	    SCTP_M_ITER);
7038	if (it == NULL) {
7039		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7040		return (-1);
7041	}
7042	memset(it, 0, sizeof(*it));
7043	it->function_assoc = af;
7044	it->function_inp = inpf;
7045	if (inpf)
7046		it->done_current_ep = 0;
7047	else
7048		it->done_current_ep = 1;
7049	it->function_atend = ef;
7050	it->pointer = argp;
7051	it->val = argi;
7052	it->pcb_flags = pcb_state;
7053	it->pcb_features = pcb_features;
7054	it->asoc_state = asoc_state;
7055	it->function_inp_end = inpe;
7056	it->no_chunk_output = chunk_output_off;
7057	it->vn = curvnet;
7058	if (s_inp) {
7059		/* Assume lock is held here */
7060		it->inp = s_inp;
7061		SCTP_INP_INCR_REF(it->inp);
7062		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7063	} else {
7064		SCTP_INP_INFO_RLOCK();
7065		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7066		if (it->inp) {
7067			SCTP_INP_INCR_REF(it->inp);
7068		}
7069		SCTP_INP_INFO_RUNLOCK();
7070		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7071	}
7072	SCTP_IPI_ITERATOR_WQ_LOCK();
7073	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7074		SCTP_IPI_ITERATOR_WQ_UNLOCK();
7075		SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__,
7076		    SCTP_BASE_VAR(sctp_pcb_initialized), it);
7077		SCTP_FREE(it, SCTP_M_ITER);
7078		return (-1);
7079	}
7080	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7081	if (sctp_it_ctl.iterator_running == 0) {
7082		sctp_wakeup_iterator();
7083	}
7084	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7085	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7086	return (0);
7087}
7088
7089/*
7090 * Atomically add flags to the sctp_flags of an inp.
7091 * To be used when the write lock of the inp is not held.
7092 */
7093void
7094sctp_pcb_add_flags(struct sctp_inpcb *inp, uint32_t flags)
7095{
7096	uint32_t old_flags, new_flags;
7097
7098	do {
7099		old_flags = inp->sctp_flags;
7100		new_flags = old_flags | flags;
7101	} while (atomic_cmpset_int(&inp->sctp_flags, old_flags, new_flags) == 0);
7102}
7103