sctp_pcb.c revision 246674
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: head/sys/netinet/sctp_pcb.c 246674 2013-02-11 13:57:03Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#include <netinet/udp.h>
50#ifdef INET6
51#include <netinet6/ip6_var.h>
52#endif
53#include <sys/sched.h>
54#include <sys/smp.h>
55#include <sys/unistd.h>
56
57
58VNET_DEFINE(struct sctp_base_info, system_base_info);
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78
79#endif
80
81void
82sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
83{
84	/*
85	 * We really don't need to lock this, but I will just because it
86	 * does not hurt.
87	 */
88	SCTP_INP_INFO_RLOCK();
89	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
90	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
91	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
92	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
93	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
94	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
95	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
96	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
97	SCTP_INP_INFO_RUNLOCK();
98}
99
100/*-
101 * Addresses are added to VRF's (Virtual Router's). For BSD we
102 * have only the default VRF 0. We maintain a hash list of
103 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
104 * these has a list of addresses. When we add a new address
105 * to a VRF we lookup the ifn/ifn_index, if the ifn does
106 * not exist we create it and add it to the list of IFN's
107 * within the VRF. Once we have the sctp_ifn, we add the
108 * address to the list. So we look something like:
109 *
110 * hash-vrf-table
111 *   vrf-> ifn-> ifn -> ifn
112 *   vrf    |
113 *    ...   +--ifa-> ifa -> ifa
114 *   vrf
115 *
116 * We keep these separate lists since the SCTP subsystem will
117 * point to these from its source address selection nets structure.
118 * When an address is deleted it does not happen right away on
119 * the SCTP side, it gets scheduled. What we do when a
120 * delete happens is immediately remove the address from
121 * the master list and decrement the refcount. As our
122 * addip iterator works through and frees the src address
123 * selection pointing to the sctp_ifa, eventually the refcount
124 * will reach 0 and we will delete it. Note that it is assumed
125 * that any locking on system level ifn/ifa is done at the
126 * caller of these functions and these routines will only
127 * lock the SCTP structures as they add or delete things.
128 *
129 * Other notes on VRF concepts.
130 *  - An endpoint can be in multiple VRF's
131 *  - An association lives within a VRF and only one VRF.
132 *  - Any incoming packet we can deduce the VRF for by
133 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
134 *  - Any downward send call or connect call must supply the
135 *    VRF via ancillary data or via some sort of set default
136 *    VRF socket option call (again for BSD no brainer since
137 *    the VRF is always 0).
138 *  - An endpoint may add multiple VRF's to it.
139 *  - Listening sockets can accept associations in any
140 *    of the VRF's they are in but the assoc will end up
141 *    in only one VRF (gotten from the packet or connect/send).
142 *
143 */
144
145struct sctp_vrf *
146sctp_allocate_vrf(int vrf_id)
147{
148	struct sctp_vrf *vrf = NULL;
149	struct sctp_vrflist *bucket;
150
151	/* First allocate the VRF structure */
152	vrf = sctp_find_vrf(vrf_id);
153	if (vrf) {
154		/* Already allocated */
155		return (vrf);
156	}
157	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
158	    SCTP_M_VRF);
159	if (vrf == NULL) {
160		/* No memory */
161#ifdef INVARIANTS
162		panic("No memory for VRF:%d", vrf_id);
163#endif
164		return (NULL);
165	}
166	/* setup the VRF */
167	memset(vrf, 0, sizeof(struct sctp_vrf));
168	vrf->vrf_id = vrf_id;
169	LIST_INIT(&vrf->ifnlist);
170	vrf->total_ifa_count = 0;
171	vrf->refcount = 0;
172	/* now also setup table ids */
173	SCTP_INIT_VRF_TABLEID(vrf);
174	/* Init the HASH of addresses */
175	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
176	    &vrf->vrf_addr_hashmark);
177	if (vrf->vrf_addr_hash == NULL) {
178		/* No memory */
179#ifdef INVARIANTS
180		panic("No memory for VRF:%d", vrf_id);
181#endif
182		SCTP_FREE(vrf, SCTP_M_VRF);
183		return (NULL);
184	}
185	/* Add it to the hash table */
186	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
187	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
188	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
189	return (vrf);
190}
191
192
193struct sctp_ifn *
194sctp_find_ifn(void *ifn, uint32_t ifn_index)
195{
196	struct sctp_ifn *sctp_ifnp;
197	struct sctp_ifnlist *hash_ifn_head;
198
199	/*
200	 * We assume the lock is held for the addresses if that's wrong
201	 * problems could occur :-)
202	 */
203	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
204	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
205		if (sctp_ifnp->ifn_index == ifn_index) {
206			return (sctp_ifnp);
207		}
208		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
209			return (sctp_ifnp);
210		}
211	}
212	return (NULL);
213}
214
215
216struct sctp_vrf *
217sctp_find_vrf(uint32_t vrf_id)
218{
219	struct sctp_vrflist *bucket;
220	struct sctp_vrf *liste;
221
222	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
223	LIST_FOREACH(liste, bucket, next_vrf) {
224		if (vrf_id == liste->vrf_id) {
225			return (liste);
226		}
227	}
228	return (NULL);
229}
230
231
232void
233sctp_free_vrf(struct sctp_vrf *vrf)
234{
235	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
236		if (vrf->vrf_addr_hash) {
237			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
238			vrf->vrf_addr_hash = NULL;
239		}
240		/* We zero'd the count */
241		LIST_REMOVE(vrf, next_vrf);
242		SCTP_FREE(vrf, SCTP_M_VRF);
243		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
244	}
245}
246
247
248void
249sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
250{
251	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
252		/* We zero'd the count */
253		if (sctp_ifnp->vrf) {
254			sctp_free_vrf(sctp_ifnp->vrf);
255		}
256		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
257		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
258	}
259}
260
261
262void
263sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
264{
265	struct sctp_ifn *sctp_ifnp;
266
267	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
268	if (sctp_ifnp != NULL) {
269		sctp_ifnp->ifn_mtu = mtu;
270	}
271}
272
273
274void
275sctp_free_ifa(struct sctp_ifa *sctp_ifap)
276{
277	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
278		/* We zero'd the count */
279		if (sctp_ifap->ifn_p) {
280			sctp_free_ifn(sctp_ifap->ifn_p);
281		}
282		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
283		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
284	}
285}
286
287
288static void
289sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
290{
291	struct sctp_ifn *found;
292
293	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
294	if (found == NULL) {
295		/* Not in the list.. sorry */
296		return;
297	}
298	if (hold_addr_lock == 0)
299		SCTP_IPI_ADDR_WLOCK();
300	LIST_REMOVE(sctp_ifnp, next_bucket);
301	LIST_REMOVE(sctp_ifnp, next_ifn);
302	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
303	    sctp_ifnp->registered_af);
304	if (hold_addr_lock == 0)
305		SCTP_IPI_ADDR_WUNLOCK();
306	/* Take away the reference, and possibly free it */
307	sctp_free_ifn(sctp_ifnp);
308}
309
310
311void
312sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
313    const char *if_name, uint32_t ifn_index)
314{
315	struct sctp_vrf *vrf;
316	struct sctp_ifa *sctp_ifap;
317
318	SCTP_IPI_ADDR_RLOCK();
319	vrf = sctp_find_vrf(vrf_id);
320	if (vrf == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
322		goto out;
323
324	}
325	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
326	if (sctp_ifap == NULL) {
327		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
328		goto out;
329	}
330	if (sctp_ifap->ifn_p == NULL) {
331		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
332		goto out;
333	}
334	if (if_name) {
335		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
337			    sctp_ifap->ifn_p->ifn_name, if_name);
338			goto out;
339		}
340	} else {
341		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
342			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
343			    sctp_ifap->ifn_p->ifn_index, ifn_index);
344			goto out;
345		}
346	}
347
348	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
349	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
350out:
351	SCTP_IPI_ADDR_RUNLOCK();
352}
353
354
355void
356sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
357    const char *if_name, uint32_t ifn_index)
358{
359	struct sctp_vrf *vrf;
360	struct sctp_ifa *sctp_ifap;
361
362	SCTP_IPI_ADDR_RLOCK();
363	vrf = sctp_find_vrf(vrf_id);
364	if (vrf == NULL) {
365		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
366		goto out;
367
368	}
369	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
370	if (sctp_ifap == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
372		goto out;
373	}
374	if (sctp_ifap->ifn_p == NULL) {
375		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
376		goto out;
377	}
378	if (if_name) {
379		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
380			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
381			    sctp_ifap->ifn_p->ifn_name, if_name);
382			goto out;
383		}
384	} else {
385		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
386			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
387			    sctp_ifap->ifn_p->ifn_index, ifn_index);
388			goto out;
389		}
390	}
391
392	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
393	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
394out:
395	SCTP_IPI_ADDR_RUNLOCK();
396}
397
398
399/*-
400 * Add an ifa to an ifn.
401 * Register the interface as necessary.
402 * NOTE: ADDR write lock MUST be held.
403 */
404static void
405sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
406{
407	int ifa_af;
408
409	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
410	sctp_ifap->ifn_p = sctp_ifnp;
411	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
412	/* update address counts */
413	sctp_ifnp->ifa_count++;
414	ifa_af = sctp_ifap->address.sa.sa_family;
415	switch (ifa_af) {
416#ifdef INET
417	case AF_INET:
418		sctp_ifnp->num_v4++;
419		break;
420#endif
421#ifdef INET6
422	case AF_INET6:
423		sctp_ifnp->num_v6++;
424		break;
425#endif
426	default:
427		break;
428	}
429	if (sctp_ifnp->ifa_count == 1) {
430		/* register the new interface */
431		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
432		sctp_ifnp->registered_af = ifa_af;
433	}
434}
435
436
437/*-
438 * Remove an ifa from its ifn.
439 * If no more addresses exist, remove the ifn too. Otherwise, re-register
440 * the interface based on the remaining address families left.
441 * NOTE: ADDR write lock MUST be held.
442 */
443static void
444sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
445{
446	LIST_REMOVE(sctp_ifap, next_ifa);
447	if (sctp_ifap->ifn_p) {
448		/* update address counts */
449		sctp_ifap->ifn_p->ifa_count--;
450		switch (sctp_ifap->address.sa.sa_family) {
451#ifdef INET
452		case AF_INET:
453			sctp_ifap->ifn_p->num_v4--;
454			break;
455#endif
456#ifdef INET6
457		case AF_INET6:
458			sctp_ifap->ifn_p->num_v6--;
459			break;
460#endif
461		default:
462			break;
463		}
464
465		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
466			/* remove the ifn, possibly freeing it */
467			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
468		} else {
469			/* re-register address family type, if needed */
470			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
471			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
472				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
474				sctp_ifap->ifn_p->registered_af = AF_INET;
475			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
476			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
477				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
478				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
479				sctp_ifap->ifn_p->registered_af = AF_INET6;
480			}
481			/* free the ifn refcount */
482			sctp_free_ifn(sctp_ifap->ifn_p);
483		}
484		sctp_ifap->ifn_p = NULL;
485	}
486}
487
488
489struct sctp_ifa *
490sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
491    uint32_t ifn_type, const char *if_name, void *ifa,
492    struct sockaddr *addr, uint32_t ifa_flags,
493    int dynamic_add)
494{
495	struct sctp_vrf *vrf;
496	struct sctp_ifn *sctp_ifnp = NULL;
497	struct sctp_ifa *sctp_ifap = NULL;
498	struct sctp_ifalist *hash_addr_head;
499	struct sctp_ifnlist *hash_ifn_head;
500	uint32_t hash_of_addr;
501	int new_ifn_af = 0;
502
503#ifdef SCTP_DEBUG
504	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
505	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
506#endif
507	SCTP_IPI_ADDR_WLOCK();
508	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
509	if (sctp_ifnp) {
510		vrf = sctp_ifnp->vrf;
511	} else {
512		vrf = sctp_find_vrf(vrf_id);
513		if (vrf == NULL) {
514			vrf = sctp_allocate_vrf(vrf_id);
515			if (vrf == NULL) {
516				SCTP_IPI_ADDR_WUNLOCK();
517				return (NULL);
518			}
519		}
520	}
521	if (sctp_ifnp == NULL) {
522		/*
523		 * build one and add it, can't hold lock until after malloc
524		 * done though.
525		 */
526		SCTP_IPI_ADDR_WUNLOCK();
527		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
528		    sizeof(struct sctp_ifn), SCTP_M_IFN);
529		if (sctp_ifnp == NULL) {
530#ifdef INVARIANTS
531			panic("No memory for IFN");
532#endif
533			return (NULL);
534		}
535		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
536		sctp_ifnp->ifn_index = ifn_index;
537		sctp_ifnp->ifn_p = ifn;
538		sctp_ifnp->ifn_type = ifn_type;
539		sctp_ifnp->refcount = 0;
540		sctp_ifnp->vrf = vrf;
541		atomic_add_int(&vrf->refcount, 1);
542		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
543		if (if_name != NULL) {
544			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
545		} else {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
547		}
548		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
549		LIST_INIT(&sctp_ifnp->ifalist);
550		SCTP_IPI_ADDR_WLOCK();
551		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
552		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
553		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
554		new_ifn_af = 1;
555	}
556	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
557	if (sctp_ifap) {
558		/* Hmm, it already exists? */
559		if ((sctp_ifap->ifn_p) &&
560		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
561			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
562			    sctp_ifap->ifn_p->ifn_name, ifn_index,
563			    (void *)sctp_ifap);
564			if (new_ifn_af) {
565				/* Remove the created one that we don't want */
566				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
567			}
568			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
569				/* easy to solve, just switch back to active */
570				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
571				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
572				sctp_ifap->ifn_p = sctp_ifnp;
573				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
574			}
575	exit_stage_left:
576			SCTP_IPI_ADDR_WUNLOCK();
577			return (sctp_ifap);
578		} else {
579			if (sctp_ifap->ifn_p) {
580				/*
581				 * The last IFN gets the address, remove the
582				 * old one
583				 */
584				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
585				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
586				    sctp_ifap->ifn_p->ifn_index, if_name,
587				    ifn_index);
588				/* remove the address from the old ifn */
589				sctp_remove_ifa_from_ifn(sctp_ifap);
590				/* move the address over to the new ifn */
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592				goto exit_stage_left;
593			} else {
594				/* repair ifnp which was NULL ? */
595				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
596				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
597				    (void *)sctp_ifnp, (void *)sctp_ifap);
598				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
599			}
600			goto exit_stage_left;
601		}
602	}
603	SCTP_IPI_ADDR_WUNLOCK();
604	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
605	if (sctp_ifap == NULL) {
606#ifdef INVARIANTS
607		panic("No memory for IFA");
608#endif
609		return (NULL);
610	}
611	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
612	sctp_ifap->ifn_p = sctp_ifnp;
613	atomic_add_int(&sctp_ifnp->refcount, 1);
614	sctp_ifap->vrf_id = vrf_id;
615	sctp_ifap->ifa = ifa;
616	memcpy(&sctp_ifap->address, addr, addr->sa_len);
617	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
618	sctp_ifap->flags = ifa_flags;
619	/* Set scope */
620	switch (sctp_ifap->address.sa.sa_family) {
621#ifdef INET
622	case AF_INET:
623		{
624			struct sockaddr_in *sin;
625
626			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
627			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
628			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
629				sctp_ifap->src_is_loop = 1;
630			}
631			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
632				sctp_ifap->src_is_priv = 1;
633			}
634			sctp_ifnp->num_v4++;
635			if (new_ifn_af)
636				new_ifn_af = AF_INET;
637			break;
638		}
639#endif
640#ifdef INET6
641	case AF_INET6:
642		{
643			/* ok to use deprecated addresses? */
644			struct sockaddr_in6 *sin6;
645
646			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
647			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
648			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
649				sctp_ifap->src_is_loop = 1;
650			}
651			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
652				sctp_ifap->src_is_priv = 1;
653			}
654			sctp_ifnp->num_v6++;
655			if (new_ifn_af)
656				new_ifn_af = AF_INET6;
657			break;
658		}
659#endif
660	default:
661		new_ifn_af = 0;
662		break;
663	}
664	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
665
666	if ((sctp_ifap->src_is_priv == 0) &&
667	    (sctp_ifap->src_is_loop == 0)) {
668		sctp_ifap->src_is_glob = 1;
669	}
670	SCTP_IPI_ADDR_WLOCK();
671	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
672	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
673	sctp_ifap->refcount = 1;
674	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
675	sctp_ifnp->ifa_count++;
676	vrf->total_ifa_count++;
677	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
678	if (new_ifn_af) {
679		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
680		sctp_ifnp->registered_af = new_ifn_af;
681	}
682	SCTP_IPI_ADDR_WUNLOCK();
683	if (dynamic_add) {
684		/*
685		 * Bump up the refcount so that when the timer completes it
686		 * will drop back down.
687		 */
688		struct sctp_laddr *wi;
689
690		atomic_add_int(&sctp_ifap->refcount, 1);
691		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
692		if (wi == NULL) {
693			/*
694			 * Gak, what can we do? We have lost an address
695			 * change can you say HOSED?
696			 */
697			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
698			/* Opps, must decrement the count */
699			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
700			    if_name);
701			return (NULL);
702		}
703		SCTP_INCR_LADDR_COUNT();
704		bzero(wi, sizeof(*wi));
705		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
706		wi->ifa = sctp_ifap;
707		wi->action = SCTP_ADD_IP_ADDRESS;
708
709		SCTP_WQ_ADDR_LOCK();
710		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
711		SCTP_WQ_ADDR_UNLOCK();
712
713		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
714		    (struct sctp_inpcb *)NULL,
715		    (struct sctp_tcb *)NULL,
716		    (struct sctp_nets *)NULL);
717	} else {
718		/* it's ready for use */
719		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
720	}
721	return (sctp_ifap);
722}
723
724void
725sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
726    uint32_t ifn_index, const char *if_name)
727{
728	struct sctp_vrf *vrf;
729	struct sctp_ifa *sctp_ifap = NULL;
730
731	SCTP_IPI_ADDR_WLOCK();
732	vrf = sctp_find_vrf(vrf_id);
733	if (vrf == NULL) {
734		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
735		goto out_now;
736	}
737#ifdef SCTP_DEBUG
738	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
739	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
740#endif
741	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
742	if (sctp_ifap) {
743		/* Validate the delete */
744		if (sctp_ifap->ifn_p) {
745			int valid = 0;
746
747			/*-
748			 * The name has priority over the ifn_index
749			 * if its given. We do this especially for
750			 * panda who might recycle indexes fast.
751			 */
752			if (if_name) {
753				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
754					/* They match its a correct delete */
755					valid = 1;
756				}
757			}
758			if (!valid) {
759				/* last ditch check ifn_index */
760				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
761					valid = 1;
762				}
763			}
764			if (!valid) {
765				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
766				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
768				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
769				SCTP_IPI_ADDR_WUNLOCK();
770				return;
771			}
772		}
773		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
774		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
775		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
776		vrf->total_ifa_count--;
777		LIST_REMOVE(sctp_ifap, next_bucket);
778		sctp_remove_ifa_from_ifn(sctp_ifap);
779	}
780#ifdef SCTP_DEBUG
781	else {
782		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
783		    ifn_index);
784		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
785	}
786#endif
787
788out_now:
789	SCTP_IPI_ADDR_WUNLOCK();
790	if (sctp_ifap) {
791		struct sctp_laddr *wi;
792
793		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
794		if (wi == NULL) {
795			/*
796			 * Gak, what can we do? We have lost an address
797			 * change can you say HOSED?
798			 */
799			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
800
801			/* Oops, must decrement the count */
802			sctp_free_ifa(sctp_ifap);
803			return;
804		}
805		SCTP_INCR_LADDR_COUNT();
806		bzero(wi, sizeof(*wi));
807		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
808		wi->ifa = sctp_ifap;
809		wi->action = SCTP_DEL_IP_ADDRESS;
810		SCTP_WQ_ADDR_LOCK();
811		/*
812		 * Should this really be a tailq? As it is we will process
813		 * the newest first :-0
814		 */
815		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
816		SCTP_WQ_ADDR_UNLOCK();
817
818		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
819		    (struct sctp_inpcb *)NULL,
820		    (struct sctp_tcb *)NULL,
821		    (struct sctp_nets *)NULL);
822	}
823	return;
824}
825
826
827static int
828sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
829{
830	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
831	int ipv4_addr_legal, ipv6_addr_legal;
832	struct sctp_vrf *vrf;
833	struct sctp_ifn *sctp_ifn;
834	struct sctp_ifa *sctp_ifa;
835
836	loopback_scope = stcb->asoc.scope.loopback_scope;
837	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
838	local_scope = stcb->asoc.scope.local_scope;
839	site_scope = stcb->asoc.scope.site_scope;
840	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
841	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
842
843	SCTP_IPI_ADDR_RLOCK();
844	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
845	if (vrf == NULL) {
846		/* no vrf, no addresses */
847		SCTP_IPI_ADDR_RUNLOCK();
848		return (0);
849	}
850	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
851		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
852			if ((loopback_scope == 0) &&
853			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
854				continue;
855			}
856			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
857				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
858				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
859					/*
860					 * We allow pending addresses, where
861					 * we have sent an asconf-add to be
862					 * considered valid.
