sctp_pcb.c revision 332172
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: stable/11/sys/netinet/sctp_pcb.c 332172 2018-04-07 14:41:44Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#if defined(INET) || defined(INET6)
50#include <netinet/udp.h>
51#endif
52#ifdef INET6
53#include <netinet6/ip6_var.h>
54#endif
55#include <sys/sched.h>
56#include <sys/smp.h>
57#include <sys/unistd.h>
58
59
60VNET_DEFINE(struct sctp_base_info, system_base_info);
61
62/* FIX: we don't handle multiple link local scopes */
63/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
64#ifdef INET6
65int
66SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
67{
68	struct sockaddr_in6 tmp_a, tmp_b;
69
70	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
71	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
72		return (0);
73	}
74	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
75	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
76		return (0);
77	}
78	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
79}
80#endif
81
82void
83sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
84{
85	/*
86	 * We really don't need to lock this, but I will just because it
87	 * does not hurt.
88	 */
89	SCTP_INP_INFO_RLOCK();
90	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
91	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
92	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
93	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
94	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
95	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
96	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
97	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
98	SCTP_INP_INFO_RUNLOCK();
99}
100
101/*-
102 * Addresses are added to VRF's (Virtual Router's). For BSD we
103 * have only the default VRF 0. We maintain a hash list of
104 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
105 * these has a list of addresses. When we add a new address
106 * to a VRF we lookup the ifn/ifn_index, if the ifn does
107 * not exist we create it and add it to the list of IFN's
108 * within the VRF. Once we have the sctp_ifn, we add the
109 * address to the list. So we look something like:
110 *
111 * hash-vrf-table
112 *   vrf-> ifn-> ifn -> ifn
113 *   vrf    |
114 *    ...   +--ifa-> ifa -> ifa
115 *   vrf
116 *
117 * We keep these separate lists since the SCTP subsystem will
118 * point to these from its source address selection nets structure.
119 * When an address is deleted it does not happen right away on
120 * the SCTP side, it gets scheduled. What we do when a
121 * delete happens is immediately remove the address from
122 * the master list and decrement the refcount. As our
123 * addip iterator works through and frees the src address
124 * selection pointing to the sctp_ifa, eventually the refcount
125 * will reach 0 and we will delete it. Note that it is assumed
126 * that any locking on system level ifn/ifa is done at the
127 * caller of these functions and these routines will only
128 * lock the SCTP structures as they add or delete things.
129 *
130 * Other notes on VRF concepts.
131 *  - An endpoint can be in multiple VRF's
132 *  - An association lives within a VRF and only one VRF.
133 *  - Any incoming packet we can deduce the VRF for by
134 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
135 *  - Any downward send call or connect call must supply the
136 *    VRF via ancillary data or via some sort of set default
137 *    VRF socket option call (again for BSD no brainer since
138 *    the VRF is always 0).
139 *  - An endpoint may add multiple VRF's to it.
140 *  - Listening sockets can accept associations in any
141 *    of the VRF's they are in but the assoc will end up
142 *    in only one VRF (gotten from the packet or connect/send).
143 *
144 */
145
146struct sctp_vrf *
147sctp_allocate_vrf(int vrf_id)
148{
149	struct sctp_vrf *vrf = NULL;
150	struct sctp_vrflist *bucket;
151
152	/* First allocate the VRF structure */
153	vrf = sctp_find_vrf(vrf_id);
154	if (vrf) {
155		/* Already allocated */
156		return (vrf);
157	}
158	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
159	    SCTP_M_VRF);
160	if (vrf == NULL) {
161		/* No memory */
162#ifdef INVARIANTS
163		panic("No memory for VRF:%d", vrf_id);
164#endif
165		return (NULL);
166	}
167	/* setup the VRF */
168	memset(vrf, 0, sizeof(struct sctp_vrf));
169	vrf->vrf_id = vrf_id;
170	LIST_INIT(&vrf->ifnlist);
171	vrf->total_ifa_count = 0;
172	vrf->refcount = 0;
173	/* now also setup table ids */
174	SCTP_INIT_VRF_TABLEID(vrf);
175	/* Init the HASH of addresses */
176	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
177	    &vrf->vrf_addr_hashmark);
178	if (vrf->vrf_addr_hash == NULL) {
179		/* No memory */
180#ifdef INVARIANTS
181		panic("No memory for VRF:%d", vrf_id);
182#endif
183		SCTP_FREE(vrf, SCTP_M_VRF);
184		return (NULL);
185	}
186	/* Add it to the hash table */
187	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
188	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
189	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
190	return (vrf);
191}
192
193
194struct sctp_ifn *
195sctp_find_ifn(void *ifn, uint32_t ifn_index)
196{
197	struct sctp_ifn *sctp_ifnp;
198	struct sctp_ifnlist *hash_ifn_head;
199
200	/*
201	 * We assume the lock is held for the addresses if that's wrong
202	 * problems could occur :-)
203	 */
204	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
205	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
206		if (sctp_ifnp->ifn_index == ifn_index) {
207			return (sctp_ifnp);
208		}
209		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
210			return (sctp_ifnp);
211		}
212	}
213	return (NULL);
214}
215
216
217struct sctp_vrf *
218sctp_find_vrf(uint32_t vrf_id)
219{
220	struct sctp_vrflist *bucket;
221	struct sctp_vrf *liste;
222
223	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
224	LIST_FOREACH(liste, bucket, next_vrf) {
225		if (vrf_id == liste->vrf_id) {
226			return (liste);
227		}
228	}
229	return (NULL);
230}
231
232
233void
234sctp_free_vrf(struct sctp_vrf *vrf)
235{
236	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
237		if (vrf->vrf_addr_hash) {
238			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
239			vrf->vrf_addr_hash = NULL;
240		}
241		/* We zero'd the count */
242		LIST_REMOVE(vrf, next_vrf);
243		SCTP_FREE(vrf, SCTP_M_VRF);
244		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
245	}
246}
247
248
249void
250sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
251{
252	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
253		/* We zero'd the count */
254		if (sctp_ifnp->vrf) {
255			sctp_free_vrf(sctp_ifnp->vrf);
256		}
257		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
258		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
259	}
260}
261
262
263void
264sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
265{
266	struct sctp_ifn *sctp_ifnp;
267
268	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
269	if (sctp_ifnp != NULL) {
270		sctp_ifnp->ifn_mtu = mtu;
271	}
272}
273
274
275void
276sctp_free_ifa(struct sctp_ifa *sctp_ifap)
277{
278	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
279		/* We zero'd the count */
280		if (sctp_ifap->ifn_p) {
281			sctp_free_ifn(sctp_ifap->ifn_p);
282		}
283		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
284		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
285	}
286}
287
288
289static void
290sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
291{
292	struct sctp_ifn *found;
293
294	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
295	if (found == NULL) {
296		/* Not in the list.. sorry */
297		return;
298	}
299	if (hold_addr_lock == 0)
300		SCTP_IPI_ADDR_WLOCK();
301	LIST_REMOVE(sctp_ifnp, next_bucket);
302	LIST_REMOVE(sctp_ifnp, next_ifn);
303	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
304	    sctp_ifnp->registered_af);
305	if (hold_addr_lock == 0)
306		SCTP_IPI_ADDR_WUNLOCK();
307	/* Take away the reference, and possibly free it */
308	sctp_free_ifn(sctp_ifnp);
309}
310
311
312void
313sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
314    const char *if_name, uint32_t ifn_index)
315{
316	struct sctp_vrf *vrf;
317	struct sctp_ifa *sctp_ifap;
318
319	SCTP_IPI_ADDR_RLOCK();
320	vrf = sctp_find_vrf(vrf_id);
321	if (vrf == NULL) {
322		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
323		goto out;
324
325	}
326	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
327	if (sctp_ifap == NULL) {
328		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
329		goto out;
330	}
331	if (sctp_ifap->ifn_p == NULL) {
332		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
333		goto out;
334	}
335	if (if_name) {
336		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
337			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
338			    sctp_ifap->ifn_p->ifn_name, if_name);
339			goto out;
340		}
341	} else {
342		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
343			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
344			    sctp_ifap->ifn_p->ifn_index, ifn_index);
345			goto out;
346		}
347	}
348
349	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
350	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
351out:
352	SCTP_IPI_ADDR_RUNLOCK();
353}
354
355
356void
357sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
358    const char *if_name, uint32_t ifn_index)
359{
360	struct sctp_vrf *vrf;
361	struct sctp_ifa *sctp_ifap;
362
363	SCTP_IPI_ADDR_RLOCK();
364	vrf = sctp_find_vrf(vrf_id);
365	if (vrf == NULL) {
366		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
367		goto out;
368
369	}
370	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
371	if (sctp_ifap == NULL) {
372		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
373		goto out;
374	}
375	if (sctp_ifap->ifn_p == NULL) {
376		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unusable\n");
377		goto out;
378	}
379	if (if_name) {
380		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
381			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
382			    sctp_ifap->ifn_p->ifn_name, if_name);
383			goto out;
384		}
385	} else {
386		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
387			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
388			    sctp_ifap->ifn_p->ifn_index, ifn_index);
389			goto out;
390		}
391	}
392
393	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
394	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
395out:
396	SCTP_IPI_ADDR_RUNLOCK();
397}
398
399
400/*-
401 * Add an ifa to an ifn.
402 * Register the interface as necessary.
403 * NOTE: ADDR write lock MUST be held.
404 */
405static void
406sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
407{
408	int ifa_af;
409
410	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
411	sctp_ifap->ifn_p = sctp_ifnp;
412	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
413	/* update address counts */
414	sctp_ifnp->ifa_count++;
415	ifa_af = sctp_ifap->address.sa.sa_family;
416	switch (ifa_af) {
417#ifdef INET
418	case AF_INET:
419		sctp_ifnp->num_v4++;
420		break;
421#endif
422#ifdef INET6
423	case AF_INET6:
424		sctp_ifnp->num_v6++;
425		break;
426#endif
427	default:
428		break;
429	}
430	if (sctp_ifnp->ifa_count == 1) {
431		/* register the new interface */
432		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
433		sctp_ifnp->registered_af = ifa_af;
434	}
435}
436
437
438/*-
439 * Remove an ifa from its ifn.
440 * If no more addresses exist, remove the ifn too. Otherwise, re-register
441 * the interface based on the remaining address families left.
442 * NOTE: ADDR write lock MUST be held.
443 */
444static void
445sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
446{
447	LIST_REMOVE(sctp_ifap, next_ifa);
448	if (sctp_ifap->ifn_p) {
449		/* update address counts */
450		sctp_ifap->ifn_p->ifa_count--;
451		switch (sctp_ifap->address.sa.sa_family) {
452#ifdef INET
453		case AF_INET:
454			sctp_ifap->ifn_p->num_v4--;
455			break;
456#endif
457#ifdef INET6
458		case AF_INET6:
459			sctp_ifap->ifn_p->num_v6--;
460			break;
461#endif
462		default:
463			break;
464		}
465
466		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
467			/* remove the ifn, possibly freeing it */
468			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
469		} else {
470			/* re-register address family type, if needed */
471			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
472			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
473				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
474				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
475				sctp_ifap->ifn_p->registered_af = AF_INET;
476			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
477			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
478				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
479				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
480				sctp_ifap->ifn_p->registered_af = AF_INET6;
481			}
482			/* free the ifn refcount */
483			sctp_free_ifn(sctp_ifap->ifn_p);
484		}
485		sctp_ifap->ifn_p = NULL;
486	}
487}
488
489
490struct sctp_ifa *
491sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
492    uint32_t ifn_type, const char *if_name, void *ifa,
493    struct sockaddr *addr, uint32_t ifa_flags,
494    int dynamic_add)
495{
496	struct sctp_vrf *vrf;
497	struct sctp_ifn *sctp_ifnp = NULL;
498	struct sctp_ifa *sctp_ifap = NULL;
499	struct sctp_ifalist *hash_addr_head;
500	struct sctp_ifnlist *hash_ifn_head;
501	uint32_t hash_of_addr;
502	int new_ifn_af = 0;
503
504#ifdef SCTP_DEBUG
505	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
506	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
507#endif
508	SCTP_IPI_ADDR_WLOCK();
509	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
510	if (sctp_ifnp) {
511		vrf = sctp_ifnp->vrf;
512	} else {
513		vrf = sctp_find_vrf(vrf_id);
514		if (vrf == NULL) {
515			vrf = sctp_allocate_vrf(vrf_id);
516			if (vrf == NULL) {
517				SCTP_IPI_ADDR_WUNLOCK();
518				return (NULL);
519			}
520		}
521	}
522	if (sctp_ifnp == NULL) {
523		/*
524		 * build one and add it, can't hold lock until after malloc
525		 * done though.
526		 */
527		SCTP_IPI_ADDR_WUNLOCK();
528		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
529		    sizeof(struct sctp_ifn), SCTP_M_IFN);
530		if (sctp_ifnp == NULL) {
531#ifdef INVARIANTS
532			panic("No memory for IFN");
533#endif
534			return (NULL);
535		}
536		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
537		sctp_ifnp->ifn_index = ifn_index;
538		sctp_ifnp->ifn_p = ifn;
539		sctp_ifnp->ifn_type = ifn_type;
540		sctp_ifnp->refcount = 0;
541		sctp_ifnp->vrf = vrf;
542		atomic_add_int(&vrf->refcount, 1);
543		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
544		if (if_name != NULL) {
545			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
546		} else {
547			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
548		}
549		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
550		LIST_INIT(&sctp_ifnp->ifalist);
551		SCTP_IPI_ADDR_WLOCK();
552		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
553		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
554		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
555		new_ifn_af = 1;
556	}
557	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
558	if (sctp_ifap) {
559		/* Hmm, it already exists? */
560		if ((sctp_ifap->ifn_p) &&
561		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
562			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
563			    sctp_ifap->ifn_p->ifn_name, ifn_index,
564			    (void *)sctp_ifap);
565			if (new_ifn_af) {
566				/* Remove the created one that we don't want */
567				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
568			}
569			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
570				/* easy to solve, just switch back to active */
571				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
572				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
573				sctp_ifap->ifn_p = sctp_ifnp;
574				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
575			}
576	exit_stage_left:
577			SCTP_IPI_ADDR_WUNLOCK();
578			return (sctp_ifap);
579		} else {
580			if (sctp_ifap->ifn_p) {
581				/*
582				 * The last IFN gets the address, remove the
583				 * old one
584				 */
585				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
586				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
587				    sctp_ifap->ifn_p->ifn_index, if_name,
588				    ifn_index);
589				/* remove the address from the old ifn */
590				sctp_remove_ifa_from_ifn(sctp_ifap);
591				/* move the address over to the new ifn */
592				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
593				goto exit_stage_left;
594			} else {
595				/* repair ifnp which was NULL ? */
596				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
597				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
598				    (void *)sctp_ifnp, (void *)sctp_ifap);
599				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
600			}
601			goto exit_stage_left;
602		}
603	}
604	SCTP_IPI_ADDR_WUNLOCK();
605	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
606	if (sctp_ifap == NULL) {
607#ifdef INVARIANTS
608		panic("No memory for IFA");
609#endif
610		return (NULL);
611	}
612	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
613	sctp_ifap->ifn_p = sctp_ifnp;
614	atomic_add_int(&sctp_ifnp->refcount, 1);
615	sctp_ifap->vrf_id = vrf_id;
616	sctp_ifap->ifa = ifa;
617	memcpy(&sctp_ifap->address, addr, addr->sa_len);
618	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
619	sctp_ifap->flags = ifa_flags;
620	/* Set scope */
621	switch (sctp_ifap->address.sa.sa_family) {
622#ifdef INET
623	case AF_INET:
624		{
625			struct sockaddr_in *sin;
626
627			sin = &sctp_ifap->address.sin;
628			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
629			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
630				sctp_ifap->src_is_loop = 1;
631			}
632			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
633				sctp_ifap->src_is_priv = 1;
634			}
635			sctp_ifnp->num_v4++;
636			if (new_ifn_af)
637				new_ifn_af = AF_INET;
638			break;
639		}
640#endif
641#ifdef INET6
642	case AF_INET6:
643		{
644			/* ok to use deprecated addresses? */
645			struct sockaddr_in6 *sin6;
646
647			sin6 = &sctp_ifap->address.sin6;
648			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
649			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
650				sctp_ifap->src_is_loop = 1;
651			}
652			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
653				sctp_ifap->src_is_priv = 1;
654			}
655			sctp_ifnp->num_v6++;
656			if (new_ifn_af)
657				new_ifn_af = AF_INET6;
658			break;
659		}
660#endif
661	default:
662		new_ifn_af = 0;
663		break;
664	}
665	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
666
667	if ((sctp_ifap->src_is_priv == 0) &&
668	    (sctp_ifap->src_is_loop == 0)) {
669		sctp_ifap->src_is_glob = 1;
670	}
671	SCTP_IPI_ADDR_WLOCK();
672	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
673	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
674	sctp_ifap->refcount = 1;
675	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
676	sctp_ifnp->ifa_count++;
677	vrf->total_ifa_count++;
678	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
679	if (new_ifn_af) {
680		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
681		sctp_ifnp->registered_af = new_ifn_af;
682	}
683	SCTP_IPI_ADDR_WUNLOCK();
684	if (dynamic_add) {
685		/*
686		 * Bump up the refcount so that when the timer completes it
687		 * will drop back down.
688		 */
689		struct sctp_laddr *wi;
690
691		atomic_add_int(&sctp_ifap->refcount, 1);
692		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
693		if (wi == NULL) {
694			/*
695			 * Gak, what can we do? We have lost an address
696			 * change can you say HOSED?
697			 */
698			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
699			/* Opps, must decrement the count */
700			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
701			    if_name);
702			return (NULL);
703		}
704		SCTP_INCR_LADDR_COUNT();
705		memset(wi, 0, sizeof(*wi));
706		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
707		wi->ifa = sctp_ifap;
708		wi->action = SCTP_ADD_IP_ADDRESS;
709
710		SCTP_WQ_ADDR_LOCK();
711		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
712		SCTP_WQ_ADDR_UNLOCK();
713
714		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
715		    (struct sctp_inpcb *)NULL,
716		    (struct sctp_tcb *)NULL,
717		    (struct sctp_nets *)NULL);
718	} else {
719		/* it's ready for use */
720		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
721	}
722	return (sctp_ifap);
723}
724
725void
726sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
727    uint32_t ifn_index, const char *if_name)
728{
729	struct sctp_vrf *vrf;
730	struct sctp_ifa *sctp_ifap = NULL;
731
732	SCTP_IPI_ADDR_WLOCK();
733	vrf = sctp_find_vrf(vrf_id);
734	if (vrf == NULL) {
735		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
736		goto out_now;
737	}
738#ifdef SCTP_DEBUG
739	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
740	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
741#endif
742	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
743	if (sctp_ifap) {
744		/* Validate the delete */
745		if (sctp_ifap->ifn_p) {
746			int valid = 0;
747
748			/*-
749			 * The name has priority over the ifn_index
750			 * if its given. We do this especially for
751			 * panda who might recycle indexes fast.
752			 */
753			if (if_name) {
754				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
755					/* They match its a correct delete */
756					valid = 1;
757				}
758			}
759			if (!valid) {
760				/* last ditch check ifn_index */
761				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
762					valid = 1;
763				}
764			}
765			if (!valid) {
766				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
767				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
768				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
769				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
770				SCTP_IPI_ADDR_WUNLOCK();
771				return;
772			}
773		}
774		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
775		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
776		/*
777		 * We don't set the flag. This means that the structure will
778		 * hang around in EP's that have bound specific to it until
779		 * they close. This gives us TCP like behavior if someone
780		 * removes an address (or for that matter adds it right
781		 * back).
782		 */
783		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
784		vrf->total_ifa_count--;
785		LIST_REMOVE(sctp_ifap, next_bucket);
786		sctp_remove_ifa_from_ifn(sctp_ifap);
787	}
788#ifdef SCTP_DEBUG
789	else {
790		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
791		    ifn_index);
792		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
793	}
794#endif
795
796out_now:
797	SCTP_IPI_ADDR_WUNLOCK();
798	if (sctp_ifap) {
799		struct sctp_laddr *wi;
800
801		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
802		if (wi == NULL) {
803			/*
804			 * Gak, what can we do? We have lost an address
805			 * change can you say HOSED?
806			 */
807			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
808
809			/* Oops, must decrement the count */
810			sctp_free_ifa(sctp_ifap);
811			return;
812		}
813		SCTP_INCR_LADDR_COUNT();
814		memset(wi, 0, sizeof(*wi));
815		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
816		wi->ifa = sctp_ifap;
817		wi->action = SCTP_DEL_IP_ADDRESS;
818		SCTP_WQ_ADDR_LOCK();
819		/*
820		 * Should this really be a tailq? As it is we will process
821		 * the newest first :-0
822		 */
823		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
824		SCTP_WQ_ADDR_UNLOCK();
825
826		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
827		    (struct sctp_inpcb *)NULL,
828		    (struct sctp_tcb *)NULL,
829		    (struct sctp_nets *)NULL);
830	}
831	return;
832}
833
834
835static int
836sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
837{
838	int loopback_scope;
839#if defined(INET)
840	int ipv4_local_scope, ipv4_addr_legal;
841#endif
842#if defined(INET6)
843	int local_scope, site_scope, ipv6_addr_legal;
844#endif
845	struct sctp_vrf *vrf;
846	struct sctp_ifn *sctp_ifn;
847	struct sctp_ifa *sctp_ifa;
848
849	loopback_scope = stcb->asoc.scope.loopback_scope;
850#if defined(INET)
851	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
852	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
853#endif
854#if defined(INET6)
855	local_scope = stcb->asoc.scope.local_scope;
856	site_scope = stcb->asoc.scope.site_scope;
857	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
858#endif
859
860	SCTP_IPI_ADDR_RLOCK();
861	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
862	if (vrf == NULL) {
863		/* no vrf, no addresses */
864		SCTP_IPI_ADDR_RUNLOCK();
865		return (0);
866	}
867	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
868		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
869			if ((loopback_scope == 0) &&
870			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
871				continue;
872			}
873			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
874				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
875				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
876					/*
877					 * We allow pending addresses, where
878					 * we have sent an asconf-add to be
879					 * considered valid.
880					 */
881					continue;
882				}
883				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
884					continue;
885				}
886				switch (sctp_ifa->address.sa.sa_family) {
887#ifdef INET
888				case AF_INET:
889					if (ipv4_addr_legal) {
890						struct sockaddr_in *sin,
891						           *rsin;
892
893						sin = &sctp_ifa->address.sin;
894						rsin = (struct sockaddr_in *)to;
895						if ((ipv4_local_scope == 0) &&
896						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
897							continue;
898						}
899						if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred,
900						    &sin->sin_addr) != 0) {
901							continue;
902						}
903						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
904							SCTP_IPI_ADDR_RUNLOCK();
905							return (1);
906						}
907					}
908					break;
909#endif
910#ifdef INET6
911				case AF_INET6:
912					if (ipv6_addr_legal) {
913						struct sockaddr_in6 *sin6,
914						            *rsin6;
915
916						sin6 = &sctp_ifa->address.sin6;
917						rsin6 = (struct sockaddr_in6 *)to;
918						if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred,
919						    &sin6->sin6_addr) != 0) {
920							continue;
921						}
922						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
923							if (local_scope == 0)
924								continue;
925							if (sin6->sin6_scope_id == 0) {
926								if (sa6_recoverscope(sin6) != 0)
927									continue;
928							}
929						}
930						if ((site_scope == 0) &&
931						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
932							continue;
933						}
934						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
935							SCTP_IPI_ADDR_RUNLOCK();
936							return (1);
937						}
938					}
939					break;
940#endif
941				default:
942					/* TSNH */
943					break;
944				}
945			}
946		}
947	} else {
948		struct sctp_laddr *laddr;
949
950		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
951			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
952				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
953				continue;
954			}
955			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
956			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
957				/*
958				 * We allow pending addresses, where we have
959				 * sent an asconf-add to be considered
960				 * valid.
961				 */
962				continue;
963			}
964			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
965				continue;
966			}
967			switch (to->sa_family) {
968#ifdef INET
969			case AF_INET:
970				{
971					struct sockaddr_in *sin, *rsin;
972
973					sin = &laddr->ifa->address.sin;
974					rsin = (struct sockaddr_in *)to;
975					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
976						SCTP_IPI_ADDR_RUNLOCK();
977						return (1);
978					}
979					break;
980				}
981#endif
982#ifdef INET6
983			case AF_INET6:
984				{
985					struct sockaddr_in6 *sin6, *rsin6;
986
987					sin6 = &laddr->ifa->address.sin6;
988					rsin6 = (struct sockaddr_in6 *)to;
989					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
990						SCTP_IPI_ADDR_RUNLOCK();
991						return (1);
992					}
993					break;
994				}
995
996#endif
997			default:
998				/* TSNH */
999				break;
1000			}
1001
1002		}
1003	}
1004	SCTP_IPI_ADDR_RUNLOCK();
1005	return (0);
1006}
1007
1008
1009static struct sctp_tcb *
1010sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1011    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1012{
1013	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1014	/*
1015	 * If we support the TCP model, then we must now dig through to see
1016	 * if we can find our endpoint in the list of tcp ep's.
1017	 */
1018	uint16_t lport, rport;
1019	struct sctppcbhead *ephead;
1020	struct sctp_inpcb *inp;
1021	struct sctp_laddr *laddr;
1022	struct sctp_tcb *stcb;
1023	struct sctp_nets *net;
1024
1025	if ((to == NULL) || (from == NULL)) {
1026		return (NULL);
1027	}
1028	switch (to->sa_family) {
1029#ifdef INET
1030	case AF_INET:
1031		if (from->sa_family == AF_INET) {
1032			lport = ((struct sockaddr_in *)to)->sin_port;
1033			rport = ((struct sockaddr_in *)from)->sin_port;
1034		} else {
1035			return (NULL);
1036		}
1037		break;
1038#endif
1039#ifdef INET6
1040	case AF_INET6:
1041		if (from->sa_family == AF_INET6) {
1042			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1043			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1044		} else {
1045			return (NULL);
1046		}
1047		break;
1048#endif
1049	default:
1050		return (NULL);
1051	}
1052	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1053	/*
1054	 * Ok now for each of the guys in this bucket we must look and see:
1055	 * - Does the remote port match. - Does there single association's
1056	 * addresses match this address (to). If so we update p_ep to point
1057	 * to this ep and return the tcb from it.
1058	 */
1059	LIST_FOREACH(inp, ephead, sctp_hash) {
1060		SCTP_INP_RLOCK(inp);
1061		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1062			SCTP_INP_RUNLOCK(inp);
1063			continue;
1064		}
1065		if (lport != inp->sctp_lport) {
1066			SCTP_INP_RUNLOCK(inp);
1067			continue;
1068		}
1069		switch (to->sa_family) {
1070#ifdef INET
1071		case AF_INET:
1072			{
1073				struct sockaddr_in *sin;
1074
1075				sin = (struct sockaddr_in *)to;
1076				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1077				    &sin->sin_addr) != 0) {
1078					SCTP_INP_RUNLOCK(inp);
1079					continue;
1080				}
1081				break;
1082			}
1083#endif
1084#ifdef INET6
1085		case AF_INET6:
1086			{
1087				struct sockaddr_in6 *sin6;
1088
1089				sin6 = (struct sockaddr_in6 *)to;
1090				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1091				    &sin6->sin6_addr) != 0) {
1092					SCTP_INP_RUNLOCK(inp);
1093					continue;
1094				}
1095				break;
1096			}
1097#endif
1098		default:
1099			SCTP_INP_RUNLOCK(inp);
1100			continue;
1101		}
1102		if (inp->def_vrf_id != vrf_id) {
1103			SCTP_INP_RUNLOCK(inp);
1104			continue;
1105		}
1106		/* check to see if the ep has one of the addresses */
1107		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1108			/* We are NOT bound all, so look further */
1109			int match = 0;
1110
1111			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1112
1113				if (laddr->ifa == NULL) {
1114					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__);
1115					continue;
1116				}
1117				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1118					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1119					continue;
1120				}
1121				if (laddr->ifa->address.sa.sa_family ==
1122				    to->sa_family) {
1123					/* see if it matches */
1124#ifdef INET
1125					if (from->sa_family == AF_INET) {
1126						struct sockaddr_in *intf_addr,
1127						           *sin;
1128
1129						intf_addr = &laddr->ifa->address.sin;
1130						sin = (struct sockaddr_in *)to;
1131						if (sin->sin_addr.s_addr ==
1132						    intf_addr->sin_addr.s_addr) {
1133							match = 1;
1134							break;
1135						}
1136					}
1137#endif
1138#ifdef INET6
1139					if (from->sa_family == AF_INET6) {
1140						struct sockaddr_in6 *intf_addr6;
1141						struct sockaddr_in6 *sin6;
1142
1143						sin6 = (struct sockaddr_in6 *)
1144						    to;
1145						intf_addr6 = &laddr->ifa->address.sin6;
1146
1147						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1148						    intf_addr6)) {
1149							match = 1;
1150							break;
1151						}
1152					}
1153#endif
1154				}
1155			}
1156			if (match == 0) {
1157				/* This endpoint does not have this address */
1158				SCTP_INP_RUNLOCK(inp);
1159				continue;
1160			}
1161		}
1162		/*
1163		 * Ok if we hit here the ep has the address, does it hold
1164		 * the tcb?
1165		 */
1166		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1167		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1168		if (stcb == NULL) {
1169			SCTP_INP_RUNLOCK(inp);
1170			continue;
1171		}
1172		SCTP_TCB_LOCK(stcb);
1173		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1174			SCTP_TCB_UNLOCK(stcb);
1175			SCTP_INP_RUNLOCK(inp);
1176			continue;
1177		}
1178		if (stcb->rport != rport) {
1179			/* remote port does not match. */
1180			SCTP_TCB_UNLOCK(stcb);
1181			SCTP_INP_RUNLOCK(inp);
1182			continue;
1183		}
1184		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1185			SCTP_TCB_UNLOCK(stcb);
1186			SCTP_INP_RUNLOCK(inp);
1187			continue;
1188		}
1189		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1190			SCTP_TCB_UNLOCK(stcb);
1191			SCTP_INP_RUNLOCK(inp);
1192			continue;
1193		}
1194		/* Does this TCB have a matching address? */
1195		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1196
1197			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1198				/* not the same family, can't be a match */
1199				continue;
1200			}
1201			switch (from->sa_family) {
1202#ifdef INET
1203			case AF_INET:
1204				{
1205					struct sockaddr_in *sin, *rsin;
1206
1207					sin = (struct sockaddr_in *)&net->ro._l_addr;
1208					rsin = (struct sockaddr_in *)from;
1209					if (sin->sin_addr.s_addr ==
1210					    rsin->sin_addr.s_addr) {
1211						/* found it */
1212						if (netp != NULL) {
1213							*netp = net;
1214						}
1215						/*
1216						 * Update the endpoint
1217						 * pointer
1218						 */
1219						*inp_p = inp;
1220						SCTP_INP_RUNLOCK(inp);
1221						return (stcb);
1222					}
1223					break;
1224				}
1225#endif
1226#ifdef INET6
1227			case AF_INET6:
1228				{
1229					struct sockaddr_in6 *sin6, *rsin6;
1230
1231					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1232					rsin6 = (struct sockaddr_in6 *)from;
1233					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1234					    rsin6)) {
1235						/* found it */
1236						if (netp != NULL) {
1237							*netp = net;
1238						}
1239						/*
1240						 * Update the endpoint
1241						 * pointer
1242						 */
1243						*inp_p = inp;
1244						SCTP_INP_RUNLOCK(inp);
1245						return (stcb);
1246					}
1247					break;
1248				}
1249#endif
1250			default:
1251				/* TSNH */
1252				break;
1253			}
1254		}
1255		SCTP_TCB_UNLOCK(stcb);
1256		SCTP_INP_RUNLOCK(inp);
1257	}
1258	return (NULL);
1259}
1260
1261
1262/*
1263 * rules for use
1264 *
1265 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1266 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1267 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1268 * NULL.
1269 */
1270
1271struct sctp_tcb *
1272sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1273    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1274{
1275	struct sctpasochead *head;
1276	struct sctp_inpcb *inp;
1277	struct sctp_tcb *stcb = NULL;
1278	struct sctp_nets *net;
1279	uint16_t rport;
1280
1281	inp = *inp_p;
1282	switch (remote->sa_family) {
1283#ifdef INET
1284	case AF_INET:
1285		rport = (((struct sockaddr_in *)remote)->sin_port);
1286		break;
1287#endif
1288#ifdef INET6
1289	case AF_INET6:
1290		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1291		break;
1292#endif
1293	default:
1294		return (NULL);
1295	}
1296	if (locked_tcb) {
1297		/*
1298		 * UN-lock so we can do proper locking here this occurs when
1299		 * called from load_addresses_from_init.
1300		 */
1301		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1302		SCTP_TCB_UNLOCK(locked_tcb);
1303	}
1304	SCTP_INP_INFO_RLOCK();
1305	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1306	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1307		/*-
1308		 * Now either this guy is our listener or it's the
1309		 * connector. If it is the one that issued the connect, then
1310		 * it's only chance is to be the first TCB in the list. If
1311		 * it is the acceptor, then do the special_lookup to hash
1312		 * and find the real inp.
1313		 */
1314		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1315			/* to is peer addr, from is my addr */
1316			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1317			    netp, inp->def_vrf_id);
1318			if ((stcb != NULL) && (locked_tcb == NULL)) {
1319				/* we have a locked tcb, lower refcount */
1320				SCTP_INP_DECR_REF(inp);
1321			}
1322			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1323				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1324				SCTP_TCB_LOCK(locked_tcb);
1325				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1326				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1327			}
1328			SCTP_INP_INFO_RUNLOCK();
1329			return (stcb);
1330		} else {
1331			SCTP_INP_WLOCK(inp);
1332			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1333				goto null_return;
1334			}
1335			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1336			if (stcb == NULL) {
1337				goto null_return;
1338			}
1339			SCTP_TCB_LOCK(stcb);
1340
1341			if (stcb->rport != rport) {
1342				/* remote port does not match. */
1343				SCTP_TCB_UNLOCK(stcb);
1344				goto null_return;
1345			}
1346			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1347				SCTP_TCB_UNLOCK(stcb);
1348				goto null_return;
1349			}
1350			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1351				SCTP_TCB_UNLOCK(stcb);
1352				goto null_return;
1353			}
1354			/* now look at the list of remote addresses */
1355			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1356#ifdef INVARIANTS
1357				if (net == (TAILQ_NEXT(net, sctp_next))) {
1358					panic("Corrupt net list");
1359				}
1360#endif
1361				if (net->ro._l_addr.sa.sa_family !=
1362				    remote->sa_family) {
1363					/* not the same family */
1364					continue;
1365				}
1366				switch (remote->sa_family) {
1367#ifdef INET
1368				case AF_INET:
1369					{
1370						struct sockaddr_in *sin,
1371						           *rsin;
1372
1373						sin = (struct sockaddr_in *)
1374						    &net->ro._l_addr;
1375						rsin = (struct sockaddr_in *)remote;
1376						if (sin->sin_addr.s_addr ==
1377						    rsin->sin_addr.s_addr) {
1378							/* found it */
1379							if (netp != NULL) {
1380								*netp = net;
1381							}
1382							if (locked_tcb == NULL) {
1383								SCTP_INP_DECR_REF(inp);
1384							} else if (locked_tcb != stcb) {
1385								SCTP_TCB_LOCK(locked_tcb);
1386							}
1387							if (locked_tcb) {
1388								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1389							}
1390							SCTP_INP_WUNLOCK(inp);
1391							SCTP_INP_INFO_RUNLOCK();
1392							return (stcb);
1393						}
1394						break;
1395					}
1396#endif
1397#ifdef INET6
1398				case AF_INET6:
1399					{
1400						struct sockaddr_in6 *sin6,
1401						            *rsin6;
1402
1403						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1404						rsin6 = (struct sockaddr_in6 *)remote;
1405						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1406						    rsin6)) {
1407							/* found it */
1408							if (netp != NULL) {
1409								*netp = net;
1410							}
1411							if (locked_tcb == NULL) {
1412								SCTP_INP_DECR_REF(inp);
1413							} else if (locked_tcb != stcb) {
1414								SCTP_TCB_LOCK(locked_tcb);
1415							}
1416							if (locked_tcb) {
1417								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1418							}
1419							SCTP_INP_WUNLOCK(inp);
1420							SCTP_INP_INFO_RUNLOCK();
1421							return (stcb);
1422						}
1423						break;
1424					}
1425#endif
1426				default:
1427					/* TSNH */
1428					break;
1429				}
1430			}
1431			SCTP_TCB_UNLOCK(stcb);
1432		}
1433	} else {
1434		SCTP_INP_WLOCK(inp);
1435		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1436			goto null_return;
1437		}
1438		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1439		    inp->sctp_hashmark)];
1440		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1441			if (stcb->rport != rport) {
1442				/* remote port does not match */
1443				continue;
1444			}
1445			SCTP_TCB_LOCK(stcb);
1446			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1447				SCTP_TCB_UNLOCK(stcb);
1448				continue;
1449			}
1450			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1451				SCTP_TCB_UNLOCK(stcb);
1452				continue;
1453			}
1454			/* now look at the list of remote addresses */
1455			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1456#ifdef INVARIANTS
1457				if (net == (TAILQ_NEXT(net, sctp_next))) {
1458					panic("Corrupt net list");
1459				}
1460#endif
1461				if (net->ro._l_addr.sa.sa_family !=
1462				    remote->sa_family) {
1463					/* not the same family */
1464					continue;
1465				}
1466				switch (remote->sa_family) {
1467#ifdef INET
1468				case AF_INET:
1469					{
1470						struct sockaddr_in *sin,
1471						           *rsin;
1472
1473						sin = (struct sockaddr_in *)
1474						    &net->ro._l_addr;
1475						rsin = (struct sockaddr_in *)remote;
1476						if (sin->sin_addr.s_addr ==
1477						    rsin->sin_addr.s_addr) {
1478							/* found it */
1479							if (netp != NULL) {
1480								*netp = net;
1481							}
1482							if (locked_tcb == NULL) {
1483								SCTP_INP_DECR_REF(inp);
1484							} else if (locked_tcb != stcb) {
1485								SCTP_TCB_LOCK(locked_tcb);
1486							}
1487							if (locked_tcb) {
1488								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1489							}
1490							SCTP_INP_WUNLOCK(inp);
1491							SCTP_INP_INFO_RUNLOCK();
1492							return (stcb);
1493						}
1494						break;
1495					}
1496#endif
1497#ifdef INET6
1498				case AF_INET6:
1499					{
1500						struct sockaddr_in6 *sin6,
1501						            *rsin6;
1502
1503						sin6 = (struct sockaddr_in6 *)
1504						    &net->ro._l_addr;
1505						rsin6 = (struct sockaddr_in6 *)remote;
1506						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1507						    rsin6)) {
1508							/* found it */
1509							if (netp != NULL) {
1510								*netp = net;
1511							}
1512							if (locked_tcb == NULL) {
1513								SCTP_INP_DECR_REF(inp);
1514							} else if (locked_tcb != stcb) {
1515								SCTP_TCB_LOCK(locked_tcb);
1516							}
1517							if (locked_tcb) {
1518								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1519							}
1520							SCTP_INP_WUNLOCK(inp);
1521							SCTP_INP_INFO_RUNLOCK();
1522							return (stcb);
1523						}
1524						break;
1525					}
1526#endif
1527				default:
1528					/* TSNH */
1529					break;
1530				}
1531			}
1532			SCTP_TCB_UNLOCK(stcb);
1533		}
1534	}
1535null_return:
1536	/* clean up for returning null */
1537	if (locked_tcb) {
1538		SCTP_TCB_LOCK(locked_tcb);
1539		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1540	}
1541	SCTP_INP_WUNLOCK(inp);
1542	SCTP_INP_INFO_RUNLOCK();
1543	/* not found */
1544	return (NULL);
1545}
1546
1547
1548/*
1549 * Find an association for a specific endpoint using the association id given
1550 * out in the COMM_UP notification
1551 */
1552struct sctp_tcb *
1553sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1554{
1555	/*
1556	 * Use my the assoc_id to find a endpoint
1557	 */
1558	struct sctpasochead *head;
1559	struct sctp_tcb *stcb;
1560	uint32_t id;
1561
1562	if (inp == NULL) {
1563		SCTP_PRINTF("TSNH ep_associd\n");
1564		return (NULL);
1565	}
1566	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1567		SCTP_PRINTF("TSNH ep_associd0\n");
1568		return (NULL);
1569	}
1570	id = (uint32_t)asoc_id;
1571	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1572	if (head == NULL) {
1573		/* invalid id TSNH */
1574		SCTP_PRINTF("TSNH ep_associd1\n");
1575		return (NULL);
1576	}
1577	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1578		if (stcb->asoc.assoc_id == id) {
1579			if (inp != stcb->sctp_ep) {
1580				/*
1581				 * some other guy has the same id active (id
1582				 * collision ??).
1583				 */
1584				SCTP_PRINTF("TSNH ep_associd2\n");
1585				continue;
1586			}
1587			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1588				continue;
1589			}
1590			if (want_lock) {
1591				SCTP_TCB_LOCK(stcb);
1592			}
1593			return (stcb);
1594		}
1595	}
1596	return (NULL);
1597}
1598
1599
1600struct sctp_tcb *
1601sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1602{
1603	struct sctp_tcb *stcb;
1604
1605	SCTP_INP_RLOCK(inp);
1606	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1607	SCTP_INP_RUNLOCK(inp);
1608	return (stcb);
1609}
1610
1611
1612/*
1613 * Endpoint probe expects that the INP_INFO is locked.
1614 */
1615static struct sctp_inpcb *
1616sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1617    uint16_t lport, uint32_t vrf_id)
1618{
1619	struct sctp_inpcb *inp;
1620	struct sctp_laddr *laddr;
1621#ifdef INET
1622	struct sockaddr_in *sin;
1623#endif
1624#ifdef INET6
1625	struct sockaddr_in6 *sin6;
1626	struct sockaddr_in6 *intf_addr6;
1627#endif
1628	int fnd;
1629
1630#ifdef INET
1631	sin = NULL;
1632#endif
1633#ifdef INET6
1634	sin6 = NULL;
1635#endif
1636	switch (nam->sa_family) {
1637#ifdef INET
1638	case AF_INET:
1639		sin = (struct sockaddr_in *)nam;
1640		break;
1641#endif
1642#ifdef INET6
1643	case AF_INET6:
1644		sin6 = (struct sockaddr_in6 *)nam;
1645		break;
1646#endif
1647	default:
1648		/* unsupported family */
1649		return (NULL);
1650	}
1651
1652	if (head == NULL)
1653		return (NULL);
1654
1655	LIST_FOREACH(inp, head, sctp_hash) {
1656		SCTP_INP_RLOCK(inp);
1657		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1658			SCTP_INP_RUNLOCK(inp);
1659			continue;
1660		}
1661		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1662		    (inp->sctp_lport == lport)) {
1663			/* got it */
1664			switch (nam->sa_family) {
1665#ifdef INET
1666			case AF_INET:
1667				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1668				    SCTP_IPV6_V6ONLY(inp)) {
1669					/*
1670					 * IPv4 on a IPv6 socket with ONLY
1671					 * IPv6 set
1672					 */
1673					SCTP_INP_RUNLOCK(inp);
1674					continue;
1675				}
1676				if (prison_check_ip4(inp->ip_inp.inp.inp_cred,
1677				    &sin->sin_addr) != 0) {
1678					SCTP_INP_RUNLOCK(inp);
1679					continue;
1680				}
1681				break;
1682#endif
1683#ifdef INET6
1684			case AF_INET6:
1685				/*
1686				 * A V6 address and the endpoint is NOT
1687				 * bound V6
1688				 */
1689				if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1690					SCTP_INP_RUNLOCK(inp);
1691					continue;
1692				}
1693				if (prison_check_ip6(inp->ip_inp.inp.inp_cred,
1694				    &sin6->sin6_addr) != 0) {
1695					SCTP_INP_RUNLOCK(inp);
1696					continue;
1697				}
1698				break;
1699#endif
1700			default:
1701				break;
1702			}
1703			/* does a VRF id match? */
1704			fnd = 0;
1705			if (inp->def_vrf_id == vrf_id)
1706				fnd = 1;
1707
1708			SCTP_INP_RUNLOCK(inp);
1709			if (!fnd)
1710				continue;
1711			return (inp);
1712		}
1713		SCTP_INP_RUNLOCK(inp);
1714	}
1715	switch (nam->sa_family) {
1716#ifdef INET
1717	case AF_INET:
1718		if (sin->sin_addr.s_addr == INADDR_ANY) {
1719			/* Can't hunt for one that has no address specified */
1720			return (NULL);
1721		}
1722		break;
1723#endif
1724#ifdef INET6
1725	case AF_INET6:
1726		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1727			/* Can't hunt for one that has no address specified */
1728			return (NULL);
1729		}
1730		break;
1731#endif
1732	default:
1733		break;
1734	}
1735	/*
1736	 * ok, not bound to all so see if we can find a EP bound to this
1737	 * address.
1738	 */
1739	LIST_FOREACH(inp, head, sctp_hash) {
1740		SCTP_INP_RLOCK(inp);
1741		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1742			SCTP_INP_RUNLOCK(inp);
1743			continue;
1744		}
1745		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1746			SCTP_INP_RUNLOCK(inp);
1747			continue;
1748		}
1749		/*
1750		 * Ok this could be a likely candidate, look at all of its
1751		 * addresses
1752		 */
1753		if (inp->sctp_lport != lport) {
1754			SCTP_INP_RUNLOCK(inp);
1755			continue;
1756		}
1757		/* does a VRF id match? */
1758		fnd = 0;
1759		if (inp->def_vrf_id == vrf_id)
1760			fnd = 1;
1761
1762		if (!fnd) {
1763			SCTP_INP_RUNLOCK(inp);
1764			continue;
1765		}
1766		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1767			if (laddr->ifa == NULL) {
1768				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1769				    __func__);
1770				continue;
1771			}
1772			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1773			    (void *)laddr->ifa);
1774			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1775				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1776				continue;
1777			}
1778			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1779				/* possible, see if it matches */
1780				switch (nam->sa_family) {
1781#ifdef INET
1782				case AF_INET:
1783					if (sin->sin_addr.s_addr ==
1784					    laddr->ifa->address.sin.sin_addr.s_addr) {
1785						SCTP_INP_RUNLOCK(inp);
1786						return (inp);
1787					}
1788					break;
1789#endif
1790#ifdef INET6
1791				case AF_INET6:
1792					intf_addr6 = &laddr->ifa->address.sin6;
1793					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1794					    intf_addr6)) {
1795						SCTP_INP_RUNLOCK(inp);
1796						return (inp);
1797					}
1798					break;
1799#endif
1800				}
1801			}
1802		}
1803		SCTP_INP_RUNLOCK(inp);
1804	}
1805	return (NULL);
1806}
1807
1808
1809static struct sctp_inpcb *
1810sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1811{
1812	struct sctppcbhead *head;
1813	struct sctp_inpcb *t_inp;
1814	int fnd;
1815
1816	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1817	    SCTP_BASE_INFO(hashmark))];
1818	LIST_FOREACH(t_inp, head, sctp_hash) {
1819		if (t_inp->sctp_lport != lport) {
1820			continue;
1821		}
1822		/* is it in the VRF in question */
1823		fnd = 0;
1824		if (t_inp->def_vrf_id == vrf_id)
1825			fnd = 1;
1826		if (!fnd)
1827			continue;
1828
1829		/* This one is in use. */
1830		/* check the v6/v4 binding issue */
1831		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1832		    SCTP_IPV6_V6ONLY(t_inp)) {
1833			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1834				/* collision in V6 space */
1835				return (t_inp);
1836			} else {
1837				/* inp is BOUND_V4 no conflict */
1838				continue;
1839			}
1840		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1841			/* t_inp is bound v4 and v6, conflict always */
1842			return (t_inp);
1843		} else {
1844			/* t_inp is bound only V4 */
1845			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1846			    SCTP_IPV6_V6ONLY(inp)) {
1847				/* no conflict */
1848				continue;
1849			}
1850			/* else fall through to conflict */
1851		}
1852		return (t_inp);
1853	}
1854	return (NULL);
1855}
1856
1857
1858int
1859sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1860{
1861	/* For 1-2-1 with port reuse */
1862	struct sctppcbhead *head;
1863	struct sctp_inpcb *tinp, *ninp;
1864
1865	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1866		/* only works with port reuse on */
1867		return (-1);
1868	}
1869	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1870		return (0);
1871	}
1872	SCTP_INP_RUNLOCK(inp);
1873	SCTP_INP_INFO_WLOCK();
1874	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1875	    SCTP_BASE_INFO(hashmark))];
1876	/* Kick out all non-listeners to the TCP hash */
1877	LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) {
1878		if (tinp->sctp_lport != inp->sctp_lport) {
1879			continue;
1880		}
1881		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1882			continue;
1883		}
1884		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1885			continue;
1886		}
1887		if (tinp->sctp_socket->so_qlimit) {
1888			continue;
1889		}
1890		SCTP_INP_WLOCK(tinp);
1891		LIST_REMOVE(tinp, sctp_hash);
1892		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1893		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1894		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1895		SCTP_INP_WUNLOCK(tinp);
1896	}
1897	SCTP_INP_WLOCK(inp);
1898	/* Pull from where he was */
1899	LIST_REMOVE(inp, sctp_hash);
1900	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1901	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1902	LIST_INSERT_HEAD(head, inp, sctp_hash);
1903	SCTP_INP_WUNLOCK(inp);
1904	SCTP_INP_RLOCK(inp);
1905	SCTP_INP_INFO_WUNLOCK();
1906	return (0);
1907}
1908
1909
1910struct sctp_inpcb *
1911sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1912    uint32_t vrf_id)
1913{
1914	/*
1915	 * First we check the hash table to see if someone has this port
1916	 * bound with just the port.
1917	 */
1918	struct sctp_inpcb *inp;
1919	struct sctppcbhead *head;
1920	int lport;
1921	unsigned int i;
1922#ifdef INET
1923	struct sockaddr_in *sin;
1924#endif
1925#ifdef INET6
1926	struct sockaddr_in6 *sin6;
1927#endif
1928
1929	switch (nam->sa_family) {
1930#ifdef INET
1931	case AF_INET:
1932		sin = (struct sockaddr_in *)nam;
1933		lport = sin->sin_port;
1934		break;
1935#endif
1936#ifdef INET6
1937	case AF_INET6:
1938		sin6 = (struct sockaddr_in6 *)nam;
1939		lport = sin6->sin6_port;
1940		break;
1941#endif
1942	default:
1943		return (NULL);
1944	}
1945	/*
1946	 * I could cheat here and just cast to one of the types but we will
1947	 * do it right. It also provides the check against an Unsupported
1948	 * type too.
1949	 */
1950	/* Find the head of the ALLADDR chain */
1951	if (have_lock == 0) {
1952		SCTP_INP_INFO_RLOCK();
1953	}
1954	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1955	    SCTP_BASE_INFO(hashmark))];
1956	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1957
1958	/*
1959	 * If the TCP model exists it could be that the main listening
1960	 * endpoint is gone but there still exists a connected socket for
1961	 * this guy. If so we can return the first one that we find. This
1962	 * may NOT be the correct one so the caller should be wary on the
1963	 * returned INP. Currently the only caller that sets find_tcp_pool
1964	 * is in bindx where we are verifying that a user CAN bind the
1965	 * address. He either has bound it already, or someone else has, or
1966	 * its open to bind, so this is good enough.