863					 */
864					continue;
865				}
866				switch (sctp_ifa->address.sa.sa_family) {
867#ifdef INET
868				case AF_INET:
869					if (ipv4_addr_legal) {
870						struct sockaddr_in *sin,
871						           *rsin;
872
873						sin = &sctp_ifa->address.sin;
874						rsin = (struct sockaddr_in *)to;
875						if ((ipv4_local_scope == 0) &&
876						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
877							continue;
878						}
879						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
880							SCTP_IPI_ADDR_RUNLOCK();
881							return (1);
882						}
883					}
884					break;
885#endif
886#ifdef INET6
887				case AF_INET6:
888					if (ipv6_addr_legal) {
889						struct sockaddr_in6 *sin6,
890						            *rsin6;
891
892						sin6 = &sctp_ifa->address.sin6;
893						rsin6 = (struct sockaddr_in6 *)to;
894						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
895							if (local_scope == 0)
896								continue;
897							if (sin6->sin6_scope_id == 0) {
898								if (sa6_recoverscope(sin6) != 0)
899									continue;
900							}
901						}
902						if ((site_scope == 0) &&
903						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
904							continue;
905						}
906						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
907							SCTP_IPI_ADDR_RUNLOCK();
908							return (1);
909						}
910					}
911					break;
912#endif
913				default:
914					/* TSNH */
915					break;
916				}
917			}
918		}
919	} else {
920		struct sctp_laddr *laddr;
921
922		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
923			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
924				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
925				continue;
926			}
927			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
928			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
929				/*
930				 * We allow pending addresses, where we have
931				 * sent an asconf-add to be considered
932				 * valid.
933				 */
934				continue;
935			}
936			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
937				continue;
938			}
939			switch (to->sa_family) {
940#ifdef INET
941			case AF_INET:
942				{
943					struct sockaddr_in *sin, *rsin;
944
945					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
946					rsin = (struct sockaddr_in *)to;
947					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
948						SCTP_IPI_ADDR_RUNLOCK();
949						return (1);
950					}
951					break;
952				}
953#endif
954#ifdef INET6
955			case AF_INET6:
956				{
957					struct sockaddr_in6 *sin6, *rsin6;
958
959					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
960					rsin6 = (struct sockaddr_in6 *)to;
961					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
962						SCTP_IPI_ADDR_RUNLOCK();
963						return (1);
964					}
965					break;
966				}
967
968#endif
969			default:
970				/* TSNH */
971				break;
972			}
973
974		}
975	}
976	SCTP_IPI_ADDR_RUNLOCK();
977	return (0);
978}
979
980
981static struct sctp_tcb *
982sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
983    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
984{
985	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
986	/*
987	 * If we support the TCP model, then we must now dig through to see
988	 * if we can find our endpoint in the list of tcp ep's.
989	 */
990	uint16_t lport, rport;
991	struct sctppcbhead *ephead;
992	struct sctp_inpcb *inp;
993	struct sctp_laddr *laddr;
994	struct sctp_tcb *stcb;
995	struct sctp_nets *net;
996
997	if ((to == NULL) || (from == NULL)) {
998		return (NULL);
999	}
1000	switch (to->sa_family) {
1001#ifdef INET
1002	case AF_INET:
1003		if (from->sa_family == AF_INET) {
1004			lport = ((struct sockaddr_in *)to)->sin_port;
1005			rport = ((struct sockaddr_in *)from)->sin_port;
1006		} else {
1007			return (NULL);
1008		}
1009		break;
1010#endif
1011#ifdef INET6
1012	case AF_INET6:
1013		if (from->sa_family == AF_INET6) {
1014			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1015			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1016		} else {
1017			return (NULL);
1018		}
1019		break;
1020#endif
1021	default:
1022		return (NULL);
1023	}
1024	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1025	/*
1026	 * Ok now for each of the guys in this bucket we must look and see:
1027	 * - Does the remote port match. - Does there single association's
1028	 * addresses match this address (to). If so we update p_ep to point
1029	 * to this ep and return the tcb from it.
1030	 */
1031	LIST_FOREACH(inp, ephead, sctp_hash) {
1032		SCTP_INP_RLOCK(inp);
1033		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1034			SCTP_INP_RUNLOCK(inp);
1035			continue;
1036		}
1037		if (lport != inp->sctp_lport) {
1038			SCTP_INP_RUNLOCK(inp);
1039			continue;
1040		}
1041		if (inp->def_vrf_id != vrf_id) {
1042			SCTP_INP_RUNLOCK(inp);
1043			continue;
1044		}
1045		/* check to see if the ep has one of the addresses */
1046		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1047			/* We are NOT bound all, so look further */
1048			int match = 0;
1049
1050			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1051
1052				if (laddr->ifa == NULL) {
1053					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1054					continue;
1055				}
1056				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1057					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1058					continue;
1059				}
1060				if (laddr->ifa->address.sa.sa_family ==
1061				    to->sa_family) {
1062					/* see if it matches */
1063#ifdef INET
1064					if (from->sa_family == AF_INET) {
1065						struct sockaddr_in *intf_addr,
1066						           *sin;
1067
1068						intf_addr = &laddr->ifa->address.sin;
1069						sin = (struct sockaddr_in *)to;
1070						if (sin->sin_addr.s_addr ==
1071						    intf_addr->sin_addr.s_addr) {
1072							match = 1;
1073							break;
1074						}
1075					}
1076#endif
1077#ifdef INET6
1078					if (from->sa_family == AF_INET6) {
1079						struct sockaddr_in6 *intf_addr6;
1080						struct sockaddr_in6 *sin6;
1081
1082						sin6 = (struct sockaddr_in6 *)
1083						    to;
1084						intf_addr6 = &laddr->ifa->address.sin6;
1085
1086						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1087						    intf_addr6)) {
1088							match = 1;
1089							break;
1090						}
1091					}
1092#endif
1093				}
1094			}
1095			if (match == 0) {
1096				/* This endpoint does not have this address */
1097				SCTP_INP_RUNLOCK(inp);
1098				continue;
1099			}
1100		}
1101		/*
1102		 * Ok if we hit here the ep has the address, does it hold
1103		 * the tcb?
1104		 */
1105		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1106		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1107		if (stcb == NULL) {
1108			SCTP_INP_RUNLOCK(inp);
1109			continue;
1110		}
1111		SCTP_TCB_LOCK(stcb);
1112		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1113			SCTP_TCB_UNLOCK(stcb);
1114			SCTP_INP_RUNLOCK(inp);
1115			continue;
1116		}
1117		if (stcb->rport != rport) {
1118			/* remote port does not match. */
1119			SCTP_TCB_UNLOCK(stcb);
1120			SCTP_INP_RUNLOCK(inp);
1121			continue;
1122		}
1123		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1124			SCTP_TCB_UNLOCK(stcb);
1125			SCTP_INP_RUNLOCK(inp);
1126			continue;
1127		}
1128		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1129			SCTP_TCB_UNLOCK(stcb);
1130			SCTP_INP_RUNLOCK(inp);
1131			continue;
1132		}
1133		/* Does this TCB have a matching address? */
1134		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1135
1136			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1137				/* not the same family, can't be a match */
1138				continue;
1139			}
1140			switch (from->sa_family) {
1141#ifdef INET
1142			case AF_INET:
1143				{
1144					struct sockaddr_in *sin, *rsin;
1145
1146					sin = (struct sockaddr_in *)&net->ro._l_addr;
1147					rsin = (struct sockaddr_in *)from;
1148					if (sin->sin_addr.s_addr ==
1149					    rsin->sin_addr.s_addr) {
1150						/* found it */
1151						if (netp != NULL) {
1152							*netp = net;
1153						}
1154						/*
1155						 * Update the endpoint
1156						 * pointer
1157						 */
1158						*inp_p = inp;
1159						SCTP_INP_RUNLOCK(inp);
1160						return (stcb);
1161					}
1162					break;
1163				}
1164#endif
1165#ifdef INET6
1166			case AF_INET6:
1167				{
1168					struct sockaddr_in6 *sin6, *rsin6;
1169
1170					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1171					rsin6 = (struct sockaddr_in6 *)from;
1172					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1173					    rsin6)) {
1174						/* found it */
1175						if (netp != NULL) {
1176							*netp = net;
1177						}
1178						/*
1179						 * Update the endpoint
1180						 * pointer
1181						 */
1182						*inp_p = inp;
1183						SCTP_INP_RUNLOCK(inp);
1184						return (stcb);
1185					}
1186					break;
1187				}
1188#endif
1189			default:
1190				/* TSNH */
1191				break;
1192			}
1193		}
1194		SCTP_TCB_UNLOCK(stcb);
1195		SCTP_INP_RUNLOCK(inp);
1196	}
1197	return (NULL);
1198}
1199
1200
1201/*
1202 * rules for use
1203 *
1204 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1205 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1206 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1207 * NULL.
1208 */
1209
1210struct sctp_tcb *
1211sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1212    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1213{
1214	struct sctpasochead *head;
1215	struct sctp_inpcb *inp;
1216	struct sctp_tcb *stcb = NULL;
1217	struct sctp_nets *net;
1218	uint16_t rport;
1219
1220	inp = *inp_p;
1221	switch (remote->sa_family) {
1222#ifdef INET
1223	case AF_INET:
1224		rport = (((struct sockaddr_in *)remote)->sin_port);
1225		break;
1226#endif
1227#ifdef INET6
1228	case AF_INET6:
1229		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1230		break;
1231#endif
1232	default:
1233		return (NULL);
1234	}
1235	if (locked_tcb) {
1236		/*
1237		 * UN-lock so we can do proper locking here this occurs when
1238		 * called from load_addresses_from_init.
1239		 */
1240		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1241		SCTP_TCB_UNLOCK(locked_tcb);
1242	}
1243	SCTP_INP_INFO_RLOCK();
1244	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1245	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1246		/*-
1247		 * Now either this guy is our listener or it's the
1248		 * connector. If it is the one that issued the connect, then
1249		 * it's only chance is to be the first TCB in the list. If
1250		 * it is the acceptor, then do the special_lookup to hash
1251		 * and find the real inp.
1252		 */
1253		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1254			/* to is peer addr, from is my addr */
1255			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1256			    netp, inp->def_vrf_id);
1257			if ((stcb != NULL) && (locked_tcb == NULL)) {
1258				/* we have a locked tcb, lower refcount */
1259				SCTP_INP_DECR_REF(inp);
1260			}
1261			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1262				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1263				SCTP_TCB_LOCK(locked_tcb);
1264				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1265				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1266			}
1267			SCTP_INP_INFO_RUNLOCK();
1268			return (stcb);
1269		} else {
1270			SCTP_INP_WLOCK(inp);
1271			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1272				goto null_return;
1273			}
1274			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1275			if (stcb == NULL) {
1276				goto null_return;
1277			}
1278			SCTP_TCB_LOCK(stcb);
1279
1280			if (stcb->rport != rport) {
1281				/* remote port does not match. */
1282				SCTP_TCB_UNLOCK(stcb);
1283				goto null_return;
1284			}
1285			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1286				SCTP_TCB_UNLOCK(stcb);
1287				goto null_return;
1288			}
1289			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1290				SCTP_TCB_UNLOCK(stcb);
1291				goto null_return;
1292			}
1293			/* now look at the list of remote addresses */
1294			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1295#ifdef INVARIANTS
1296				if (net == (TAILQ_NEXT(net, sctp_next))) {
1297					panic("Corrupt net list");
1298				}
1299#endif
1300				if (net->ro._l_addr.sa.sa_family !=
1301				    remote->sa_family) {
1302					/* not the same family */
1303					continue;
1304				}
1305				switch (remote->sa_family) {
1306#ifdef INET
1307				case AF_INET:
1308					{
1309						struct sockaddr_in *sin,
1310						           *rsin;
1311
1312						sin = (struct sockaddr_in *)
1313						    &net->ro._l_addr;
1314						rsin = (struct sockaddr_in *)remote;
1315						if (sin->sin_addr.s_addr ==
1316						    rsin->sin_addr.s_addr) {
1317							/* found it */
1318							if (netp != NULL) {
1319								*netp = net;
1320							}
1321							if (locked_tcb == NULL) {
1322								SCTP_INP_DECR_REF(inp);
1323							} else if (locked_tcb != stcb) {
1324								SCTP_TCB_LOCK(locked_tcb);
1325							}
1326							if (locked_tcb) {
1327								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1328							}
1329							SCTP_INP_WUNLOCK(inp);
1330							SCTP_INP_INFO_RUNLOCK();
1331							return (stcb);
1332						}
1333						break;
1334					}
1335#endif
1336#ifdef INET6
1337				case AF_INET6:
1338					{
1339						struct sockaddr_in6 *sin6,
1340						            *rsin6;
1341
1342						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1343						rsin6 = (struct sockaddr_in6 *)remote;
1344						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1345						    rsin6)) {
1346							/* found it */
1347							if (netp != NULL) {
1348								*netp = net;
1349							}
1350							if (locked_tcb == NULL) {
1351								SCTP_INP_DECR_REF(inp);
1352							} else if (locked_tcb != stcb) {
1353								SCTP_TCB_LOCK(locked_tcb);
1354							}
1355							if (locked_tcb) {
1356								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1357							}
1358							SCTP_INP_WUNLOCK(inp);
1359							SCTP_INP_INFO_RUNLOCK();
1360							return (stcb);
1361						}
1362						break;
1363					}
1364#endif
1365				default:
1366					/* TSNH */
1367					break;
1368				}
1369			}
1370			SCTP_TCB_UNLOCK(stcb);
1371		}
1372	} else {
1373		SCTP_INP_WLOCK(inp);
1374		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1375			goto null_return;
1376		}
1377		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1378		    inp->sctp_hashmark)];
1379		if (head == NULL) {
1380			goto null_return;
1381		}
1382		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1383			if (stcb->rport != rport) {
1384				/* remote port does not match */
1385				continue;
1386			}
1387			SCTP_TCB_LOCK(stcb);
1388			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1389				SCTP_TCB_UNLOCK(stcb);
1390				continue;
1391			}
1392			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1393				SCTP_TCB_UNLOCK(stcb);
1394				continue;
1395			}
1396			/* now look at the list of remote addresses */
1397			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1398#ifdef INVARIANTS
1399				if (net == (TAILQ_NEXT(net, sctp_next))) {
1400					panic("Corrupt net list");
1401				}
1402#endif
1403				if (net->ro._l_addr.sa.sa_family !=
1404				    remote->sa_family) {
1405					/* not the same family */
1406					continue;
1407				}
1408				switch (remote->sa_family) {
1409#ifdef INET
1410				case AF_INET:
1411					{
1412						struct sockaddr_in *sin,
1413						           *rsin;
1414
1415						sin = (struct sockaddr_in *)
1416						    &net->ro._l_addr;
1417						rsin = (struct sockaddr_in *)remote;
1418						if (sin->sin_addr.s_addr ==
1419						    rsin->sin_addr.s_addr) {
1420							/* found it */
1421							if (netp != NULL) {
1422								*netp = net;
1423							}
1424							if (locked_tcb == NULL) {
1425								SCTP_INP_DECR_REF(inp);
1426							} else if (locked_tcb != stcb) {
1427								SCTP_TCB_LOCK(locked_tcb);
1428							}
1429							if (locked_tcb) {
1430								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1431							}
1432							SCTP_INP_WUNLOCK(inp);
1433							SCTP_INP_INFO_RUNLOCK();
1434							return (stcb);
1435						}
1436						break;
1437					}
1438#endif
1439#ifdef INET6
1440				case AF_INET6:
1441					{
1442						struct sockaddr_in6 *sin6,
1443						            *rsin6;
1444
1445						sin6 = (struct sockaddr_in6 *)
1446						    &net->ro._l_addr;
1447						rsin6 = (struct sockaddr_in6 *)remote;
1448						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1449						    rsin6)) {
1450							/* found it */
1451							if (netp != NULL) {
1452								*netp = net;
1453							}
1454							if (locked_tcb == NULL) {
1455								SCTP_INP_DECR_REF(inp);
1456							} else if (locked_tcb != stcb) {
1457								SCTP_TCB_LOCK(locked_tcb);
1458							}
1459							if (locked_tcb) {
1460								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1461							}
1462							SCTP_INP_WUNLOCK(inp);
1463							SCTP_INP_INFO_RUNLOCK();
1464							return (stcb);
1465						}
1466						break;
1467					}
1468#endif
1469				default:
1470					/* TSNH */
1471					break;
1472				}
1473			}
1474			SCTP_TCB_UNLOCK(stcb);
1475		}
1476	}
1477null_return:
1478	/* clean up for returning null */
1479	if (locked_tcb) {
1480		SCTP_TCB_LOCK(locked_tcb);
1481		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1482	}
1483	SCTP_INP_WUNLOCK(inp);
1484	SCTP_INP_INFO_RUNLOCK();
1485	/* not found */
1486	return (NULL);
1487}
1488
1489
1490/*
1491 * Find an association for a specific endpoint using the association id given
1492 * out in the COMM_UP notification
1493 */
1494struct sctp_tcb *
1495sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1496{
1497	/*
1498	 * Use my the assoc_id to find a endpoint
1499	 */
1500	struct sctpasochead *head;
1501	struct sctp_tcb *stcb;
1502	uint32_t id;
1503
1504	if (inp == NULL) {
1505		SCTP_PRINTF("TSNH ep_associd\n");
1506		return (NULL);
1507	}
1508	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1509		SCTP_PRINTF("TSNH ep_associd0\n");
1510		return (NULL);
1511	}
1512	id = (uint32_t) asoc_id;
1513	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1514	if (head == NULL) {
1515		/* invalid id TSNH */
1516		SCTP_PRINTF("TSNH ep_associd1\n");
1517		return (NULL);
1518	}
1519	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1520		if (stcb->asoc.assoc_id == id) {
1521			if (inp != stcb->sctp_ep) {
1522				/*
1523				 * some other guy has the same id active (id
1524				 * collision ??).
1525				 */
1526				SCTP_PRINTF("TSNH ep_associd2\n");
1527				continue;
1528			}
1529			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1530				continue;
1531			}
1532			if (want_lock) {
1533				SCTP_TCB_LOCK(stcb);
1534			}
1535			return (stcb);
1536		}
1537	}
1538	return (NULL);
1539}
1540
1541
1542struct sctp_tcb *
1543sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1544{
1545	struct sctp_tcb *stcb;
1546
1547	SCTP_INP_RLOCK(inp);
1548	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1549	SCTP_INP_RUNLOCK(inp);
1550	return (stcb);
1551}
1552
1553
1554/*
1555 * Endpoint probe expects that the INP_INFO is locked.
1556 */
1557static struct sctp_inpcb *
1558sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1559    uint16_t lport, uint32_t vrf_id)
1560{
1561	struct sctp_inpcb *inp;
1562	struct sctp_laddr *laddr;
1563
1564#ifdef INET
1565	struct sockaddr_in *sin;
1566
1567#endif
1568#ifdef INET6
1569	struct sockaddr_in6 *sin6;
1570	struct sockaddr_in6 *intf_addr6;
1571
1572#endif
1573	int fnd;
1574
1575#ifdef INET
1576	sin = NULL;
1577#endif
1578#ifdef INET6
1579	sin6 = NULL;
1580#endif
1581	switch (nam->sa_family) {
1582#ifdef INET
1583	case AF_INET:
1584		sin = (struct sockaddr_in *)nam;
1585		break;
1586#endif
1587#ifdef INET6
1588	case AF_INET6:
1589		sin6 = (struct sockaddr_in6 *)nam;
1590		break;
1591#endif
1592	default:
1593		/* unsupported family */
1594		return (NULL);
1595	}
1596
1597	if (head == NULL)
1598		return (NULL);
1599
1600	LIST_FOREACH(inp, head, sctp_hash) {
1601		SCTP_INP_RLOCK(inp);
1602		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1603			SCTP_INP_RUNLOCK(inp);
1604			continue;
1605		}
1606		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1607		    (inp->sctp_lport == lport)) {
1608			/* got it */
1609#ifdef INET
1610			if ((nam->sa_family == AF_INET) &&
1611			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1612			    SCTP_IPV6_V6ONLY(inp)) {
1613				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1614				SCTP_INP_RUNLOCK(inp);
1615				continue;
1616			}
1617#endif
1618#ifdef INET6
1619			/* A V6 address and the endpoint is NOT bound V6 */
1620			if (nam->sa_family == AF_INET6 &&
1621			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1622				SCTP_INP_RUNLOCK(inp);
1623				continue;
1624			}
1625#endif
1626			/* does a VRF id match? */
1627			fnd = 0;
1628			if (inp->def_vrf_id == vrf_id)
1629				fnd = 1;
1630
1631			SCTP_INP_RUNLOCK(inp);
1632			if (!fnd)
1633				continue;
1634			return (inp);
1635		}
1636		SCTP_INP_RUNLOCK(inp);
1637	}
1638	switch (nam->sa_family) {
1639#ifdef INET
1640	case AF_INET:
1641		if (sin->sin_addr.s_addr == INADDR_ANY) {
1642			/* Can't hunt for one that has no address specified */
1643			return (NULL);
1644		}
1645		break;
1646#endif
1647#ifdef INET6
1648	case AF_INET6:
1649		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1650			/* Can't hunt for one that has no address specified */
1651			return (NULL);
1652		}
1653		break;
1654#endif
1655	default:
1656		break;
1657	}
1658	/*
1659	 * ok, not bound to all so see if we can find a EP bound to this
1660	 * address.
1661	 */
1662	LIST_FOREACH(inp, head, sctp_hash) {
1663		SCTP_INP_RLOCK(inp);
1664		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1665			SCTP_INP_RUNLOCK(inp);
1666			continue;
1667		}
1668		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1669			SCTP_INP_RUNLOCK(inp);
1670			continue;
1671		}
1672		/*
1673		 * Ok this could be a likely candidate, look at all of its
1674		 * addresses
1675		 */
1676		if (inp->sctp_lport != lport) {
1677			SCTP_INP_RUNLOCK(inp);
1678			continue;
1679		}
1680		/* does a VRF id match? */
1681		fnd = 0;
1682		if (inp->def_vrf_id == vrf_id)
1683			fnd = 1;
1684
1685		if (!fnd) {
1686			SCTP_INP_RUNLOCK(inp);
1687			continue;
1688		}
1689		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1690			if (laddr->ifa == NULL) {
1691				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1692				    __FUNCTION__);
1693				continue;
1694			}
1695			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1696			    (void *)laddr->ifa);
1697			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1698				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1699				continue;
1700			}
1701			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1702				/* possible, see if it matches */
1703				switch (nam->sa_family) {
1704#ifdef INET
1705				case AF_INET:
1706					if (sin->sin_addr.s_addr ==
1707					    laddr->ifa->address.sin.sin_addr.s_addr) {
1708						SCTP_INP_RUNLOCK(inp);
1709						return (inp);
1710					}
1711					break;
1712#endif
1713#ifdef INET6
1714				case AF_INET6:
1715					intf_addr6 = &laddr->ifa->address.sin6;
1716					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1717					    intf_addr6)) {
1718						SCTP_INP_RUNLOCK(inp);
1719						return (inp);
1720					}
1721					break;
1722#endif
1723				}
1724			}
1725		}
1726		SCTP_INP_RUNLOCK(inp);
1727	}
1728	return (NULL);
1729}
1730
1731
1732static struct sctp_inpcb *
1733sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1734{
1735	struct sctppcbhead *head;
1736	struct sctp_inpcb *t_inp;
1737	int fnd;
1738
1739	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1740	    SCTP_BASE_INFO(hashmark))];
1741	LIST_FOREACH(t_inp, head, sctp_hash) {
1742		if (t_inp->sctp_lport != lport) {
1743			continue;
1744		}
1745		/* is it in the VRF in question */
1746		fnd = 0;
1747		if (t_inp->def_vrf_id == vrf_id)
1748			fnd = 1;
1749		if (!fnd)
1750			continue;
1751
1752		/* This one is in use. */
1753		/* check the v6/v4 binding issue */
1754		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1755		    SCTP_IPV6_V6ONLY(t_inp)) {
1756			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1757				/* collision in V6 space */
1758				return (t_inp);
1759			} else {
1760				/* inp is BOUND_V4 no conflict */
1761				continue;
1762			}
1763		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1764			/* t_inp is bound v4 and v6, conflict always */
1765			return (t_inp);
1766		} else {
1767			/* t_inp is bound only V4 */
1768			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1769			    SCTP_IPV6_V6ONLY(inp)) {
1770				/* no conflict */
1771				continue;
1772			}
1773			/* else fall through to conflict */
1774		}
1775		return (t_inp);
1776	}
1777	return (NULL);
1778}
1779
1780
1781int
1782sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1783{
1784	/* For 1-2-1 with port reuse */
1785	struct sctppcbhead *head;
1786	struct sctp_inpcb *tinp;
1787
1788	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1789		/* only works with port reuse on */
1790		return (-1);
1791	}
1792	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1793		return (0);
1794	}
1795	SCTP_INP_RUNLOCK(inp);
1796	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1797	    SCTP_BASE_INFO(hashmark))];
1798	/* Kick out all non-listeners to the TCP hash */
1799	LIST_FOREACH(tinp, head, sctp_hash) {
1800		if (tinp->sctp_lport != inp->sctp_lport) {
1801			continue;
1802		}
1803		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1804			continue;
1805		}
1806		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1807			continue;
1808		}
1809		if (tinp->sctp_socket->so_qlimit) {
1810			continue;
1811		}
1812		SCTP_INP_WLOCK(tinp);
1813		LIST_REMOVE(tinp, sctp_hash);
1814		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1815		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1816		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1817		SCTP_INP_WUNLOCK(tinp);
1818	}
1819	SCTP_INP_WLOCK(inp);
1820	/* Pull from where he was */
1821	LIST_REMOVE(inp, sctp_hash);
1822	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1823	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1824	LIST_INSERT_HEAD(head, inp, sctp_hash);
1825	SCTP_INP_WUNLOCK(inp);
1826	SCTP_INP_RLOCK(inp);
1827	return (0);
1828}
1829
1830
1831struct sctp_inpcb *
1832sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1833    uint32_t vrf_id)
1834{
1835	/*
1836	 * First we check the hash table to see if someone has this port
1837	 * bound with just the port.
1838	 */
1839	struct sctp_inpcb *inp;
1840	struct sctppcbhead *head;
1841	int lport;
1842	unsigned int i;
1843
1844#ifdef INET
1845	struct sockaddr_in *sin;
1846
1847#endif
1848#ifdef INET6
1849	struct sockaddr_in6 *sin6;
1850
1851#endif
1852
1853	switch (nam->sa_family) {
1854#ifdef INET
1855	case AF_INET:
1856		sin = (struct sockaddr_in *)nam;
1857		lport = sin->sin_port;
1858		break;
1859#endif
1860#ifdef INET6
1861	case AF_INET6:
1862		sin6 = (struct sockaddr_in6 *)nam;
1863		lport = sin6->sin6_port;
1864		break;
1865#endif
1866	default:
1867		return (NULL);
1868	}
1869	/*
1870	 * I could cheat here and just cast to one of the types but we will
1871	 * do it right. It also provides the check against an Unsupported
1872	 * type too.
1873	 */
1874	/* Find the head of the ALLADDR chain */
1875	if (have_lock == 0) {
1876		SCTP_INP_INFO_RLOCK();
1877	}
1878	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1879	    SCTP_BASE_INFO(hashmark))];
1880	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1881
1882	/*
1883	 * If the TCP model exists it could be that the main listening
1884	 * endpoint is gone but there still exists a connected socket for
1885	 * this guy. If so we can return the first one that we find. This
1886	 * may NOT be the correct one so the caller should be wary on the
1887	 * returned INP. Currently the only caller that sets find_tcp_pool
1888	 * is in bindx where we are verifying that a user CAN bind the
1889	 * address. He either has bound it already, or someone else has, or
1890	 * its open to bind, so this is good enough.
1891	 */
1892	if (inp == NULL && find_tcp_pool) {
1893		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1894			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1895			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1896			if (inp) {
1897				break;
1898			}
1899		}
1900	}
1901	if (inp) {
1902		SCTP_INP_INCR_REF(inp);
1903	}
1904	if (have_lock == 0) {
1905		SCTP_INP_INFO_RUNLOCK();
1906	}
1907	return (inp);
1908}
1909
1910
1911/*
1912 * Find an association for an endpoint with the pointer to whom you want to
1913 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1914 * need to change the *to to some other struct like a mbuf...
1915 */
1916struct sctp_tcb *
1917sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1918    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1919    uint32_t vrf_id)
1920{
1921	struct sctp_inpcb *inp = NULL;
1922	struct sctp_tcb *stcb;
1923
1924	SCTP_INP_INFO_RLOCK();
1925	if (find_tcp_pool) {
1926		if (inp_p != NULL) {
1927			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1928			    vrf_id);
1929		} else {
1930			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1931			    vrf_id);
1932		}
1933		if (stcb != NULL) {
1934			SCTP_INP_INFO_RUNLOCK();
1935			return (stcb);
1936		}
1937	}
1938	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1939	if (inp_p != NULL) {
1940		*inp_p = inp;
1941	}
1942	SCTP_INP_INFO_RUNLOCK();
1943	if (inp == NULL) {
1944		return (NULL);
1945	}
1946	/*
1947	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1948	 * we now place the source address or from in the to of the find
1949	 * endpoint call. Since in reality this chain is used from the
1950	 * inbound packet side.
1951	 */
1952	if (inp_p != NULL) {
1953		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1954		    NULL);
1955	} else {
1956		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
1957		    NULL);
1958	}
1959	return (stcb);
1960}
1961
1962
1963/*
1964 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1965 * find all addresses that the sender has specified in any address list. Each
1966 * address will be used to lookup the TCB and see if one exits.
1967 */
1968static struct sctp_tcb *
1969sctp_findassociation_special_addr(struct mbuf *m, int offset,
1970    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1971    struct sockaddr *dst)
1972{
1973	struct sctp_paramhdr *phdr, parm_buf;
1974	struct sctp_tcb *stcb;
1975	uint32_t ptype, plen;
1976
1977#ifdef INET
1978	struct sockaddr_in sin4;
1979
1980#endif
1981#ifdef INET6
1982	struct sockaddr_in6 sin6;
1983
1984#endif
1985
1986#ifdef INET
1987	memset(&sin4, 0, sizeof(sin4));
1988	sin4.sin_len = sizeof(sin4);
1989	sin4.sin_family = AF_INET;
1990	sin4.sin_port = sh->src_port;
1991#endif
1992#ifdef INET6
1993	memset(&sin6, 0, sizeof(sin6));
1994	sin6.sin6_len = sizeof(sin6);
1995	sin6.sin6_family = AF_INET6;
1996	sin6.sin6_port = sh->src_port;
1997#endif
1998
1999	stcb = NULL;
2000	offset += sizeof(struct sctp_init_chunk);
2001
2002	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2003	while (phdr != NULL) {
2004		/* now we must see if we want the parameter */
2005		ptype = ntohs(phdr->param_type);
2006		plen = ntohs(phdr->param_length);
2007		if (plen == 0) {
2008			break;
2009		}
2010#ifdef INET
2011		if (ptype == SCTP_IPV4_ADDRESS &&
2012		    plen == sizeof(struct sctp_ipv4addr_param)) {
2013			/* Get the rest of the address */
2014			struct sctp_ipv4addr_param ip4_parm, *p4;
2015
2016			phdr = sctp_get_next_param(m, offset,
2017			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2018			if (phdr == NULL) {
2019				return (NULL);
2020			}
2021			p4 = (struct sctp_ipv4addr_param *)phdr;
2022			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2023			/* look it up */
2024			stcb = sctp_findassociation_ep_addr(inp_p,
2025			    (struct sockaddr *)&sin4, netp, dst, NULL);
2026			if (stcb != NULL) {
2027				return (stcb);
2028			}
2029		}
2030#endif
2031#ifdef INET6
2032		if (ptype == SCTP_IPV6_ADDRESS &&
2033		    plen == sizeof(struct sctp_ipv6addr_param)) {
2034			/* Get the rest of the address */
2035			struct sctp_ipv6addr_param ip6_parm, *p6;
2036
2037			phdr = sctp_get_next_param(m, offset,
2038			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2039			if (phdr == NULL) {
2040				return (NULL);
2041			}
2042			p6 = (struct sctp_ipv6addr_param *)phdr;
2043			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2044			/* look it up */
2045			stcb = sctp_findassociation_ep_addr(inp_p,
2046			    (struct sockaddr *)&sin6, netp, dst, NULL);
2047			if (stcb != NULL) {
2048				return (stcb);
2049			}
2050		}
2051#endif
2052		offset += SCTP_SIZE32(plen);
2053		phdr = sctp_get_next_param(m, offset, &parm_buf,
2054		    sizeof(parm_buf));
2055	}
2056	return (NULL);
2057}
2058
2059static struct sctp_tcb *
2060sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2061    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2062    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2063{
2064	/*
2065	 * Use my vtag to hash. If we find it we then verify the source addr
2066	 * is in the assoc. If all goes well we save a bit on rec of a
2067	 * packet.
2068	 */
2069	struct sctpasochead *head;
2070	struct sctp_nets *net;
2071	struct sctp_tcb *stcb;
2072
2073	SCTP_INP_INFO_RLOCK();
2074	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2075	    SCTP_BASE_INFO(hashasocmark))];
2076	if (head == NULL) {
2077		/* invalid vtag */
2078		SCTP_INP_INFO_RUNLOCK();
2079		return (NULL);
2080	}
2081	LIST_FOREACH(stcb, head, sctp_asocs) {
2082		SCTP_INP_RLOCK(stcb->sctp_ep);
2083		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2084			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2085			continue;
2086		}
2087		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2088			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2089			continue;
2090		}
2091		SCTP_TCB_LOCK(stcb);
2092		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2093		if (stcb->asoc.my_vtag == vtag) {
2094			/* candidate */
2095			if (stcb->rport != rport) {
2096				SCTP_TCB_UNLOCK(stcb);
2097				continue;
2098			}
2099			if (stcb->sctp_ep->sctp_lport != lport) {
2100				SCTP_TCB_UNLOCK(stcb);
2101				continue;
2102			}
2103			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2104				SCTP_TCB_UNLOCK(stcb);
2105				continue;
2106			}
2107			/* RRS:Need toaddr check here */
2108			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2109				/* Endpoint does not own this address */
2110				SCTP_TCB_UNLOCK(stcb);
2111				continue;
2112			}
2113			if (remote_tag) {
2114				/*
2115				 * If we have both vtags that's all we match
2116				 * on
2117				 */
2118				if (stcb->asoc.peer_vtag == remote_tag) {
2119					/*
2120					 * If both tags match we consider it
2121					 * conclusive and check NO
2122					 * source/destination addresses
2123					 */
2124					goto conclusive;
2125				}
2126			}
2127			if (skip_src_check) {
2128		conclusive:
2129				if (from) {
2130					*netp = sctp_findnet(stcb, from);
2131				} else {
2132					*netp = NULL;	/* unknown */
2133				}
2134				if (inp_p)
2135					*inp_p = stcb->sctp_ep;
2136				SCTP_INP_INFO_RUNLOCK();
2137				return (stcb);
2138			}
2139			net = sctp_findnet(stcb, from);
2140			if (net) {
2141				/* yep its him. */
2142				*netp = net;
2143				SCTP_STAT_INCR(sctps_vtagexpress);
2144				*inp_p = stcb->sctp_ep;
2145				SCTP_INP_INFO_RUNLOCK();
2146				return (stcb);
2147			} else {
2148				/*
2149				 * not him, this should only happen in rare
2150				 * cases so I peg it.
2151				 */
2152				SCTP_STAT_INCR(sctps_vtagbogus);
2153			}
2154		}
2155		SCTP_TCB_UNLOCK(stcb);
2156	}
2157	SCTP_INP_INFO_RUNLOCK();
2158	return (NULL);
2159}
2160
2161
2162/*
2163 * Find an association with the pointer to the inbound IP packet. This can be
2164 * a IPv4 or IPv6 packet.
2165 */
2166struct sctp_tcb *
2167sctp_findassociation_addr(struct mbuf *m, int offset,
2168    struct sockaddr *src, struct sockaddr *dst,
2169    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2170    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2171{
2172	int find_tcp_pool;
2173	struct sctp_tcb *stcb;
2174	struct sctp_inpcb *inp;
2175
2176	if (sh->v_tag) {
2177		/* we only go down this path if vtag is non-zero */
2178		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2179		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2180		if (stcb) {
2181			return (stcb);
2182		}
2183	}
2184	find_tcp_pool = 0;
2185	if ((ch->chunk_type != SCTP_INITIATION) &&
2186	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2187	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2188	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2189		/* Other chunk types go to the tcp pool. */
2190		find_tcp_pool = 1;
2191	}
2192	if (inp_p) {
2193		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2194		    find_tcp_pool, vrf_id);
2195		inp = *inp_p;
2196	} else {
2197		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2198		    find_tcp_pool, vrf_id);
2199	}
2200	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2201	if (stcb == NULL && inp) {
2202		/* Found a EP but not this address */
2203		if ((ch->chunk_type == SCTP_INITIATION) ||
2204		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2205			/*-
2206			 * special hook, we do NOT return linp or an
2207			 * association that is linked to an existing
2208			 * association that is under the TCP pool (i.e. no
2209			 * listener exists). The endpoint finding routine
2210			 * will always find a listener before examining the
2211			 * TCP pool.
2212			 */
2213			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2214				if (inp_p) {
2215					*inp_p = NULL;
2216				}
2217				return (NULL);
2218			}
2219			stcb = sctp_findassociation_special_addr(m,
2220			    offset, sh, &inp, netp, dst);
2221			if (inp_p != NULL) {
2222				*inp_p = inp;
2223			}
2224		}
2225	}
2226	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2227	return (stcb);
2228}
2229
2230/*
2231 * lookup an association by an ASCONF lookup address.
2232 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2233 */
2234struct sctp_tcb *
2235sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2236    struct sockaddr *dst, struct sctphdr *sh,
2237    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2238{
2239	struct sctp_tcb *stcb;
2240	struct sockaddr_storage remote_store;
2241	struct sctp_paramhdr parm_buf, *phdr;
2242	int ptype;
2243	int zero_address = 0;
2244
2245#ifdef INET
2246	struct sockaddr_in *sin;
2247
2248#endif
2249#ifdef INET6
2250	struct sockaddr_in6 *sin6;
2251
2252#endif
2253
2254	memset(&remote_store, 0, sizeof(remote_store));
2255	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2256	    &parm_buf, sizeof(struct sctp_paramhdr));
2257	if (phdr == NULL) {
2258		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2259		    __FUNCTION__);
2260		return NULL;
2261	}
2262	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2263	/* get the correlation address */
2264	switch (ptype) {
2265#ifdef INET6
2266	case SCTP_IPV6_ADDRESS:
2267		{
2268			/* ipv6 address param */
2269			struct sctp_ipv6addr_param *p6, p6_buf;
2270
2271			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2272				return NULL;
2273			}
2274			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2275			    offset + sizeof(struct sctp_asconf_chunk),
2276			    &p6_buf.ph, sizeof(*p6));
2277			if (p6 == NULL) {
2278				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2279				    __FUNCTION__);
2280				return (NULL);
2281			}
2282			sin6 = (struct sockaddr_in6 *)&remote_store;
2283			sin6->sin6_family = AF_INET6;
2284			sin6->sin6_len = sizeof(*sin6);
2285			sin6->sin6_port = sh->src_port;
2286			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2287			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2288				zero_address = 1;
2289			break;
2290		}
2291#endif
2292#ifdef INET
2293	case SCTP_IPV4_ADDRESS:
2294		{
2295			/* ipv4 address param */
2296			struct sctp_ipv4addr_param *p4, p4_buf;
2297
2298			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2299				return NULL;
2300			}
2301			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2302			    offset + sizeof(struct sctp_asconf_chunk),
2303			    &p4_buf.ph, sizeof(*p4));
2304			if (p4 == NULL) {
2305				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2306				    __FUNCTION__);
2307				return (NULL);
2308			}
2309			sin = (struct sockaddr_in *)&remote_store;
2310			sin->sin_family = AF_INET;
2311			sin->sin_len = sizeof(*sin);
2312			sin->sin_port = sh->src_port;
2313			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2314			if (sin->sin_addr.s_addr == INADDR_ANY)
2315				zero_address = 1;
2316			break;
2317		}
2318#endif
2319	default:
2320		/* invalid address param type */
2321		return NULL;
2322	}
2323
2324	if (zero_address) {
2325		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2326		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2327		if (stcb != NULL) {
2328			SCTP_INP_DECR_REF(*inp_p);
2329		}
2330	} else {
2331		stcb = sctp_findassociation_ep_addr(inp_p,
2332		    (struct sockaddr *)&remote_store, netp,
2333		    dst, NULL);
2334	}
2335	return (stcb);
2336}
2337
2338
2339/*
2340 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2341 * addresses. This way if we don't get a bind we by default pick a ephemeral
2342 * port with all addresses bound.
2343 */
2344int
2345sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2346{
2347	/*
2348	 * we get called when a new endpoint starts up. We need to allocate
2349	 * the sctp_inpcb structure from the zone and init it. Mark it as
2350	 * unbound and find a port that we can use as an ephemeral with
2351	 * INADDR_ANY. If the user binds later no problem we can then add in
2352	 * the specific addresses. And setup the default parameters for the
2353	 * EP.