1967	 */
1968	if (inp == NULL && find_tcp_pool) {
1969		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1970			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1971			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1972			if (inp) {
1973				break;
1974			}
1975		}
1976	}
1977	if (inp) {
1978		SCTP_INP_INCR_REF(inp);
1979	}
1980	if (have_lock == 0) {
1981		SCTP_INP_INFO_RUNLOCK();
1982	}
1983	return (inp);
1984}
1985
1986
1987/*
1988 * Find an association for an endpoint with the pointer to whom you want to
1989 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1990 * need to change the *to to some other struct like a mbuf...
1991 */
1992struct sctp_tcb *
1993sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1994    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1995    uint32_t vrf_id)
1996{
1997	struct sctp_inpcb *inp = NULL;
1998	struct sctp_tcb *stcb;
1999
2000	SCTP_INP_INFO_RLOCK();
2001	if (find_tcp_pool) {
2002		if (inp_p != NULL) {
2003			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
2004			    vrf_id);
2005		} else {
2006			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
2007			    vrf_id);
2008		}
2009		if (stcb != NULL) {
2010			SCTP_INP_INFO_RUNLOCK();
2011			return (stcb);
2012		}
2013	}
2014	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
2015	if (inp_p != NULL) {
2016		*inp_p = inp;
2017	}
2018	SCTP_INP_INFO_RUNLOCK();
2019	if (inp == NULL) {
2020		return (NULL);
2021	}
2022	/*
2023	 * ok, we have an endpoint, now lets find the assoc for it (if any)
2024	 * we now place the source address or from in the to of the find
2025	 * endpoint call. Since in reality this chain is used from the
2026	 * inbound packet side.
2027	 */
2028	if (inp_p != NULL) {
2029		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
2030		    NULL);
2031	} else {
2032		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
2033		    NULL);
2034	}
2035	return (stcb);
2036}
2037
2038
2039/*
2040 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
2041 * find all addresses that the sender has specified in any address list. Each
2042 * address will be used to lookup the TCB and see if one exits.
2043 */
2044static struct sctp_tcb *
2045sctp_findassociation_special_addr(struct mbuf *m, int offset,
2046    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
2047    struct sockaddr *dst)
2048{
2049	struct sctp_paramhdr *phdr, param_buf;
2050#if defined(INET) || defined(INET6)
2051	struct sctp_tcb *stcb;
2052	uint16_t ptype;
2053#endif
2054	uint16_t plen;
2055#ifdef INET
2056	struct sockaddr_in sin4;
2057#endif
2058#ifdef INET6
2059	struct sockaddr_in6 sin6;
2060#endif
2061
2062#ifdef INET
2063	memset(&sin4, 0, sizeof(sin4));
2064	sin4.sin_len = sizeof(sin4);
2065	sin4.sin_family = AF_INET;
2066	sin4.sin_port = sh->src_port;
2067#endif
2068#ifdef INET6
2069	memset(&sin6, 0, sizeof(sin6));
2070	sin6.sin6_len = sizeof(sin6);
2071	sin6.sin6_family = AF_INET6;
2072	sin6.sin6_port = sh->src_port;
2073#endif
2074
2075	offset += sizeof(struct sctp_init_chunk);
2076
2077	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
2078	while (phdr != NULL) {
2079		/* now we must see if we want the parameter */
2080#if defined(INET) || defined(INET6)
2081		ptype = ntohs(phdr->param_type);
2082#endif
2083		plen = ntohs(phdr->param_length);
2084		if (plen == 0) {
2085			break;
2086		}
2087#ifdef INET
2088		if (ptype == SCTP_IPV4_ADDRESS &&
2089		    plen == sizeof(struct sctp_ipv4addr_param)) {
2090			/* Get the rest of the address */
2091			struct sctp_ipv4addr_param ip4_param, *p4;
2092
2093			phdr = sctp_get_next_param(m, offset,
2094			    (struct sctp_paramhdr *)&ip4_param, sizeof(ip4_param));
2095			if (phdr == NULL) {
2096				return (NULL);
2097			}
2098			p4 = (struct sctp_ipv4addr_param *)phdr;
2099			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2100			/* look it up */
2101			stcb = sctp_findassociation_ep_addr(inp_p,
2102			    (struct sockaddr *)&sin4, netp, dst, NULL);
2103			if (stcb != NULL) {
2104				return (stcb);
2105			}
2106		}
2107#endif
2108#ifdef INET6
2109		if (ptype == SCTP_IPV6_ADDRESS &&
2110		    plen == sizeof(struct sctp_ipv6addr_param)) {
2111			/* Get the rest of the address */
2112			struct sctp_ipv6addr_param ip6_param, *p6;
2113
2114			phdr = sctp_get_next_param(m, offset,
2115			    (struct sctp_paramhdr *)&ip6_param, sizeof(ip6_param));
2116			if (phdr == NULL) {
2117				return (NULL);
2118			}
2119			p6 = (struct sctp_ipv6addr_param *)phdr;
2120			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2121			/* look it up */
2122			stcb = sctp_findassociation_ep_addr(inp_p,
2123			    (struct sockaddr *)&sin6, netp, dst, NULL);
2124			if (stcb != NULL) {
2125				return (stcb);
2126			}
2127		}
2128#endif
2129		offset += SCTP_SIZE32(plen);
2130		phdr = sctp_get_next_param(m, offset, &param_buf,
2131		    sizeof(param_buf));
2132	}
2133	return (NULL);
2134}
2135
2136static struct sctp_tcb *
2137sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2138    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2139    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2140{
2141	/*
2142	 * Use my vtag to hash. If we find it we then verify the source addr
2143	 * is in the assoc. If all goes well we save a bit on rec of a
2144	 * packet.
2145	 */
2146	struct sctpasochead *head;
2147	struct sctp_nets *net;
2148	struct sctp_tcb *stcb;
2149
2150	SCTP_INP_INFO_RLOCK();
2151	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2152	    SCTP_BASE_INFO(hashasocmark))];
2153	LIST_FOREACH(stcb, head, sctp_asocs) {
2154		SCTP_INP_RLOCK(stcb->sctp_ep);
2155		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2156			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2157			continue;
2158		}
2159		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2160			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2161			continue;
2162		}
2163		SCTP_TCB_LOCK(stcb);
2164		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2165		if (stcb->asoc.my_vtag == vtag) {
2166			/* candidate */
2167			if (stcb->rport != rport) {
2168				SCTP_TCB_UNLOCK(stcb);
2169				continue;
2170			}
2171			if (stcb->sctp_ep->sctp_lport != lport) {
2172				SCTP_TCB_UNLOCK(stcb);
2173				continue;
2174			}
2175			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2176				SCTP_TCB_UNLOCK(stcb);
2177				continue;
2178			}
2179			/* RRS:Need toaddr check here */
2180			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2181				/* Endpoint does not own this address */
2182				SCTP_TCB_UNLOCK(stcb);
2183				continue;
2184			}
2185			if (remote_tag) {
2186				/*
2187				 * If we have both vtags that's all we match
2188				 * on
2189				 */
2190				if (stcb->asoc.peer_vtag == remote_tag) {
2191					/*
2192					 * If both tags match we consider it
2193					 * conclusive and check NO
2194					 * source/destination addresses
2195					 */
2196					goto conclusive;
2197				}
2198			}
2199			if (skip_src_check) {
2200		conclusive:
2201				if (from) {
2202					*netp = sctp_findnet(stcb, from);
2203				} else {
2204					*netp = NULL;	/* unknown */
2205				}
2206				if (inp_p)
2207					*inp_p = stcb->sctp_ep;
2208				SCTP_INP_INFO_RUNLOCK();
2209				return (stcb);
2210			}
2211			net = sctp_findnet(stcb, from);
2212			if (net) {
2213				/* yep its him. */
2214				*netp = net;
2215				SCTP_STAT_INCR(sctps_vtagexpress);
2216				*inp_p = stcb->sctp_ep;
2217				SCTP_INP_INFO_RUNLOCK();
2218				return (stcb);
2219			} else {
2220				/*
2221				 * not him, this should only happen in rare
2222				 * cases so I peg it.
2223				 */
2224				SCTP_STAT_INCR(sctps_vtagbogus);
2225			}
2226		}
2227		SCTP_TCB_UNLOCK(stcb);
2228	}
2229	SCTP_INP_INFO_RUNLOCK();
2230	return (NULL);
2231}
2232
2233
2234/*
2235 * Find an association with the pointer to the inbound IP packet. This can be
2236 * a IPv4 or IPv6 packet.
2237 */
2238struct sctp_tcb *
2239sctp_findassociation_addr(struct mbuf *m, int offset,
2240    struct sockaddr *src, struct sockaddr *dst,
2241    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2242    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2243{
2244	struct sctp_tcb *stcb;
2245	struct sctp_inpcb *inp;
2246
2247	if (sh->v_tag) {
2248		/* we only go down this path if vtag is non-zero */
2249		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2250		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2251		if (stcb) {
2252			return (stcb);
2253		}
2254	}
2255	if (inp_p) {
2256		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2257		    1, vrf_id);
2258		inp = *inp_p;
2259	} else {
2260		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2261		    1, vrf_id);
2262	}
2263	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2264	if (stcb == NULL && inp) {
2265		/* Found a EP but not this address */
2266		if ((ch->chunk_type == SCTP_INITIATION) ||
2267		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2268			/*-
2269			 * special hook, we do NOT return linp or an
2270			 * association that is linked to an existing
2271			 * association that is under the TCP pool (i.e. no
2272			 * listener exists). The endpoint finding routine
2273			 * will always find a listener before examining the
2274			 * TCP pool.
2275			 */
2276			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2277				if (inp_p) {
2278					*inp_p = NULL;
2279				}
2280				return (NULL);
2281			}
2282			stcb = sctp_findassociation_special_addr(m,
2283			    offset, sh, &inp, netp, dst);
2284			if (inp_p != NULL) {
2285				*inp_p = inp;
2286			}
2287		}
2288	}
2289	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2290	return (stcb);
2291}
2292
2293/*
2294 * lookup an association by an ASCONF lookup address.
2295 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2296 */
2297struct sctp_tcb *
2298sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2299    struct sockaddr *dst, struct sctphdr *sh,
2300    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2301{
2302	struct sctp_tcb *stcb;
2303	union sctp_sockstore remote_store;
2304	struct sctp_paramhdr param_buf, *phdr;
2305	int ptype;
2306	int zero_address = 0;
2307#ifdef INET
2308	struct sockaddr_in *sin;
2309#endif
2310#ifdef INET6
2311	struct sockaddr_in6 *sin6;
2312#endif
2313
2314	memset(&remote_store, 0, sizeof(remote_store));
2315	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2316	    &param_buf, sizeof(struct sctp_paramhdr));
2317	if (phdr == NULL) {
2318		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2319		    __func__);
2320		return NULL;
2321	}
2322	ptype = (int)((uint32_t)ntohs(phdr->param_type));
2323	/* get the correlation address */
2324	switch (ptype) {
2325#ifdef INET6
2326	case SCTP_IPV6_ADDRESS:
2327		{
2328			/* ipv6 address param */
2329			struct sctp_ipv6addr_param *p6, p6_buf;
2330
2331			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2332				return NULL;
2333			}
2334			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2335			    offset + sizeof(struct sctp_asconf_chunk),
2336			    &p6_buf.ph, sizeof(p6_buf));
2337			if (p6 == NULL) {
2338				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2339				    __func__);
2340				return (NULL);
2341			}
2342			sin6 = &remote_store.sin6;
2343			sin6->sin6_family = AF_INET6;
2344			sin6->sin6_len = sizeof(*sin6);
2345			sin6->sin6_port = sh->src_port;
2346			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2347			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2348				zero_address = 1;
2349			break;
2350		}
2351#endif
2352#ifdef INET
2353	case SCTP_IPV4_ADDRESS:
2354		{
2355			/* ipv4 address param */
2356			struct sctp_ipv4addr_param *p4, p4_buf;
2357
2358			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2359				return NULL;
2360			}
2361			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2362			    offset + sizeof(struct sctp_asconf_chunk),
2363			    &p4_buf.ph, sizeof(p4_buf));
2364			if (p4 == NULL) {
2365				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2366				    __func__);
2367				return (NULL);
2368			}
2369			sin = &remote_store.sin;
2370			sin->sin_family = AF_INET;
2371			sin->sin_len = sizeof(*sin);
2372			sin->sin_port = sh->src_port;
2373			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2374			if (sin->sin_addr.s_addr == INADDR_ANY)
2375				zero_address = 1;
2376			break;
2377		}
2378#endif
2379	default:
2380		/* invalid address param type */
2381		return NULL;
2382	}
2383
2384	if (zero_address) {
2385		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2386		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2387		if (stcb != NULL) {
2388			SCTP_INP_DECR_REF(*inp_p);
2389		}
2390	} else {
2391		stcb = sctp_findassociation_ep_addr(inp_p,
2392		    &remote_store.sa, netp,
2393		    dst, NULL);
2394	}
2395	return (stcb);
2396}
2397
2398
2399/*
2400 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2401 * addresses. This way if we don't get a bind we by default pick a ephemeral
2402 * port with all addresses bound.
2403 */
2404int
2405sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2406{
2407	/*
2408	 * we get called when a new endpoint starts up. We need to allocate
2409	 * the sctp_inpcb structure from the zone and init it. Mark it as
2410	 * unbound and find a port that we can use as an ephemeral with
2411	 * INADDR_ANY. If the user binds later no problem we can then add in
2412	 * the specific addresses. And setup the default parameters for the
2413	 * EP.
2414	 */
2415	int i, error;
2416	struct sctp_inpcb *inp;
2417	struct sctp_pcb *m;
2418	struct timeval time;
2419	sctp_sharedkey_t *null_key;
2420
2421	error = 0;
2422
2423	SCTP_INP_INFO_WLOCK();
2424	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2425	if (inp == NULL) {
2426		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2427		SCTP_INP_INFO_WUNLOCK();
2428		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2429		return (ENOBUFS);
2430	}
2431	/* zap it */
2432	memset(inp, 0, sizeof(*inp));
2433
2434	/* bump generations */
2435	/* setup socket pointers */
2436	inp->sctp_socket = so;
2437	inp->ip_inp.inp.inp_socket = so;
2438	inp->ip_inp.inp.inp_cred = crhold(so->so_cred);
2439#ifdef INET6
2440	if (INP_SOCKAF(so) == AF_INET6) {
2441		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2442			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2443		}
2444		if (MODULE_GLOBAL(ip6_v6only)) {
2445			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2446		}
2447	}
2448#endif
2449	inp->sctp_associd_counter = 1;
2450	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2451	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2452	inp->max_cwnd = 0;
2453	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2454	inp->ecn_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_ecn_enable);
2455	inp->prsctp_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pr_enable);
2456	inp->auth_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_auth_enable);
2457	inp->asconf_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_asconf_enable);
2458	inp->reconfig_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_reconfig_enable);
2459	inp->nrsack_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_nrsack_enable);
2460	inp->pktdrop_supported = (uint8_t)SCTP_BASE_SYSCTL(sctp_pktdrop_enable);
2461	inp->idata_supported = 0;
2462
2463	inp->fibnum = so->so_fibnum;
2464	/* init the small hash table we use to track asocid <-> tcb */
2465	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2466	if (inp->sctp_asocidhash == NULL) {
2467		crfree(inp->ip_inp.inp.inp_cred);
2468		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2469		SCTP_INP_INFO_WUNLOCK();
2470		return (ENOBUFS);
2471	}
2472	SCTP_INCR_EP_COUNT();
2473	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2474	SCTP_INP_INFO_WUNLOCK();
2475
2476	so->so_pcb = (caddr_t)inp;
2477
2478	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2479		/* UDP style socket */
2480		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2481		    SCTP_PCB_FLAGS_UNBOUND);
2482		/* Be sure it is NON-BLOCKING IO for UDP */
2483		/* SCTP_SET_SO_NBIO(so); */
2484	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2485		/* TCP style socket */
2486		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2487		    SCTP_PCB_FLAGS_UNBOUND);
2488		/* Be sure we have blocking IO by default */
2489		SCTP_CLEAR_SO_NBIO(so);
2490	} else {
2491		/*
2492		 * unsupported socket type (RAW, etc)- in case we missed it
2493		 * in protosw
2494		 */
2495		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2496		so->so_pcb = NULL;
2497		crfree(inp->ip_inp.inp.inp_cred);
2498		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2499		return (EOPNOTSUPP);
2500	}
2501	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2502		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2503		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2504	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2505		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2506		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2507	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2508		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2509		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2510	}
2511	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2512	    &inp->sctp_hashmark);
2513	if (inp->sctp_tcbhash == NULL) {
2514		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2515		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2516		so->so_pcb = NULL;
2517		crfree(inp->ip_inp.inp.inp_cred);
2518		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2519		return (ENOBUFS);
2520	}
2521	inp->def_vrf_id = vrf_id;
2522
2523	SCTP_INP_INFO_WLOCK();
2524	SCTP_INP_LOCK_INIT(inp);
2525	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2526	SCTP_INP_READ_INIT(inp);
2527	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2528	/* lock the new ep */
2529	SCTP_INP_WLOCK(inp);
2530
2531	/* add it to the info area */
2532	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2533	SCTP_INP_INFO_WUNLOCK();
2534
2535	TAILQ_INIT(&inp->read_queue);
2536	LIST_INIT(&inp->sctp_addr_list);
2537
2538	LIST_INIT(&inp->sctp_asoc_list);
2539
2540#ifdef SCTP_TRACK_FREED_ASOCS
2541	/* TEMP CODE */
2542	LIST_INIT(&inp->sctp_asoc_free_list);
2543#endif
2544	/* Init the timer structure for signature change */
2545	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2546	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2547
2548	/* now init the actual endpoint default data */
2549	m = &inp->sctp_ep;
2550
2551	/* setup the base timeout information */
2552	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2553	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2554	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2555	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2556	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2557	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2558	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2559	/* all max/min max are in ms */
2560	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2561	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2562	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2563	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2564	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2565	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2566	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2567	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2568	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2569	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2570	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2571	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2572	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2573
2574	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2575	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2576	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2577	/* number of streams to pre-open on a association */
2578	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2579
2580	/* Add adaptation cookie */
2581	m->adaptation_layer_indicator = 0;
2582	m->adaptation_layer_indicator_provided = 0;
2583
2584	/* seed random number generator */
2585	m->random_counter = 1;
2586	m->store_at = SCTP_SIGNATURE_SIZE;
2587	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2588	sctp_fill_random_store(m);
2589
2590	/* Minimum cookie size */
2591	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2592	    sizeof(struct sctp_state_cookie);
2593	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2594
2595	/* Setup the initial secret */
2596	(void)SCTP_GETTIME_TIMEVAL(&time);
2597	m->time_of_secret_change = time.tv_sec;
2598
2599	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2600		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2601	}
2602	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2603
2604	/* How long is a cookie good for ? */
2605	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2606	/*
2607	 * Initialize authentication parameters
2608	 */
2609	m->local_hmacs = sctp_default_supported_hmaclist();
2610	m->local_auth_chunks = sctp_alloc_chunklist();
2611	if (inp->asconf_supported) {
2612		sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks);
2613		sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks);
2614	}
2615	m->default_dscp = 0;
2616#ifdef INET6
2617	m->default_flowlabel = 0;
2618#endif
2619	m->port = 0;		/* encapsulation disabled by default */
2620	LIST_INIT(&m->shared_keys);
2621	/* add default NULL key as key id 0 */
2622	null_key = sctp_alloc_sharedkey();
2623	sctp_insert_sharedkey(&m->shared_keys, null_key);
2624	SCTP_INP_WUNLOCK(inp);
2625#ifdef SCTP_LOG_CLOSING
2626	sctp_log_closing(inp, NULL, 12);
2627#endif
2628	return (error);
2629}
2630
2631
2632void
2633sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2634    struct sctp_tcb *stcb)
2635{
2636	struct sctp_nets *net;
2637	uint16_t lport, rport;
2638	struct sctppcbhead *head;
2639	struct sctp_laddr *laddr, *oladdr;
2640
2641	atomic_add_int(&stcb->asoc.refcnt, 1);
2642	SCTP_TCB_UNLOCK(stcb);
2643	SCTP_INP_INFO_WLOCK();
2644	SCTP_INP_WLOCK(old_inp);
2645	SCTP_INP_WLOCK(new_inp);
2646	SCTP_TCB_LOCK(stcb);
2647	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2648
2649	new_inp->sctp_ep.time_of_secret_change =
2650	    old_inp->sctp_ep.time_of_secret_change;
2651	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2652	    sizeof(old_inp->sctp_ep.secret_key));
2653	new_inp->sctp_ep.current_secret_number =
2654	    old_inp->sctp_ep.current_secret_number;
2655	new_inp->sctp_ep.last_secret_number =
2656	    old_inp->sctp_ep.last_secret_number;
2657	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2658
2659	/* make it so new data pours into the new socket */
2660	stcb->sctp_socket = new_inp->sctp_socket;
2661	stcb->sctp_ep = new_inp;
2662
2663	/* Copy the port across */
2664	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2665	rport = stcb->rport;
2666	/* Pull the tcb from the old association */
2667	LIST_REMOVE(stcb, sctp_tcbhash);
2668	LIST_REMOVE(stcb, sctp_tcblist);
2669	if (stcb->asoc.in_asocid_hash) {
2670		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2671	}
2672	/* Now insert the new_inp into the TCP connected hash */
2673	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2674
2675	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2676	/* Its safe to access */
2677	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2678
2679	/* Now move the tcb into the endpoint list */
2680	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2681	/*
2682	 * Question, do we even need to worry about the ep-hash since we
2683	 * only have one connection? Probably not :> so lets get rid of it
2684	 * and not suck up any kernel memory in that.
2685	 */
2686	if (stcb->asoc.in_asocid_hash) {
2687		struct sctpasochead *lhd;
2688
2689		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2690		    new_inp->hashasocidmark)];
2691		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2692	}
2693	/* Ok. Let's restart timer. */
2694	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2695		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2696		    stcb, net);
2697	}
2698
2699	SCTP_INP_INFO_WUNLOCK();
2700	if (new_inp->sctp_tcbhash != NULL) {
2701		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2702		new_inp->sctp_tcbhash = NULL;
2703	}
2704	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2705		/* Subset bound, so copy in the laddr list from the old_inp */
2706		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2707			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2708			if (laddr == NULL) {
2709				/*
2710				 * Gak, what can we do? This assoc is really
2711				 * HOSED. We probably should send an abort
2712				 * here.
2713				 */
2714				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2715				continue;
2716			}
2717			SCTP_INCR_LADDR_COUNT();
2718			memset(laddr, 0, sizeof(*laddr));
2719			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2720			laddr->ifa = oladdr->ifa;
2721			atomic_add_int(&laddr->ifa->refcount, 1);
2722			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2723			    sctp_nxt_addr);
2724			new_inp->laddr_count++;
2725			if (oladdr == stcb->asoc.last_used_address) {
2726				stcb->asoc.last_used_address = laddr;
2727			}
2728		}
2729	}
2730	/*
2731	 * Now any running timers need to be adjusted since we really don't
2732	 * care if they are running or not just blast in the new_inp into
2733	 * all of them.
2734	 */
2735
2736	stcb->asoc.dack_timer.ep = (void *)new_inp;
2737	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2738	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2739	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2740	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2741	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2742	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2743	/* now what about the nets? */
2744	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2745		net->pmtu_timer.ep = (void *)new_inp;
2746		net->hb_timer.ep = (void *)new_inp;
2747		net->rxt_timer.ep = (void *)new_inp;
2748	}
2749	SCTP_INP_WUNLOCK(new_inp);
2750	SCTP_INP_WUNLOCK(old_inp);
2751}
2752
2753/*
2754 * insert an laddr entry with the given ifa for the desired list
2755 */
2756static int
2757sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
2758{
2759	struct sctp_laddr *laddr;
2760
2761	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2762	if (laddr == NULL) {
2763		/* out of memory? */
2764		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2765		return (EINVAL);
2766	}
2767	SCTP_INCR_LADDR_COUNT();
2768	memset(laddr, 0, sizeof(*laddr));
2769	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2770	laddr->ifa = ifa;
2771	laddr->action = act;
2772	atomic_add_int(&ifa->refcount, 1);
2773	/* insert it */
2774	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
2775
2776	return (0);
2777}
2778
2779/*
2780 * Remove an laddr entry from the local address list (on an assoc)
2781 */
2782static void
2783sctp_remove_laddr(struct sctp_laddr *laddr)
2784{
2785
2786	/* remove from the list */
2787	LIST_REMOVE(laddr, sctp_nxt_addr);
2788	sctp_free_ifa(laddr->ifa);
2789	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
2790	SCTP_DECR_LADDR_COUNT();
2791}
2792
2793
2794
2795/* sctp_ifap is used to bypass normal local address validation checks */
2796int
2797sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2798    struct sctp_ifa *sctp_ifap, struct thread *p)
2799{
2800	/* bind a ep to a socket address */
2801	struct sctppcbhead *head;
2802	struct sctp_inpcb *inp, *inp_tmp;
2803	struct inpcb *ip_inp;
2804	int port_reuse_active = 0;
2805	int bindall;
2806	uint16_t lport;
2807	int error;
2808	uint32_t vrf_id;
2809
2810	lport = 0;
2811	bindall = 1;
2812	inp = (struct sctp_inpcb *)so->so_pcb;
2813	ip_inp = (struct inpcb *)so->so_pcb;
2814#ifdef SCTP_DEBUG
2815	if (addr) {
2816		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2817		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2818		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2819		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2820	}
2821#endif
2822	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2823		/* already did a bind, subsequent binds NOT allowed ! */
2824		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2825		return (EINVAL);
2826	}
2827#ifdef INVARIANTS
2828	if (p == NULL)
2829		panic("null proc/thread");
2830#endif
2831	if (addr != NULL) {
2832		switch (addr->sa_family) {
2833#ifdef INET
2834		case AF_INET:
2835			{
2836				struct sockaddr_in *sin;
2837
2838				/* IPV6_V6ONLY socket? */
2839				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2840					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2841					return (EINVAL);
2842				}
2843				if (addr->sa_len != sizeof(*sin)) {
2844					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2845					return (EINVAL);
2846				}
2847				sin = (struct sockaddr_in *)addr;
2848				lport = sin->sin_port;
2849				/*
2850				 * For LOOPBACK the prison_local_ip4() call
2851				 * will transmute the ip address to the
2852				 * proper value.