2354	 */
2355	int i, error;
2356	struct sctp_inpcb *inp;
2357	struct sctp_pcb *m;
2358	struct timeval time;
2359	sctp_sharedkey_t *null_key;
2360
2361	error = 0;
2362
2363	SCTP_INP_INFO_WLOCK();
2364	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2365	if (inp == NULL) {
2366		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2367		SCTP_INP_INFO_WUNLOCK();
2368		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2369		return (ENOBUFS);
2370	}
2371	/* zap it */
2372	bzero(inp, sizeof(*inp));
2373
2374	/* bump generations */
2375	/* setup socket pointers */
2376	inp->sctp_socket = so;
2377	inp->ip_inp.inp.inp_socket = so;
2378#ifdef INET6
2379	if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2380		inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2381	}
2382#endif
2383	inp->sctp_associd_counter = 1;
2384	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2385	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2386	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2387	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2388	/* init the small hash table we use to track asocid <-> tcb */
2389	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2390	if (inp->sctp_asocidhash == NULL) {
2391		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2392		SCTP_INP_INFO_WUNLOCK();
2393		return (ENOBUFS);
2394	}
2395#ifdef IPSEC
2396	{
2397		struct inpcbpolicy *pcb_sp = NULL;
2398
2399		error = ipsec_init_policy(so, &pcb_sp);
2400		/* Arrange to share the policy */
2401		inp->ip_inp.inp.inp_sp = pcb_sp;
2402		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2403	}
2404	if (error != 0) {
2405		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2406		SCTP_INP_INFO_WUNLOCK();
2407		return error;
2408	}
2409#endif				/* IPSEC */
2410	SCTP_INCR_EP_COUNT();
2411	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2412	SCTP_INP_INFO_WUNLOCK();
2413
2414	so->so_pcb = (caddr_t)inp;
2415
2416	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2417		/* UDP style socket */
2418		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2419		    SCTP_PCB_FLAGS_UNBOUND);
2420		/* Be sure it is NON-BLOCKING IO for UDP */
2421		/* SCTP_SET_SO_NBIO(so); */
2422	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2423		/* TCP style socket */
2424		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2425		    SCTP_PCB_FLAGS_UNBOUND);
2426		/* Be sure we have blocking IO by default */
2427		SCTP_CLEAR_SO_NBIO(so);
2428	} else {
2429		/*
2430		 * unsupported socket type (RAW, etc)- in case we missed it
2431		 * in protosw
2432		 */
2433		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2434		so->so_pcb = NULL;
2435		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2436		return (EOPNOTSUPP);
2437	}
2438	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2439		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2440		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2441	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2442		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2443		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2444	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2445		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2446		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2447	}
2448	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2449	    &inp->sctp_hashmark);
2450	if (inp->sctp_tcbhash == NULL) {
2451		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2452		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2453		so->so_pcb = NULL;
2454		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2455		return (ENOBUFS);
2456	}
2457	inp->def_vrf_id = vrf_id;
2458
2459	SCTP_INP_INFO_WLOCK();
2460	SCTP_INP_LOCK_INIT(inp);
2461	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2462	SCTP_INP_READ_INIT(inp);
2463	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2464	/* lock the new ep */
2465	SCTP_INP_WLOCK(inp);
2466
2467	/* add it to the info area */
2468	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2469	SCTP_INP_INFO_WUNLOCK();
2470
2471	TAILQ_INIT(&inp->read_queue);
2472	LIST_INIT(&inp->sctp_addr_list);
2473
2474	LIST_INIT(&inp->sctp_asoc_list);
2475
2476#ifdef SCTP_TRACK_FREED_ASOCS
2477	/* TEMP CODE */
2478	LIST_INIT(&inp->sctp_asoc_free_list);
2479#endif
2480	/* Init the timer structure for signature change */
2481	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2482	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2483
2484	/* now init the actual endpoint default data */
2485	m = &inp->sctp_ep;
2486
2487	/* setup the base timeout information */
2488	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2489	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2490	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2491	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2492	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2493	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2494	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2495	/* all max/min max are in ms */
2496	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2497	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2498	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2499	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2500	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2501
2502	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2503
2504	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2505	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2506	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2507	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2508	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2509	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2510	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2511	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2512
2513	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2514	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2515	/* number of streams to pre-open on a association */
2516	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2517
2518	/* Add adaptation cookie */
2519	m->adaptation_layer_indicator = 0x504C5253;
2520
2521	/* seed random number generator */
2522	m->random_counter = 1;
2523	m->store_at = SCTP_SIGNATURE_SIZE;
2524	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2525	sctp_fill_random_store(m);
2526
2527	/* Minimum cookie size */
2528	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2529	    sizeof(struct sctp_state_cookie);
2530	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2531
2532	/* Setup the initial secret */
2533	(void)SCTP_GETTIME_TIMEVAL(&time);
2534	m->time_of_secret_change = time.tv_sec;
2535
2536	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2537		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2538	}
2539	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2540
2541	/* How long is a cookie good for ? */
2542	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2543	/*
2544	 * Initialize authentication parameters
2545	 */
2546	m->local_hmacs = sctp_default_supported_hmaclist();
2547	m->local_auth_chunks = sctp_alloc_chunklist();
2548	m->default_dscp = 0;
2549#ifdef INET6
2550	m->default_flowlabel = 0;
2551#endif
2552	m->port = 0;		/* encapsulation disabled by default */
2553	sctp_auth_set_default_chunks(m->local_auth_chunks);
2554	LIST_INIT(&m->shared_keys);
2555	/* add default NULL key as key id 0 */
2556	null_key = sctp_alloc_sharedkey();
2557	sctp_insert_sharedkey(&m->shared_keys, null_key);
2558	SCTP_INP_WUNLOCK(inp);
2559#ifdef SCTP_LOG_CLOSING
2560	sctp_log_closing(inp, NULL, 12);
2561#endif
2562	return (error);
2563}
2564
2565
2566void
2567sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2568    struct sctp_tcb *stcb)
2569{
2570	struct sctp_nets *net;
2571	uint16_t lport, rport;
2572	struct sctppcbhead *head;
2573	struct sctp_laddr *laddr, *oladdr;
2574
2575	atomic_add_int(&stcb->asoc.refcnt, 1);
2576	SCTP_TCB_UNLOCK(stcb);
2577	SCTP_INP_INFO_WLOCK();
2578	SCTP_INP_WLOCK(old_inp);
2579	SCTP_INP_WLOCK(new_inp);
2580	SCTP_TCB_LOCK(stcb);
2581	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2582
2583	new_inp->sctp_ep.time_of_secret_change =
2584	    old_inp->sctp_ep.time_of_secret_change;
2585	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2586	    sizeof(old_inp->sctp_ep.secret_key));
2587	new_inp->sctp_ep.current_secret_number =
2588	    old_inp->sctp_ep.current_secret_number;
2589	new_inp->sctp_ep.last_secret_number =
2590	    old_inp->sctp_ep.last_secret_number;
2591	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2592
2593	/* make it so new data pours into the new socket */
2594	stcb->sctp_socket = new_inp->sctp_socket;
2595	stcb->sctp_ep = new_inp;
2596
2597	/* Copy the port across */
2598	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2599	rport = stcb->rport;
2600	/* Pull the tcb from the old association */
2601	LIST_REMOVE(stcb, sctp_tcbhash);
2602	LIST_REMOVE(stcb, sctp_tcblist);
2603	if (stcb->asoc.in_asocid_hash) {
2604		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2605	}
2606	/* Now insert the new_inp into the TCP connected hash */
2607	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2608
2609	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2610	/* Its safe to access */
2611	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2612
2613	/* Now move the tcb into the endpoint list */
2614	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2615	/*
2616	 * Question, do we even need to worry about the ep-hash since we
2617	 * only have one connection? Probably not :> so lets get rid of it
2618	 * and not suck up any kernel memory in that.
2619	 */
2620	if (stcb->asoc.in_asocid_hash) {
2621		struct sctpasochead *lhd;
2622
2623		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2624		    new_inp->hashasocidmark)];
2625		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2626	}
2627	/* Ok. Let's restart timer. */
2628	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2629		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2630		    stcb, net);
2631	}
2632
2633	SCTP_INP_INFO_WUNLOCK();
2634	if (new_inp->sctp_tcbhash != NULL) {
2635		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2636		new_inp->sctp_tcbhash = NULL;
2637	}
2638	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2639		/* Subset bound, so copy in the laddr list from the old_inp */
2640		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2641			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2642			if (laddr == NULL) {
2643				/*
2644				 * Gak, what can we do? This assoc is really
2645				 * HOSED. We probably should send an abort
2646				 * here.
2647				 */
2648				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2649				continue;
2650			}
2651			SCTP_INCR_LADDR_COUNT();
2652			bzero(laddr, sizeof(*laddr));
2653			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2654			laddr->ifa = oladdr->ifa;
2655			atomic_add_int(&laddr->ifa->refcount, 1);
2656			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2657			    sctp_nxt_addr);
2658			new_inp->laddr_count++;
2659			if (oladdr == stcb->asoc.last_used_address) {
2660				stcb->asoc.last_used_address = laddr;
2661			}
2662		}
2663	}
2664	/*
2665	 * Now any running timers need to be adjusted since we really don't
2666	 * care if they are running or not just blast in the new_inp into
2667	 * all of them.
2668	 */
2669
2670	stcb->asoc.dack_timer.ep = (void *)new_inp;
2671	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2672	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2673	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2674	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2675	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2676	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2677	/* now what about the nets? */
2678	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2679		net->pmtu_timer.ep = (void *)new_inp;
2680		net->hb_timer.ep = (void *)new_inp;
2681		net->rxt_timer.ep = (void *)new_inp;
2682	}
2683	SCTP_INP_WUNLOCK(new_inp);
2684	SCTP_INP_WUNLOCK(old_inp);
2685}
2686
2687
2688
2689
2690/* sctp_ifap is used to bypass normal local address validation checks */
2691int
2692sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2693    struct sctp_ifa *sctp_ifap, struct thread *p)
2694{
2695	/* bind a ep to a socket address */
2696	struct sctppcbhead *head;
2697	struct sctp_inpcb *inp, *inp_tmp;
2698	struct inpcb *ip_inp;
2699	int port_reuse_active = 0;
2700	int bindall;
2701	uint16_t lport;
2702	int error;
2703	uint32_t vrf_id;
2704
2705	lport = 0;
2706	error = 0;
2707	bindall = 1;
2708	inp = (struct sctp_inpcb *)so->so_pcb;
2709	ip_inp = (struct inpcb *)so->so_pcb;
2710#ifdef SCTP_DEBUG
2711	if (addr) {
2712		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2713		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2714		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2715		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2716	}
2717#endif
2718	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2719		/* already did a bind, subsequent binds NOT allowed ! */
2720		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2721		return (EINVAL);
2722	}
2723#ifdef INVARIANTS
2724	if (p == NULL)
2725		panic("null proc/thread");
2726#endif
2727	if (addr != NULL) {
2728		switch (addr->sa_family) {
2729#ifdef INET
2730		case AF_INET:
2731			{
2732				struct sockaddr_in *sin;
2733
2734				/* IPV6_V6ONLY socket? */
2735				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2736					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2737					return (EINVAL);
2738				}
2739				if (addr->sa_len != sizeof(*sin)) {
2740					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2741					return (EINVAL);
2742				}
2743				sin = (struct sockaddr_in *)addr;
2744				lport = sin->sin_port;
2745				/*
2746				 * For LOOPBACK the prison_local_ip4() call
2747				 * will transmute the ip address to the
2748				 * proper value.
2749				 */
2750				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2751					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2752					return (error);
2753				}
2754				if (sin->sin_addr.s_addr != INADDR_ANY) {
2755					bindall = 0;
2756				}
2757				break;
2758			}
2759#endif
2760#ifdef INET6
2761		case AF_INET6:
2762			{
2763				/*
2764				 * Only for pure IPv6 Address. (No IPv4
2765				 * Mapped!)
2766				 */
2767				struct sockaddr_in6 *sin6;
2768
2769				sin6 = (struct sockaddr_in6 *)addr;
2770
2771				if (addr->sa_len != sizeof(*sin6)) {
2772					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2773					return (EINVAL);
2774				}
2775				lport = sin6->sin6_port;
2776				/*
2777				 * For LOOPBACK the prison_local_ip6() call
2778				 * will transmute the ipv6 address to the
2779				 * proper value.
2780				 */
2781				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2782				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2783					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2784					return (error);
2785				}
2786				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2787					bindall = 0;
2788					/* KAME hack: embed scopeid */
2789					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2790						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2791						return (EINVAL);
2792					}
2793				}
2794				/* this must be cleared for ifa_ifwithaddr() */
2795				sin6->sin6_scope_id = 0;
2796				break;
2797			}
2798#endif
2799		default:
2800			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2801			return (EAFNOSUPPORT);
2802		}
2803	}
2804	SCTP_INP_INFO_WLOCK();
2805	SCTP_INP_WLOCK(inp);
2806	/* Setup a vrf_id to be the default for the non-bind-all case. */
2807	vrf_id = inp->def_vrf_id;
2808
2809	/* increase our count due to the unlock we do */
2810	SCTP_INP_INCR_REF(inp);
2811	if (lport) {
2812		/*
2813		 * Did the caller specify a port? if so we must see if an ep
2814		 * already has this one bound.
2815		 */
2816		/* got to be root to get at low ports */
2817		if (ntohs(lport) < IPPORT_RESERVED) {
2818			if (p && (error =
2819			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2820			    )) {
2821				SCTP_INP_DECR_REF(inp);
2822				SCTP_INP_WUNLOCK(inp);
2823				SCTP_INP_INFO_WUNLOCK();
2824				return (error);
2825			}
2826		}
2827		if (p == NULL) {
2828			SCTP_INP_DECR_REF(inp);
2829			SCTP_INP_WUNLOCK(inp);
2830			SCTP_INP_INFO_WUNLOCK();
2831			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2832			return (error);
2833		}
2834		SCTP_INP_WUNLOCK(inp);
2835		if (bindall) {
2836			vrf_id = inp->def_vrf_id;
2837			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2838			if (inp_tmp != NULL) {
2839				/*
2840				 * lock guy returned and lower count note
2841				 * that we are not bound so inp_tmp should
2842				 * NEVER be inp. And it is this inp
2843				 * (inp_tmp) that gets the reference bump,
2844				 * so we must lower it.
2845				 */
2846				SCTP_INP_DECR_REF(inp_tmp);
2847				/* unlock info */
2848				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2849				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2850					/*
2851					 * Ok, must be one-2-one and
2852					 * allowing port re-use
2853					 */
2854					port_reuse_active = 1;
2855					goto continue_anyway;
2856				}
2857				SCTP_INP_DECR_REF(inp);
2858				SCTP_INP_INFO_WUNLOCK();
2859				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2860				return (EADDRINUSE);
2861			}
2862		} else {
2863			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2864			if (inp_tmp != NULL) {
2865				/*
2866				 * lock guy returned and lower count note
2867				 * that we are not bound so inp_tmp should
2868				 * NEVER be inp. And it is this inp
2869				 * (inp_tmp) that gets the reference bump,
2870				 * so we must lower it.
2871				 */
2872				SCTP_INP_DECR_REF(inp_tmp);
2873				/* unlock info */
2874				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2875				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2876					/*
2877					 * Ok, must be one-2-one and
2878					 * allowing port re-use
2879					 */
2880					port_reuse_active = 1;
2881					goto continue_anyway;
2882				}
2883				SCTP_INP_DECR_REF(inp);
2884				SCTP_INP_INFO_WUNLOCK();
2885				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2886				return (EADDRINUSE);
2887			}
2888		}
2889continue_anyway:
2890		SCTP_INP_WLOCK(inp);
2891		if (bindall) {
2892			/* verify that no lport is not used by a singleton */
2893			if ((port_reuse_active == 0) &&
2894			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2895				/* Sorry someone already has this one bound */
2896				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2897				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2898					port_reuse_active = 1;
2899				} else {
2900					SCTP_INP_DECR_REF(inp);
2901					SCTP_INP_WUNLOCK(inp);
2902					SCTP_INP_INFO_WUNLOCK();
2903					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2904					return (EADDRINUSE);
2905				}
2906			}
2907		}
2908	} else {
2909		uint16_t first, last, candidate;
2910		uint16_t count;
2911		int done;
2912
2913		if (ip_inp->inp_flags & INP_HIGHPORT) {
2914			first = MODULE_GLOBAL(ipport_hifirstauto);
2915			last = MODULE_GLOBAL(ipport_hilastauto);
2916		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2917			if (p && (error =
2918			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2919			    )) {
2920				SCTP_INP_DECR_REF(inp);
2921				SCTP_INP_WUNLOCK(inp);
2922				SCTP_INP_INFO_WUNLOCK();
2923				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2924				return (error);
2925			}
2926			first = MODULE_GLOBAL(ipport_lowfirstauto);
2927			last = MODULE_GLOBAL(ipport_lowlastauto);
2928		} else {
2929			first = MODULE_GLOBAL(ipport_firstauto);
2930			last = MODULE_GLOBAL(ipport_lastauto);
2931		}
2932		if (first > last) {
2933			uint16_t temp;
2934
2935			temp = first;
2936			first = last;
2937			last = temp;
2938		}
2939		count = last - first + 1;	/* number of candidates */
2940		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2941
2942		done = 0;
2943		while (!done) {
2944			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2945				done = 1;
2946			}
2947			if (!done) {
2948				if (--count == 0) {
2949					SCTP_INP_DECR_REF(inp);
2950					SCTP_INP_WUNLOCK(inp);
2951					SCTP_INP_INFO_WUNLOCK();
2952					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2953					return (EADDRINUSE);
2954				}
2955				if (candidate == last)
2956					candidate = first;
2957				else
2958					candidate = candidate + 1;
2959			}
2960		}
2961		lport = htons(candidate);
2962	}
2963	SCTP_INP_DECR_REF(inp);
2964	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2965	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2966		/*
2967		 * this really should not happen. The guy did a non-blocking
2968		 * bind and then did a close at the same time.
2969		 */
2970		SCTP_INP_WUNLOCK(inp);
2971		SCTP_INP_INFO_WUNLOCK();
2972		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2973		return (EINVAL);
2974	}
2975	/* ok we look clear to give out this port, so lets setup the binding */
2976	if (bindall) {
2977		/* binding to all addresses, so just set in the proper flags */
2978		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2979		/* set the automatic addr changes from kernel flag */
2980		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2981			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2982			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2983		} else {
2984			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2985			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2986		}
2987		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2988			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2989		} else {
2990			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2991		}
2992		/*
2993		 * set the automatic mobility_base from kernel flag (by
2994		 * micchie)
2995		 */
2996		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
2997			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
2998			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2999		} else {
3000			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3001			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3002		}
3003		/*
3004		 * set the automatic mobility_fasthandoff from kernel flag
3005		 * (by micchie)
3006		 */
3007		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3008			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3009			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3010		} else {
3011			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3012			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3013		}
3014	} else {
3015		/*
3016		 * bind specific, make sure flags is off and add a new
3017		 * address structure to the sctp_addr_list inside the ep
3018		 * structure.
3019		 *
3020		 * We will need to allocate one and insert it at the head. The
3021		 * socketopt call can just insert new addresses in there as
3022		 * well. It will also have to do the embed scope kame hack
3023		 * too (before adding).
3024		 */
3025		struct sctp_ifa *ifa;
3026		struct sockaddr_storage store_sa;
3027
3028		memset(&store_sa, 0, sizeof(store_sa));
3029		switch (addr->sa_family) {
3030#ifdef INET
3031		case AF_INET:
3032			{
3033				struct sockaddr_in *sin;
3034
3035				sin = (struct sockaddr_in *)&store_sa;
3036				memcpy(sin, addr, sizeof(struct sockaddr_in));
3037				sin->sin_port = 0;
3038				break;
3039			}
3040#endif
3041#ifdef INET6
3042		case AF_INET6:
3043			{
3044				struct sockaddr_in6 *sin6;
3045
3046				sin6 = (struct sockaddr_in6 *)&store_sa;
3047				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3048				sin6->sin6_port = 0;
3049				break;
3050			}
3051#endif
3052		default:
3053			break;
3054		}
3055		/*
3056		 * first find the interface with the bound address need to
3057		 * zero out the port to find the address! yuck! can't do
3058		 * this earlier since need port for sctp_pcb_findep()
3059		 */
3060		if (sctp_ifap != NULL) {
3061			ifa = sctp_ifap;
3062		} else {
3063			/*
3064			 * Note for BSD we hit here always other O/S's will
3065			 * pass things in via the sctp_ifap argument
3066			 * (Panda).
3067			 */
3068			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3069			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3070		}
3071		if (ifa == NULL) {
3072			/* Can't find an interface with that address */
3073			SCTP_INP_WUNLOCK(inp);
3074			SCTP_INP_INFO_WUNLOCK();
3075			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3076			return (EADDRNOTAVAIL);
3077		}
3078#ifdef INET6
3079		if (addr->sa_family == AF_INET6) {
3080			/* GAK, more FIXME IFA lock? */
3081			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3082				/* Can't bind a non-existent addr. */
3083				SCTP_INP_WUNLOCK(inp);
3084				SCTP_INP_INFO_WUNLOCK();
3085				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3086				return (EINVAL);
3087			}
3088		}
3089#endif
3090		/* we're not bound all */
3091		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3092		/* allow bindx() to send ASCONF's for binding changes */
3093		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3094		/* clear automatic addr changes from kernel flag */
3095		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3096
3097		/* add this address to the endpoint list */
3098		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3099		if (error != 0) {
3100			SCTP_INP_WUNLOCK(inp);
3101			SCTP_INP_INFO_WUNLOCK();
3102			return (error);
3103		}
3104		inp->laddr_count++;
3105	}
3106	/* find the bucket */
3107	if (port_reuse_active) {
3108		/* Put it into tcp 1-2-1 hash */
3109		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3110		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3111	} else {
3112		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3113	}
3114	/* put it in the bucket */
3115	LIST_INSERT_HEAD(head, inp, sctp_hash);
3116	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3117	    (void *)head, ntohs(lport), port_reuse_active);
3118	/* set in the port */
3119	inp->sctp_lport = lport;
3120
3121	/* turn off just the unbound flag */
3122	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3123	SCTP_INP_WUNLOCK(inp);
3124	SCTP_INP_INFO_WUNLOCK();
3125	return (0);
3126}
3127
3128
3129static void
3130sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3131{
3132	struct sctp_iterator *it, *nit;
3133
3134	/*
3135	 * We enter with the only the ITERATOR_LOCK in place and a write
3136	 * lock on the inp_info stuff.
3137	 */
3138	it = sctp_it_ctl.cur_it;
3139	if (it && (it->vn != curvnet)) {
3140		/* Its not looking at our VNET */
3141		return;
3142	}
3143	if (it && (it->inp == inp)) {
3144		/*
3145		 * This is tricky and we hold the iterator lock, but when it
3146		 * returns and gets the lock (when we release it) the
3147		 * iterator will try to operate on inp. We need to stop that
3148		 * from happening. But of course the iterator has a
3149		 * reference on the stcb and inp. We can mark it and it will
3150		 * stop.
3151		 *
3152		 * If its a single iterator situation, we set the end iterator
3153		 * flag. Otherwise we set the iterator to go to the next
3154		 * inp.
3155		 *
3156		 */
3157		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3158			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3159		} else {
3160			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3161		}
3162	}
3163	/*
3164	 * Now go through and remove any single reference to our inp that
3165	 * may be still pending on the list
3166	 */
3167	SCTP_IPI_ITERATOR_WQ_LOCK();
3168	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3169		if (it->vn != curvnet) {
3170			continue;
3171		}
3172		if (it->inp == inp) {
3173			/* This one points to me is it inp specific? */
3174			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3175				/* Remove and free this one */
3176				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3177				    it, sctp_nxt_itr);
3178				if (it->function_atend != NULL) {
3179					(*it->function_atend) (it->pointer, it->val);
3180				}
3181				SCTP_FREE(it, SCTP_M_ITER);
3182			} else {
3183				it->inp = LIST_NEXT(it->inp, sctp_list);
3184				if (it->inp) {
3185					SCTP_INP_INCR_REF(it->inp);
3186				}
3187			}
3188			/*
3189			 * When its put in the refcnt is incremented so decr
3190			 * it
3191			 */
3192			SCTP_INP_DECR_REF(inp);
3193		}
3194	}
3195	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3196}
3197
3198/* release sctp_inpcb unbind the port */
3199void
3200sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3201{
3202	/*
3203	 * Here we free a endpoint. We must find it (if it is in the Hash
3204	 * table) and remove it from there. Then we must also find it in the
3205	 * overall list and remove it from there. After all removals are
3206	 * complete then any timer has to be stopped. Then start the actual
3207	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3208	 * all associations. d) finally the ep itself.
3209	 */
3210	struct sctp_tcb *asoc, *nasoc;
3211	struct sctp_laddr *laddr, *nladdr;
3212	struct inpcb *ip_pcb;
3213	struct socket *so;
3214	int being_refed = 0;
3215	struct sctp_queued_to_read *sq, *nsq;
3216	int cnt;
3217	sctp_sharedkey_t *shared_key, *nshared_key;
3218
3219
3220#ifdef SCTP_LOG_CLOSING
3221	sctp_log_closing(inp, NULL, 0);
3222#endif
3223	SCTP_ITERATOR_LOCK();
3224	/* mark any iterators on the list or being processed */
3225	sctp_iterator_inp_being_freed(inp);
3226	SCTP_ITERATOR_UNLOCK();
3227	so = inp->sctp_socket;
3228	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3229		/* been here before.. eeks.. get out of here */
3230		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3231#ifdef SCTP_LOG_CLOSING
3232		sctp_log_closing(inp, NULL, 1);
3233#endif
3234		return;
3235	}
3236	SCTP_ASOC_CREATE_LOCK(inp);
3237	SCTP_INP_INFO_WLOCK();
3238
3239	SCTP_INP_WLOCK(inp);
3240	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3241		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3242		/* socket is gone, so no more wakeups allowed */
3243		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3244		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3245		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3246
3247	}
3248	/* First time through we have the socket lock, after that no more. */
3249	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3250	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3251
3252	if (inp->control) {
3253		sctp_m_freem(inp->control);
3254		inp->control = NULL;
3255	}
3256	if (inp->pkt) {
3257		sctp_m_freem(inp->pkt);
3258		inp->pkt = NULL;
3259	}
3260	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3261					 * here but I will be nice :> (i.e.
3262					 * ip_pcb = ep;) */
3263	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3264		int cnt_in_sd;
3265
3266		cnt_in_sd = 0;
3267		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3268			SCTP_TCB_LOCK(asoc);
3269			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3270				/* Skip guys being freed */
3271				cnt_in_sd++;
3272				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3273					/*
3274					 * Special case - we did not start a
3275					 * kill timer on the asoc due to it
3276					 * was not closed. So go ahead and
3277					 * start it now.
3278					 */
3279					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3280					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3281				}
3282				SCTP_TCB_UNLOCK(asoc);
3283				continue;
3284			}
3285			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3286			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3287			    (asoc->asoc.total_output_queue_size == 0)) {
3288				/*
3289				 * If we have data in queue, we don't want
3290				 * to just free since the app may have done,
3291				 * send()/close or connect/send/close. And
3292				 * it wants the data to get across first.
3293				 */
3294				/* Just abandon things in the front states */
3295				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3296				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3297					cnt_in_sd++;
3298				}
3299				continue;
3300			}
3301			/* Disconnect the socket please */
3302			asoc->sctp_socket = NULL;
3303			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3304			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3305			    (asoc->asoc.control_pdapi) ||
3306			    (asoc->asoc.size_on_all_streams > 0) ||
3307			    (so && (so->so_rcv.sb_cc > 0))) {
3308				/* Left with Data unread */
3309				struct mbuf *op_err;
3310
3311				op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3312				    0, M_NOWAIT, 1, MT_DATA);
3313				if (op_err) {
3314					/* Fill in the user initiated abort */
3315					struct sctp_paramhdr *ph;
3316
3317					SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3318					ph = mtod(op_err, struct sctp_paramhdr *);
3319					ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3320					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3321				}
3322				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3323				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3324				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3325				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3326				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3327					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3328				}
3329				if (sctp_free_assoc(inp, asoc,
3330				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3331					cnt_in_sd++;
3332				}
3333				continue;
3334			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3335				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3336			    (asoc->asoc.stream_queue_cnt == 0)) {
3337				if (asoc->asoc.locked_on_sending) {
3338					goto abort_anyway;
3339				}
3340				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3341				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3342					struct sctp_nets *netp;
3343
3344					/*
3345					 * there is nothing queued to send,
3346					 * so I send shutdown
3347					 */
3348					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3349					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3350						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3351					}
3352					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3353					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3354					sctp_stop_timers_for_shutdown(asoc);
3355					if (asoc->asoc.alternate) {
3356						netp = asoc->asoc.alternate;
3357					} else {
3358						netp = asoc->asoc.primary_destination;
3359					}
3360					sctp_send_shutdown(asoc, netp);
3361					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3362					    netp);
3363					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3364					    asoc->asoc.primary_destination);
3365					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3366				}
3367			} else {
3368				/* mark into shutdown pending */
3369				struct sctp_stream_queue_pending *sp;
3370
3371				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3372				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3373				    asoc->asoc.primary_destination);
3374				if (asoc->asoc.locked_on_sending) {
3375					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3376					    sctp_streamhead);
3377					if (sp == NULL) {
3378						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3379						    (void *)asoc->asoc.locked_on_sending,
3380						    asoc->asoc.locked_on_sending->stream_no);
3381					} else {
3382						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3383							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3384					}
3385				}
3386				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3387				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3388				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3389					struct mbuf *op_err;
3390
3391			abort_anyway:
3392					op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3393					    0, M_NOWAIT, 1, MT_DATA);
3394					if (op_err) {
3395						/*
3396						 * Fill in the user
3397						 * initiated abort
3398						 */
3399						struct sctp_paramhdr *ph;
3400
3401						SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3402						ph = mtod(op_err, struct sctp_paramhdr *);
3403						ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3404						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3405					}
3406					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3407					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3408					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3409					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3410					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3411						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3412					}
3413					if (sctp_free_assoc(inp, asoc,
3414					    SCTP_PCBFREE_NOFORCE,
3415					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3416						cnt_in_sd++;
3417					}
3418					continue;
3419				} else {
3420					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3421				}
3422			}
3423			cnt_in_sd++;
3424			SCTP_TCB_UNLOCK(asoc);
3425		}
3426		/* now is there some left in our SHUTDOWN state? */
3427		if (cnt_in_sd) {
3428#ifdef SCTP_LOG_CLOSING
3429			sctp_log_closing(inp, NULL, 2);
3430#endif
3431			inp->sctp_socket = NULL;
3432			SCTP_INP_WUNLOCK(inp);
3433			SCTP_ASOC_CREATE_UNLOCK(inp);
3434			SCTP_INP_INFO_WUNLOCK();
3435			return;
3436		}
3437	}
3438	inp->sctp_socket = NULL;
3439	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3440	    SCTP_PCB_FLAGS_UNBOUND) {
3441		/*
3442		 * ok, this guy has been bound. It's port is somewhere in
3443		 * the SCTP_BASE_INFO(hash table). Remove it!
3444		 */
3445		LIST_REMOVE(inp, sctp_hash);
3446		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3447	}
3448	/*
3449	 * If there is a timer running to kill us, forget it, since it may
3450	 * have a contest on the INP lock.. which would cause us to die ...
3451	 */
3452	cnt = 0;
3453	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3454		SCTP_TCB_LOCK(asoc);
3455		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3456			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3457				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3458				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3459			}
3460			cnt++;
3461			SCTP_TCB_UNLOCK(asoc);
3462			continue;
3463		}
3464		/* Free associations that are NOT killing us */
3465		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3466		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3467			struct mbuf *op_err;
3468
3469			op_err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr),
3470			    0, M_NOWAIT, 1, MT_DATA);
3471			if (op_err) {
3472				/* Fill in the user initiated abort */
3473				struct sctp_paramhdr *ph;
3474
3475				SCTP_BUF_LEN(op_err) = sizeof(struct sctp_paramhdr);
3476				ph = mtod(op_err, struct sctp_paramhdr *);
3477				ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
3478				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3479			}
3480			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3481			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3482			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3483		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3484			cnt++;
3485			SCTP_TCB_UNLOCK(asoc);
3486			continue;
3487		}
3488		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3489		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3490			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3491		}
3492		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3493			cnt++;
3494		}
3495	}
3496	if (cnt) {
3497		/* Ok we have someone out there that will kill us */
3498		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3499#ifdef SCTP_LOG_CLOSING
3500		sctp_log_closing(inp, NULL, 3);
3501#endif
3502		SCTP_INP_WUNLOCK(inp);
3503		SCTP_ASOC_CREATE_UNLOCK(inp);
3504		SCTP_INP_INFO_WUNLOCK();
3505		return;
3506	}
3507	if (SCTP_INP_LOCK_CONTENDED(inp))
3508		being_refed++;
3509	if (SCTP_INP_READ_CONTENDED(inp))
3510		being_refed++;
3511	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3512		being_refed++;
3513
3514	if ((inp->refcount) ||
3515	    (being_refed) ||
3516	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3517		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3518#ifdef SCTP_LOG_CLOSING
3519		sctp_log_closing(inp, NULL, 4);
3520#endif
3521		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3522		SCTP_INP_WUNLOCK(inp);
3523		SCTP_ASOC_CREATE_UNLOCK(inp);
3524		SCTP_INP_INFO_WUNLOCK();
3525		return;
3526	}
3527	inp->sctp_ep.signature_change.type = 0;
3528	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3529	/*
3530	 * Remove it from the list .. last thing we need a lock for.
3531	 */
3532	LIST_REMOVE(inp, sctp_list);
3533	SCTP_INP_WUNLOCK(inp);
3534	SCTP_ASOC_CREATE_UNLOCK(inp);
3535	SCTP_INP_INFO_WUNLOCK();
3536	/*
3537	 * Now we release all locks. Since this INP cannot be found anymore
3538	 * except possibly by the kill timer that might be running. We call
3539	 * the drain function here. It should hit the case were it sees the
3540	 * ACTIVE flag cleared and exit out freeing us to proceed and
3541	 * destroy everything.
3542	 */
3543	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3544		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3545	} else {
3546		/* Probably un-needed */
3547		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3548	}
3549
3550#ifdef SCTP_LOG_CLOSING
3551	sctp_log_closing(inp, NULL, 5);
3552#endif
3553
3554
3555	if ((inp->sctp_asocidhash) != NULL) {
3556		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3557		inp->sctp_asocidhash = NULL;
3558	}
3559	/* sa_ignore FREED_MEMORY */
3560	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3561		/* Its only abandoned if it had data left */
3562		if (sq->length)
3563			SCTP_STAT_INCR(sctps_left_abandon);
3564
3565		TAILQ_REMOVE(&inp->read_queue, sq, next);
3566		sctp_free_remote_addr(sq->whoFrom);
3567		if (so)
3568			so->so_rcv.sb_cc -= sq->length;
3569		if (sq->data) {
3570			sctp_m_freem(sq->data);
3571			sq->data = NULL;
3572		}
3573		/*
3574		 * no need to free the net count, since at this point all
3575		 * assoc's are gone.
3576		 */
3577		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3578		SCTP_DECR_READQ_COUNT();
3579	}
3580	/* Now the sctp_pcb things */
3581	/*
3582	 * free each asoc if it is not already closed/free. we can't use the
3583	 * macro here since le_next will get freed as part of the
3584	 * sctp_free_assoc() call.
3585	 */
3586	if (so) {
3587#ifdef IPSEC
3588		ipsec_delete_pcbpolicy(ip_pcb);
3589#endif				/* IPSEC */
3590
3591		/* Unlocks not needed since the socket is gone now */
3592	}
3593	if (ip_pcb->inp_options) {
3594		(void)sctp_m_free(ip_pcb->inp_options);
3595		ip_pcb->inp_options = 0;
3596	}
3597#ifdef INET6
3598	if (ip_pcb->inp_vflag & INP_IPV6) {
3599		struct in6pcb *in6p;
3600
3601		in6p = (struct in6pcb *)inp;
3602		ip6_freepcbopts(in6p->in6p_outputopts);
3603	}
3604#endif				/* INET6 */
3605	ip_pcb->inp_vflag = 0;
3606	/* free up authentication fields */
3607	if (inp->sctp_ep.local_auth_chunks != NULL)
3608		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3609	if (inp->sctp_ep.local_hmacs != NULL)
3610		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3611
3612	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3613		LIST_REMOVE(shared_key, next);
3614		sctp_free_sharedkey(shared_key);
3615		/* sa_ignore FREED_MEMORY */
3616	}
3617
3618	/*
3619	 * if we have an address list the following will free the list of
3620	 * ifaddr's that are set into this ep. Again macro limitations here,
3621	 * since the LIST_FOREACH could be a bad idea.
3622	 */
3623	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3624		sctp_remove_laddr(laddr);
3625	}
3626
3627#ifdef SCTP_TRACK_FREED_ASOCS
3628	/* TEMP CODE */
3629	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3630		LIST_REMOVE(asoc, sctp_tcblist);
3631		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3632		SCTP_DECR_ASOC_COUNT();
3633	}
3634	/* *** END TEMP CODE *** */
3635#endif
3636	/* Now lets see about freeing the EP hash table. */
3637	if (inp->sctp_tcbhash != NULL) {
3638		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3639		inp->sctp_tcbhash = NULL;
3640	}
3641	/* Now we must put the ep memory back into the zone pool */
3642	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3643	SCTP_INP_LOCK_DESTROY(inp);
3644	SCTP_INP_READ_DESTROY(inp);
3645	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3646	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3647	SCTP_DECR_EP_COUNT();
3648}
3649
3650
3651struct sctp_nets *
3652sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3653{
3654	struct sctp_nets *net;
3655
3656	/* locate the address */
3657	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3658		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3659			return (net);
3660	}
3661	return (NULL);
3662}
3663
3664
3665int
3666sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3667{
3668	struct sctp_ifa *sctp_ifa;
3669
3670	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3671	if (sctp_ifa) {
3672		return (1);
3673	} else {
3674		return (0);
3675	}
3676}
3677
3678/*
3679 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3680 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3681 * stats of stuff.
3682 */
3683int
3684sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3685    struct sctp_nets **netp, int set_scope, int from)
3686{
3687	/*
3688	 * The following is redundant to the same lines in the
3689	 * sctp_aloc_assoc() but is needed since others call the add address
3690	 * function
3691	 */
3692	struct sctp_nets *net, *netfirst;
3693	int addr_inscope;
3694
3695	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3696	    from);
3697	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3698
3699	netfirst = sctp_findnet(stcb, newaddr);
3700	if (netfirst) {
3701		/*
3702		 * Lie and return ok, we don't want to make the association
3703		 * go away for this behavior. It will happen in the TCP
3704		 * model in a connected socket. It does not reach the hash
3705		 * table until after the association is built so it can't be
3706		 * found. Mark as reachable, since the initial creation will
3707		 * have been cleared and the NOT_IN_ASSOC flag will have
3708		 * been added... and we don't want to end up removing it
3709		 * back out.
3710		 */
3711		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3712			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3713			    SCTP_ADDR_UNCONFIRMED);
3714		} else {
3715			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3716		}
3717
3718		return (0);
3719	}
3720	addr_inscope = 1;
3721	switch (newaddr->sa_family) {
3722#ifdef INET
3723	case AF_INET:
3724		{
3725			struct sockaddr_in *sin;
3726
3727			sin = (struct sockaddr_in *)newaddr;
3728			if (sin->sin_addr.s_addr == 0) {
3729				/* Invalid address */
3730				return (-1);
3731			}
3732			/* zero out the bzero area */
3733			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3734
3735			/* assure len is set */
3736			sin->sin_len = sizeof(struct sockaddr_in);
3737			if (set_scope) {
3738#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3739				stcb->ipv4_local_scope = 1;
3740#else
3741				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3742					stcb->asoc.scope.ipv4_local_scope = 1;
3743				}
3744#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3745			} else {
3746				/* Validate the address is in scope */
3747				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3748				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3749					addr_inscope = 0;
3750				}
3751			}
3752			break;
3753		}
3754#endif
3755#ifdef INET6
3756	case AF_INET6:
3757		{
3758			struct sockaddr_in6 *sin6;
3759
3760			sin6 = (struct sockaddr_in6 *)newaddr;
3761			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3762				/* Invalid address */
3763				return (-1);
3764			}
3765			/* assure len is set */
3766			sin6->sin6_len = sizeof(struct sockaddr_in6);
3767			if (set_scope) {
3768				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3769					stcb->asoc.scope.loopback_scope = 1;
3770					stcb->asoc.scope.local_scope = 0;
3771					stcb->asoc.scope.ipv4_local_scope = 1;
3772					stcb->asoc.scope.site_scope = 1;
3773				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3774					/*
3775					 * If the new destination is a
3776					 * LINK_LOCAL we must have common
3777					 * site scope. Don't set the local
3778					 * scope since we may not share all
3779					 * links, only loopback can do this.
3780					 * Links on the local network would
3781					 * also be on our private network
3782					 * for v4 too.
3783					 */
3784					stcb->asoc.scope.ipv4_local_scope = 1;
3785					stcb->asoc.scope.site_scope = 1;
3786				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3787					/*
3788					 * If the new destination is
3789					 * SITE_LOCAL then we must have site
3790					 * scope in common.
3791					 */
3792					stcb->asoc.scope.site_scope = 1;
3793				}
3794			} else {
3795				/* Validate the address is in scope */
3796				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3797				    (stcb->asoc.scope.loopback_scope == 0)) {
3798					addr_inscope = 0;
3799				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3800				    (stcb->asoc.scope.local_scope == 0)) {
3801					addr_inscope = 0;
3802				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3803				    (stcb->asoc.scope.site_scope == 0)) {
3804					addr_inscope = 0;
3805				}
3806			}
3807			break;
3808		}
3809#endif
3810	default:
3811		/* not supported family type */
3812		return (-1);
3813	}
3814	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3815	if (net == NULL) {
3816		return (-1);
3817	}
3818	SCTP_INCR_RADDR_COUNT();
3819	bzero(net, sizeof(struct sctp_nets));
3820	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3821	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3822	switch (newaddr->sa_family) {
3823#ifdef INET
3824	case AF_INET:
3825		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3826		break;
3827#endif
3828#ifdef INET6
3829	case AF_INET6:
3830		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3831		break;
3832#endif
3833	default:
3834		break;
3835	}
3836	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3837	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3838		stcb->asoc.scope.loopback_scope = 1;
3839		stcb->asoc.scope.ipv4_local_scope = 1;
3840		stcb->asoc.scope.local_scope = 0;
3841		stcb->asoc.scope.site_scope = 1;
3842		addr_inscope = 1;
3843	}
3844	net->failure_threshold = stcb->asoc.def_net_failure;
3845	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3846	if (addr_inscope == 0) {
3847		net->dest_state = (SCTP_ADDR_REACHABLE |
3848		    SCTP_ADDR_OUT_OF_SCOPE);
3849	} else {
3850		if (from == SCTP_ADDR_IS_CONFIRMED)
3851			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3852			net->dest_state = SCTP_ADDR_REACHABLE;
3853		else
3854			net->dest_state = SCTP_ADDR_REACHABLE |
3855			    SCTP_ADDR_UNCONFIRMED;
3856	}
3857	/*
3858	 * We set this to 0, the timer code knows that this means its an
3859	 * initial value
3860	 */
3861	net->rto_needed = 1;
3862	net->RTO = 0;
3863	net->RTO_measured = 0;
3864	stcb->asoc.numnets++;
3865	net->ref_count = 1;
3866	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3867	net->port = stcb->asoc.port;
3868	net->dscp = stcb->asoc.default_dscp;
3869#ifdef INET6
3870	net->flowlabel = stcb->asoc.default_flowlabel;
3871#endif
3872	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3873		net->dest_state |= SCTP_ADDR_NOHB;
3874	} else {
3875		net->dest_state &= ~SCTP_ADDR_NOHB;
3876	}
3877	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3878		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3879	} else {
3880		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3881	}
3882	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3883	/* Init the timer structure */
3884	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3885	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3886	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3887
3888	/* Now generate a route for this guy */
3889#ifdef INET6
3890	/* KAME hack: embed scopeid */
3891	if (newaddr->sa_family == AF_INET6) {
3892		struct sockaddr_in6 *sin6;
3893
3894		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3895		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3896		sin6->sin6_scope_id = 0;
3897	}
3898#endif
3899	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3900
3901	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3902		/* Get source address */
3903		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3904		    stcb,
3905		    (sctp_route_t *) & net->ro,
3906		    net,
3907		    0,
3908		    stcb->asoc.vrf_id);
3909		/* Now get the interface MTU */
3910		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3911			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3912		}
3913		if (net->mtu > 0) {
3914			uint32_t rmtu;
3915
3916			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3917			if (rmtu == 0) {
3918				/*
3919				 * Start things off to match mtu of
3920				 * interface please.
3921				 */
3922				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3923				    net->ro.ro_rt, net->mtu);
3924			} else {
3925				/*
3926				 * we take the route mtu over the interface,
3927				 * since the route may be leading out the
3928				 * loopback, or a different interface.