2853				 */
2854				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2855					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2856					return (error);
2857				}
2858				if (sin->sin_addr.s_addr != INADDR_ANY) {
2859					bindall = 0;
2860				}
2861				break;
2862			}
2863#endif
2864#ifdef INET6
2865		case AF_INET6:
2866			{
2867				/*
2868				 * Only for pure IPv6 Address. (No IPv4
2869				 * Mapped!)
2870				 */
2871				struct sockaddr_in6 *sin6;
2872
2873				sin6 = (struct sockaddr_in6 *)addr;
2874
2875				if (addr->sa_len != sizeof(*sin6)) {
2876					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2877					return (EINVAL);
2878				}
2879				lport = sin6->sin6_port;
2880				/*
2881				 * For LOOPBACK the prison_local_ip6() call
2882				 * will transmute the ipv6 address to the
2883				 * proper value.
2884				 */
2885				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2886				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2887					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2888					return (error);
2889				}
2890				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2891					bindall = 0;
2892					/* KAME hack: embed scopeid */
2893					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2894						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2895						return (EINVAL);
2896					}
2897				}
2898				/* this must be cleared for ifa_ifwithaddr() */
2899				sin6->sin6_scope_id = 0;
2900				break;
2901			}
2902#endif
2903		default:
2904			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2905			return (EAFNOSUPPORT);
2906		}
2907	}
2908	SCTP_INP_INFO_WLOCK();
2909	SCTP_INP_WLOCK(inp);
2910	/* Setup a vrf_id to be the default for the non-bind-all case. */
2911	vrf_id = inp->def_vrf_id;
2912
2913	/* increase our count due to the unlock we do */
2914	SCTP_INP_INCR_REF(inp);
2915	if (lport) {
2916		/*
2917		 * Did the caller specify a port? if so we must see if an ep
2918		 * already has this one bound.
2919		 */
2920		/* got to be root to get at low ports */
2921		if (ntohs(lport) < IPPORT_RESERVED) {
2922			if (p && (error =
2923			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2924			    )) {
2925				SCTP_INP_DECR_REF(inp);
2926				SCTP_INP_WUNLOCK(inp);
2927				SCTP_INP_INFO_WUNLOCK();
2928				return (error);
2929			}
2930		}
2931		SCTP_INP_WUNLOCK(inp);
2932		if (bindall) {
2933			vrf_id = inp->def_vrf_id;
2934			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2935			if (inp_tmp != NULL) {
2936				/*
2937				 * lock guy returned and lower count note
2938				 * that we are not bound so inp_tmp should
2939				 * NEVER be inp. And it is this inp
2940				 * (inp_tmp) that gets the reference bump,
2941				 * so we must lower it.
2942				 */
2943				SCTP_INP_DECR_REF(inp_tmp);
2944				/* unlock info */
2945				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2946				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2947					/*
2948					 * Ok, must be one-2-one and
2949					 * allowing port re-use
2950					 */
2951					port_reuse_active = 1;
2952					goto continue_anyway;
2953				}
2954				SCTP_INP_DECR_REF(inp);
2955				SCTP_INP_INFO_WUNLOCK();
2956				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2957				return (EADDRINUSE);
2958			}
2959		} else {
2960			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2961			if (inp_tmp != NULL) {
2962				/*
2963				 * lock guy returned and lower count note
2964				 * that we are not bound so inp_tmp should
2965				 * NEVER be inp. And it is this inp
2966				 * (inp_tmp) that gets the reference bump,
2967				 * so we must lower it.
2968				 */
2969				SCTP_INP_DECR_REF(inp_tmp);
2970				/* unlock info */
2971				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2972				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2973					/*
2974					 * Ok, must be one-2-one and
2975					 * allowing port re-use
2976					 */
2977					port_reuse_active = 1;
2978					goto continue_anyway;
2979				}
2980				SCTP_INP_DECR_REF(inp);
2981				SCTP_INP_INFO_WUNLOCK();
2982				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2983				return (EADDRINUSE);
2984			}
2985		}
2986continue_anyway:
2987		SCTP_INP_WLOCK(inp);
2988		if (bindall) {
2989			/* verify that no lport is not used by a singleton */
2990			if ((port_reuse_active == 0) &&
2991			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2992				/* Sorry someone already has this one bound */
2993				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2994				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2995					port_reuse_active = 1;
2996				} else {
2997					SCTP_INP_DECR_REF(inp);
2998					SCTP_INP_WUNLOCK(inp);
2999					SCTP_INP_INFO_WUNLOCK();
3000					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3001					return (EADDRINUSE);
3002				}
3003			}
3004		}
3005	} else {
3006		uint16_t first, last, candidate;
3007		uint16_t count;
3008		int done;
3009
3010		if (ip_inp->inp_flags & INP_HIGHPORT) {
3011			first = MODULE_GLOBAL(ipport_hifirstauto);
3012			last = MODULE_GLOBAL(ipport_hilastauto);
3013		} else if (ip_inp->inp_flags & INP_LOWPORT) {
3014			if (p && (error =
3015			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
3016			    )) {
3017				SCTP_INP_DECR_REF(inp);
3018				SCTP_INP_WUNLOCK(inp);
3019				SCTP_INP_INFO_WUNLOCK();
3020				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
3021				return (error);
3022			}
3023			first = MODULE_GLOBAL(ipport_lowfirstauto);
3024			last = MODULE_GLOBAL(ipport_lowlastauto);
3025		} else {
3026			first = MODULE_GLOBAL(ipport_firstauto);
3027			last = MODULE_GLOBAL(ipport_lastauto);
3028		}
3029		if (first > last) {
3030			uint16_t temp;
3031
3032			temp = first;
3033			first = last;
3034			last = temp;
3035		}
3036		count = last - first + 1;	/* number of candidates */
3037		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
3038
3039		done = 0;
3040		while (!done) {
3041			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
3042				done = 1;
3043			}
3044			if (!done) {
3045				if (--count == 0) {
3046					SCTP_INP_DECR_REF(inp);
3047					SCTP_INP_WUNLOCK(inp);
3048					SCTP_INP_INFO_WUNLOCK();
3049					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
3050					return (EADDRINUSE);
3051				}
3052				if (candidate == last)
3053					candidate = first;
3054				else
3055					candidate = candidate + 1;
3056			}
3057		}
3058		lport = htons(candidate);
3059	}
3060	SCTP_INP_DECR_REF(inp);
3061	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
3062	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3063		/*
3064		 * this really should not happen. The guy did a non-blocking
3065		 * bind and then did a close at the same time.
3066		 */
3067		SCTP_INP_WUNLOCK(inp);
3068		SCTP_INP_INFO_WUNLOCK();
3069		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3070		return (EINVAL);
3071	}
3072	/* ok we look clear to give out this port, so lets setup the binding */
3073	if (bindall) {
3074		/* binding to all addresses, so just set in the proper flags */
3075		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3076		/* set the automatic addr changes from kernel flag */
3077		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3078			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3079			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3080		} else {
3081			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3082			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3083		}
3084		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3085			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3086		} else {
3087			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3088		}
3089		/*
3090		 * set the automatic mobility_base from kernel flag (by
3091		 * micchie)
3092		 */
3093		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3094			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3095			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3096		} else {
3097			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3098			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3099		}
3100		/*
3101		 * set the automatic mobility_fasthandoff from kernel flag
3102		 * (by micchie)
3103		 */
3104		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3105			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3106			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3107		} else {
3108			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3109			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3110		}
3111	} else {
3112		/*
3113		 * bind specific, make sure flags is off and add a new
3114		 * address structure to the sctp_addr_list inside the ep
3115		 * structure.
3116		 *
3117		 * We will need to allocate one and insert it at the head.
3118		 * The socketopt call can just insert new addresses in there
3119		 * as well. It will also have to do the embed scope kame
3120		 * hack too (before adding).
3121		 */
3122		struct sctp_ifa *ifa;
3123		union sctp_sockstore store;
3124
3125		memset(&store, 0, sizeof(store));
3126		switch (addr->sa_family) {
3127#ifdef INET
3128		case AF_INET:
3129			memcpy(&store.sin, addr, sizeof(struct sockaddr_in));
3130			store.sin.sin_port = 0;
3131			break;
3132#endif
3133#ifdef INET6
3134		case AF_INET6:
3135			memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6));
3136			store.sin6.sin6_port = 0;
3137			break;
3138#endif
3139		default:
3140			break;
3141		}
3142		/*
3143		 * first find the interface with the bound address need to
3144		 * zero out the port to find the address! yuck! can't do
3145		 * this earlier since need port for sctp_pcb_findep()
3146		 */
3147		if (sctp_ifap != NULL) {
3148			ifa = sctp_ifap;
3149		} else {
3150			/*
3151			 * Note for BSD we hit here always other O/S's will
3152			 * pass things in via the sctp_ifap argument
3153			 * (Panda).
3154			 */
3155			ifa = sctp_find_ifa_by_addr(&store.sa,
3156			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3157		}
3158		if (ifa == NULL) {
3159			/* Can't find an interface with that address */
3160			SCTP_INP_WUNLOCK(inp);
3161			SCTP_INP_INFO_WUNLOCK();
3162			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3163			return (EADDRNOTAVAIL);
3164		}
3165#ifdef INET6
3166		if (addr->sa_family == AF_INET6) {
3167			/* GAK, more FIXME IFA lock? */
3168			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3169				/* Can't bind a non-existent addr. */
3170				SCTP_INP_WUNLOCK(inp);
3171				SCTP_INP_INFO_WUNLOCK();
3172				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3173				return (EINVAL);
3174			}
3175		}
3176#endif
3177		/* we're not bound all */
3178		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3179		/* allow bindx() to send ASCONF's for binding changes */
3180		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3181		/* clear automatic addr changes from kernel flag */
3182		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3183
3184		/* add this address to the endpoint list */
3185		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3186		if (error != 0) {
3187			SCTP_INP_WUNLOCK(inp);
3188			SCTP_INP_INFO_WUNLOCK();
3189			return (error);
3190		}
3191		inp->laddr_count++;
3192	}
3193	/* find the bucket */
3194	if (port_reuse_active) {
3195		/* Put it into tcp 1-2-1 hash */
3196		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3197		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3198	} else {
3199		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3200	}
3201	/* put it in the bucket */
3202	LIST_INSERT_HEAD(head, inp, sctp_hash);
3203	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3204	    (void *)head, ntohs(lport), port_reuse_active);
3205	/* set in the port */
3206	inp->sctp_lport = lport;
3207
3208	/* turn off just the unbound flag */
3209	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3210	SCTP_INP_WUNLOCK(inp);
3211	SCTP_INP_INFO_WUNLOCK();
3212	return (0);
3213}
3214
3215
3216static void
3217sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3218{
3219	struct sctp_iterator *it, *nit;
3220
3221	/*
3222	 * We enter with the only the ITERATOR_LOCK in place and a write
3223	 * lock on the inp_info stuff.
3224	 */
3225	it = sctp_it_ctl.cur_it;
3226	if (it && (it->vn != curvnet)) {
3227		/* Its not looking at our VNET */
3228		return;
3229	}
3230	if (it && (it->inp == inp)) {
3231		/*
3232		 * This is tricky and we hold the iterator lock, but when it
3233		 * returns and gets the lock (when we release it) the
3234		 * iterator will try to operate on inp. We need to stop that
3235		 * from happening. But of course the iterator has a
3236		 * reference on the stcb and inp. We can mark it and it will
3237		 * stop.
3238		 *
3239		 * If its a single iterator situation, we set the end
3240		 * iterator flag. Otherwise we set the iterator to go to the
3241		 * next inp.
3242		 *
3243		 */
3244		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3245			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3246		} else {
3247			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3248		}
3249	}
3250	/*
3251	 * Now go through and remove any single reference to our inp that
3252	 * may be still pending on the list
3253	 */
3254	SCTP_IPI_ITERATOR_WQ_LOCK();
3255	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3256		if (it->vn != curvnet) {
3257			continue;
3258		}
3259		if (it->inp == inp) {
3260			/* This one points to me is it inp specific? */
3261			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3262				/* Remove and free this one */
3263				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3264				    it, sctp_nxt_itr);
3265				if (it->function_atend != NULL) {
3266					(*it->function_atend) (it->pointer, it->val);
3267				}
3268				SCTP_FREE(it, SCTP_M_ITER);
3269			} else {
3270				it->inp = LIST_NEXT(it->inp, sctp_list);
3271				if (it->inp) {
3272					SCTP_INP_INCR_REF(it->inp);
3273				}
3274			}
3275			/*
3276			 * When its put in the refcnt is incremented so decr
3277			 * it
3278			 */
3279			SCTP_INP_DECR_REF(inp);
3280		}
3281	}
3282	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3283}
3284
3285/* release sctp_inpcb unbind the port */
3286void
3287sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3288{
3289	/*
3290	 * Here we free a endpoint. We must find it (if it is in the Hash
3291	 * table) and remove it from there. Then we must also find it in the
3292	 * overall list and remove it from there. After all removals are
3293	 * complete then any timer has to be stopped. Then start the actual
3294	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3295	 * all associations. d) finally the ep itself.
3296	 */
3297	struct sctp_tcb *asoc, *nasoc;
3298	struct sctp_laddr *laddr, *nladdr;
3299	struct inpcb *ip_pcb;
3300	struct socket *so;
3301	int being_refed = 0;
3302	struct sctp_queued_to_read *sq, *nsq;
3303	int cnt;
3304	sctp_sharedkey_t *shared_key, *nshared_key;
3305
3306
3307#ifdef SCTP_LOG_CLOSING
3308	sctp_log_closing(inp, NULL, 0);
3309#endif
3310	SCTP_ITERATOR_LOCK();
3311	/* mark any iterators on the list or being processed */
3312	sctp_iterator_inp_being_freed(inp);
3313	SCTP_ITERATOR_UNLOCK();
3314	so = inp->sctp_socket;
3315	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3316		/* been here before.. eeks.. get out of here */
3317		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3318#ifdef SCTP_LOG_CLOSING
3319		sctp_log_closing(inp, NULL, 1);
3320#endif
3321		return;
3322	}
3323	SCTP_ASOC_CREATE_LOCK(inp);
3324	SCTP_INP_INFO_WLOCK();
3325
3326	SCTP_INP_WLOCK(inp);
3327	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3328		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3329		/* socket is gone, so no more wakeups allowed */
3330		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3331		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3332		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3333
3334	}
3335	/* First time through we have the socket lock, after that no more. */
3336	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3337	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3338
3339	if (inp->control) {
3340		sctp_m_freem(inp->control);
3341		inp->control = NULL;
3342	}
3343	if (inp->pkt) {
3344		sctp_m_freem(inp->pkt);
3345		inp->pkt = NULL;
3346	}
3347	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3348					 * here but I will be nice :> (i.e.
3349					 * ip_pcb = ep;) */
3350	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3351		int cnt_in_sd;
3352
3353		cnt_in_sd = 0;
3354		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3355			SCTP_TCB_LOCK(asoc);
3356			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3357				/* Skip guys being freed */
3358				cnt_in_sd++;
3359				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3360					/*
3361					 * Special case - we did not start a
3362					 * kill timer on the asoc due to it
3363					 * was not closed. So go ahead and
3364					 * start it now.
3365					 */
3366					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3367					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3368				}
3369				SCTP_TCB_UNLOCK(asoc);
3370				continue;
3371			}
3372			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3373			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3374			    (asoc->asoc.total_output_queue_size == 0)) {
3375				/*
3376				 * If we have data in queue, we don't want
3377				 * to just free since the app may have done,
3378				 * send()/close or connect/send/close. And
3379				 * it wants the data to get across first.
3380				 */
3381				/* Just abandon things in the front states */
3382				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3383				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3384					cnt_in_sd++;
3385				}
3386				continue;
3387			}
3388			/* Disconnect the socket please */
3389			asoc->sctp_socket = NULL;
3390			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3391			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3392			    (asoc->asoc.control_pdapi) ||
3393			    (asoc->asoc.size_on_all_streams > 0) ||
3394			    (so && (so->so_rcv.sb_cc > 0))) {
3395				/* Left with Data unread */
3396				struct mbuf *op_err;
3397
3398				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3399				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3400				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3401				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3402				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3403				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3404					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3405				}
3406				if (sctp_free_assoc(inp, asoc,
3407				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3408					cnt_in_sd++;
3409				}
3410				continue;
3411			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3412				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3413			    (asoc->asoc.stream_queue_cnt == 0)) {
3414				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3415					goto abort_anyway;
3416				}
3417				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3418				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3419					struct sctp_nets *netp;
3420
3421					/*
3422					 * there is nothing queued to send,
3423					 * so I send shutdown
3424					 */
3425					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3426					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3427						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3428					}
3429					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3430					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3431					sctp_stop_timers_for_shutdown(asoc);
3432					if (asoc->asoc.alternate) {
3433						netp = asoc->asoc.alternate;
3434					} else {
3435						netp = asoc->asoc.primary_destination;
3436					}
3437					sctp_send_shutdown(asoc, netp);
3438					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3439					    netp);
3440					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3441					    asoc->asoc.primary_destination);
3442					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3443				}
3444			} else {
3445				/* mark into shutdown pending */
3446				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3447				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3448				    asoc->asoc.primary_destination);
3449				if ((*asoc->asoc.ss_functions.sctp_ss_is_user_msgs_incomplete) (asoc, &asoc->asoc)) {
3450					asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3451				}
3452				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3453				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3454				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3455					struct mbuf *op_err;
3456
3457			abort_anyway:
3458					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3459					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3460					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3461					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3462					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3463					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3464						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3465					}
3466					if (sctp_free_assoc(inp, asoc,
3467					    SCTP_PCBFREE_NOFORCE,
3468					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3469						cnt_in_sd++;
3470					}
3471					continue;
3472				} else {
3473					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3474				}
3475			}
3476			cnt_in_sd++;
3477			SCTP_TCB_UNLOCK(asoc);
3478		}
3479		/* now is there some left in our SHUTDOWN state? */
3480		if (cnt_in_sd) {
3481#ifdef SCTP_LOG_CLOSING
3482			sctp_log_closing(inp, NULL, 2);
3483#endif
3484			inp->sctp_socket = NULL;
3485			SCTP_INP_WUNLOCK(inp);
3486			SCTP_ASOC_CREATE_UNLOCK(inp);
3487			SCTP_INP_INFO_WUNLOCK();
3488			return;
3489		}
3490	}
3491	inp->sctp_socket = NULL;
3492	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3493	    SCTP_PCB_FLAGS_UNBOUND) {
3494		/*
3495		 * ok, this guy has been bound. It's port is somewhere in
3496		 * the SCTP_BASE_INFO(hash table). Remove it!
3497		 */
3498		LIST_REMOVE(inp, sctp_hash);
3499		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3500	}
3501	/*
3502	 * If there is a timer running to kill us, forget it, since it may
3503	 * have a contest on the INP lock.. which would cause us to die ...
3504	 */
3505	cnt = 0;
3506	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3507		SCTP_TCB_LOCK(asoc);
3508		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3509			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3510				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3511				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3512			}
3513			cnt++;
3514			SCTP_TCB_UNLOCK(asoc);
3515			continue;
3516		}
3517		/* Free associations that are NOT killing us */
3518		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3519		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3520			struct mbuf *op_err;
3521
3522			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3523			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3524			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3525			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3526		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3527			cnt++;
3528			SCTP_TCB_UNLOCK(asoc);
3529			continue;
3530		}
3531		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3532		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3533			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3534		}
3535		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE,
3536		    SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3537			cnt++;
3538		}
3539	}
3540	if (cnt) {
3541		/* Ok we have someone out there that will kill us */
3542		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3543#ifdef SCTP_LOG_CLOSING
3544		sctp_log_closing(inp, NULL, 3);
3545#endif
3546		SCTP_INP_WUNLOCK(inp);
3547		SCTP_ASOC_CREATE_UNLOCK(inp);
3548		SCTP_INP_INFO_WUNLOCK();
3549		return;
3550	}
3551	if (SCTP_INP_LOCK_CONTENDED(inp))
3552		being_refed++;
3553	if (SCTP_INP_READ_CONTENDED(inp))
3554		being_refed++;
3555	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3556		being_refed++;
3557
3558	if ((inp->refcount) ||
3559	    (being_refed) ||
3560	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3561		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3562#ifdef SCTP_LOG_CLOSING
3563		sctp_log_closing(inp, NULL, 4);
3564#endif
3565		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3566		SCTP_INP_WUNLOCK(inp);
3567		SCTP_ASOC_CREATE_UNLOCK(inp);
3568		SCTP_INP_INFO_WUNLOCK();
3569		return;
3570	}
3571	inp->sctp_ep.signature_change.type = 0;
3572	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3573	/*
3574	 * Remove it from the list .. last thing we need a lock for.
3575	 */
3576	LIST_REMOVE(inp, sctp_list);
3577	SCTP_INP_WUNLOCK(inp);
3578	SCTP_ASOC_CREATE_UNLOCK(inp);
3579	SCTP_INP_INFO_WUNLOCK();
3580	/*
3581	 * Now we release all locks. Since this INP cannot be found anymore
3582	 * except possibly by the kill timer that might be running. We call
3583	 * the drain function here. It should hit the case were it sees the
3584	 * ACTIVE flag cleared and exit out freeing us to proceed and
3585	 * destroy everything.
3586	 */
3587	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3588		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3589	} else {
3590		/* Probably un-needed */
3591		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3592	}
3593
3594#ifdef SCTP_LOG_CLOSING
3595	sctp_log_closing(inp, NULL, 5);
3596#endif
3597
3598
3599	if ((inp->sctp_asocidhash) != NULL) {
3600		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3601		inp->sctp_asocidhash = NULL;
3602	}
3603	/* sa_ignore FREED_MEMORY */
3604	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3605		/* Its only abandoned if it had data left */
3606		if (sq->length)
3607			SCTP_STAT_INCR(sctps_left_abandon);
3608
3609		TAILQ_REMOVE(&inp->read_queue, sq, next);
3610		sctp_free_remote_addr(sq->whoFrom);
3611		if (so)
3612			so->so_rcv.sb_cc -= sq->length;
3613		if (sq->data) {
3614			sctp_m_freem(sq->data);
3615			sq->data = NULL;
3616		}
3617		/*
3618		 * no need to free the net count, since at this point all
3619		 * assoc's are gone.
3620		 */
3621		sctp_free_a_readq(NULL, sq);
3622	}
3623	/* Now the sctp_pcb things */
3624	/*
3625	 * free each asoc if it is not already closed/free. we can't use the
3626	 * macro here since le_next will get freed as part of the
3627	 * sctp_free_assoc() call.
3628	 */
3629	if (ip_pcb->inp_options) {
3630		(void)sctp_m_free(ip_pcb->inp_options);
3631		ip_pcb->inp_options = 0;
3632	}
3633#ifdef INET6
3634	if (ip_pcb->inp_vflag & INP_IPV6) {
3635		struct in6pcb *in6p;
3636
3637		in6p = (struct in6pcb *)inp;
3638		ip6_freepcbopts(in6p->in6p_outputopts);
3639	}
3640#endif				/* INET6 */
3641	ip_pcb->inp_vflag = 0;
3642	/* free up authentication fields */
3643	if (inp->sctp_ep.local_auth_chunks != NULL)
3644		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3645	if (inp->sctp_ep.local_hmacs != NULL)
3646		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3647
3648	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3649		LIST_REMOVE(shared_key, next);
3650		sctp_free_sharedkey(shared_key);
3651		/* sa_ignore FREED_MEMORY */
3652	}
3653
3654	/*
3655	 * if we have an address list the following will free the list of
3656	 * ifaddr's that are set into this ep. Again macro limitations here,
3657	 * since the LIST_FOREACH could be a bad idea.
3658	 */
3659	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3660		sctp_remove_laddr(laddr);
3661	}
3662
3663#ifdef SCTP_TRACK_FREED_ASOCS
3664	/* TEMP CODE */
3665	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3666		LIST_REMOVE(asoc, sctp_tcblist);
3667		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3668		SCTP_DECR_ASOC_COUNT();
3669	}
3670	/* *** END TEMP CODE *** */
3671#endif
3672	/* Now lets see about freeing the EP hash table. */
3673	if (inp->sctp_tcbhash != NULL) {
3674		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3675		inp->sctp_tcbhash = NULL;
3676	}
3677	/* Now we must put the ep memory back into the zone pool */
3678	crfree(inp->ip_inp.inp.inp_cred);
3679	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3680	SCTP_INP_LOCK_DESTROY(inp);
3681	SCTP_INP_READ_DESTROY(inp);
3682	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3683	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3684	SCTP_DECR_EP_COUNT();
3685}
3686
3687
3688struct sctp_nets *
3689sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3690{
3691	struct sctp_nets *net;
3692
3693	/* locate the address */
3694	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3695		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3696			return (net);
3697	}
3698	return (NULL);
3699}
3700
3701
3702int
3703sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3704{
3705	struct sctp_ifa *sctp_ifa;
3706
3707	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3708	if (sctp_ifa) {
3709		return (1);
3710	} else {
3711		return (0);
3712	}
3713}
3714
3715/*
3716 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3717 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3718 * stats of stuff.
3719 */
3720int
3721sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3722    struct sctp_nets **netp, uint16_t port, int set_scope, int from)
3723{
3724	/*
3725	 * The following is redundant to the same lines in the
3726	 * sctp_aloc_assoc() but is needed since others call the add address
3727	 * function
3728	 */
3729	struct sctp_nets *net, *netfirst;
3730	int addr_inscope;
3731
3732	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3733	    from);
3734	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3735
3736	netfirst = sctp_findnet(stcb, newaddr);
3737	if (netfirst) {
3738		/*
3739		 * Lie and return ok, we don't want to make the association
3740		 * go away for this behavior. It will happen in the TCP
3741		 * model in a connected socket. It does not reach the hash
3742		 * table until after the association is built so it can't be
3743		 * found. Mark as reachable, since the initial creation will
3744		 * have been cleared and the NOT_IN_ASSOC flag will have
3745		 * been added... and we don't want to end up removing it
3746		 * back out.
3747		 */
3748		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3749			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3750			    SCTP_ADDR_UNCONFIRMED);
3751		} else {
3752			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3753		}
3754
3755		return (0);
3756	}
3757	addr_inscope = 1;
3758	switch (newaddr->sa_family) {
3759#ifdef INET
3760	case AF_INET:
3761		{
3762			struct sockaddr_in *sin;
3763
3764			sin = (struct sockaddr_in *)newaddr;
3765			if (sin->sin_addr.s_addr == 0) {
3766				/* Invalid address */
3767				return (-1);
3768			}
3769			/* zero out the zero area */
3770			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3771
3772			/* assure len is set */
3773			sin->sin_len = sizeof(struct sockaddr_in);
3774			if (set_scope) {
3775				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3776					stcb->asoc.scope.ipv4_local_scope = 1;
3777				}
3778			} else {
3779				/* Validate the address is in scope */
3780				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3781				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3782					addr_inscope = 0;
3783				}
3784			}
3785			break;
3786		}
3787#endif
3788#ifdef INET6
3789	case AF_INET6:
3790		{
3791			struct sockaddr_in6 *sin6;
3792
3793			sin6 = (struct sockaddr_in6 *)newaddr;
3794			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3795				/* Invalid address */
3796				return (-1);
3797			}
3798			/* assure len is set */
3799			sin6->sin6_len = sizeof(struct sockaddr_in6);
3800			if (set_scope) {
3801				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3802					stcb->asoc.scope.loopback_scope = 1;
3803					stcb->asoc.scope.local_scope = 0;
3804					stcb->asoc.scope.ipv4_local_scope = 1;
3805					stcb->asoc.scope.site_scope = 1;
3806				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3807					/*
3808					 * If the new destination is a
3809					 * LINK_LOCAL we must have common
3810					 * site scope. Don't set the local
3811					 * scope since we may not share all
3812					 * links, only loopback can do this.
3813					 * Links on the local network would
3814					 * also be on our private network
3815					 * for v4 too.
3816					 */
3817					stcb->asoc.scope.ipv4_local_scope = 1;
3818					stcb->asoc.scope.site_scope = 1;
3819				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3820					/*
3821					 * If the new destination is
3822					 * SITE_LOCAL then we must have site
3823					 * scope in common.
3824					 */
3825					stcb->asoc.scope.site_scope = 1;
3826				}
3827			} else {
3828				/* Validate the address is in scope */
3829				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3830				    (stcb->asoc.scope.loopback_scope == 0)) {
3831					addr_inscope = 0;
3832				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3833				    (stcb->asoc.scope.local_scope == 0)) {
3834					addr_inscope = 0;
3835				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3836				    (stcb->asoc.scope.site_scope == 0)) {
3837					addr_inscope = 0;
3838				}
3839			}
3840			break;
3841		}
3842#endif
3843	default:
3844		/* not supported family type */
3845		return (-1);
3846	}
3847	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3848	if (net == NULL) {
3849		return (-1);
3850	}
3851	SCTP_INCR_RADDR_COUNT();
3852	memset(net, 0, sizeof(struct sctp_nets));
3853	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3854	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3855	switch (newaddr->sa_family) {
3856#ifdef INET
3857	case AF_INET:
3858		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3859		break;
3860#endif
3861#ifdef INET6
3862	case AF_INET6:
3863		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3864		break;
3865#endif
3866	default:
3867		break;
3868	}
3869	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3870	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3871		stcb->asoc.scope.loopback_scope = 1;
3872		stcb->asoc.scope.ipv4_local_scope = 1;
3873		stcb->asoc.scope.local_scope = 0;
3874		stcb->asoc.scope.site_scope = 1;
3875		addr_inscope = 1;
3876	}
3877	net->failure_threshold = stcb->asoc.def_net_failure;
3878	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3879	if (addr_inscope == 0) {
3880		net->dest_state = (SCTP_ADDR_REACHABLE |
3881		    SCTP_ADDR_OUT_OF_SCOPE);
3882	} else {
3883		if (from == SCTP_ADDR_IS_CONFIRMED)
3884			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3885			net->dest_state = SCTP_ADDR_REACHABLE;
3886		else
3887			net->dest_state = SCTP_ADDR_REACHABLE |
3888			    SCTP_ADDR_UNCONFIRMED;
3889	}
3890	/*
3891	 * We set this to 0, the timer code knows that this means its an
3892	 * initial value
3893	 */
3894	net->rto_needed = 1;
3895	net->RTO = 0;
3896	net->RTO_measured = 0;
3897	stcb->asoc.numnets++;
3898	net->ref_count = 1;
3899	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3900	net->port = port;
3901	net->dscp = stcb->asoc.default_dscp;
3902#ifdef INET6
3903	net->flowlabel = stcb->asoc.default_flowlabel;
3904#endif
3905	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3906		net->dest_state |= SCTP_ADDR_NOHB;
3907	} else {
3908		net->dest_state &= ~SCTP_ADDR_NOHB;
3909	}
3910	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3911		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3912	} else {
3913		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3914	}
3915	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3916	/* Init the timer structure */
3917	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3918	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3919	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3920
3921	/* Now generate a route for this guy */
3922#ifdef INET6
3923	/* KAME hack: embed scopeid */
3924	if (newaddr->sa_family == AF_INET6) {
3925		struct sockaddr_in6 *sin6;
3926
3927		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3928		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3929		sin6->sin6_scope_id = 0;
3930	}
3931#endif
3932	SCTP_RTALLOC((sctp_route_t *)&net->ro,
3933	    stcb->asoc.vrf_id,
3934	    stcb->sctp_ep->fibnum);
3935
3936	net->src_addr_selected = 0;
3937	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3938		/* Get source address */
3939		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3940		    stcb,
3941		    (sctp_route_t *)&net->ro,
3942		    net,
3943		    0,
3944		    stcb->asoc.vrf_id);
3945		if (net->ro._s_addr != NULL) {
3946			uint32_t imtu, rmtu, hcmtu;
3947
3948			net->src_addr_selected = 1;
3949			/* Now get the interface MTU */
3950			if (net->ro._s_addr->ifn_p != NULL) {
3951				imtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3952			} else {
3953				imtu = 0;
3954			}
3955			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3956			hcmtu = sctp_hc_get_mtu(&net->ro._l_addr, stcb->sctp_ep->fibnum);
3957			net->mtu = sctp_min_mtu(hcmtu, rmtu, imtu);
3958			if (rmtu == 0) {
3959				/*
3960				 * Start things off to match mtu of
3961				 * interface please.
3962				 */
3963				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3964				    net->ro.ro_rt, net->mtu);
3965			}
3966		}
3967	}
3968	if (net->mtu == 0) {
3969		switch (newaddr->sa_family) {
3970#ifdef INET
3971		case AF_INET:
3972			net->mtu = SCTP_DEFAULT_MTU;
3973			break;
3974#endif
3975#ifdef INET6
3976		case AF_INET6:
3977			net->mtu = 1280;
3978			break;
3979#endif
3980		default:
3981			break;
3982		}
3983	}
3984#if defined(INET) || defined(INET6)
3985	if (net->port) {
3986		net->mtu -= (uint32_t)sizeof(struct udphdr);
3987	}
3988#endif
3989	if (from == SCTP_ALLOC_ASOC) {
3990		stcb->asoc.smallest_mtu = net->mtu;
3991	}
3992	if (stcb->asoc.smallest_mtu > net->mtu) {
3993		sctp_pathmtu_adjustment(stcb, net->mtu);
3994	}
3995#ifdef INET6
3996	if (newaddr->sa_family == AF_INET6) {
3997		struct sockaddr_in6 *sin6;
3998
3999		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
4000		(void)sa6_recoverscope(sin6);
4001	}
4002#endif
4003
4004	/* JRS - Use the congestion control given in the CC module */
4005	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
4006		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
4007
4008	/*
4009	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
4010	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
4011	 */
4012	net->find_pseudo_cumack = 1;
4013	net->find_rtx_pseudo_cumack = 1;
4014	/* Choose an initial flowid. */
4015	net->flowid = stcb->asoc.my_vtag ^
4016	    ntohs(stcb->rport) ^
4017	    ntohs(stcb->sctp_ep->sctp_lport);
4018	net->flowtype = M_HASHTYPE_OPAQUE_HASH;
4019	if (netp) {
4020		*netp = net;
4021	}
4022	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
4023	if (net->ro.ro_rt == NULL) {
4024		/* Since we have no route put it at the back */
4025		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4026	} else if (netfirst == NULL) {
4027		/* We are the first one in the pool. */
4028		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4029	} else if (netfirst->ro.ro_rt == NULL) {
4030		/*
4031		 * First one has NO route. Place this one ahead of the first
4032		 * one.
4033		 */
4034		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4035	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4036		/*
4037		 * This one has a different interface than the one at the
4038		 * top of the list. Place it ahead.
4039		 */
4040		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4041	} else {
4042		/*
4043		 * Ok we have the same interface as the first one. Move
4044		 * forward until we find either a) one with a NULL route...
4045		 * insert ahead of that b) one with a different ifp.. insert
4046		 * after that. c) end of the list.. insert at the tail.
4047		 */
4048		struct sctp_nets *netlook;
4049
4050		do {
4051			netlook = TAILQ_NEXT(netfirst, sctp_next);
4052			if (netlook == NULL) {
4053				/* End of the list */
4054				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4055				break;
4056			} else if (netlook->ro.ro_rt == NULL) {
4057				/* next one has NO route */
4058				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4059				break;
4060			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4061				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4062				    net, sctp_next);
4063				break;
4064			}
4065			/* Shift forward */
4066			netfirst = netlook;
4067		} while (netlook != NULL);
4068	}
4069
4070	/* got to have a primary set */
4071	if (stcb->asoc.primary_destination == 0) {
4072		stcb->asoc.primary_destination = net;
4073	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4074		    (net->ro.ro_rt) &&
4075	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4076		/* No route to current primary adopt new primary */
4077		stcb->asoc.primary_destination = net;
4078	}
4079	/* Validate primary is first */
4080	net = TAILQ_FIRST(&stcb->asoc.nets);
4081	if ((net != stcb->asoc.primary_destination) &&
4082	    (stcb->asoc.primary_destination)) {
4083		/*
4084		 * first one on the list is NOT the primary sctp_cmpaddr()
4085		 * is much more efficient if the primary is the first on the
4086		 * list, make it so.
4087		 */
4088		TAILQ_REMOVE(&stcb->asoc.nets,
4089		    stcb->asoc.primary_destination, sctp_next);
4090		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4091		    stcb->asoc.primary_destination, sctp_next);
4092	}
4093	return (0);
4094}
4095
4096
4097static uint32_t
4098sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4099{
4100	uint32_t id;
4101	struct sctpasochead *head;
4102	struct sctp_tcb *lstcb;
4103
4104	SCTP_INP_WLOCK(inp);
4105try_again:
4106	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4107		/* TSNH */
4108		SCTP_INP_WUNLOCK(inp);
4109		return (0);
4110	}
4111	/*
4112	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4113	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4114	 */
4115	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4116		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4117	}
4118	id = inp->sctp_associd_counter;
4119	inp->sctp_associd_counter++;
4120	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t)id, 0);
4121	if (lstcb) {
4122		goto try_again;
4123	}
4124	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4125	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4126	stcb->asoc.in_asocid_hash = 1;
4127	SCTP_INP_WUNLOCK(inp);
4128	return id;
4129}
4130
4131/*
4132 * allocate an association and add it to the endpoint. The caller must be
4133 * careful to add all additional addresses once they are know right away or
4134 * else the assoc will be may experience a blackout scenario.
4135 */
4136struct sctp_tcb *
4137sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4138    int *error, uint32_t override_tag, uint32_t vrf_id,
4139    uint16_t o_streams, uint16_t port,
4140    struct thread *p
4141)
4142{
4143	/* note the p argument is only valid in unbound sockets */
4144
4145	struct sctp_tcb *stcb;
4146	struct sctp_association *asoc;
4147	struct sctpasochead *head;
4148	uint16_t rport;
4149	int err;
4150
4151	/*
4152	 * Assumption made here: Caller has done a
4153	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4154	 * address does not exist already.
4155	 */
4156	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4157		/* Hit max assoc, sorry no more */
4158		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4159		*error = ENOBUFS;
4160		return (NULL);
4161	}
4162	if (firstaddr == NULL) {
4163		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4164		*error = EINVAL;
4165		return (NULL);
4166	}
4167	SCTP_INP_RLOCK(inp);
4168	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4169	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4170	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4171		/*
4172		 * If its in the TCP pool, its NOT allowed to create an
4173		 * association. The parent listener needs to call
4174		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4175		 * off, or connected one does this.. its an error.
4176		 */
4177		SCTP_INP_RUNLOCK(inp);
4178		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4179		*error = EINVAL;
4180		return (NULL);
4181	}
4182	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4183	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4184		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4185		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4186			SCTP_INP_RUNLOCK(inp);
4187			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4188			*error = EINVAL;
4189			return (NULL);
4190		}
4191	}
4192	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4193#ifdef SCTP_DEBUG
4194	if (firstaddr) {
4195		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4196		switch (firstaddr->sa_family) {
4197#ifdef INET
4198		case AF_INET:
4199			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4200			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4201			break;
4202#endif
4203#ifdef INET6
4204		case AF_INET6:
4205			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4206			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4207			break;
4208#endif
4209		default:
4210			break;
4211		}
4212	} else {
4213		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4214	}
4215#endif				/* SCTP_DEBUG */
4216	switch (firstaddr->sa_family) {
4217#ifdef INET
4218	case AF_INET:
4219		{
4220			struct sockaddr_in *sin;
4221
4222			sin = (struct sockaddr_in *)firstaddr;
4223			if ((ntohs(sin->sin_port) == 0) ||
4224			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4225			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4226			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4227				/* Invalid address */
4228				SCTP_INP_RUNLOCK(inp);
4229				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4230				*error = EINVAL;
4231				return (NULL);
4232			}
4233			rport = sin->sin_port;
4234			break;
4235		}
4236#endif
4237#ifdef INET6
4238	case AF_INET6:
4239		{
4240			struct sockaddr_in6 *sin6;
4241
4242			sin6 = (struct sockaddr_in6 *)firstaddr;
4243			if ((ntohs(sin6->sin6_port) == 0) ||
4244			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4245			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4246				/* Invalid address */
4247				SCTP_INP_RUNLOCK(inp);
4248				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4249				*error = EINVAL;
4250				return (NULL);
4251			}
4252			rport = sin6->sin6_port;
4253			break;
4254		}
4255#endif
4256	default:
4257		/* not supported family type */
4258		SCTP_INP_RUNLOCK(inp);
4259		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4260		*error = EINVAL;
4261		return (NULL);
4262	}
4263	SCTP_INP_RUNLOCK(inp);
4264	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4265		/*
4266		 * If you have not performed a bind, then we need to do the
4267		 * ephemeral bind for you.
4268		 */
4269		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4270		    (struct sockaddr *)NULL,
4271		    (struct sctp_ifa *)NULL,
4272		    p
4273		    ))) {
4274			/* bind error, probably perm */
4275			*error = err;
4276			return (NULL);
4277		}
4278	}
4279	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4280	if (stcb == NULL) {
4281		/* out of memory? */
4282		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4283		*error = ENOMEM;
4284		return (NULL);
4285	}
4286	SCTP_INCR_ASOC_COUNT();
4287
4288	memset(stcb, 0, sizeof(*stcb));
4289	asoc = &stcb->asoc;
4290
4291	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4292	SCTP_TCB_LOCK_INIT(stcb);
4293	SCTP_TCB_SEND_LOCK_INIT(stcb);
4294	stcb->rport = rport;
4295	/* setup back pointer's */
4296	stcb->sctp_ep = inp;
4297	stcb->sctp_socket = inp->sctp_socket;
4298	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) {
4299		/* failed */
4300		SCTP_TCB_LOCK_DESTROY(stcb);
4301		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4302		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4303		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4304		SCTP_DECR_ASOC_COUNT();
4305		*error = err;
4306		return (NULL);
4307	}
4308	/* and the port */
4309	SCTP_INP_INFO_WLOCK();
4310	SCTP_INP_WLOCK(inp);
4311	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4312		/* inpcb freed while alloc going on */
4313		SCTP_TCB_LOCK_DESTROY(stcb);
4314		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4315		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4316		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4317		SCTP_INP_WUNLOCK(inp);
4318		SCTP_INP_INFO_WUNLOCK();
4319		SCTP_DECR_ASOC_COUNT();
4320		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4321		*error = EINVAL;
4322		return (NULL);
4323	}
4324	SCTP_TCB_LOCK(stcb);
4325
4326	/* now that my_vtag is set, add it to the hash */
4327	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4328	/* put it in the bucket in the vtag hash of assoc's for the system */
4329	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4330	SCTP_INP_INFO_WUNLOCK();
4331
4332	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, port, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4333		/* failure.. memory error? */
4334		if (asoc->strmout) {
4335			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4336			asoc->strmout = NULL;
4337		}
4338		if (asoc->mapping_array) {
4339			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4340			asoc->mapping_array = NULL;
4341		}
4342		if (asoc->nr_mapping_array) {
4343			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4344			asoc->nr_mapping_array = NULL;
4345		}
4346		SCTP_DECR_ASOC_COUNT();
4347		SCTP_TCB_UNLOCK(stcb);
4348		SCTP_TCB_LOCK_DESTROY(stcb);
4349		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4350		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4351		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4352		SCTP_INP_WUNLOCK(inp);
4353		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4354		*error = ENOBUFS;
4355		return (NULL);
4356	}
4357	/* Init all the timers */
4358	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4359	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4360	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4361	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4362	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4363	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4364	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4365
4366	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4367	/* now file the port under the hash as well */
4368	if (inp->sctp_tcbhash != NULL) {
4369		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4370		    inp->sctp_hashmark)];
4371		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4372	}
4373	SCTP_INP_WUNLOCK(inp);
4374	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4375	return (stcb);
4376}
4377
4378
4379void
4380sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4381{
4382	struct sctp_association *asoc;
4383
4384	asoc = &stcb->asoc;
4385	asoc->numnets--;
4386	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4387	if (net == asoc->primary_destination) {
4388		/* Reset primary */
4389		struct sctp_nets *lnet;
4390
4391		lnet = TAILQ_FIRST(&asoc->nets);
4392		/*
4393		 * Mobility adaptation Ideally, if deleted destination is
4394		 * the primary, it becomes a fast retransmission trigger by
4395		 * the subsequent SET PRIMARY. (by micchie)
4396		 */
4397		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4398		    SCTP_MOBILITY_BASE) ||
4399		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4400		    SCTP_MOBILITY_FASTHANDOFF)) {
4401			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4402			if (asoc->deleted_primary != NULL) {
4403				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4404				goto out;
4405			}
4406			asoc->deleted_primary = net;
4407			atomic_add_int(&net->ref_count, 1);
4408			memset(&net->lastsa, 0, sizeof(net->lastsa));
4409			memset(&net->lastsv, 0, sizeof(net->lastsv));
4410			sctp_mobility_feature_on(stcb->sctp_ep,
4411			    SCTP_MOBILITY_PRIM_DELETED);
4412			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4413			    stcb->sctp_ep, stcb, NULL);
4414		}
4415out:
4416		/* Try to find a confirmed primary */
4417		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4418	}
4419	if (net == asoc->last_data_chunk_from) {
4420		/* Reset primary */
4421		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4422	}
4423	if (net == asoc->last_control_chunk_from) {
4424		/* Clear net */
4425		asoc->last_control_chunk_from = NULL;
4426	}
4427	if (net == stcb->asoc.alternate) {
4428		sctp_free_remote_addr(stcb->asoc.alternate);
4429		stcb->asoc.alternate = NULL;
4430	}
4431	sctp_free_remote_addr(net);
4432}
4433
4434/*
4435 * remove a remote endpoint address from an association, it will fail if the
4436 * address does not exist.
4437 */
4438int
4439sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4440{
4441	/*
4442	 * Here we need to remove a remote address. This is quite simple, we
4443	 * first find it in the list of address for the association
4444	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4445	 * on that item. Note we do not allow it to be removed if there are
4446	 * no other addresses.
4447	 */
4448	struct sctp_association *asoc;
4449	struct sctp_nets *net, *nnet;
4450
4451	asoc = &stcb->asoc;
4452
4453	/* locate the address */
4454	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4455		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4456			continue;
4457		}
4458		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4459		    remaddr)) {
4460			/* we found the guy */
4461			if (asoc->numnets < 2) {
4462				/* Must have at LEAST two remote addresses */
4463				return (-1);
4464			} else {
4465				sctp_remove_net(stcb, net);
4466				return (0);
4467			}
4468		}
4469	}
4470	/* not found. */
4471	return (-2);
4472}
4473
4474void
4475sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4476{
4477	struct sctpvtaghead *chain;
4478	struct sctp_tagblock *twait_block;
4479	int found = 0;
4480	int i;
4481
4482	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4483	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4484		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4485			if ((twait_block->vtag_block[i].v_tag == tag) &&
4486			    (twait_block->vtag_block[i].lport == lport) &&
4487			    (twait_block->vtag_block[i].rport == rport)) {
4488				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4489				twait_block->vtag_block[i].v_tag = 0;
4490				twait_block->vtag_block[i].lport = 0;
4491				twait_block->vtag_block[i].rport = 0;
4492				found = 1;
4493				break;
4494			}
4495		}
4496		if (found)
4497			break;
4498	}
4499}
4500
4501int
4502sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4503{
4504	struct sctpvtaghead *chain;
4505	struct sctp_tagblock *twait_block;
4506	int found = 0;
4507	int i;
4508
4509	SCTP_INP_INFO_WLOCK();
4510	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4511	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4512		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4513			if ((twait_block->vtag_block[i].v_tag == tag) &&
4514			    (twait_block->vtag_block[i].lport == lport) &&
4515			    (twait_block->vtag_block[i].rport == rport)) {
4516				found = 1;
4517				break;
4518			}
4519		}
4520		if (found)
4521			break;
4522	}
4523	SCTP_INP_INFO_WUNLOCK();
4524	return (found);
4525}
4526
4527
4528void
4529sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4530{
4531	struct sctpvtaghead *chain;
4532	struct sctp_tagblock *twait_block;
4533	struct timeval now;
4534	int set, i;
4535
4536	if (time == 0) {
4537		/* Its disabled */
4538		return;
4539	}
4540	(void)SCTP_GETTIME_TIMEVAL(&now);
4541	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4542	set = 0;
4543	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4544		/* Block(s) present, lets find space, and expire on the fly */
4545		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4546			if ((twait_block->vtag_block[i].v_tag == 0) &&
4547			    !set) {
4548				twait_block->vtag_block[i].tv_sec_at_expire =
4549				    now.tv_sec + time;
4550				twait_block->vtag_block[i].v_tag = tag;
4551				twait_block->vtag_block[i].lport = lport;
4552				twait_block->vtag_block[i].rport = rport;
4553				set = 1;
4554			} else if ((twait_block->vtag_block[i].v_tag) &&
4555			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4556				/* Audit expires this guy */
4557				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4558				twait_block->vtag_block[i].v_tag = 0;
4559				twait_block->vtag_block[i].lport = 0;
4560				twait_block->vtag_block[i].rport = 0;
4561				if (set == 0) {
4562					/* Reuse it for my new tag */
4563					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4564					twait_block->vtag_block[i].v_tag = tag;
4565					twait_block->vtag_block[i].lport = lport;
4566					twait_block->vtag_block[i].rport = rport;
4567					set = 1;
4568				}
4569			}
4570		}
4571		if (set) {
4572			/*
4573			 * We only do up to the block where we can place our
4574			 * tag for audits
4575			 */
4576			break;
4577		}
4578	}
4579	/* Need to add a new block to chain */
4580	if (!set) {
4581		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4582		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4583		if (twait_block == NULL) {
4584#ifdef INVARIANTS
4585			panic("Can not alloc tagblock");
4586#endif
4587			return;
4588		}
4589		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4590		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4591		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4592		twait_block->vtag_block[0].v_tag = tag;
4593		twait_block->vtag_block[0].lport = lport;
4594		twait_block->vtag_block[0].rport = rport;
4595	}
4596}
4597
4598void
4599sctp_clean_up_stream(struct sctp_tcb *stcb, struct sctp_readhead *rh)
4600{
4601	struct sctp_tmit_chunk *chk, *nchk;
4602	struct sctp_queued_to_read *control, *ncontrol;
4603
4604	TAILQ_FOREACH_SAFE(control, rh, next_instrm, ncontrol) {
4605		TAILQ_REMOVE(rh, control, next_instrm);
4606		control->on_strm_q = 0;
4607		if (control->on_read_q == 0) {
4608			sctp_free_remote_addr(control->whoFrom);
4609			if (control->data) {
4610				sctp_m_freem(control->data);
4611				control->data = NULL;
4612			}
4613		}
4614		/* Reassembly free? */
4615		TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
4616			TAILQ_REMOVE(&control->reasm, chk, sctp_next);
4617			if (chk->data) {
4618				sctp_m_freem(chk->data);
4619				chk->data = NULL;
4620			}
4621			if (chk->holds_key_ref)
4622				sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4623			sctp_free_remote_addr(chk->whoTo);
4624			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4625			SCTP_DECR_CHK_COUNT();
4626			/* sa_ignore FREED_MEMORY */
4627		}
4628		/*
4629		 * We don't free the address here since all the net's were
4630		 * freed above.
4631		 */
4632		if (control->on_read_q == 0) {
4633			sctp_free_a_readq(stcb, control);
4634		}
4635	}
4636}
4637
4638
4639/*-
4640 * Free the association after un-hashing the remote port. This
4641 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4642 * expect that the input to this function IS a locked TCB.