3929				 */
3930				net->mtu = rmtu;
3931			}
3932		}
3933	}
3934	if (net->mtu == 0) {
3935		switch (newaddr->sa_family) {
3936#ifdef INET
3937		case AF_INET:
3938			net->mtu = SCTP_DEFAULT_MTU;
3939			break;
3940#endif
3941#ifdef INET6
3942		case AF_INET6:
3943			net->mtu = 1280;
3944			break;
3945#endif
3946		default:
3947			break;
3948		}
3949	}
3950	if (net->port) {
3951		net->mtu -= (uint32_t) sizeof(struct udphdr);
3952	}
3953	if (from == SCTP_ALLOC_ASOC) {
3954		stcb->asoc.smallest_mtu = net->mtu;
3955	}
3956	if (stcb->asoc.smallest_mtu > net->mtu) {
3957		stcb->asoc.smallest_mtu = net->mtu;
3958	}
3959#ifdef INET6
3960	if (newaddr->sa_family == AF_INET6) {
3961		struct sockaddr_in6 *sin6;
3962
3963		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3964		(void)sa6_recoverscope(sin6);
3965	}
3966#endif
3967
3968	/* JRS - Use the congestion control given in the CC module */
3969	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
3970		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
3971
3972	/*
3973	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3974	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3975	 */
3976	net->find_pseudo_cumack = 1;
3977	net->find_rtx_pseudo_cumack = 1;
3978	net->src_addr_selected = 0;
3979	/* Choose an initial flowid. */
3980	net->flowid = stcb->asoc.my_vtag ^
3981	    ntohs(stcb->rport) ^
3982	    ntohs(stcb->sctp_ep->sctp_lport);
3983#ifdef INVARIANTS
3984	net->flowidset = 1;
3985#endif
3986	if (netp) {
3987		*netp = net;
3988	}
3989	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3990	if (net->ro.ro_rt == NULL) {
3991		/* Since we have no route put it at the back */
3992		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3993	} else if (netfirst == NULL) {
3994		/* We are the first one in the pool. */
3995		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3996	} else if (netfirst->ro.ro_rt == NULL) {
3997		/*
3998		 * First one has NO route. Place this one ahead of the first
3999		 * one.
4000		 */
4001		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4002	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4003		/*
4004		 * This one has a different interface than the one at the
4005		 * top of the list. Place it ahead.
4006		 */
4007		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4008	} else {
4009		/*
4010		 * Ok we have the same interface as the first one. Move
4011		 * forward until we find either a) one with a NULL route...
4012		 * insert ahead of that b) one with a different ifp.. insert
4013		 * after that. c) end of the list.. insert at the tail.
4014		 */
4015		struct sctp_nets *netlook;
4016
4017		do {
4018			netlook = TAILQ_NEXT(netfirst, sctp_next);
4019			if (netlook == NULL) {
4020				/* End of the list */
4021				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4022				break;
4023			} else if (netlook->ro.ro_rt == NULL) {
4024				/* next one has NO route */
4025				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4026				break;
4027			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4028				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4029				    net, sctp_next);
4030				break;
4031			}
4032			/* Shift forward */
4033			netfirst = netlook;
4034		} while (netlook != NULL);
4035	}
4036
4037	/* got to have a primary set */
4038	if (stcb->asoc.primary_destination == 0) {
4039		stcb->asoc.primary_destination = net;
4040	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4041		    (net->ro.ro_rt) &&
4042	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4043		/* No route to current primary adopt new primary */
4044		stcb->asoc.primary_destination = net;
4045	}
4046	/* Validate primary is first */
4047	net = TAILQ_FIRST(&stcb->asoc.nets);
4048	if ((net != stcb->asoc.primary_destination) &&
4049	    (stcb->asoc.primary_destination)) {
4050		/*
4051		 * first one on the list is NOT the primary sctp_cmpaddr()
4052		 * is much more efficient if the primary is the first on the
4053		 * list, make it so.
4054		 */
4055		TAILQ_REMOVE(&stcb->asoc.nets,
4056		    stcb->asoc.primary_destination, sctp_next);
4057		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4058		    stcb->asoc.primary_destination, sctp_next);
4059	}
4060	return (0);
4061}
4062
4063
4064static uint32_t
4065sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4066{
4067	uint32_t id;
4068	struct sctpasochead *head;
4069	struct sctp_tcb *lstcb;
4070
4071	SCTP_INP_WLOCK(inp);
4072try_again:
4073	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4074		/* TSNH */
4075		SCTP_INP_WUNLOCK(inp);
4076		return (0);
4077	}
4078	/*
4079	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4080	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4081	 */
4082	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4083		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4084	}
4085	id = inp->sctp_associd_counter;
4086	inp->sctp_associd_counter++;
4087	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4088	if (lstcb) {
4089		goto try_again;
4090	}
4091	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4092	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4093	stcb->asoc.in_asocid_hash = 1;
4094	SCTP_INP_WUNLOCK(inp);
4095	return id;
4096}
4097
4098/*
4099 * allocate an association and add it to the endpoint. The caller must be
4100 * careful to add all additional addresses once they are know right away or
4101 * else the assoc will be may experience a blackout scenario.
4102 */
4103struct sctp_tcb *
4104sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4105    int *error, uint32_t override_tag, uint32_t vrf_id,
4106    struct thread *p
4107)
4108{
4109	/* note the p argument is only valid in unbound sockets */
4110
4111	struct sctp_tcb *stcb;
4112	struct sctp_association *asoc;
4113	struct sctpasochead *head;
4114	uint16_t rport;
4115	int err;
4116
4117	/*
4118	 * Assumption made here: Caller has done a
4119	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4120	 * address does not exist already.
4121	 */
4122	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4123		/* Hit max assoc, sorry no more */
4124		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4125		*error = ENOBUFS;
4126		return (NULL);
4127	}
4128	if (firstaddr == NULL) {
4129		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4130		*error = EINVAL;
4131		return (NULL);
4132	}
4133	SCTP_INP_RLOCK(inp);
4134	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4135	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4136	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4137		/*
4138		 * If its in the TCP pool, its NOT allowed to create an
4139		 * association. The parent listener needs to call
4140		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4141		 * off, or connected one does this.. its an error.
4142		 */
4143		SCTP_INP_RUNLOCK(inp);
4144		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4145		*error = EINVAL;
4146		return (NULL);
4147	}
4148	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4149	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4150		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4151		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4152			SCTP_INP_RUNLOCK(inp);
4153			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4154			*error = EINVAL;
4155			return (NULL);
4156		}
4157	}
4158	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4159#ifdef SCTP_DEBUG
4160	if (firstaddr) {
4161		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4162		switch (firstaddr->sa_family) {
4163#ifdef INET
4164		case AF_INET:
4165			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4166			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4167			break;
4168#endif
4169#ifdef INET6
4170		case AF_INET6:
4171			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4172			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4173			break;
4174#endif
4175		default:
4176			break;
4177		}
4178	} else {
4179		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4180	}
4181#endif				/* SCTP_DEBUG */
4182	switch (firstaddr->sa_family) {
4183#ifdef INET
4184	case AF_INET:
4185		{
4186			struct sockaddr_in *sin;
4187
4188			sin = (struct sockaddr_in *)firstaddr;
4189			if ((ntohs(sin->sin_port) == 0) ||
4190			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4191			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4192			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4193				/* Invalid address */
4194				SCTP_INP_RUNLOCK(inp);
4195				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4196				*error = EINVAL;
4197				return (NULL);
4198			}
4199			rport = sin->sin_port;
4200			break;
4201		}
4202#endif
4203#ifdef INET6
4204	case AF_INET6:
4205		{
4206			struct sockaddr_in6 *sin6;
4207
4208			sin6 = (struct sockaddr_in6 *)firstaddr;
4209			if ((ntohs(sin6->sin6_port) == 0) ||
4210			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4211			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4212				/* Invalid address */
4213				SCTP_INP_RUNLOCK(inp);
4214				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4215				*error = EINVAL;
4216				return (NULL);
4217			}
4218			rport = sin6->sin6_port;
4219			break;
4220		}
4221#endif
4222	default:
4223		/* not supported family type */
4224		SCTP_INP_RUNLOCK(inp);
4225		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4226		*error = EINVAL;
4227		return (NULL);
4228	}
4229	SCTP_INP_RUNLOCK(inp);
4230	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4231		/*
4232		 * If you have not performed a bind, then we need to do the
4233		 * ephemeral bind for you.
4234		 */
4235		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4236		    (struct sockaddr *)NULL,
4237		    (struct sctp_ifa *)NULL,
4238		    p
4239		    ))) {
4240			/* bind error, probably perm */
4241			*error = err;
4242			return (NULL);
4243		}
4244	}
4245	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4246	if (stcb == NULL) {
4247		/* out of memory? */
4248		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4249		*error = ENOMEM;
4250		return (NULL);
4251	}
4252	SCTP_INCR_ASOC_COUNT();
4253
4254	bzero(stcb, sizeof(*stcb));
4255	asoc = &stcb->asoc;
4256
4257	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4258	SCTP_TCB_LOCK_INIT(stcb);
4259	SCTP_TCB_SEND_LOCK_INIT(stcb);
4260	stcb->rport = rport;
4261	/* setup back pointer's */
4262	stcb->sctp_ep = inp;
4263	stcb->sctp_socket = inp->sctp_socket;
4264	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4265		/* failed */
4266		SCTP_TCB_LOCK_DESTROY(stcb);
4267		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4268		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4269		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4270		SCTP_DECR_ASOC_COUNT();
4271		*error = err;
4272		return (NULL);
4273	}
4274	/* and the port */
4275	SCTP_INP_INFO_WLOCK();
4276	SCTP_INP_WLOCK(inp);
4277	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4278		/* inpcb freed while alloc going on */
4279		SCTP_TCB_LOCK_DESTROY(stcb);
4280		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4281		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4282		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4283		SCTP_INP_WUNLOCK(inp);
4284		SCTP_INP_INFO_WUNLOCK();
4285		SCTP_DECR_ASOC_COUNT();
4286		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4287		*error = EINVAL;
4288		return (NULL);
4289	}
4290	SCTP_TCB_LOCK(stcb);
4291
4292	/* now that my_vtag is set, add it to the hash */
4293	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4294	/* put it in the bucket in the vtag hash of assoc's for the system */
4295	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4296	SCTP_INP_INFO_WUNLOCK();
4297
4298	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4299		/* failure.. memory error? */
4300		if (asoc->strmout) {
4301			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4302			asoc->strmout = NULL;
4303		}
4304		if (asoc->mapping_array) {
4305			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4306			asoc->mapping_array = NULL;
4307		}
4308		if (asoc->nr_mapping_array) {
4309			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4310			asoc->nr_mapping_array = NULL;
4311		}
4312		SCTP_DECR_ASOC_COUNT();
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_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4319		*error = ENOBUFS;
4320		return (NULL);
4321	}
4322	/* Init all the timers */
4323	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4324	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4325	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4326	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4327	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4328	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4329	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4330
4331	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4332	/* now file the port under the hash as well */
4333	if (inp->sctp_tcbhash != NULL) {
4334		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4335		    inp->sctp_hashmark)];
4336		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4337	}
4338	SCTP_INP_WUNLOCK(inp);
4339	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4340	return (stcb);
4341}
4342
4343
4344void
4345sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4346{
4347	struct sctp_association *asoc;
4348
4349	asoc = &stcb->asoc;
4350	asoc->numnets--;
4351	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4352	if (net == asoc->primary_destination) {
4353		/* Reset primary */
4354		struct sctp_nets *lnet;
4355
4356		lnet = TAILQ_FIRST(&asoc->nets);
4357		/*
4358		 * Mobility adaptation Ideally, if deleted destination is
4359		 * the primary, it becomes a fast retransmission trigger by
4360		 * the subsequent SET PRIMARY. (by micchie)
4361		 */
4362		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4363		    SCTP_MOBILITY_BASE) ||
4364		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4365		    SCTP_MOBILITY_FASTHANDOFF)) {
4366			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4367			if (asoc->deleted_primary != NULL) {
4368				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4369				goto out;
4370			}
4371			asoc->deleted_primary = net;
4372			atomic_add_int(&net->ref_count, 1);
4373			memset(&net->lastsa, 0, sizeof(net->lastsa));
4374			memset(&net->lastsv, 0, sizeof(net->lastsv));
4375			sctp_mobility_feature_on(stcb->sctp_ep,
4376			    SCTP_MOBILITY_PRIM_DELETED);
4377			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4378			    stcb->sctp_ep, stcb, NULL);
4379		}
4380out:
4381		/* Try to find a confirmed primary */
4382		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4383	}
4384	if (net == asoc->last_data_chunk_from) {
4385		/* Reset primary */
4386		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4387	}
4388	if (net == asoc->last_control_chunk_from) {
4389		/* Clear net */
4390		asoc->last_control_chunk_from = NULL;
4391	}
4392	if (net == stcb->asoc.alternate) {
4393		sctp_free_remote_addr(stcb->asoc.alternate);
4394		stcb->asoc.alternate = NULL;
4395	}
4396	sctp_free_remote_addr(net);
4397}
4398
4399/*
4400 * remove a remote endpoint address from an association, it will fail if the
4401 * address does not exist.
4402 */
4403int
4404sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4405{
4406	/*
4407	 * Here we need to remove a remote address. This is quite simple, we
4408	 * first find it in the list of address for the association
4409	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4410	 * on that item. Note we do not allow it to be removed if there are
4411	 * no other addresses.
4412	 */
4413	struct sctp_association *asoc;
4414	struct sctp_nets *net, *nnet;
4415
4416	asoc = &stcb->asoc;
4417
4418	/* locate the address */
4419	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4420		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4421			continue;
4422		}
4423		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4424		    remaddr)) {
4425			/* we found the guy */
4426			if (asoc->numnets < 2) {
4427				/* Must have at LEAST two remote addresses */
4428				return (-1);
4429			} else {
4430				sctp_remove_net(stcb, net);
4431				return (0);
4432			}
4433		}
4434	}
4435	/* not found. */
4436	return (-2);
4437}
4438
4439void
4440sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4441{
4442	struct sctpvtaghead *chain;
4443	struct sctp_tagblock *twait_block;
4444	int found = 0;
4445	int i;
4446
4447	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4448	if (!LIST_EMPTY(chain)) {
4449		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4450			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4451				if ((twait_block->vtag_block[i].v_tag == tag) &&
4452				    (twait_block->vtag_block[i].lport == lport) &&
4453				    (twait_block->vtag_block[i].rport == rport)) {
4454					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4455					twait_block->vtag_block[i].v_tag = 0;
4456					twait_block->vtag_block[i].lport = 0;
4457					twait_block->vtag_block[i].rport = 0;
4458					found = 1;
4459					break;
4460				}
4461			}
4462			if (found)
4463				break;
4464		}
4465	}
4466}
4467
4468int
4469sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4470{
4471	struct sctpvtaghead *chain;
4472	struct sctp_tagblock *twait_block;
4473	int found = 0;
4474	int i;
4475
4476	SCTP_INP_INFO_WLOCK();
4477	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4478	if (!LIST_EMPTY(chain)) {
4479		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4480			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4481				if ((twait_block->vtag_block[i].v_tag == tag) &&
4482				    (twait_block->vtag_block[i].lport == lport) &&
4483				    (twait_block->vtag_block[i].rport == rport)) {
4484					found = 1;
4485					break;
4486				}
4487			}
4488			if (found)
4489				break;
4490		}
4491	}
4492	SCTP_INP_INFO_WUNLOCK();
4493	return (found);
4494}
4495
4496
4497void
4498sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4499{
4500	struct sctpvtaghead *chain;
4501	struct sctp_tagblock *twait_block;
4502	struct timeval now;
4503	int set, i;
4504
4505	if (time == 0) {
4506		/* Its disabled */
4507		return;
4508	}
4509	(void)SCTP_GETTIME_TIMEVAL(&now);
4510	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4511	set = 0;
4512	if (!LIST_EMPTY(chain)) {
4513		/* Block(s) present, lets find space, and expire on the fly */
4514		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4515			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4516				if ((twait_block->vtag_block[i].v_tag == 0) &&
4517				    !set) {
4518					twait_block->vtag_block[i].tv_sec_at_expire =
4519					    now.tv_sec + time;
4520					twait_block->vtag_block[i].v_tag = tag;
4521					twait_block->vtag_block[i].lport = lport;
4522					twait_block->vtag_block[i].rport = rport;
4523					set = 1;
4524				} else if ((twait_block->vtag_block[i].v_tag) &&
4525				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4526					/* Audit expires this guy */
4527					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4528					twait_block->vtag_block[i].v_tag = 0;
4529					twait_block->vtag_block[i].lport = 0;
4530					twait_block->vtag_block[i].rport = 0;
4531					if (set == 0) {
4532						/* Reuse it for my new tag */
4533						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4534						twait_block->vtag_block[i].v_tag = tag;
4535						twait_block->vtag_block[i].lport = lport;
4536						twait_block->vtag_block[i].rport = rport;
4537						set = 1;
4538					}
4539				}
4540			}
4541			if (set) {
4542				/*
4543				 * We only do up to the block where we can
4544				 * place our tag for audits
4545				 */
4546				break;
4547			}
4548		}
4549	}
4550	/* Need to add a new block to chain */
4551	if (!set) {
4552		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4553		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4554		if (twait_block == NULL) {
4555#ifdef INVARIANTS
4556			panic("Can not alloc tagblock");
4557#endif
4558			return;
4559		}
4560		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4561		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4562		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4563		twait_block->vtag_block[0].v_tag = tag;
4564		twait_block->vtag_block[0].lport = lport;
4565		twait_block->vtag_block[0].rport = rport;
4566	}
4567}
4568
4569
4570
4571/*-
4572 * Free the association after un-hashing the remote port. This
4573 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4574 * expect that the input to this function IS a locked TCB.
4575 * It will return 0, if it did NOT destroy the association (instead
4576 * it unlocks it. It will return NON-zero if it either destroyed the
4577 * association OR the association is already destroyed.
4578 */
4579int
4580sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4581{
4582	int i;
4583	struct sctp_association *asoc;
4584	struct sctp_nets *net, *nnet;
4585	struct sctp_laddr *laddr, *naddr;
4586	struct sctp_tmit_chunk *chk, *nchk;
4587	struct sctp_asconf_addr *aparam, *naparam;
4588	struct sctp_asconf_ack *aack, *naack;
4589	struct sctp_stream_reset_list *strrst, *nstrrst;
4590	struct sctp_queued_to_read *sq, *nsq;
4591	struct sctp_stream_queue_pending *sp, *nsp;
4592	sctp_sharedkey_t *shared_key, *nshared_key;
4593	struct socket *so;
4594
4595	/* first, lets purge the entry from the hash table. */
4596
4597#ifdef SCTP_LOG_CLOSING
4598	sctp_log_closing(inp, stcb, 6);
4599#endif
4600	if (stcb->asoc.state == 0) {
4601#ifdef SCTP_LOG_CLOSING
4602		sctp_log_closing(inp, NULL, 7);
4603#endif
4604		/* there is no asoc, really TSNH :-0 */
4605		return (1);
4606	}
4607	if (stcb->asoc.alternate) {
4608		sctp_free_remote_addr(stcb->asoc.alternate);
4609		stcb->asoc.alternate = NULL;
4610	}
4611	/* TEMP CODE */
4612	if (stcb->freed_from_where == 0) {
4613		/* Only record the first place free happened from */
4614		stcb->freed_from_where = from_location;
4615	}
4616	/* TEMP CODE */
4617
4618	asoc = &stcb->asoc;
4619	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4620	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4621		/* nothing around */
4622		so = NULL;
4623	else
4624		so = inp->sctp_socket;
4625
4626	/*
4627	 * We used timer based freeing if a reader or writer is in the way.
4628	 * So we first check if we are actually being called from a timer,
4629	 * if so we abort early if a reader or writer is still in the way.
4630	 */
4631	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4632	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4633		/*
4634		 * is it the timer driving us? if so are the reader/writers
4635		 * gone?
4636		 */
4637		if (stcb->asoc.refcnt) {
4638			/* nope, reader or writer in the way */
4639			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4640			/* no asoc destroyed */
4641			SCTP_TCB_UNLOCK(stcb);
4642#ifdef SCTP_LOG_CLOSING
4643			sctp_log_closing(inp, stcb, 8);
4644#endif
4645			return (0);
4646		}
4647	}
4648	/* now clean up any other timers */
4649	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4650	asoc->dack_timer.self = NULL;
4651	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4652	/*-
4653	 * For stream reset we don't blast this unless
4654	 * it is a str-reset timer, it might be the
4655	 * free-asoc timer which we DON'T want to
4656	 * disturb.
4657	 */
4658	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4659		asoc->strreset_timer.self = NULL;
4660	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4661	asoc->asconf_timer.self = NULL;
4662	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4663	asoc->autoclose_timer.self = NULL;
4664	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4665	asoc->shut_guard_timer.self = NULL;
4666	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4667	asoc->delayed_event_timer.self = NULL;
4668	/* Mobility adaptation */
4669	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4670	asoc->delete_prim_timer.self = NULL;
4671	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4672		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4673		net->rxt_timer.self = NULL;
4674		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4675		net->pmtu_timer.self = NULL;
4676		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4677		net->hb_timer.self = NULL;
4678	}
4679	/* Now the read queue needs to be cleaned up (only once) */
4680	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4681		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4682		SCTP_INP_READ_LOCK(inp);
4683		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4684			if (sq->stcb == stcb) {
4685				sq->do_not_ref_stcb = 1;
4686				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4687				/*
4688				 * If there is no end, there never will be
4689				 * now.
4690				 */
4691				if (sq->end_added == 0) {
4692					/* Held for PD-API clear that. */
4693					sq->pdapi_aborted = 1;
4694					sq->held_length = 0;
4695					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4696						/*
4697						 * Need to add a PD-API
4698						 * aborted indication.
4699						 * Setting the control_pdapi
4700						 * assures that it will be
4701						 * added right after this
4702						 * msg.
4703						 */
4704						uint32_t strseq;
4705
4706						stcb->asoc.control_pdapi = sq;
4707						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4708						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4709						    stcb,
4710						    SCTP_PARTIAL_DELIVERY_ABORTED,
4711						    (void *)&strseq,
4712						    SCTP_SO_LOCKED);
4713						stcb->asoc.control_pdapi = NULL;
4714					}
4715				}
4716				/* Add an end to wake them */
4717				sq->end_added = 1;
4718			}
4719		}
4720		SCTP_INP_READ_UNLOCK(inp);
4721		if (stcb->block_entry) {
4722			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4723			stcb->block_entry->error = ECONNRESET;
4724			stcb->block_entry = NULL;
4725		}
4726	}
4727	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4728		/*
4729		 * Someone holds a reference OR the socket is unaccepted
4730		 * yet.
4731		 */
4732		if ((stcb->asoc.refcnt) ||
4733		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4734		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4735			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4736			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4737		}
4738		SCTP_TCB_UNLOCK(stcb);
4739		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4740		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4741			/* nothing around */
4742			so = NULL;
4743		if (so) {
4744			/* Wake any reader/writers */
4745			sctp_sorwakeup(inp, so);
4746			sctp_sowwakeup(inp, so);
4747		}
4748#ifdef SCTP_LOG_CLOSING
4749		sctp_log_closing(inp, stcb, 9);
4750#endif
4751		/* no asoc destroyed */
4752		return (0);
4753	}
4754#ifdef SCTP_LOG_CLOSING
4755	sctp_log_closing(inp, stcb, 10);
4756#endif
4757	/*
4758	 * When I reach here, no others want to kill the assoc yet.. and I
4759	 * own the lock. Now its possible an abort comes in when I do the
4760	 * lock exchange below to grab all the locks to do the final take
4761	 * out. to prevent this we increment the count, which will start a
4762	 * timer and blow out above thus assuring us that we hold exclusive
4763	 * killing of the asoc. Note that after getting back the TCB lock we
4764	 * will go ahead and increment the counter back up and stop any
4765	 * timer a passing stranger may have started :-S
4766	 */
4767	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4768		atomic_add_int(&stcb->asoc.refcnt, 1);
4769
4770		SCTP_TCB_UNLOCK(stcb);
4771		SCTP_INP_INFO_WLOCK();
4772		SCTP_INP_WLOCK(inp);
4773		SCTP_TCB_LOCK(stcb);
4774	}
4775	/* Double check the GONE flag */
4776	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4777	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4778		/* nothing around */
4779		so = NULL;
4780
4781	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4782	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4783		/*
4784		 * For TCP type we need special handling when we are
4785		 * connected. We also include the peel'ed off ones to.
4786		 */
4787		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4788			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4789			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4790			if (so) {
4791				SOCK_LOCK(so);
4792				if (so->so_rcv.sb_cc == 0) {
4793					so->so_state &= ~(SS_ISCONNECTING |
4794					    SS_ISDISCONNECTING |
4795					    SS_ISCONFIRMING |
4796					    SS_ISCONNECTED);
4797				}
4798				socantrcvmore_locked(so);
4799				sctp_sowwakeup(inp, so);
4800				sctp_sorwakeup(inp, so);
4801				SCTP_SOWAKEUP(so);
4802			}
4803		}
4804	}
4805	/*
4806	 * Make it invalid too, that way if its about to run it will abort
4807	 * and return.
4808	 */
4809	/* re-increment the lock */
4810	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4811		atomic_add_int(&stcb->asoc.refcnt, -1);
4812	}
4813	if (stcb->asoc.refcnt) {
4814		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4815		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4816		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4817			SCTP_INP_INFO_WUNLOCK();
4818			SCTP_INP_WUNLOCK(inp);
4819		}
4820		SCTP_TCB_UNLOCK(stcb);
4821		return (0);
4822	}
4823	asoc->state = 0;
4824	if (inp->sctp_tcbhash) {
4825		LIST_REMOVE(stcb, sctp_tcbhash);
4826	}
4827	if (stcb->asoc.in_asocid_hash) {
4828		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4829	}
4830	/* Now lets remove it from the list of ALL associations in the EP */
4831	LIST_REMOVE(stcb, sctp_tcblist);
4832	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4833		SCTP_INP_INCR_REF(inp);
4834		SCTP_INP_WUNLOCK(inp);
4835	}
4836	/* pull from vtag hash */
4837	LIST_REMOVE(stcb, sctp_asocs);
4838	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4839	    inp->sctp_lport, stcb->rport);
4840
4841	/*
4842	 * Now restop the timers to be sure this is paranoia at is finest!
4843	 */
4844	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4845	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4846	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4847	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4848	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4849	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4850	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4851	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4852		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4853		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4854		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4855	}
4856
4857	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4858	/*
4859	 * The chunk lists and such SHOULD be empty but we check them just
4860	 * in case.
4861	 */
4862	/* anything on the wheel needs to be removed */
4863	for (i = 0; i < asoc->streamoutcnt; i++) {
4864		struct sctp_stream_out *outs;
4865
4866		outs = &asoc->strmout[i];
4867		/* now clean up any chunks here */
4868		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4869			TAILQ_REMOVE(&outs->outqueue, sp, next);
4870			sctp_free_spbufspace(stcb, asoc, sp);
4871			if (sp->data) {
4872				if (so) {
4873					/* Still an open socket - report */
4874					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4875					    0, (void *)sp, SCTP_SO_LOCKED);
4876				}
4877				if (sp->data) {
4878					sctp_m_freem(sp->data);
4879					sp->data = NULL;
4880					sp->tail_mbuf = NULL;
4881					sp->length = 0;
4882				}
4883			}
4884			if (sp->net) {
4885				sctp_free_remote_addr(sp->net);
4886				sp->net = NULL;
4887			}
4888			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4889		}
4890	}
4891	/* sa_ignore FREED_MEMORY */
4892	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4893		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4894		SCTP_FREE(strrst, SCTP_M_STRESET);
4895	}
4896	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4897		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4898		if (sq->data) {
4899			sctp_m_freem(sq->data);
4900			sq->data = NULL;
4901		}
4902		sctp_free_remote_addr(sq->whoFrom);
4903		sq->whoFrom = NULL;
4904		sq->stcb = NULL;
4905		/* Free the ctl entry */
4906		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4907		SCTP_DECR_READQ_COUNT();
4908		/* sa_ignore FREED_MEMORY */
4909	}
4910	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4911		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4912		if (chk->data) {
4913			sctp_m_freem(chk->data);
4914			chk->data = NULL;
4915		}
4916		if (chk->holds_key_ref)
4917			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4918		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4919		SCTP_DECR_CHK_COUNT();
4920		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4921		asoc->free_chunk_cnt--;
4922		/* sa_ignore FREED_MEMORY */
4923	}
4924	/* pending send queue SHOULD be empty */
4925	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4926		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4927			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4928#ifdef INVARIANTS
4929		} else {
4930			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4931#endif
4932		}
4933		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4934		if (chk->data) {
4935			if (so) {
4936				/* Still a socket? */
4937				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4938				    0, chk, SCTP_SO_LOCKED);
4939			}
4940			if (chk->data) {
4941				sctp_m_freem(chk->data);
4942				chk->data = NULL;
4943			}
4944		}
4945		if (chk->holds_key_ref)
4946			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4947		if (chk->whoTo) {
4948			sctp_free_remote_addr(chk->whoTo);
4949			chk->whoTo = NULL;
4950		}
4951		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4952		SCTP_DECR_CHK_COUNT();
4953		/* sa_ignore FREED_MEMORY */
4954	}
4955	/* sent queue SHOULD be empty */
4956	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4957		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
4958			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4959				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4960#ifdef INVARIANTS
4961			} else {
4962				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4963#endif
4964			}
4965		}
4966		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4967		if (chk->data) {
4968			if (so) {
4969				/* Still a socket? */
4970				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
4971				    0, chk, SCTP_SO_LOCKED);
4972			}
4973			if (chk->data) {
4974				sctp_m_freem(chk->data);
4975				chk->data = NULL;
4976			}
4977		}
4978		if (chk->holds_key_ref)
4979			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4980		sctp_free_remote_addr(chk->whoTo);
4981		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4982		SCTP_DECR_CHK_COUNT();
4983		/* sa_ignore FREED_MEMORY */
4984	}
4985#ifdef INVARIANTS
4986	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
4987		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
4988			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
4989		}
4990	}
4991#endif
4992	/* control queue MAY not be empty */
4993	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
4994		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4995		if (chk->data) {
4996			sctp_m_freem(chk->data);
4997			chk->data = NULL;
4998		}
4999		if (chk->holds_key_ref)
5000			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5001		sctp_free_remote_addr(chk->whoTo);
5002		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5003		SCTP_DECR_CHK_COUNT();
5004		/* sa_ignore FREED_MEMORY */
5005	}
5006	/* ASCONF queue MAY not be empty */
5007	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5008		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5009		if (chk->data) {
5010			sctp_m_freem(chk->data);
5011			chk->data = NULL;
5012		}
5013		if (chk->holds_key_ref)
5014			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5015		sctp_free_remote_addr(chk->whoTo);
5016		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5017		SCTP_DECR_CHK_COUNT();
5018		/* sa_ignore FREED_MEMORY */
5019	}
5020	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5021		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5022		if (chk->data) {
5023			sctp_m_freem(chk->data);
5024			chk->data = NULL;
5025		}
5026		if (chk->holds_key_ref)
5027			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5028		sctp_free_remote_addr(chk->whoTo);
5029		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5030		SCTP_DECR_CHK_COUNT();
5031		/* sa_ignore FREED_MEMORY */
5032	}
5033
5034	if (asoc->mapping_array) {
5035		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5036		asoc->mapping_array = NULL;
5037	}
5038	if (asoc->nr_mapping_array) {
5039		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5040		asoc->nr_mapping_array = NULL;
5041	}
5042	/* the stream outs */
5043	if (asoc->strmout) {
5044		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5045		asoc->strmout = NULL;
5046	}
5047	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5048	if (asoc->strmin) {
5049		struct sctp_queued_to_read *ctl, *nctl;
5050
5051		for (i = 0; i < asoc->streamincnt; i++) {
5052			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5053				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5054				sctp_free_remote_addr(ctl->whoFrom);
5055				if (ctl->data) {
5056					sctp_m_freem(ctl->data);
5057					ctl->data = NULL;
5058				}
5059				/*
5060				 * We don't free the address here since all
5061				 * the net's were freed above.
5062				 */
5063				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5064				SCTP_DECR_READQ_COUNT();
5065			}
5066		}
5067		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5068		asoc->strmin = NULL;
5069	}
5070	asoc->streamincnt = 0;
5071	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5072#ifdef INVARIANTS
5073		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5074			panic("no net's left alloc'ed, or list points to itself");
5075		}
5076#endif
5077		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5078		sctp_free_remote_addr(net);
5079	}
5080	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5081		/* sa_ignore FREED_MEMORY */
5082		sctp_remove_laddr(laddr);
5083	}
5084
5085	/* pending asconf (address) parameters */
5086	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5087		/* sa_ignore FREED_MEMORY */
5088		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5089		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5090	}
5091	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5092		/* sa_ignore FREED_MEMORY */
5093		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5094		if (aack->data != NULL) {
5095			sctp_m_freem(aack->data);
5096		}
5097		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5098	}
5099	/* clean up auth stuff */
5100	if (asoc->local_hmacs)
5101		sctp_free_hmaclist(asoc->local_hmacs);
5102	if (asoc->peer_hmacs)
5103		sctp_free_hmaclist(asoc->peer_hmacs);
5104
5105	if (asoc->local_auth_chunks)
5106		sctp_free_chunklist(asoc->local_auth_chunks);
5107	if (asoc->peer_auth_chunks)
5108		sctp_free_chunklist(asoc->peer_auth_chunks);
5109
5110	sctp_free_authinfo(&asoc->authinfo);
5111
5112	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5113		LIST_REMOVE(shared_key, next);
5114		sctp_free_sharedkey(shared_key);
5115		/* sa_ignore FREED_MEMORY */
5116	}
5117
5118	/* Insert new items here :> */
5119
5120	/* Get rid of LOCK */
5121	SCTP_TCB_LOCK_DESTROY(stcb);
5122	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5123	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5124		SCTP_INP_INFO_WUNLOCK();
5125		SCTP_INP_RLOCK(inp);
5126	}
5127#ifdef SCTP_TRACK_FREED_ASOCS
5128	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5129		/* now clean up the tasoc itself */
5130		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5131		SCTP_DECR_ASOC_COUNT();
5132	} else {
5133		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5134	}
5135#else
5136	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5137	SCTP_DECR_ASOC_COUNT();
5138#endif
5139	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5140		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5141			/*
5142			 * If its NOT the inp_free calling us AND sctp_close
5143			 * as been called, we call back...
5144			 */
5145			SCTP_INP_RUNLOCK(inp);
5146			/*
5147			 * This will start the kill timer (if we are the
5148			 * last one) since we hold an increment yet. But
5149			 * this is the only safe way to do this since
5150			 * otherwise if the socket closes at the same time
5151			 * we are here we might collide in the cleanup.
5152			 */
5153			sctp_inpcb_free(inp,
5154			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5155			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5156			SCTP_INP_DECR_REF(inp);
5157			goto out_of;
5158		} else {
5159			/* The socket is still open. */
5160			SCTP_INP_DECR_REF(inp);
5161		}
5162	}
5163	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5164		SCTP_INP_RUNLOCK(inp);
5165	}
5166out_of:
5167	/* destroyed the asoc */
5168#ifdef SCTP_LOG_CLOSING
5169	sctp_log_closing(inp, NULL, 11);
5170#endif
5171	return (1);
5172}
5173
5174
5175
5176/*
5177 * determine if a destination is "reachable" based upon the addresses bound
5178 * to the current endpoint (e.g. only v4 or v6 currently bound)
5179 */
5180/*
5181 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5182 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5183 * types bound as its endpoint
5184 */
5185int
5186sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5187{
5188	struct sctp_inpcb *inp;
5189	int answer;
5190
5191	/*
5192	 * No locks here, the TCB, in all cases is already locked and an
5193	 * assoc is up. There is either a INP lock by the caller applied (in
5194	 * asconf case when deleting an address) or NOT in the HB case,
5195	 * however if HB then the INP increment is up and the INP will not
5196	 * be removed (on top of the fact that we have a TCB lock). So we
5197	 * only want to read the sctp_flags, which is either bound-all or
5198	 * not.. no protection needed since once an assoc is up you can't be
5199	 * changing your binding.
5200	 */
5201	inp = stcb->sctp_ep;
5202	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5203		/* if bound all, destination is not restricted */
5204		/*
5205		 * RRS: Question during lock work: Is this correct? If you
5206		 * are bound-all you still might need to obey the V4--V6
5207		 * flags??? IMO this bound-all stuff needs to be removed!
5208		 */
5209		return (1);
5210	}
5211	/* NOTE: all "scope" checks are done when local addresses are added */
5212	switch (destaddr->sa_family) {
5213#ifdef INET6
5214	case AF_INET6:
5215		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5216		break;
5217#endif
5218#ifdef INET
5219	case AF_INET:
5220		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5221		break;
5222#endif
5223	default:
5224		/* invalid family, so it's unreachable */
5225		answer = 0;
5226		break;
5227	}
5228	return (answer);
5229}
5230
5231/*
5232 * update the inp_vflags on an endpoint
5233 */
5234static void
5235sctp_update_ep_vflag(struct sctp_inpcb *inp)
5236{
5237	struct sctp_laddr *laddr;
5238
5239	/* first clear the flag */
5240	inp->ip_inp.inp.inp_vflag = 0;
5241	/* set the flag based on addresses on the ep list */
5242	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5243		if (laddr->ifa == NULL) {
5244			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5245			    __FUNCTION__);
5246			continue;
5247		}
5248		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5249			continue;
5250		}
5251		switch (laddr->ifa->address.sa.sa_family) {
5252#ifdef INET6
5253		case AF_INET6:
5254			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5255			break;
5256#endif
5257#ifdef INET
5258		case AF_INET:
5259			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5260			break;
5261#endif
5262		default:
5263			break;
5264		}
5265	}
5266}
5267
5268/*
5269 * Add the address to the endpoint local address list There is nothing to be
5270 * done if we are bound to all addresses
5271 */
5272void
5273sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5274{
5275	struct sctp_laddr *laddr;
5276	int fnd, error = 0;
5277
5278	fnd = 0;
5279
5280	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5281		/* You are already bound to all. You have it already */
5282		return;
5283	}
5284#ifdef INET6
5285	if (ifa->address.sa.sa_family == AF_INET6) {
5286		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5287			/* Can't bind a non-useable addr. */
5288			return;
5289		}
5290	}
5291#endif
5292	/* first, is it already present? */
5293	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5294		if (laddr->ifa == ifa) {
5295			fnd = 1;
5296			break;
5297		}
5298	}
5299
5300	if (fnd == 0) {
5301		/* Not in the ep list */
5302		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5303		if (error != 0)
5304			return;
5305		inp->laddr_count++;
5306		/* update inp_vflag flags */
5307		switch (ifa->address.sa.sa_family) {
5308#ifdef INET6
5309		case AF_INET6:
5310			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5311			break;
5312#endif
5313#ifdef INET
5314		case AF_INET:
5315			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5316			break;
5317#endif
5318		default:
5319			break;
5320		}
5321	}
5322	return;
5323}
5324
5325
5326/*
5327 * select a new (hopefully reachable) destination net (should only be used
5328 * when we deleted an ep addr that is the only usable source address to reach
5329 * the destination net)
5330 */
5331static void
5332sctp_select_primary_destination(struct sctp_tcb *stcb)
5333{
5334	struct sctp_nets *net;
5335
5336	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5337		/* for now, we'll just pick the first reachable one we find */
5338		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5339			continue;
5340		if (sctp_destination_is_reachable(stcb,
5341		    (struct sockaddr *)&net->ro._l_addr)) {
5342			/* found a reachable destination */
5343			stcb->asoc.primary_destination = net;
5344		}
5345	}
5346	/* I can't there from here! ...we're gonna die shortly... */
5347}
5348
5349
5350/*
5351 * Delete the address from the endpoint local address list There is nothing
5352 * to be done if we are bound to all addresses
5353 */
5354void
5355sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5356{
5357	struct sctp_laddr *laddr;
5358	int fnd;
5359
5360	fnd = 0;
5361	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5362		/* You are already bound to all. You have it already */
5363		return;
5364	}
5365	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5366		if (laddr->ifa == ifa) {
5367			fnd = 1;
5368			break;
5369		}
5370	}
5371	if (fnd && (inp->laddr_count < 2)) {
5372		/* can't delete unless there are at LEAST 2 addresses */
5373		return;
5374	}
5375	if (fnd) {
5376		/*
5377		 * clean up any use of this address go through our
5378		 * associations and clear any last_used_address that match
5379		 * this one for each assoc, see if a new primary_destination
5380		 * is needed
5381		 */
5382		struct sctp_tcb *stcb;
5383
5384		/* clean up "next_addr_touse" */
5385		if (inp->next_addr_touse == laddr)
5386			/* delete this address */
5387			inp->next_addr_touse = NULL;
5388
5389		/* clean up "last_used_address" */
5390		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5391			struct sctp_nets *net;
5392
5393			SCTP_TCB_LOCK(stcb);
5394			if (stcb->asoc.last_used_address == laddr)
5395				/* delete this address */
5396				stcb->asoc.last_used_address = NULL;
5397			/*
5398			 * Now spin through all the nets and purge any ref
5399			 * to laddr
5400			 */
5401			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5402				if (net->ro._s_addr &&
5403				    (net->ro._s_addr->ifa == laddr->ifa)) {
5404					/* Yep, purge src address selected */
5405					sctp_rtentry_t *rt;
5406
5407					/* delete this address if cached */
5408					rt = net->ro.ro_rt;
5409					if (rt != NULL) {
5410						RTFREE(rt);
5411						net->ro.ro_rt = NULL;
5412					}
5413					sctp_free_ifa(net->ro._s_addr);
5414					net->ro._s_addr = NULL;
5415					net->src_addr_selected = 0;
5416				}
5417			}
5418			SCTP_TCB_UNLOCK(stcb);
5419		}		/* for each tcb */
5420		/* remove it from the ep list */
5421		sctp_remove_laddr(laddr);
5422		inp->laddr_count--;
5423		/* update inp_vflag flags */
5424		sctp_update_ep_vflag(inp);
5425	}
5426	return;
5427}
5428
5429/*
5430 * Add the address to the TCB local address restricted list.
5431 * This is a "pending" address list (eg. addresses waiting for an
5432 * ASCONF-ACK response) and cannot be used as a valid source address.
5433 */
5434void
5435sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5436{
5437	struct sctp_laddr *laddr;
5438	struct sctpladdr *list;
5439
5440	/*
5441	 * Assumes TCB is locked.. and possibly the INP. May need to
5442	 * confirm/fix that if we need it and is not the case.