4643 * It will return 0, if it did NOT destroy the association (instead
4644 * it unlocks it. It will return NON-zero if it either destroyed the
4645 * association OR the association is already destroyed.
4646 */
4647int
4648sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4649{
4650	int i;
4651	struct sctp_association *asoc;
4652	struct sctp_nets *net, *nnet;
4653	struct sctp_laddr *laddr, *naddr;
4654	struct sctp_tmit_chunk *chk, *nchk;
4655	struct sctp_asconf_addr *aparam, *naparam;
4656	struct sctp_asconf_ack *aack, *naack;
4657	struct sctp_stream_reset_list *strrst, *nstrrst;
4658	struct sctp_queued_to_read *sq, *nsq;
4659	struct sctp_stream_queue_pending *sp, *nsp;
4660	sctp_sharedkey_t *shared_key, *nshared_key;
4661	struct socket *so;
4662
4663	/* first, lets purge the entry from the hash table. */
4664
4665#ifdef SCTP_LOG_CLOSING
4666	sctp_log_closing(inp, stcb, 6);
4667#endif
4668	if (stcb->asoc.state == 0) {
4669#ifdef SCTP_LOG_CLOSING
4670		sctp_log_closing(inp, NULL, 7);
4671#endif
4672		/* there is no asoc, really TSNH :-0 */
4673		return (1);
4674	}
4675	if (stcb->asoc.alternate) {
4676		sctp_free_remote_addr(stcb->asoc.alternate);
4677		stcb->asoc.alternate = NULL;
4678	}
4679	/* TEMP CODE */
4680	if (stcb->freed_from_where == 0) {
4681		/* Only record the first place free happened from */
4682		stcb->freed_from_where = from_location;
4683	}
4684	/* TEMP CODE */
4685
4686	asoc = &stcb->asoc;
4687	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4688	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4689		/* nothing around */
4690		so = NULL;
4691	else
4692		so = inp->sctp_socket;
4693
4694	/*
4695	 * We used timer based freeing if a reader or writer is in the way.
4696	 * So we first check if we are actually being called from a timer,
4697	 * if so we abort early if a reader or writer is still in the way.
4698	 */
4699	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4700	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4701		/*
4702		 * is it the timer driving us? if so are the reader/writers
4703		 * gone?
4704		 */
4705		if (stcb->asoc.refcnt) {
4706			/* nope, reader or writer in the way */
4707			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4708			/* no asoc destroyed */
4709			SCTP_TCB_UNLOCK(stcb);
4710#ifdef SCTP_LOG_CLOSING
4711			sctp_log_closing(inp, stcb, 8);
4712#endif
4713			return (0);
4714		}
4715	}
4716	/* now clean up any other timers */
4717	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4718	asoc->dack_timer.self = NULL;
4719	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4720	/*-
4721	 * For stream reset we don't blast this unless
4722	 * it is a str-reset timer, it might be the
4723	 * free-asoc timer which we DON'T want to
4724	 * disturb.
4725	 */
4726	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4727		asoc->strreset_timer.self = NULL;
4728	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4729	asoc->asconf_timer.self = NULL;
4730	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4731	asoc->autoclose_timer.self = NULL;
4732	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4733	asoc->shut_guard_timer.self = NULL;
4734	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4735	asoc->delayed_event_timer.self = NULL;
4736	/* Mobility adaptation */
4737	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4738	asoc->delete_prim_timer.self = NULL;
4739	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4740		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4741		net->rxt_timer.self = NULL;
4742		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4743		net->pmtu_timer.self = NULL;
4744		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4745		net->hb_timer.self = NULL;
4746	}
4747	/* Now the read queue needs to be cleaned up (only once) */
4748	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4749		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4750		SCTP_INP_READ_LOCK(inp);
4751		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4752			if (sq->stcb == stcb) {
4753				sq->do_not_ref_stcb = 1;
4754				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4755				/*
4756				 * If there is no end, there never will be
4757				 * now.
4758				 */
4759				if (sq->end_added == 0) {
4760					/* Held for PD-API clear that. */
4761					sq->pdapi_aborted = 1;
4762					sq->held_length = 0;
4763					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4764						/*
4765						 * Need to add a PD-API
4766						 * aborted indication.
4767						 * Setting the control_pdapi
4768						 * assures that it will be
4769						 * added right after this
4770						 * msg.
4771						 */
4772						uint32_t strseq;
4773
4774						stcb->asoc.control_pdapi = sq;
4775						strseq = (sq->sinfo_stream << 16) | (sq->mid & 0x0000ffff);
4776						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4777						    stcb,
4778						    SCTP_PARTIAL_DELIVERY_ABORTED,
4779						    (void *)&strseq,
4780						    SCTP_SO_LOCKED);
4781						stcb->asoc.control_pdapi = NULL;
4782					}
4783				}
4784				/* Add an end to wake them */
4785				sq->end_added = 1;
4786			}
4787		}
4788		SCTP_INP_READ_UNLOCK(inp);
4789		if (stcb->block_entry) {
4790			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4791			stcb->block_entry->error = ECONNRESET;
4792			stcb->block_entry = NULL;
4793		}
4794	}
4795	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4796		/*
4797		 * Someone holds a reference OR the socket is unaccepted
4798		 * yet.
4799		 */
4800		if ((stcb->asoc.refcnt) ||
4801		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4802		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4803			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4804			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4805		}
4806		SCTP_TCB_UNLOCK(stcb);
4807		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4808		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4809			/* nothing around */
4810			so = NULL;
4811		if (so) {
4812			/* Wake any reader/writers */
4813			sctp_sorwakeup(inp, so);
4814			sctp_sowwakeup(inp, so);
4815		}
4816#ifdef SCTP_LOG_CLOSING
4817		sctp_log_closing(inp, stcb, 9);
4818#endif
4819		/* no asoc destroyed */
4820		return (0);
4821	}
4822#ifdef SCTP_LOG_CLOSING
4823	sctp_log_closing(inp, stcb, 10);
4824#endif
4825	/*
4826	 * When I reach here, no others want to kill the assoc yet.. and I
4827	 * own the lock. Now its possible an abort comes in when I do the
4828	 * lock exchange below to grab all the locks to do the final take
4829	 * out. to prevent this we increment the count, which will start a
4830	 * timer and blow out above thus assuring us that we hold exclusive
4831	 * killing of the asoc. Note that after getting back the TCB lock we
4832	 * will go ahead and increment the counter back up and stop any
4833	 * timer a passing stranger may have started :-S
4834	 */
4835	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4836		atomic_add_int(&stcb->asoc.refcnt, 1);
4837
4838		SCTP_TCB_UNLOCK(stcb);
4839		SCTP_INP_INFO_WLOCK();
4840		SCTP_INP_WLOCK(inp);
4841		SCTP_TCB_LOCK(stcb);
4842	}
4843	/* Double check the GONE flag */
4844	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4845	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4846		/* nothing around */
4847		so = NULL;
4848
4849	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4850	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4851		/*
4852		 * For TCP type we need special handling when we are
4853		 * connected. We also include the peel'ed off ones to.
4854		 */
4855		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4856			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4857			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4858			if (so) {
4859				SOCK_LOCK(so);
4860				if (so->so_rcv.sb_cc == 0) {
4861					so->so_state &= ~(SS_ISCONNECTING |
4862					    SS_ISDISCONNECTING |
4863					    SS_ISCONFIRMING |
4864					    SS_ISCONNECTED);
4865				}
4866				socantrcvmore_locked(so);
4867				socantsendmore(so);
4868				sctp_sowwakeup(inp, so);
4869				sctp_sorwakeup(inp, so);
4870				SCTP_SOWAKEUP(so);
4871			}
4872		}
4873	}
4874	/*
4875	 * Make it invalid too, that way if its about to run it will abort
4876	 * and return.
4877	 */
4878	/* re-increment the lock */
4879	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4880		atomic_add_int(&stcb->asoc.refcnt, -1);
4881	}
4882	if (stcb->asoc.refcnt) {
4883		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4884		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4885		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4886			SCTP_INP_INFO_WUNLOCK();
4887			SCTP_INP_WUNLOCK(inp);
4888		}
4889		SCTP_TCB_UNLOCK(stcb);
4890		return (0);
4891	}
4892	asoc->state = 0;
4893	if (inp->sctp_tcbhash) {
4894		LIST_REMOVE(stcb, sctp_tcbhash);
4895	}
4896	if (stcb->asoc.in_asocid_hash) {
4897		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4898	}
4899	/* Now lets remove it from the list of ALL associations in the EP */
4900	LIST_REMOVE(stcb, sctp_tcblist);
4901	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4902		SCTP_INP_INCR_REF(inp);
4903		SCTP_INP_WUNLOCK(inp);
4904	}
4905	/* pull from vtag hash */
4906	LIST_REMOVE(stcb, sctp_asocs);
4907	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4908	    inp->sctp_lport, stcb->rport);
4909
4910	/*
4911	 * Now restop the timers to be sure this is paranoia at is finest!
4912	 */
4913	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4914	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4915	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4916	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4917	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4918	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4919	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4920	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4921		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4922		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4923		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4924	}
4925
4926	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4927	/*
4928	 * The chunk lists and such SHOULD be empty but we check them just
4929	 * in case.
4930	 */
4931	/* anything on the wheel needs to be removed */
4932	for (i = 0; i < asoc->streamoutcnt; i++) {
4933		struct sctp_stream_out *outs;
4934
4935		outs = &asoc->strmout[i];
4936		/* now clean up any chunks here */
4937		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4938			atomic_subtract_int(&asoc->stream_queue_cnt, 1);
4939			TAILQ_REMOVE(&outs->outqueue, sp, next);
4940			stcb->asoc.ss_functions.sctp_ss_remove_from_stream(stcb, asoc, outs, sp, 0);
4941			sctp_free_spbufspace(stcb, asoc, sp);
4942			if (sp->data) {
4943				if (so) {
4944					/* Still an open socket - report */
4945					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4946					    0, (void *)sp, SCTP_SO_LOCKED);
4947				}
4948				if (sp->data) {
4949					sctp_m_freem(sp->data);
4950					sp->data = NULL;
4951					sp->tail_mbuf = NULL;
4952					sp->length = 0;
4953				}
4954			}
4955			if (sp->net) {
4956				sctp_free_remote_addr(sp->net);
4957				sp->net = NULL;
4958			}
4959			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4960		}
4961	}
4962	/* sa_ignore FREED_MEMORY */
4963	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4964		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4965		SCTP_FREE(strrst, SCTP_M_STRESET);
4966	}
4967	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4968		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4969		if (sq->data) {
4970			sctp_m_freem(sq->data);
4971			sq->data = NULL;
4972		}
4973		sctp_free_remote_addr(sq->whoFrom);
4974		sq->whoFrom = NULL;
4975		sq->stcb = NULL;
4976		/* Free the ctl entry */
4977		sctp_free_a_readq(stcb, sq);
4978		/* sa_ignore FREED_MEMORY */
4979	}
4980	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4981		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4982		if (chk->data) {
4983			sctp_m_freem(chk->data);
4984			chk->data = NULL;
4985		}
4986		if (chk->holds_key_ref)
4987			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4988		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4989		SCTP_DECR_CHK_COUNT();
4990		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4991		asoc->free_chunk_cnt--;
4992		/* sa_ignore FREED_MEMORY */
4993	}
4994	/* pending send queue SHOULD be empty */
4995	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4996		if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
4997			asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
4998#ifdef INVARIANTS
4999		} else {
5000			panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
5001#endif
5002		}
5003		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
5004		if (chk->data) {
5005			if (so) {
5006				/* Still a socket? */
5007				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
5008				    0, chk, SCTP_SO_LOCKED);
5009			}
5010			if (chk->data) {
5011				sctp_m_freem(chk->data);
5012				chk->data = NULL;
5013			}
5014		}
5015		if (chk->holds_key_ref)
5016			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5017		if (chk->whoTo) {
5018			sctp_free_remote_addr(chk->whoTo);
5019			chk->whoTo = NULL;
5020		}
5021		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5022		SCTP_DECR_CHK_COUNT();
5023		/* sa_ignore FREED_MEMORY */
5024	}
5025	/* sent queue SHOULD be empty */
5026	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
5027		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
5028			if (asoc->strmout[chk->rec.data.sid].chunks_on_queues > 0) {
5029				asoc->strmout[chk->rec.data.sid].chunks_on_queues--;
5030#ifdef INVARIANTS
5031			} else {
5032				panic("No chunks on the queues for sid %u.", chk->rec.data.sid);
5033#endif
5034			}
5035		}
5036		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
5037		if (chk->data) {
5038			if (so) {
5039				/* Still a socket? */
5040				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
5041				    0, chk, SCTP_SO_LOCKED);
5042			}
5043			if (chk->data) {
5044				sctp_m_freem(chk->data);
5045				chk->data = NULL;
5046			}
5047		}
5048		if (chk->holds_key_ref)
5049			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5050		sctp_free_remote_addr(chk->whoTo);
5051		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5052		SCTP_DECR_CHK_COUNT();
5053		/* sa_ignore FREED_MEMORY */
5054	}
5055#ifdef INVARIANTS
5056	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
5057		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
5058			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
5059		}
5060	}
5061#endif
5062	/* control queue MAY not be empty */
5063	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
5064		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
5065		if (chk->data) {
5066			sctp_m_freem(chk->data);
5067			chk->data = NULL;
5068		}
5069		if (chk->holds_key_ref)
5070			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5071		sctp_free_remote_addr(chk->whoTo);
5072		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5073		SCTP_DECR_CHK_COUNT();
5074		/* sa_ignore FREED_MEMORY */
5075	}
5076	/* ASCONF queue MAY not be empty */
5077	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5078		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5079		if (chk->data) {
5080			sctp_m_freem(chk->data);
5081			chk->data = NULL;
5082		}
5083		if (chk->holds_key_ref)
5084			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5085		sctp_free_remote_addr(chk->whoTo);
5086		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5087		SCTP_DECR_CHK_COUNT();
5088		/* sa_ignore FREED_MEMORY */
5089	}
5090	if (asoc->mapping_array) {
5091		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5092		asoc->mapping_array = NULL;
5093	}
5094	if (asoc->nr_mapping_array) {
5095		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5096		asoc->nr_mapping_array = NULL;
5097	}
5098	/* the stream outs */
5099	if (asoc->strmout) {
5100		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5101		asoc->strmout = NULL;
5102	}
5103	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5104	if (asoc->strmin) {
5105		for (i = 0; i < asoc->streamincnt; i++) {
5106			sctp_clean_up_stream(stcb, &asoc->strmin[i].inqueue);
5107			sctp_clean_up_stream(stcb, &asoc->strmin[i].uno_inqueue);
5108		}
5109		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5110		asoc->strmin = NULL;
5111	}
5112	asoc->streamincnt = 0;
5113	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5114#ifdef INVARIANTS
5115		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5116			panic("no net's left alloc'ed, or list points to itself");
5117		}
5118#endif
5119		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5120		sctp_free_remote_addr(net);
5121	}
5122	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5123		/* sa_ignore FREED_MEMORY */
5124		sctp_remove_laddr(laddr);
5125	}
5126
5127	/* pending asconf (address) parameters */
5128	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5129		/* sa_ignore FREED_MEMORY */
5130		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5131		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5132	}
5133	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5134		/* sa_ignore FREED_MEMORY */
5135		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5136		if (aack->data != NULL) {
5137			sctp_m_freem(aack->data);
5138		}
5139		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5140	}
5141	/* clean up auth stuff */
5142	if (asoc->local_hmacs)
5143		sctp_free_hmaclist(asoc->local_hmacs);
5144	if (asoc->peer_hmacs)
5145		sctp_free_hmaclist(asoc->peer_hmacs);
5146
5147	if (asoc->local_auth_chunks)
5148		sctp_free_chunklist(asoc->local_auth_chunks);
5149	if (asoc->peer_auth_chunks)
5150		sctp_free_chunklist(asoc->peer_auth_chunks);
5151
5152	sctp_free_authinfo(&asoc->authinfo);
5153
5154	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5155		LIST_REMOVE(shared_key, next);
5156		sctp_free_sharedkey(shared_key);
5157		/* sa_ignore FREED_MEMORY */
5158	}
5159
5160	/* Insert new items here :> */
5161
5162	/* Get rid of LOCK */
5163	SCTP_TCB_UNLOCK(stcb);
5164	SCTP_TCB_LOCK_DESTROY(stcb);
5165	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5166	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5167		SCTP_INP_INFO_WUNLOCK();
5168		SCTP_INP_RLOCK(inp);
5169	}
5170#ifdef SCTP_TRACK_FREED_ASOCS
5171	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5172		/* now clean up the tasoc itself */
5173		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5174		SCTP_DECR_ASOC_COUNT();
5175	} else {
5176		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5177	}
5178#else
5179	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5180	SCTP_DECR_ASOC_COUNT();
5181#endif
5182	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5183		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5184			/*
5185			 * If its NOT the inp_free calling us AND sctp_close
5186			 * as been called, we call back...
5187			 */
5188			SCTP_INP_RUNLOCK(inp);
5189			/*
5190			 * This will start the kill timer (if we are the
5191			 * last one) since we hold an increment yet. But
5192			 * this is the only safe way to do this since
5193			 * otherwise if the socket closes at the same time
5194			 * we are here we might collide in the cleanup.
5195			 */
5196			sctp_inpcb_free(inp,
5197			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5198			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5199			SCTP_INP_DECR_REF(inp);
5200			goto out_of;
5201		} else {
5202			/* The socket is still open. */
5203			SCTP_INP_DECR_REF(inp);
5204		}
5205	}
5206	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5207		SCTP_INP_RUNLOCK(inp);
5208	}
5209out_of:
5210	/* destroyed the asoc */
5211#ifdef SCTP_LOG_CLOSING
5212	sctp_log_closing(inp, NULL, 11);
5213#endif
5214	return (1);
5215}
5216
5217
5218
5219/*
5220 * determine if a destination is "reachable" based upon the addresses bound
5221 * to the current endpoint (e.g. only v4 or v6 currently bound)
5222 */
5223/*
5224 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5225 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5226 * types bound as its endpoint
5227 */
5228int
5229sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5230{
5231	struct sctp_inpcb *inp;
5232	int answer;
5233
5234	/*
5235	 * No locks here, the TCB, in all cases is already locked and an
5236	 * assoc is up. There is either a INP lock by the caller applied (in
5237	 * asconf case when deleting an address) or NOT in the HB case,
5238	 * however if HB then the INP increment is up and the INP will not
5239	 * be removed (on top of the fact that we have a TCB lock). So we
5240	 * only want to read the sctp_flags, which is either bound-all or
5241	 * not.. no protection needed since once an assoc is up you can't be
5242	 * changing your binding.
5243	 */
5244	inp = stcb->sctp_ep;
5245	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5246		/* if bound all, destination is not restricted */
5247		/*
5248		 * RRS: Question during lock work: Is this correct? If you
5249		 * are bound-all you still might need to obey the V4--V6
5250		 * flags??? IMO this bound-all stuff needs to be removed!
5251		 */
5252		return (1);
5253	}
5254	/* NOTE: all "scope" checks are done when local addresses are added */
5255	switch (destaddr->sa_family) {
5256#ifdef INET6
5257	case AF_INET6:
5258		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5259		break;
5260#endif
5261#ifdef INET
5262	case AF_INET:
5263		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5264		break;
5265#endif
5266	default:
5267		/* invalid family, so it's unreachable */
5268		answer = 0;
5269		break;
5270	}
5271	return (answer);
5272}
5273
5274/*
5275 * update the inp_vflags on an endpoint
5276 */
5277static void
5278sctp_update_ep_vflag(struct sctp_inpcb *inp)
5279{
5280	struct sctp_laddr *laddr;
5281
5282	/* first clear the flag */
5283	inp->ip_inp.inp.inp_vflag = 0;
5284	/* set the flag based on addresses on the ep list */
5285	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5286		if (laddr->ifa == NULL) {
5287			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5288			    __func__);
5289			continue;
5290		}
5291		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5292			continue;
5293		}
5294		switch (laddr->ifa->address.sa.sa_family) {
5295#ifdef INET6
5296		case AF_INET6:
5297			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5298			break;
5299#endif
5300#ifdef INET
5301		case AF_INET:
5302			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5303			break;
5304#endif
5305		default:
5306			break;
5307		}
5308	}
5309}
5310
5311/*
5312 * Add the address to the endpoint local address list There is nothing to be
5313 * done if we are bound to all addresses
5314 */
5315void
5316sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5317{
5318	struct sctp_laddr *laddr;
5319	struct sctp_tcb *stcb;
5320	int fnd, error = 0;
5321
5322	fnd = 0;
5323
5324	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5325		/* You are already bound to all. You have it already */
5326		return;
5327	}
5328#ifdef INET6
5329	if (ifa->address.sa.sa_family == AF_INET6) {
5330		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5331			/* Can't bind a non-useable addr. */
5332			return;
5333		}
5334	}
5335#endif
5336	/* first, is it already present? */
5337	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5338		if (laddr->ifa == ifa) {
5339			fnd = 1;
5340			break;
5341		}
5342	}
5343
5344	if (fnd == 0) {
5345		/* Not in the ep list */
5346		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5347		if (error != 0)
5348			return;
5349		inp->laddr_count++;
5350		/* update inp_vflag flags */
5351		switch (ifa->address.sa.sa_family) {
5352#ifdef INET6
5353		case AF_INET6:
5354			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5355			break;
5356#endif
5357#ifdef INET
5358		case AF_INET:
5359			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5360			break;
5361#endif
5362		default:
5363			break;
5364		}
5365		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5366			sctp_add_local_addr_restricted(stcb, ifa);
5367		}
5368	}
5369	return;
5370}
5371
5372
5373/*
5374 * select a new (hopefully reachable) destination net (should only be used
5375 * when we deleted an ep addr that is the only usable source address to reach
5376 * the destination net)
5377 */
5378static void
5379sctp_select_primary_destination(struct sctp_tcb *stcb)
5380{
5381	struct sctp_nets *net;
5382
5383	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5384		/* for now, we'll just pick the first reachable one we find */
5385		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5386			continue;
5387		if (sctp_destination_is_reachable(stcb,
5388		    (struct sockaddr *)&net->ro._l_addr)) {
5389			/* found a reachable destination */
5390			stcb->asoc.primary_destination = net;
5391		}
5392	}
5393	/* I can't there from here! ...we're gonna die shortly... */
5394}
5395
5396
5397/*
5398 * Delete the address from the endpoint local address list. There is nothing
5399 * to be done if we are bound to all addresses
5400 */
5401void
5402sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5403{
5404	struct sctp_laddr *laddr;
5405	int fnd;
5406
5407	fnd = 0;
5408	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5409		/* You are already bound to all. You have it already */
5410		return;
5411	}
5412	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5413		if (laddr->ifa == ifa) {
5414			fnd = 1;
5415			break;
5416		}
5417	}
5418	if (fnd && (inp->laddr_count < 2)) {
5419		/* can't delete unless there are at LEAST 2 addresses */
5420		return;
5421	}
5422	if (fnd) {
5423		/*
5424		 * clean up any use of this address go through our
5425		 * associations and clear any last_used_address that match
5426		 * this one for each assoc, see if a new primary_destination
5427		 * is needed
5428		 */
5429		struct sctp_tcb *stcb;
5430
5431		/* clean up "next_addr_touse" */
5432		if (inp->next_addr_touse == laddr)
5433			/* delete this address */
5434			inp->next_addr_touse = NULL;
5435
5436		/* clean up "last_used_address" */
5437		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5438			struct sctp_nets *net;
5439
5440			SCTP_TCB_LOCK(stcb);
5441			if (stcb->asoc.last_used_address == laddr)
5442				/* delete this address */
5443				stcb->asoc.last_used_address = NULL;
5444			/*
5445			 * Now spin through all the nets and purge any ref
5446			 * to laddr
5447			 */
5448			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5449				if (net->ro._s_addr == laddr->ifa) {
5450					/* Yep, purge src address selected */
5451					sctp_rtentry_t *rt;
5452
5453					/* delete this address if cached */
5454					rt = net->ro.ro_rt;
5455					if (rt != NULL) {
5456						RTFREE(rt);
5457						net->ro.ro_rt = NULL;
5458					}
5459					sctp_free_ifa(net->ro._s_addr);
5460					net->ro._s_addr = NULL;
5461					net->src_addr_selected = 0;
5462				}
5463			}
5464			SCTP_TCB_UNLOCK(stcb);
5465		}		/* for each tcb */
5466		/* remove it from the ep list */
5467		sctp_remove_laddr(laddr);
5468		inp->laddr_count--;
5469		/* update inp_vflag flags */
5470		sctp_update_ep_vflag(inp);
5471	}
5472	return;
5473}
5474
5475/*
5476 * Add the address to the TCB local address restricted list.
5477 * This is a "pending" address list (eg. addresses waiting for an
5478 * ASCONF-ACK response) and cannot be used as a valid source address.
5479 */
5480void
5481sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5482{
5483	struct sctp_laddr *laddr;
5484	struct sctpladdr *list;
5485
5486	/*
5487	 * Assumes TCB is locked.. and possibly the INP. May need to
5488	 * confirm/fix that if we need it and is not the case.
5489	 */
5490	list = &stcb->asoc.sctp_restricted_addrs;
5491
5492#ifdef INET6
5493	if (ifa->address.sa.sa_family == AF_INET6) {
5494		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5495			/* Can't bind a non-existent addr. */
5496			return;
5497		}
5498	}
5499#endif
5500	/* does the address already exist? */
5501	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5502		if (laddr->ifa == ifa) {
5503			return;
5504		}
5505	}
5506
5507	/* add to the list */
5508	(void)sctp_insert_laddr(list, ifa, 0);
5509	return;
5510}
5511
5512/*
5513 * Remove a local address from the TCB local address restricted list
5514 */
5515void
5516sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5517{
5518	struct sctp_inpcb *inp;
5519	struct sctp_laddr *laddr;
5520
5521	/*
5522	 * This is called by asconf work. It is assumed that a) The TCB is
5523	 * locked and b) The INP is locked. This is true in as much as I can
5524	 * trace through the entry asconf code where I did these locks.
5525	 * Again, the ASCONF code is a bit different in that it does lock
5526	 * the INP during its work often times. This must be since we don't
5527	 * want other proc's looking up things while what they are looking
5528	 * up is changing :-D
5529	 */
5530
5531	inp = stcb->sctp_ep;
5532	/* if subset bound and don't allow ASCONF's, can't delete last */
5533	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5534	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5535		if (stcb->sctp_ep->laddr_count < 2) {
5536			/* can't delete last address */
5537			return;
5538		}
5539	}
5540	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5541		/* remove the address if it exists */
5542		if (laddr->ifa == NULL)
5543			continue;
5544		if (laddr->ifa == ifa) {
5545			sctp_remove_laddr(laddr);
5546			return;
5547		}
5548	}
5549
5550	/* address not found! */
5551	return;
5552}
5553
5554/*
5555 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5556 */
5557/* sysctl */
5558static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5559static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5560
5561
5562
5563#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5564struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5565int *sctp_cpuarry = NULL;
5566void
5567sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5568{
5569	/* Queue a packet to a processor for the specified core */
5570	struct sctp_mcore_queue *qent;
5571	struct sctp_mcore_ctrl *wkq;
5572	int need_wake = 0;
5573
5574	if (sctp_mcore_workers == NULL) {
5575		/* Something went way bad during setup */
5576		sctp_input_with_port(m, off, 0);
5577		return;
5578	}
5579	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5580	    (sizeof(struct sctp_mcore_queue)),
5581	    SCTP_M_MCORE);
5582	if (qent == NULL) {
5583		/* This is trouble  */
5584		sctp_input_with_port(m, off, 0);
5585		return;
5586	}
5587	qent->vn = curvnet;
5588	qent->m = m;
5589	qent->off = off;
5590	qent->v6 = 0;
5591	wkq = &sctp_mcore_workers[cpu_to_use];
5592	SCTP_MCORE_QLOCK(wkq);
5593
5594	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5595	if (wkq->running == 0) {
5596		need_wake = 1;
5597	}
5598	SCTP_MCORE_QUNLOCK(wkq);
5599	if (need_wake) {
5600		wakeup(&wkq->running);
5601	}
5602}
5603
5604static void
5605sctp_mcore_thread(void *arg)
5606{
5607
5608	struct sctp_mcore_ctrl *wkq;
5609	struct sctp_mcore_queue *qent;
5610
5611	wkq = (struct sctp_mcore_ctrl *)arg;
5612	struct mbuf *m;
5613	int off, v6;
5614
5615	/* Wait for first tickle */
5616	SCTP_MCORE_LOCK(wkq);
5617	wkq->running = 0;
5618	msleep(&wkq->running,
5619	    &wkq->core_mtx,
5620	    0, "wait for pkt", 0);
5621	SCTP_MCORE_UNLOCK(wkq);
5622
5623	/* Bind to our cpu */
5624	thread_lock(curthread);
5625	sched_bind(curthread, wkq->cpuid);
5626	thread_unlock(curthread);
5627
5628	/* Now lets start working */
5629	SCTP_MCORE_LOCK(wkq);
5630	/* Now grab lock and go */
5631	for (;;) {
5632		SCTP_MCORE_QLOCK(wkq);
5633skip_sleep:
5634		wkq->running = 1;
5635		qent = TAILQ_FIRST(&wkq->que);
5636		if (qent) {
5637			TAILQ_REMOVE(&wkq->que, qent, next);
5638			SCTP_MCORE_QUNLOCK(wkq);
5639			CURVNET_SET(qent->vn);
5640			m = qent->m;
5641			off = qent->off;
5642			v6 = qent->v6;
5643			SCTP_FREE(qent, SCTP_M_MCORE);
5644			if (v6 == 0) {
5645				sctp_input_with_port(m, off, 0);
5646			} else {
5647				SCTP_PRINTF("V6 not yet supported\n");
5648				sctp_m_freem(m);
5649			}
5650			CURVNET_RESTORE();
5651			SCTP_MCORE_QLOCK(wkq);
5652		}
5653		wkq->running = 0;
5654		if (!TAILQ_EMPTY(&wkq->que)) {
5655			goto skip_sleep;
5656		}
5657		SCTP_MCORE_QUNLOCK(wkq);
5658		msleep(&wkq->running,
5659		    &wkq->core_mtx,
5660		    0, "wait for pkt", 0);
5661	}
5662}
5663
5664static void
5665sctp_startup_mcore_threads(void)
5666{
5667	int i, cpu;
5668
5669	if (mp_ncpus == 1)
5670		return;
5671
5672	if (sctp_mcore_workers != NULL) {
5673		/*
5674		 * Already been here in some previous vnet?
5675		 */
5676		return;
5677	}
5678	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5679	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5680	    SCTP_M_MCORE);
5681	if (sctp_mcore_workers == NULL) {
5682		/* TSNH I hope */
5683		return;
5684	}
5685	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5686	    sizeof(struct sctp_mcore_ctrl)));
5687	/* Init the structures */
5688	for (i = 0; i <= mp_maxid; i++) {
5689		TAILQ_INIT(&sctp_mcore_workers[i].que);
5690		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5691		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5692		sctp_mcore_workers[i].cpuid = i;
5693	}
5694	if (sctp_cpuarry == NULL) {
5695		SCTP_MALLOC(sctp_cpuarry, int *,
5696		    (mp_ncpus * sizeof(int)),
5697		    SCTP_M_MCORE);
5698		i = 0;
5699		CPU_FOREACH(cpu) {
5700			sctp_cpuarry[i] = cpu;
5701			i++;
5702		}
5703	}
5704	/* Now start them all */
5705	CPU_FOREACH(cpu) {
5706		(void)kproc_create(sctp_mcore_thread,
5707		    (void *)&sctp_mcore_workers[cpu],
5708		    &sctp_mcore_workers[cpu].thread_proc,
5709		    RFPROC,
5710		    SCTP_KTHREAD_PAGES,
5711		    SCTP_MCORE_NAME);
5712
5713	}
5714}
5715#endif
5716
5717void
5718sctp_pcb_init()
5719{
5720	/*
5721	 * SCTP initialization for the PCB structures should be called by
5722	 * the sctp_init() function.
5723	 */
5724	int i;
5725	struct timeval tv;
5726
5727	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5728		/* error I was called twice */
5729		return;
5730	}
5731	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5732
5733#if defined(SCTP_LOCAL_TRACE_BUF)
5734	memset(&SCTP_BASE_SYSCTL(sctp_log), 0, sizeof(struct sctp_log));
5735#endif
5736#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5737	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5738	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5739	    SCTP_M_MCORE);
5740#endif
5741	(void)SCTP_GETTIME_TIMEVAL(&tv);
5742#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5743	memset(SCTP_BASE_STATS, 0, sizeof(struct sctpstat) * (mp_maxid + 1));
5744	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t)tv.tv_sec;
5745	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t)tv.tv_usec;
5746#else
5747	memset(&SCTP_BASE_STATS, 0, sizeof(struct sctpstat));
5748	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t)tv.tv_sec;
5749	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t)tv.tv_usec;
5750#endif
5751	/* init the empty list of (All) Endpoints */
5752	LIST_INIT(&SCTP_BASE_INFO(listhead));
5753
5754
5755	/* init the hash table of endpoints */
5756	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5757	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5758	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5759	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5760	    &SCTP_BASE_INFO(hashasocmark));
5761	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5762	    &SCTP_BASE_INFO(hashmark));
5763	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5764	    &SCTP_BASE_INFO(hashtcpmark));
5765	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5766
5767
5768	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5769	    &SCTP_BASE_INFO(hashvrfmark));
5770
5771	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5772	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5773	/* init the zones */
5774	/*
5775	 * FIX ME: Should check for NULL returns, but if it does fail we are
5776	 * doomed to panic anyways... add later maybe.
5777	 */
5778	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5779	    sizeof(struct sctp_inpcb), maxsockets);
5780
5781	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5782	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5783
5784	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5785	    sizeof(struct sctp_laddr),
5786	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5787
5788	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5789	    sizeof(struct sctp_nets),
5790	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5791
5792	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5793	    sizeof(struct sctp_tmit_chunk),
5794	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5795
5796	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5797	    sizeof(struct sctp_queued_to_read),
5798	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5799
5800	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5801	    sizeof(struct sctp_stream_queue_pending),
5802	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5803
5804	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5805	    sizeof(struct sctp_asconf),
5806	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5807
5808	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5809	    sizeof(struct sctp_asconf_ack),
5810	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5811
5812
5813	/* Master Lock INIT for info structure */
5814	SCTP_INP_INFO_LOCK_INIT();
5815	SCTP_STATLOG_INIT_LOCK();
5816
5817	SCTP_IPI_COUNT_INIT();
5818	SCTP_IPI_ADDR_INIT();
5819#ifdef SCTP_PACKET_LOGGING
5820	SCTP_IP_PKTLOG_INIT();
5821#endif
5822	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5823
5824	SCTP_WQ_ADDR_INIT();
5825	/* not sure if we need all the counts */
5826	SCTP_BASE_INFO(ipi_count_ep) = 0;
5827	/* assoc/tcb zone info */
5828	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5829	/* local addrlist zone info */
5830	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5831	/* remote addrlist zone info */
5832	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5833	/* chunk info */
5834	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5835
5836	/* socket queue zone info */
5837	SCTP_BASE_INFO(ipi_count_readq) = 0;
5838
5839	/* stream out queue cont */
5840	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5841
5842	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5843	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5844
5845	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5846
5847	/* Init the TIMEWAIT list */
5848	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5849		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5850	}
5851	sctp_startup_iterator();
5852
5853#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5854	sctp_startup_mcore_threads();
5855#endif
5856
5857	/*
5858	 * INIT the default VRF which for BSD is the only one, other O/S's
5859	 * may have more. But initially they must start with one and then
5860	 * add the VRF's as addresses are added.
5861	 */
5862	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5863}
5864
5865/*
5866 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5867 */
5868void
5869sctp_pcb_finish(void)
5870{
5871	struct sctp_vrflist *vrf_bucket;
5872	struct sctp_vrf *vrf, *nvrf;
5873	struct sctp_ifn *ifn, *nifn;
5874	struct sctp_ifa *ifa, *nifa;
5875	struct sctpvtaghead *chain;
5876	struct sctp_tagblock *twait_block, *prev_twait_block;
5877	struct sctp_laddr *wi, *nwi;
5878	int i;
5879	struct sctp_iterator *it, *nit;
5880
5881	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
5882		SCTP_PRINTF("%s: race condition on teardown.\n", __func__);
5883		return;
5884	}
5885	SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
5886	/*
5887	 * In FreeBSD the iterator thread never exits but we do clean up.
5888	 * The only way FreeBSD reaches here is if we have VRF's but we
5889	 * still add the ifdef to make it compile on old versions.
5890	 */
5891retry:
5892	SCTP_IPI_ITERATOR_WQ_LOCK();
5893	/*
5894	 * sctp_iterator_worker() might be working on an it entry without
5895	 * holding the lock.  We won't find it on the list either and
5896	 * continue and free/destroy it.  While holding the lock, spin, to
5897	 * avoid the race condition as sctp_iterator_worker() will have to
5898	 * wait to re-aquire the lock.
5899	 */
5900	if (sctp_it_ctl.iterator_running != 0 || sctp_it_ctl.cur_it != NULL) {
5901		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5902		SCTP_PRINTF("%s: Iterator running while we held the lock. Retry. "
5903		    "cur_it=%p\n", __func__, sctp_it_ctl.cur_it);
5904		DELAY(10);
5905		goto retry;
5906	}
5907	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5908		if (it->vn != curvnet) {
5909			continue;
5910		}
5911		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5912		if (it->function_atend != NULL) {
5913			(*it->function_atend) (it->pointer, it->val);
5914		}
5915		SCTP_FREE(it, SCTP_M_ITER);
5916	}
5917	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5918	SCTP_ITERATOR_LOCK();
5919	if ((sctp_it_ctl.cur_it) &&
5920	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5921		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5922	}
5923	SCTP_ITERATOR_UNLOCK();
5924	SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer));
5925	SCTP_WQ_ADDR_LOCK();
5926	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5927		LIST_REMOVE(wi, sctp_nxt_addr);
5928		SCTP_DECR_LADDR_COUNT();
5929		if (wi->action == SCTP_DEL_IP_ADDRESS) {
5930			SCTP_FREE(wi->ifa, SCTP_M_IFA);
5931		}
5932		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5933	}
5934	SCTP_WQ_ADDR_UNLOCK();
5935
5936	/*
5937	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5938	 * destroyed first).
5939	 */
5940	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5941	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5942		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5943			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5944				/* free the ifa */
5945				LIST_REMOVE(ifa, next_bucket);
5946				LIST_REMOVE(ifa, next_ifa);
5947				SCTP_FREE(ifa, SCTP_M_IFA);
5948			}
5949			/* free the ifn */
5950			LIST_REMOVE(ifn, next_bucket);
5951			LIST_REMOVE(ifn, next_ifn);
5952			SCTP_FREE(ifn, SCTP_M_IFN);
5953		}
5954		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5955		/* free the vrf */
5956		LIST_REMOVE(vrf, next_vrf);
5957		SCTP_FREE(vrf, SCTP_M_VRF);
5958	}
5959	/* free the vrf hashes */
5960	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5961	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5962
5963	/*
5964	 * free the TIMEWAIT list elements malloc'd in the function
5965	 * sctp_add_vtag_to_timewait()...
5966	 */
5967	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5968		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5969		if (!LIST_EMPTY(chain)) {
5970			prev_twait_block = NULL;
5971			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5972				if (prev_twait_block) {
5973					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5974				}
5975				prev_twait_block = twait_block;
5976			}
5977			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5978		}
5979	}
5980
5981	/* free the locks and mutexes */
5982#ifdef SCTP_PACKET_LOGGING
5983	SCTP_IP_PKTLOG_DESTROY();
5984#endif
5985	SCTP_IPI_ADDR_DESTROY();
5986	SCTP_STATLOG_DESTROY();
5987	SCTP_INP_INFO_LOCK_DESTROY();
5988
5989	SCTP_WQ_ADDR_DESTROY();
5990
5991	/* Get rid of other stuff too. */
5992	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5993		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5994	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5995		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5996	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5997		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5998
5999	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
6000	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
6001	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
6002	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
6003	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
6004	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
6005	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
6006	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
6007	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
6008#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
6009	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
6010#endif
6011}
6012
6013
6014int
6015sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
6016    int offset, int limit,
6017    struct sockaddr *src, struct sockaddr *dst,
6018    struct sockaddr *altsa, uint16_t port)
6019{
6020	/*
6021	 * grub through the INIT pulling addresses and loading them to the
6022	 * nets structure in the asoc. The from address in the mbuf should
6023	 * also be loaded (if it is not already). This routine can be called
6024	 * with either INIT or INIT-ACK's as long as the m points to the IP
6025	 * packet and the offset points to the beginning of the parameters.
6026	 */
6027	struct sctp_inpcb *inp;
6028	struct sctp_nets *net, *nnet, *net_tmp;
6029	struct sctp_paramhdr *phdr, param_buf;
6030	struct sctp_tcb *stcb_tmp;
6031	uint16_t ptype, plen;
6032	struct sockaddr *sa;
6033	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6034	struct sctp_auth_random *p_random = NULL;
6035	uint16_t random_len = 0;
6036	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6037	struct sctp_auth_hmac_algo *hmacs = NULL;
6038	uint16_t hmacs_len = 0;
6039	uint8_t saw_asconf = 0;
6040	uint8_t saw_asconf_ack = 0;
6041	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6042	struct sctp_auth_chunk_list *chunks = NULL;
6043	uint16_t num_chunks = 0;
6044	sctp_key_t *new_key;
6045	uint32_t keylen;
6046	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6047	uint8_t peer_supports_ecn;
6048	uint8_t peer_supports_prsctp;
6049	uint8_t peer_supports_auth;
6050	uint8_t peer_supports_asconf;
6051	uint8_t peer_supports_asconf_ack;
6052	uint8_t peer_supports_reconfig;
6053	uint8_t peer_supports_nrsack;
6054	uint8_t peer_supports_pktdrop;
6055	uint8_t peer_supports_idata;
6056#ifdef INET
6057	struct sockaddr_in sin;
6058#endif
6059#ifdef INET6
6060	struct sockaddr_in6 sin6;
6061#endif
6062
6063	/* First get the destination address setup too. */
6064#ifdef INET
6065	memset(&sin, 0, sizeof(sin));
6066	sin.sin_family = AF_INET;
6067	sin.sin_len = sizeof(sin);
6068	sin.sin_port = stcb->rport;
6069#endif
6070#ifdef INET6
6071	memset(&sin6, 0, sizeof(sin6));
6072	sin6.sin6_family = AF_INET6;
6073	sin6.sin6_len = sizeof(struct sockaddr_in6);
6074	sin6.sin6_port = stcb->rport;
6075#endif
6076	if (altsa) {
6077		sa = altsa;
6078	} else {
6079		sa = src;
6080	}
6081	peer_supports_idata = 0;
6082	peer_supports_ecn = 0;
6083	peer_supports_prsctp = 0;
6084	peer_supports_auth = 0;
6085	peer_supports_asconf = 0;
6086	peer_supports_reconfig = 0;
6087	peer_supports_nrsack = 0;
6088	peer_supports_pktdrop = 0;
6089	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6090		/* mark all addresses that we have currently on the list */
6091		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6092	}
6093	/* does the source address already exist? if so skip it */
6094	inp = stcb->sctp_ep;
6095	atomic_add_int(&stcb->asoc.refcnt, 1);
6096	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6097	atomic_add_int(&stcb->asoc.refcnt, -1);
6098
6099	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6100		/* we must add the source address */
6101		/* no scope set here since we have a tcb already. */
6102		switch (sa->sa_family) {
6103#ifdef INET
6104		case AF_INET:
6105			if (stcb->asoc.scope.ipv4_addr_legal) {
6106				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6107					return (-1);
6108				}
6109			}
6110			break;
6111#endif
6112#ifdef INET6
6113		case AF_INET6:
6114			if (stcb->asoc.scope.ipv6_addr_legal) {
6115				if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6116					return (-2);
6117				}
6118			}
6119			break;
6120#endif
6121		default:
6122			break;
6123		}
6124	} else {
6125		if (net_tmp != NULL && stcb_tmp == stcb) {
6126			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6127		} else if (stcb_tmp != stcb) {
6128			/* It belongs to another association? */
6129			if (stcb_tmp)
6130				SCTP_TCB_UNLOCK(stcb_tmp);
6131			return (-3);
6132		}
6133	}
6134	if (stcb->asoc.state == 0) {
6135		/* the assoc was freed? */
6136		return (-4);
6137	}
6138	/* now we must go through each of the params. */
6139	phdr = sctp_get_next_param(m, offset, &param_buf, sizeof(param_buf));
6140	while (phdr) {
6141		ptype = ntohs(phdr->param_type);
6142		plen = ntohs(phdr->param_length);
6143		/*
6144		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6145		 * (uint32_t)ptype, (int)plen);
6146		 */
6147		if (offset + plen > limit) {
6148			break;
6149		}
6150		if (plen == 0) {
6151			break;
6152		}
6153#ifdef INET
6154		if (ptype == SCTP_IPV4_ADDRESS) {
6155			if (stcb->asoc.scope.ipv4_addr_legal) {
6156				struct sctp_ipv4addr_param *p4, p4_buf;
6157
6158				/* ok get the v4 address and check/add */
6159				phdr = sctp_get_next_param(m, offset,
6160				    (struct sctp_paramhdr *)&p4_buf,
6161				    sizeof(p4_buf));
6162				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6163				    phdr == NULL) {
6164					return (-5);
6165				}
6166				p4 = (struct sctp_ipv4addr_param *)phdr;
6167				sin.sin_addr.s_addr = p4->addr;
6168				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6169					/* Skip multi-cast addresses */
6170					goto next_param;
6171				}
6172				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6173				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6174					goto next_param;
6175				}
6176				sa = (struct sockaddr *)&sin;
6177				inp = stcb->sctp_ep;
6178				atomic_add_int(&stcb->asoc.refcnt, 1);
6179				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6180				    dst, stcb);
6181				atomic_add_int(&stcb->asoc.refcnt, -1);
6182
6183				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6184				    inp == NULL) {
6185					/* we must add the source address */
6186					/*
6187					 * no scope set since we have a tcb
6188					 * already
6189					 */
6190
6191					/*
6192					 * we must validate the state again
6193					 * here
6194					 */
6195			add_it_now:
6196					if (stcb->asoc.state == 0) {
6197						/* the assoc was freed? */
6198						return (-7);
6199					}
6200					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6201						return (-8);
6202					}
6203				} else if (stcb_tmp == stcb) {
6204					if (stcb->asoc.state == 0) {
6205						/* the assoc was freed? */
6206						return (-10);
6207					}
6208					if (net != NULL) {
6209						/* clear flag */
6210						net->dest_state &=
6211						    ~SCTP_ADDR_NOT_IN_ASSOC;
6212					}
6213				} else {
6214					/*
6215					 * strange, address is in another
6216					 * assoc? straighten out locks.
6217					 */
6218					if (stcb_tmp) {
6219						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6220							struct mbuf *op_err;
6221							char msg[SCTP_DIAG_INFO_LEN];
6222
6223							/*
6224							 * in setup state we
6225							 * abort this guy
6226							 */
6227							snprintf(msg, sizeof(msg),
6228							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6229							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6230							    msg);
6231							sctp_abort_an_association(stcb_tmp->sctp_ep,
6232							    stcb_tmp, op_err,
6233							    SCTP_SO_NOT_LOCKED);
6234							goto add_it_now;
6235						}
6236						SCTP_TCB_UNLOCK(stcb_tmp);
6237					}
6238					if (stcb->asoc.state == 0) {
6239						/* the assoc was freed? */
6240						return (-12);
6241					}
6242					return (-13);
6243				}
6244			}
6245		} else
6246#endif
6247#ifdef INET6
6248		if (ptype == SCTP_IPV6_ADDRESS) {
6249			if (stcb->asoc.scope.ipv6_addr_legal) {
6250				/* ok get the v6 address and check/add */
6251				struct sctp_ipv6addr_param *p6, p6_buf;
6252
6253				phdr = sctp_get_next_param(m, offset,
6254				    (struct sctp_paramhdr *)&p6_buf,
6255				    sizeof(p6_buf));
6256				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6257				    phdr == NULL) {
6258					return (-14);
6259				}
6260				p6 = (struct sctp_ipv6addr_param *)phdr;
6261				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6262				    sizeof(p6->addr));
6263				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6264					/* Skip multi-cast addresses */
6265					goto next_param;
6266				}
6267				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6268					/*
6269					 * Link local make no sense without
6270					 * scope
6271					 */
6272					goto next_param;
6273				}
6274				sa = (struct sockaddr *)&sin6;
6275				inp = stcb->sctp_ep;
6276				atomic_add_int(&stcb->asoc.refcnt, 1);
6277				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6278				    dst, stcb);
6279				atomic_add_int(&stcb->asoc.refcnt, -1);
6280				if (stcb_tmp == NULL &&
6281				    (inp == stcb->sctp_ep || inp == NULL)) {
6282					/*
6283					 * we must validate the state again
6284					 * here
6285					 */
6286			add_it_now6:
6287					if (stcb->asoc.state == 0) {
6288						/* the assoc was freed? */
6289						return (-16);
6290					}
6291					/*
6292					 * we must add the address, no scope
6293					 * set
6294					 */
6295					if (sctp_add_remote_addr(stcb, sa, NULL, port, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6296						return (-17);
6297					}
6298				} else if (stcb_tmp == stcb) {
6299					/*
6300					 * we must validate the state again
6301					 * here
6302					 */
6303					if (stcb->asoc.state == 0) {
6304						/* the assoc was freed? */
6305						return (-19);
6306					}
6307					if (net != NULL) {
6308						/* clear flag */
6309						net->dest_state &=
6310						    ~SCTP_ADDR_NOT_IN_ASSOC;
6311					}
6312				} else {
6313					/*
6314					 * strange, address is in another
6315					 * assoc? straighten out locks.
6316					 */
6317					if (stcb_tmp) {
6318						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6319							struct mbuf *op_err;
6320							char msg[SCTP_DIAG_INFO_LEN];
6321
6322							/*
6323							 * in setup state we
6324							 * abort this guy
6325							 */
6326							snprintf(msg, sizeof(msg),
6327							    "%s:%d at %s", __FILE__, __LINE__, __func__);
6328							op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code),
6329							    msg);
6330							sctp_abort_an_association(stcb_tmp->sctp_ep,
6331							    stcb_tmp, op_err,
6332							    SCTP_SO_NOT_LOCKED);
6333							goto add_it_now6;
6334						}
6335						SCTP_TCB_UNLOCK(stcb_tmp);
6336					}
6337					if (stcb->asoc.state == 0) {
6338						/* the assoc was freed? */
6339						return (-21);
6340					}
6341					return (-22);
6342				}
6343			}
6344		} else
6345#endif
6346		if (ptype == SCTP_ECN_CAPABLE) {
6347			peer_supports_ecn = 1;
6348		} else if (ptype == SCTP_ULP_ADAPTATION) {
6349			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6350				struct sctp_adaptation_layer_indication ai,
6351				                                *aip;
6352
6353				phdr = sctp_get_next_param(m, offset,
6354				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6355				aip = (struct sctp_adaptation_layer_indication *)phdr;
6356				if (aip) {
6357					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6358					stcb->asoc.adaptation_needed = 1;
6359				}
6360			}
6361		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6362			struct sctp_asconf_addr_param lstore, *fee;
6363			int lptype;
6364			struct sockaddr *lsa = NULL;
6365#ifdef INET
6366			struct sctp_asconf_addrv4_param *fii;
6367#endif
6368
6369			if (stcb->asoc.asconf_supported == 0) {
6370				return (-100);
6371			}
6372			if (plen > sizeof(lstore)) {
6373				return (-23);
6374			}
6375			phdr = sctp_get_next_param(m, offset,
6376			    (struct sctp_paramhdr *)&lstore,
6377			    plen);
6378			if (phdr == NULL) {
6379				return (-24);
6380			}
6381			fee = (struct sctp_asconf_addr_param *)phdr;
6382			lptype = ntohs(fee->addrp.ph.param_type);
6383			switch (lptype) {
6384#ifdef INET
6385			case SCTP_IPV4_ADDRESS:
6386				if (plen !=
6387				    sizeof(struct sctp_asconf_addrv4_param)) {
6388					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6389					    (int)sizeof(struct sctp_asconf_addrv4_param),
6390					    plen);
6391				} else {
6392					fii = (struct sctp_asconf_addrv4_param *)fee;
6393					sin.sin_addr.s_addr = fii->addrp.addr;
6394					lsa = (struct sockaddr *)&sin;
6395				}
6396				break;
6397#endif
6398#ifdef INET6
6399			case SCTP_IPV6_ADDRESS:
6400				if (plen !=
6401				    sizeof(struct sctp_asconf_addr_param)) {
6402					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6403					    (int)sizeof(struct sctp_asconf_addr_param),
6404					    plen);
6405				} else {
6406					memcpy(sin6.sin6_addr.s6_addr,
6407					    fee->addrp.addr,
6408					    sizeof(fee->addrp.addr));
6409					lsa = (struct sockaddr *)&sin6;
6410				}
6411				break;
6412#endif
6413			default:
6414				break;
6415			}
6416			if (lsa) {
6417				(void)sctp_set_primary_addr(stcb, sa, NULL);
6418			}
6419		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6420			stcb->asoc.peer_supports_nat = 1;
6421		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6422			/* Peer supports pr-sctp */
6423			peer_supports_prsctp = 1;
6424		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6425			/* A supported extension chunk */
6426			struct sctp_supported_chunk_types_param *pr_supported;
6427			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6428			int num_ent, i;
6429
6430			if (plen > sizeof(local_store)) {
6431				return (-35);
6432			}
6433			phdr = sctp_get_next_param(m, offset,
6434			    (struct sctp_paramhdr *)&local_store, plen);
6435			if (phdr == NULL) {
6436				return (-25);
6437			}
6438			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6439			num_ent = plen - sizeof(struct sctp_paramhdr);
6440			for (i = 0; i < num_ent; i++) {
6441				switch (pr_supported->chunk_types[i]) {
6442				case SCTP_ASCONF:
6443					peer_supports_asconf = 1;
6444					break;
6445				case SCTP_ASCONF_ACK:
6446					peer_supports_asconf_ack = 1;
6447					break;
6448				case SCTP_FORWARD_CUM_TSN:
6449					peer_supports_prsctp = 1;
6450					break;
6451				case SCTP_PACKET_DROPPED:
6452					peer_supports_pktdrop = 1;
6453					break;
6454				case SCTP_NR_SELECTIVE_ACK:
6455					peer_supports_nrsack = 1;
6456					break;
6457				case SCTP_STREAM_RESET:
6458					peer_supports_reconfig = 1;
6459					break;
6460				case SCTP_AUTHENTICATION:
6461					peer_supports_auth = 1;
6462					break;
6463				case SCTP_IDATA:
6464					peer_supports_idata = 1;
6465					break;
6466				default:
6467					/* one I have not learned yet */
6468					break;
6469
6470				}
6471			}
6472		} else if (ptype == SCTP_RANDOM) {
6473			if (plen > sizeof(random_store))
6474				break;
6475			if (got_random) {
6476				/* already processed a RANDOM */
6477				goto next_param;
6478			}
6479			phdr = sctp_get_next_param(m, offset,
6480			    (struct sctp_paramhdr *)random_store,
6481			    plen);
6482			if (phdr == NULL)
6483				return (-26);
6484			p_random = (struct sctp_auth_random *)phdr;
6485			random_len = plen - sizeof(*p_random);
6486			/* enforce the random length */
6487			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6488				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6489				return (-27);
6490			}
6491			got_random = 1;
6492		} else if (ptype == SCTP_HMAC_LIST) {
6493			uint16_t num_hmacs;
6494			uint16_t i;
6495
6496			if (plen > sizeof(hmacs_store))
6497				break;
6498			if (got_hmacs) {
6499				/* already processed a HMAC list */
6500				goto next_param;
6501			}
6502			phdr = sctp_get_next_param(m, offset,
6503			    (struct sctp_paramhdr *)hmacs_store,
6504			    plen);
6505			if (phdr == NULL)
6506				return (-28);
6507			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6508			hmacs_len = plen - sizeof(*hmacs);
6509			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6510			/* validate the hmac list */
6511			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6512				return (-29);
6513			}
6514			if (stcb->asoc.peer_hmacs != NULL)
6515				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6516			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6517			if (stcb->asoc.peer_hmacs != NULL) {
6518				for (i = 0; i < num_hmacs; i++) {
6519					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6520					    ntohs(hmacs->hmac_ids[i]));
6521				}
6522			}
6523			got_hmacs = 1;
6524		} else if (ptype == SCTP_CHUNK_LIST) {
6525			int i;
6526
6527			if (plen > sizeof(chunks_store))
6528				break;
6529			if (got_chklist) {
6530				/* already processed a Chunks list */
6531				goto next_param;
6532			}
6533			phdr = sctp_get_next_param(m, offset,
6534			    (struct sctp_paramhdr *)chunks_store,
6535			    plen);
6536			if (phdr == NULL)
6537				return (-30);
6538			chunks = (struct sctp_auth_chunk_list *)phdr;
6539			num_chunks = plen - sizeof(*chunks);
6540			if (stcb->asoc.peer_auth_chunks != NULL)
6541				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6542			else
6543				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6544			for (i = 0; i < num_chunks; i++) {
6545				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6546				    stcb->asoc.peer_auth_chunks);
6547				/* record asconf/asconf-ack if listed */
6548				if (chunks->chunk_types[i] == SCTP_ASCONF)
6549					saw_asconf = 1;
6550				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6551					saw_asconf_ack = 1;
6552
6553			}
6554			got_chklist = 1;
6555		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6556			    (ptype == SCTP_STATE_COOKIE) ||
6557			    (ptype == SCTP_UNRECOG_PARAM) ||
6558			    (ptype == SCTP_COOKIE_PRESERVE) ||
6559			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6560			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6561			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6562			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6563		    (ptype == SCTP_SUCCESS_REPORT)) {
6564			 /* don't care */ ;
6565		} else {
6566			if ((ptype & 0x8000) == 0x0000) {
6567				/*
6568				 * must stop processing the rest of the
6569				 * param's. Any report bits were handled
6570				 * with the call to
6571				 * sctp_arethere_unrecognized_parameters()
6572				 * when the INIT or INIT-ACK was first seen.
6573				 */
6574				break;
6575			}
6576		}
6577
6578next_param:
6579		offset += SCTP_SIZE32(plen);
6580		if (offset >= limit) {
6581			break;
6582		}
6583		phdr = sctp_get_next_param(m, offset, &param_buf,
6584		    sizeof(param_buf));
6585	}
6586	/* Now check to see if we need to purge any addresses */
6587	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6588		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6589		    SCTP_ADDR_NOT_IN_ASSOC) {
6590			/* This address has been removed from the asoc */
6591			/* remove and free it */
6592			stcb->asoc.numnets--;
6593			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6594			sctp_free_remote_addr(net);
6595			if (net == stcb->asoc.primary_destination) {
6596				stcb->asoc.primary_destination = NULL;
6597				sctp_select_primary_destination(stcb);
6598			}
6599		}
6600	}
6601	if ((stcb->asoc.ecn_supported == 1) &&
6602	    (peer_supports_ecn == 0)) {
6603		stcb->asoc.ecn_supported = 0;
6604	}
6605	if ((stcb->asoc.prsctp_supported == 1) &&
6606	    (peer_supports_prsctp == 0)) {
6607		stcb->asoc.prsctp_supported = 0;
6608	}
6609	if ((stcb->asoc.auth_supported == 1) &&
6610	    ((peer_supports_auth == 0) ||
6611	    (got_random == 0) || (got_hmacs == 0))) {
6612		stcb->asoc.auth_supported = 0;
6613	}
6614	if ((stcb->asoc.asconf_supported == 1) &&
6615	    ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) ||
6616	    (stcb->asoc.auth_supported == 0) ||
6617	    (saw_asconf == 0) || (saw_asconf_ack == 0))) {
6618		stcb->asoc.asconf_supported = 0;
6619	}
6620	if ((stcb->asoc.reconfig_supported == 1) &&
6621	    (peer_supports_reconfig == 0)) {
6622		stcb->asoc.reconfig_supported = 0;
6623	}
6624	if ((stcb->asoc.idata_supported == 1) &&
6625	    (peer_supports_idata == 0)) {
6626		stcb->asoc.idata_supported = 0;
6627	}
6628	if ((stcb->asoc.nrsack_supported == 1) &&
6629	    (peer_supports_nrsack == 0)) {
6630		stcb->asoc.nrsack_supported = 0;
6631	}
6632	if ((stcb->asoc.pktdrop_supported == 1) &&
6633	    (peer_supports_pktdrop == 0)) {
6634		stcb->asoc.pktdrop_supported = 0;
6635	}
6636	/* validate authentication required parameters */
6637	if ((peer_supports_auth == 0) && (got_chklist == 1)) {
6638		/* peer does not support auth but sent a chunks list? */
6639		return (-31);
6640	}
6641	if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) {
6642		/* peer supports asconf but not auth? */
6643		return (-32);
6644	} else if ((peer_supports_asconf == 1) &&
6645		    (peer_supports_auth == 1) &&
6646	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6647		return (-33);
6648	}
6649	/* concatenate the full random key */
6650	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6651	if (chunks != NULL) {
6652		keylen += sizeof(*chunks) + num_chunks;
6653	}
6654	new_key = sctp_alloc_key(keylen);
6655	if (new_key != NULL) {
6656		/* copy in the RANDOM */
6657		if (p_random != NULL) {
6658			keylen = sizeof(*p_random) + random_len;
6659			memcpy(new_key->key, p_random, keylen);
6660		}
6661		/* append in the AUTH chunks */
6662		if (chunks != NULL) {
6663			memcpy(new_key->key + keylen, chunks,
6664			    sizeof(*chunks) + num_chunks);
6665			keylen += sizeof(*chunks) + num_chunks;
6666		}
6667		/* append in the HMACs */
6668		if (hmacs != NULL) {
6669			memcpy(new_key->key + keylen, hmacs,
6670			    sizeof(*hmacs) + hmacs_len);
6671		}
6672	} else {
6673		/* failed to get memory for the key */
6674		return (-34);
6675	}
6676	if (stcb->asoc.authinfo.peer_random != NULL)
6677		sctp_free_key(stcb->asoc.authinfo.peer_random);
6678	stcb->asoc.authinfo.peer_random = new_key;
6679	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6680	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6681
6682	return (0);
6683}
6684
6685int
6686sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6687    struct sctp_nets *net)
6688{
6689	/* make sure the requested primary address exists in the assoc */
6690	if (net == NULL && sa)
6691		net = sctp_findnet(stcb, sa);
6692
6693	if (net == NULL) {
6694		/* didn't find the requested primary address! */
6695		return (-1);
6696	} else {
6697		/* set the primary address */
6698		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6699			/* Must be confirmed, so queue to set */
6700			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6701			return (0);
6702		}
6703		stcb->asoc.primary_destination = net;
6704		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6705			sctp_free_remote_addr(stcb->asoc.alternate);
6706			stcb->asoc.alternate = NULL;
6707		}
6708		net = TAILQ_FIRST(&stcb->asoc.nets);
6709		if (net != stcb->asoc.primary_destination) {
6710			/*
6711			 * first one on the list is NOT the primary
6712			 * sctp_cmpaddr() is much more efficient if the
6713			 * primary is the first on the list, make it so.
6714			 */
6715			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6716			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6717		}
6718		return (0);
6719	}
6720}
6721
6722int
6723sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6724{
6725	/*
6726	 * This function serves two purposes. It will see if a TAG can be
6727	 * re-used and return 1 for yes it is ok and 0 for don't use that
6728	 * tag. A secondary function it will do is purge out old tags that
6729	 * can be removed.
6730	 */
6731	struct sctpvtaghead *chain;
6732	struct sctp_tagblock *twait_block;
6733	struct sctpasochead *head;
6734	struct sctp_tcb *stcb;
6735	int i;
6736
6737	SCTP_INP_INFO_RLOCK();
6738	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6739	    SCTP_BASE_INFO(hashasocmark))];
6740	LIST_FOREACH(stcb, head, sctp_asocs) {
6741		/*
6742		 * We choose not to lock anything here. TCB's can't be
6743		 * removed since we have the read lock, so they can't be
6744		 * freed on us, same thing for the INP. I may be wrong with
6745		 * this assumption, but we will go with it for now :-)
6746		 */
6747		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6748			continue;
6749		}
6750		if (stcb->asoc.my_vtag == tag) {
6751			/* candidate */
6752			if (stcb->rport != rport) {
6753				continue;
6754			}
6755			if (stcb->sctp_ep->sctp_lport != lport) {
6756				continue;
6757			}
6758			/* Its a used tag set */
6759			SCTP_INP_INFO_RUNLOCK();
6760			return (0);
6761		}
6762	}
6763	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6764	/* Now what about timed wait ? */
6765	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6766		/*
6767		 * Block(s) are present, lets see if we have this tag in the
6768		 * list
6769		 */
6770		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6771			if (twait_block->vtag_block[i].v_tag == 0) {
6772				/* not used */
6773				continue;
6774			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6775			    now->tv_sec) {
6776				/* Audit expires this guy */
6777				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6778				twait_block->vtag_block[i].v_tag = 0;
6779				twait_block->vtag_block[i].lport = 0;
6780				twait_block->vtag_block[i].rport = 0;
6781			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6782				    (twait_block->vtag_block[i].lport == lport) &&
6783			    (twait_block->vtag_block[i].rport == rport)) {
6784				/* Bad tag, sorry :< */
6785				SCTP_INP_INFO_RUNLOCK();
6786				return (0);
6787			}
6788		}
6789	}
6790	SCTP_INP_INFO_RUNLOCK();
6791	return (1);
6792}
6793
6794static void
6795sctp_drain_mbufs(struct sctp_tcb *stcb)
6796{
6797	/*
6798	 * We must hunt this association for MBUF's past the cumack (i.e.
6799	 * out of order data that we can renege on).
6800	 */
6801	struct sctp_association *asoc;
6802	struct sctp_tmit_chunk *chk, *nchk;
6803	uint32_t cumulative_tsn_p1;
6804	struct sctp_queued_to_read *control, *ncontrol;
6805	int cnt, strmat;
6806	uint32_t gap, i;
6807	int fnd = 0;
6808
6809	/* We look for anything larger than the cum-ack + 1 */
6810
6811	asoc = &stcb->asoc;
6812	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6813		/* none we can reneg on. */
6814		return;
6815	}
6816	SCTP_STAT_INCR(sctps_protocol_drains_done);
6817	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6818	cnt = 0;
6819	/* Ok that was fun, now we will drain all the inbound streams? */
6820	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6821		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].inqueue, next_instrm, ncontrol) {
6822#ifdef INVARIANTS
6823			if (control->on_strm_q != SCTP_ON_ORDERED) {
6824				panic("Huh control: %p on_q: %d -- not ordered?",
6825				    control, control->on_strm_q);
6826			}
6827#endif
6828			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6829				/* Yep it is above cum-ack */
6830				cnt++;
6831				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6832				KASSERT(control->length > 0, ("control has zero length"));
6833				if (asoc->size_on_all_streams >= control->length) {
6834					asoc->size_on_all_streams -= control->length;
6835				} else {
6836#ifdef INVARIANTS
6837					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6838#else
6839					asoc->size_on_all_streams = 0;
6840#endif
6841				}
6842				sctp_ucount_decr(asoc->cnt_on_all_streams);
6843				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6844				if (control->on_read_q) {
6845					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6846					control->on_read_q = 0;
6847				}
6848				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, control, next_instrm);
6849				control->on_strm_q = 0;
6850				if (control->data) {
6851					sctp_m_freem(control->data);
6852					control->data = NULL;
6853				}
6854				sctp_free_remote_addr(control->whoFrom);
6855				/* Now its reasm? */
6856				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6857					cnt++;
6858					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6859					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6860					if (asoc->size_on_reasm_queue >= chk->send_size) {
6861						asoc->size_on_reasm_queue -= chk->send_size;
6862					} else {
6863#ifdef INVARIANTS
6864						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6865#else
6866						asoc->size_on_reasm_queue = 0;
6867#endif
6868					}
6869					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6870					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6871					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6872					if (chk->data) {
6873						sctp_m_freem(chk->data);
6874						chk->data = NULL;
6875					}
6876					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6877				}
6878				sctp_free_a_readq(stcb, control);
6879			}
6880		}
6881		TAILQ_FOREACH_SAFE(control, &asoc->strmin[strmat].uno_inqueue, next_instrm, ncontrol) {
6882#ifdef INVARIANTS
6883			if (control->on_strm_q != SCTP_ON_UNORDERED) {
6884				panic("Huh control: %p on_q: %d -- not unordered?",
6885				    control, control->on_strm_q);
6886			}
6887#endif
6888			if (SCTP_TSN_GT(control->sinfo_tsn, cumulative_tsn_p1)) {
6889				/* Yep it is above cum-ack */
6890				cnt++;
6891				SCTP_CALC_TSN_TO_GAP(gap, control->sinfo_tsn, asoc->mapping_array_base_tsn);
6892				KASSERT(control->length > 0, ("control has zero length"));
6893				if (asoc->size_on_all_streams >= control->length) {
6894					asoc->size_on_all_streams -= control->length;
6895				} else {
6896#ifdef INVARIANTS
6897					panic("size_on_all_streams = %u smaller than control length %u", asoc->size_on_all_streams, control->length);
6898#else
6899					asoc->size_on_all_streams = 0;
6900#endif
6901				}
6902				sctp_ucount_decr(asoc->cnt_on_all_streams);
6903				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6904				if (control->on_read_q) {
6905					TAILQ_REMOVE(&stcb->sctp_ep->read_queue, control, next);
6906					control->on_read_q = 0;
6907				}
6908				TAILQ_REMOVE(&asoc->strmin[strmat].uno_inqueue, control, next_instrm);
6909				control->on_strm_q = 0;
6910				if (control->data) {
6911					sctp_m_freem(control->data);
6912					control->data = NULL;
6913				}
6914				sctp_free_remote_addr(control->whoFrom);
6915				/* Now its reasm? */
6916				TAILQ_FOREACH_SAFE(chk, &control->reasm, sctp_next, nchk) {
6917					cnt++;
6918					SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.tsn, asoc->mapping_array_base_tsn);
6919					KASSERT(chk->send_size > 0, ("chunk has zero length"));
6920					if (asoc->size_on_reasm_queue >= chk->send_size) {
6921						asoc->size_on_reasm_queue -= chk->send_size;
6922					} else {
6923#ifdef INVARIANTS
6924						panic("size_on_reasm_queue = %u smaller than chunk length %u", asoc->size_on_reasm_queue, chk->send_size);
6925#else
6926						asoc->size_on_reasm_queue = 0;
6927#endif
6928					}
6929					sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6930					SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6931					TAILQ_REMOVE(&control->reasm, chk, sctp_next);
6932					if (chk->data) {
6933						sctp_m_freem(chk->data);
6934						chk->data = NULL;
6935					}
6936					sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6937				}
6938				sctp_free_a_readq(stcb, control);
6939			}
6940		}
6941	}
6942	if (cnt) {
6943		/* We must back down to see what the new highest is */
6944		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6945			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6946			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6947				asoc->highest_tsn_inside_map = i;
6948				fnd = 1;
6949				break;
6950			}
6951		}
6952		if (!fnd) {
6953			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6954		}
6955		/*
6956		 * Question, should we go through the delivery queue? The
6957		 * only reason things are on here is the app not reading OR
6958		 * a p-d-api up. An attacker COULD send enough in to
6959		 * initiate the PD-API and then send a bunch of stuff to
6960		 * other streams... these would wind up on the delivery
6961		 * queue.. and then we would not get to them. But in order
6962		 * to do this I then have to back-track and un-deliver
6963		 * sequence numbers in streams.. el-yucko. I think for now
6964		 * we will NOT look at the delivery queue and leave it to be
6965		 * something to consider later. An alternative would be to
6966		 * abort the P-D-API with a notification and then deliver
6967		 * the data.... Or another method might be to keep track of
6968		 * how many times the situation occurs and if we see a
6969		 * possible attack underway just abort the association.
6970		 */
6971#ifdef SCTP_DEBUG
6972		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6973#endif
6974		/*
6975		 * Now do we need to find a new
6976		 * asoc->highest_tsn_inside_map?
6977		 */
6978		asoc->last_revoke_count = cnt;
6979		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6980		/* sa_ignore NO_NULL_CHK */
6981		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6982		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6983	}
6984	/*
6985	 * Another issue, in un-setting the TSN's in the mapping array we
6986	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6987	 * two things to occur. It may cause us to do extra work in checking
6988	 * for our mapping array movement. More importantly it may cause us
6989	 * to SACK every datagram. This may not be a bad thing though since
6990	 * we will recover once we get our cum-ack above and all this stuff
6991	 * we dumped recovered.
6992	 */
6993}
6994
6995void
6996sctp_drain()
6997{
6998	/*
6999	 * We must walk the PCB lists for ALL associations here. The system
7000	 * is LOW on MBUF's and needs help. This is where reneging will
7001	 * occur. We really hope this does NOT happen!
7002	 */
7003	VNET_ITERATOR_DECL(vnet_iter);
7004	VNET_LIST_RLOCK_NOSLEEP();
7005	VNET_FOREACH(vnet_iter) {
7006		CURVNET_SET(vnet_iter);
7007		struct sctp_inpcb *inp;
7008		struct sctp_tcb *stcb;
7009
7010		SCTP_STAT_INCR(sctps_protocol_drain_calls);
7011		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
7012#ifdef VIMAGE
7013			continue;
7014#else
7015			return;
7016#endif
7017		}
7018		SCTP_INP_INFO_RLOCK();
7019		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
7020			/* For each endpoint */
7021			SCTP_INP_RLOCK(inp);
7022			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
7023				/* For each association */
7024				SCTP_TCB_LOCK(stcb);
7025				sctp_drain_mbufs(stcb);
7026				SCTP_TCB_UNLOCK(stcb);
7027			}
7028			SCTP_INP_RUNLOCK(inp);
7029		}
7030		SCTP_INP_INFO_RUNLOCK();
7031		CURVNET_RESTORE();
7032	}
7033	VNET_LIST_RUNLOCK_NOSLEEP();
7034}
7035
7036/*
7037 * start a new iterator
7038 * iterates through all endpoints and associations based on the pcb_state
7039 * flags and asoc_state.  "af" (mandatory) is executed for all matching
7040 * assocs and "ef" (optional) is executed when the iterator completes.
7041 * "inpf" (optional) is executed for each new endpoint as it is being
7042 * iterated through. inpe (optional) is called when the inp completes
7043 * its way through all the stcbs.
7044 */
7045int
7046sctp_initiate_iterator(inp_func inpf,
7047    asoc_func af,
7048    inp_func inpe,
7049    uint32_t pcb_state,
7050    uint32_t pcb_features,
7051    uint32_t asoc_state,
7052    void *argp,
7053    uint32_t argi,
7054    end_func ef,
7055    struct sctp_inpcb *s_inp,
7056    uint8_t chunk_output_off)
7057{
7058	struct sctp_iterator *it = NULL;
7059
7060	if (af == NULL) {
7061		return (-1);
7062	}
7063	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7064		SCTP_PRINTF("%s: abort on initialize being %d\n", __func__,
7065		    SCTP_BASE_VAR(sctp_pcb_initialized));
7066		return (-1);
7067	}
7068	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
7069	    SCTP_M_ITER);
7070	if (it == NULL) {
7071		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
7072		return (ENOMEM);
7073	}
7074	memset(it, 0, sizeof(*it));
7075	it->function_assoc = af;
7076	it->function_inp = inpf;
7077	if (inpf)
7078		it->done_current_ep = 0;
7079	else
7080		it->done_current_ep = 1;
7081	it->function_atend = ef;
7082	it->pointer = argp;
7083	it->val = argi;
7084	it->pcb_flags = pcb_state;
7085	it->pcb_features = pcb_features;
7086	it->asoc_state = asoc_state;
7087	it->function_inp_end = inpe;
7088	it->no_chunk_output = chunk_output_off;
7089	it->vn = curvnet;
7090	if (s_inp) {
7091		/* Assume lock is held here */
7092		it->inp = s_inp;
7093		SCTP_INP_INCR_REF(it->inp);
7094		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
7095	} else {
7096		SCTP_INP_INFO_RLOCK();
7097		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
7098		if (it->inp) {
7099			SCTP_INP_INCR_REF(it->inp);
7100		}
7101		SCTP_INP_INFO_RUNLOCK();
7102		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
7103
7104	}
7105	SCTP_IPI_ITERATOR_WQ_LOCK();
7106	if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) {
7107		SCTP_IPI_ITERATOR_WQ_UNLOCK();
7108		SCTP_PRINTF("%s: rollback on initialize being %d it=%p\n", __func__,
7109		    SCTP_BASE_VAR(sctp_pcb_initialized), it);
7110		SCTP_FREE(it, SCTP_M_ITER);
7111		return (-1);
7112	}
7113	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
7114	if (sctp_it_ctl.iterator_running == 0) {
7115		sctp_wakeup_iterator();
7116	}
7117	SCTP_IPI_ITERATOR_WQ_UNLOCK();
7118	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
7119	return (0);
7120}
7121