5443	 */
5444	list = &stcb->asoc.sctp_restricted_addrs;
5445
5446#ifdef INET6
5447	if (ifa->address.sa.sa_family == AF_INET6) {
5448		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5449			/* Can't bind a non-existent addr. */
5450			return;
5451		}
5452	}
5453#endif
5454	/* does the address already exist? */
5455	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5456		if (laddr->ifa == ifa) {
5457			return;
5458		}
5459	}
5460
5461	/* add to the list */
5462	(void)sctp_insert_laddr(list, ifa, 0);
5463	return;
5464}
5465
5466/*
5467 * insert an laddr entry with the given ifa for the desired list
5468 */
5469int
5470sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5471{
5472	struct sctp_laddr *laddr;
5473
5474	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5475	if (laddr == NULL) {
5476		/* out of memory? */
5477		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5478		return (EINVAL);
5479	}
5480	SCTP_INCR_LADDR_COUNT();
5481	bzero(laddr, sizeof(*laddr));
5482	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5483	laddr->ifa = ifa;
5484	laddr->action = act;
5485	atomic_add_int(&ifa->refcount, 1);
5486	/* insert it */
5487	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5488
5489	return (0);
5490}
5491
5492/*
5493 * Remove an laddr entry from the local address list (on an assoc)
5494 */
5495void
5496sctp_remove_laddr(struct sctp_laddr *laddr)
5497{
5498
5499	/* remove from the list */
5500	LIST_REMOVE(laddr, sctp_nxt_addr);
5501	sctp_free_ifa(laddr->ifa);
5502	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5503	SCTP_DECR_LADDR_COUNT();
5504}
5505
5506/*
5507 * Remove a local address from the TCB local address restricted list
5508 */
5509void
5510sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5511{
5512	struct sctp_inpcb *inp;
5513	struct sctp_laddr *laddr;
5514
5515	/*
5516	 * This is called by asconf work. It is assumed that a) The TCB is
5517	 * locked and b) The INP is locked. This is true in as much as I can
5518	 * trace through the entry asconf code where I did these locks.
5519	 * Again, the ASCONF code is a bit different in that it does lock
5520	 * the INP during its work often times. This must be since we don't
5521	 * want other proc's looking up things while what they are looking
5522	 * up is changing :-D
5523	 */
5524
5525	inp = stcb->sctp_ep;
5526	/* if subset bound and don't allow ASCONF's, can't delete last */
5527	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5528	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5529		if (stcb->sctp_ep->laddr_count < 2) {
5530			/* can't delete last address */
5531			return;
5532		}
5533	}
5534	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5535		/* remove the address if it exists */
5536		if (laddr->ifa == NULL)
5537			continue;
5538		if (laddr->ifa == ifa) {
5539			sctp_remove_laddr(laddr);
5540			return;
5541		}
5542	}
5543
5544	/* address not found! */
5545	return;
5546}
5547
5548/*
5549 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5550 */
5551/* sysctl */
5552static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5553static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5554
5555
5556
5557#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5558struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5559int *sctp_cpuarry = NULL;
5560void
5561sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5562{
5563	/* Queue a packet to a processor for the specified core */
5564	struct sctp_mcore_queue *qent;
5565	struct sctp_mcore_ctrl *wkq;
5566	int need_wake = 0;
5567
5568	if (sctp_mcore_workers == NULL) {
5569		/* Something went way bad during setup */
5570		sctp_input_with_port(m, off, 0);
5571		return;
5572	}
5573	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5574	    (sizeof(struct sctp_mcore_queue)),
5575	    SCTP_M_MCORE);
5576	if (qent == NULL) {
5577		/* This is trouble  */
5578		sctp_input_with_port(m, off, 0);
5579		return;
5580	}
5581	qent->vn = curvnet;
5582	qent->m = m;
5583	qent->off = off;
5584	qent->v6 = 0;
5585	wkq = &sctp_mcore_workers[cpu_to_use];
5586	SCTP_MCORE_QLOCK(wkq);
5587
5588	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5589	if (wkq->running == 0) {
5590		need_wake = 1;
5591	}
5592	SCTP_MCORE_QUNLOCK(wkq);
5593	if (need_wake) {
5594		wakeup(&wkq->running);
5595	}
5596}
5597
5598static void
5599sctp_mcore_thread(void *arg)
5600{
5601
5602	struct sctp_mcore_ctrl *wkq;
5603	struct sctp_mcore_queue *qent;
5604
5605	wkq = (struct sctp_mcore_ctrl *)arg;
5606	struct mbuf *m;
5607	int off, v6;
5608
5609	/* Wait for first tickle */
5610	SCTP_MCORE_LOCK(wkq);
5611	wkq->running = 0;
5612	msleep(&wkq->running,
5613	    &wkq->core_mtx,
5614	    0, "wait for pkt", 0);
5615	SCTP_MCORE_UNLOCK(wkq);
5616
5617	/* Bind to our cpu */
5618	thread_lock(curthread);
5619	sched_bind(curthread, wkq->cpuid);
5620	thread_unlock(curthread);
5621
5622	/* Now lets start working */
5623	SCTP_MCORE_LOCK(wkq);
5624	/* Now grab lock and go */
5625	for (;;) {
5626		SCTP_MCORE_QLOCK(wkq);
5627skip_sleep:
5628		wkq->running = 1;
5629		qent = TAILQ_FIRST(&wkq->que);
5630		if (qent) {
5631			TAILQ_REMOVE(&wkq->que, qent, next);
5632			SCTP_MCORE_QUNLOCK(wkq);
5633			CURVNET_SET(qent->vn);
5634			m = qent->m;
5635			off = qent->off;
5636			v6 = qent->v6;
5637			SCTP_FREE(qent, SCTP_M_MCORE);
5638			if (v6 == 0) {
5639				sctp_input_with_port(m, off, 0);
5640			} else {
5641				SCTP_PRINTF("V6 not yet supported\n");
5642				sctp_m_freem(m);
5643			}
5644			CURVNET_RESTORE();
5645			SCTP_MCORE_QLOCK(wkq);
5646		}
5647		wkq->running = 0;
5648		if (!TAILQ_EMPTY(&wkq->que)) {
5649			goto skip_sleep;
5650		}
5651		SCTP_MCORE_QUNLOCK(wkq);
5652		msleep(&wkq->running,
5653		    &wkq->core_mtx,
5654		    0, "wait for pkt", 0);
5655	}
5656}
5657
5658static void
5659sctp_startup_mcore_threads(void)
5660{
5661	int i, cpu;
5662
5663	if (mp_ncpus == 1)
5664		return;
5665
5666	if (sctp_mcore_workers != NULL) {
5667		/*
5668		 * Already been here in some previous vnet?
5669		 */
5670		return;
5671	}
5672	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5673	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5674	    SCTP_M_MCORE);
5675	if (sctp_mcore_workers == NULL) {
5676		/* TSNH I hope */
5677		return;
5678	}
5679	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5680	    sizeof(struct sctp_mcore_ctrl)));
5681	/* Init the structures */
5682	for (i = 0; i <= mp_maxid; i++) {
5683		TAILQ_INIT(&sctp_mcore_workers[i].que);
5684		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5685		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5686		sctp_mcore_workers[i].cpuid = i;
5687	}
5688	if (sctp_cpuarry == NULL) {
5689		SCTP_MALLOC(sctp_cpuarry, int *,
5690		    (mp_ncpus * sizeof(int)),
5691		    SCTP_M_MCORE);
5692		i = 0;
5693		CPU_FOREACH(cpu) {
5694			sctp_cpuarry[i] = cpu;
5695			i++;
5696		}
5697	}
5698	/* Now start them all */
5699	CPU_FOREACH(cpu) {
5700		(void)kproc_create(sctp_mcore_thread,
5701		    (void *)&sctp_mcore_workers[cpu],
5702		    &sctp_mcore_workers[cpu].thread_proc,
5703		    RFPROC,
5704		    SCTP_KTHREAD_PAGES,
5705		    SCTP_MCORE_NAME);
5706
5707	}
5708}
5709
5710#endif
5711
5712void
5713sctp_pcb_init()
5714{
5715	/*
5716	 * SCTP initialization for the PCB structures should be called by
5717	 * the sctp_init() funciton.
5718	 */
5719	int i;
5720	struct timeval tv;
5721
5722	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5723		/* error I was called twice */
5724		return;
5725	}
5726	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5727
5728#if defined(SCTP_LOCAL_TRACE_BUF)
5729	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5730#endif
5731#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5732	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5733	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5734	    SCTP_M_MCORE);
5735#endif
5736	(void)SCTP_GETTIME_TIMEVAL(&tv);
5737#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5738	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5739	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5740	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5741#else
5742	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5743	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5744	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5745#endif
5746	/* init the empty list of (All) Endpoints */
5747	LIST_INIT(&SCTP_BASE_INFO(listhead));
5748
5749
5750	/* init the hash table of endpoints */
5751	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5752	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5753	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5754	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5755	    &SCTP_BASE_INFO(hashasocmark));
5756	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5757	    &SCTP_BASE_INFO(hashmark));
5758	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5759	    &SCTP_BASE_INFO(hashtcpmark));
5760	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5761
5762
5763	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5764	    &SCTP_BASE_INFO(hashvrfmark));
5765
5766	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5767	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5768	/* init the zones */
5769	/*
5770	 * FIX ME: Should check for NULL returns, but if it does fail we are
5771	 * doomed to panic anyways... add later maybe.
5772	 */
5773	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5774	    sizeof(struct sctp_inpcb), maxsockets);
5775
5776	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5777	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5778
5779	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5780	    sizeof(struct sctp_laddr),
5781	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5782
5783	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5784	    sizeof(struct sctp_nets),
5785	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5786
5787	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5788	    sizeof(struct sctp_tmit_chunk),
5789	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5790
5791	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5792	    sizeof(struct sctp_queued_to_read),
5793	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5794
5795	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5796	    sizeof(struct sctp_stream_queue_pending),
5797	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5798
5799	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5800	    sizeof(struct sctp_asconf),
5801	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5802
5803	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5804	    sizeof(struct sctp_asconf_ack),
5805	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5806
5807
5808	/* Master Lock INIT for info structure */
5809	SCTP_INP_INFO_LOCK_INIT();
5810	SCTP_STATLOG_INIT_LOCK();
5811
5812	SCTP_IPI_COUNT_INIT();
5813	SCTP_IPI_ADDR_INIT();
5814#ifdef SCTP_PACKET_LOGGING
5815	SCTP_IP_PKTLOG_INIT();
5816#endif
5817	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5818
5819	SCTP_WQ_ADDR_INIT();
5820	/* not sure if we need all the counts */
5821	SCTP_BASE_INFO(ipi_count_ep) = 0;
5822	/* assoc/tcb zone info */
5823	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5824	/* local addrlist zone info */
5825	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5826	/* remote addrlist zone info */
5827	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5828	/* chunk info */
5829	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5830
5831	/* socket queue zone info */
5832	SCTP_BASE_INFO(ipi_count_readq) = 0;
5833
5834	/* stream out queue cont */
5835	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5836
5837	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5838	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5839
5840	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5841
5842	/* Init the TIMEWAIT list */
5843	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5844		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5845	}
5846
5847	sctp_startup_iterator();
5848
5849#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5850	sctp_startup_mcore_threads();
5851#endif
5852
5853	/*
5854	 * INIT the default VRF which for BSD is the only one, other O/S's
5855	 * may have more. But initially they must start with one and then
5856	 * add the VRF's as addresses are added.
5857	 */
5858	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5859}
5860
5861/*
5862 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5863 */
5864void
5865sctp_pcb_finish(void)
5866{
5867	struct sctp_vrflist *vrf_bucket;
5868	struct sctp_vrf *vrf, *nvrf;
5869	struct sctp_ifn *ifn, *nifn;
5870	struct sctp_ifa *ifa, *nifa;
5871	struct sctpvtaghead *chain;
5872	struct sctp_tagblock *twait_block, *prev_twait_block;
5873	struct sctp_laddr *wi, *nwi;
5874	int i;
5875
5876	/*
5877	 * Free BSD the it thread never exits but we do clean up. The only
5878	 * way freebsd reaches here if we have VRF's but we still add the
5879	 * ifdef to make it compile on old versions.
5880	 */
5881	{
5882		struct sctp_iterator *it, *nit;
5883
5884		SCTP_IPI_ITERATOR_WQ_LOCK();
5885		TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5886			if (it->vn != curvnet) {
5887				continue;
5888			}
5889			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5890			if (it->function_atend != NULL) {
5891				(*it->function_atend) (it->pointer, it->val);
5892			}
5893			SCTP_FREE(it, SCTP_M_ITER);
5894		}
5895		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5896		SCTP_ITERATOR_LOCK();
5897		if ((sctp_it_ctl.cur_it) &&
5898		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5899			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5900		}
5901		SCTP_ITERATOR_UNLOCK();
5902	}
5903
5904	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5905	SCTP_WQ_ADDR_LOCK();
5906	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5907		LIST_REMOVE(wi, sctp_nxt_addr);
5908		SCTP_DECR_LADDR_COUNT();
5909		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5910	}
5911	SCTP_WQ_ADDR_UNLOCK();
5912
5913	/*
5914	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5915	 * destroyed first).
5916	 */
5917	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5918	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5919		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5920			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5921				/* free the ifa */
5922				LIST_REMOVE(ifa, next_bucket);
5923				LIST_REMOVE(ifa, next_ifa);
5924				SCTP_FREE(ifa, SCTP_M_IFA);
5925			}
5926			/* free the ifn */
5927			LIST_REMOVE(ifn, next_bucket);
5928			LIST_REMOVE(ifn, next_ifn);
5929			SCTP_FREE(ifn, SCTP_M_IFN);
5930		}
5931		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5932		/* free the vrf */
5933		LIST_REMOVE(vrf, next_vrf);
5934		SCTP_FREE(vrf, SCTP_M_VRF);
5935	}
5936	/* free the vrf hashes */
5937	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5938	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5939
5940	/*
5941	 * free the TIMEWAIT list elements malloc'd in the function
5942	 * sctp_add_vtag_to_timewait()...
5943	 */
5944	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5945		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5946		if (!LIST_EMPTY(chain)) {
5947			prev_twait_block = NULL;
5948			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5949				if (prev_twait_block) {
5950					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5951				}
5952				prev_twait_block = twait_block;
5953			}
5954			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5955		}
5956	}
5957
5958	/* free the locks and mutexes */
5959#ifdef SCTP_PACKET_LOGGING
5960	SCTP_IP_PKTLOG_DESTROY();
5961#endif
5962	SCTP_IPI_ADDR_DESTROY();
5963	SCTP_STATLOG_DESTROY();
5964	SCTP_INP_INFO_LOCK_DESTROY();
5965
5966	SCTP_WQ_ADDR_DESTROY();
5967
5968	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5969	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5970	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5971	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5972	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5973	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5974	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5975	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5976	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5977	/* Get rid of other stuff to */
5978	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5979		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5980	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5981		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5982	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5983		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5984#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5985	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5986#endif
5987}
5988
5989
5990int
5991sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5992    int offset, int limit,
5993    struct sockaddr *src, struct sockaddr *dst,
5994    struct sockaddr *altsa)
5995{
5996	/*
5997	 * grub through the INIT pulling addresses and loading them to the
5998	 * nets structure in the asoc. The from address in the mbuf should
5999	 * also be loaded (if it is not already). This routine can be called
6000	 * with either INIT or INIT-ACK's as long as the m points to the IP
6001	 * packet and the offset points to the beginning of the parameters.
6002	 */
6003	struct sctp_inpcb *inp;
6004	struct sctp_nets *net, *nnet, *net_tmp;
6005	struct sctp_paramhdr *phdr, parm_buf;
6006	struct sctp_tcb *stcb_tmp;
6007	uint16_t ptype, plen;
6008	struct sockaddr *sa;
6009	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6010	struct sctp_auth_random *p_random = NULL;
6011	uint16_t random_len = 0;
6012	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6013	struct sctp_auth_hmac_algo *hmacs = NULL;
6014	uint16_t hmacs_len = 0;
6015	uint8_t saw_asconf = 0;
6016	uint8_t saw_asconf_ack = 0;
6017	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6018	struct sctp_auth_chunk_list *chunks = NULL;
6019	uint16_t num_chunks = 0;
6020	sctp_key_t *new_key;
6021	uint32_t keylen;
6022	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6023	uint8_t ecn_allowed;
6024
6025#ifdef INET
6026	struct sockaddr_in sin;
6027
6028#endif
6029#ifdef INET6
6030	struct sockaddr_in6 sin6;
6031
6032#endif
6033
6034	/* First get the destination address setup too. */
6035#ifdef INET
6036	memset(&sin, 0, sizeof(sin));
6037	sin.sin_family = AF_INET;
6038	sin.sin_len = sizeof(sin);
6039	sin.sin_port = stcb->rport;
6040#endif
6041#ifdef INET6
6042	memset(&sin6, 0, sizeof(sin6));
6043	sin6.sin6_family = AF_INET6;
6044	sin6.sin6_len = sizeof(struct sockaddr_in6);
6045	sin6.sin6_port = stcb->rport;
6046#endif
6047	if (altsa) {
6048		sa = altsa;
6049	} else {
6050		sa = src;
6051	}
6052	/* Turn off ECN until we get through all params */
6053	ecn_allowed = 0;
6054	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6055		/* mark all addresses that we have currently on the list */
6056		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6057	}
6058	/* does the source address already exist? if so skip it */
6059	inp = stcb->sctp_ep;
6060	atomic_add_int(&stcb->asoc.refcnt, 1);
6061	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6062	atomic_add_int(&stcb->asoc.refcnt, -1);
6063
6064	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6065		/* we must add the source address */
6066		/* no scope set here since we have a tcb already. */
6067		switch (sa->sa_family) {
6068#ifdef INET
6069		case AF_INET:
6070			if (stcb->asoc.scope.ipv4_addr_legal) {
6071				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6072					return (-1);
6073				}
6074			}
6075			break;
6076#endif
6077#ifdef INET6
6078		case AF_INET6:
6079			if (stcb->asoc.scope.ipv6_addr_legal) {
6080				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6081					return (-2);
6082				}
6083			}
6084			break;
6085#endif
6086		default:
6087			break;
6088		}
6089	} else {
6090		if (net_tmp != NULL && stcb_tmp == stcb) {
6091			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6092		} else if (stcb_tmp != stcb) {
6093			/* It belongs to another association? */
6094			if (stcb_tmp)
6095				SCTP_TCB_UNLOCK(stcb_tmp);
6096			return (-3);
6097		}
6098	}
6099	if (stcb->asoc.state == 0) {
6100		/* the assoc was freed? */
6101		return (-4);
6102	}
6103	/*
6104	 * peer must explicitly turn this on. This may have been initialized
6105	 * to be "on" in order to allow local addr changes while INIT's are
6106	 * in flight.
6107	 */
6108	stcb->asoc.peer_supports_asconf = 0;
6109	/* now we must go through each of the params. */
6110	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6111	while (phdr) {
6112		ptype = ntohs(phdr->param_type);
6113		plen = ntohs(phdr->param_length);
6114		/*
6115		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6116		 * (uint32_t)ptype, (int)plen);
6117		 */
6118		if (offset + plen > limit) {
6119			break;
6120		}
6121		if (plen == 0) {
6122			break;
6123		}
6124#ifdef INET
6125		if (ptype == SCTP_IPV4_ADDRESS) {
6126			if (stcb->asoc.scope.ipv4_addr_legal) {
6127				struct sctp_ipv4addr_param *p4, p4_buf;
6128
6129				/* ok get the v4 address and check/add */
6130				phdr = sctp_get_next_param(m, offset,
6131				    (struct sctp_paramhdr *)&p4_buf,
6132				    sizeof(p4_buf));
6133				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6134				    phdr == NULL) {
6135					return (-5);
6136				}
6137				p4 = (struct sctp_ipv4addr_param *)phdr;
6138				sin.sin_addr.s_addr = p4->addr;
6139				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6140					/* Skip multi-cast addresses */
6141					goto next_param;
6142				}
6143				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6144				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6145					goto next_param;
6146				}
6147				sa = (struct sockaddr *)&sin;
6148				inp = stcb->sctp_ep;
6149				atomic_add_int(&stcb->asoc.refcnt, 1);
6150				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6151				    dst, stcb);
6152				atomic_add_int(&stcb->asoc.refcnt, -1);
6153
6154				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6155				    inp == NULL) {
6156					/* we must add the source address */
6157					/*
6158					 * no scope set since we have a tcb
6159					 * already
6160					 */
6161
6162					/*
6163					 * we must validate the state again
6164					 * here
6165					 */
6166			add_it_now:
6167					if (stcb->asoc.state == 0) {
6168						/* the assoc was freed? */
6169						return (-7);
6170					}
6171					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6172						return (-8);
6173					}
6174				} else if (stcb_tmp == stcb) {
6175					if (stcb->asoc.state == 0) {
6176						/* the assoc was freed? */
6177						return (-10);
6178					}
6179					if (net != NULL) {
6180						/* clear flag */
6181						net->dest_state &=
6182						    ~SCTP_ADDR_NOT_IN_ASSOC;
6183					}
6184				} else {
6185					/*
6186					 * strange, address is in another
6187					 * assoc? straighten out locks.
6188					 */
6189					if (stcb_tmp) {
6190						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6191							/*
6192							 * in setup state we
6193							 * abort this guy
6194							 */
6195							sctp_abort_an_association(stcb_tmp->sctp_ep,
6196							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6197							goto add_it_now;
6198						}
6199						SCTP_TCB_UNLOCK(stcb_tmp);
6200					}
6201					if (stcb->asoc.state == 0) {
6202						/* the assoc was freed? */
6203						return (-12);
6204					}
6205					return (-13);
6206				}
6207			}
6208		} else
6209#endif
6210#ifdef INET6
6211		if (ptype == SCTP_IPV6_ADDRESS) {
6212			if (stcb->asoc.scope.ipv6_addr_legal) {
6213				/* ok get the v6 address and check/add */
6214				struct sctp_ipv6addr_param *p6, p6_buf;
6215
6216				phdr = sctp_get_next_param(m, offset,
6217				    (struct sctp_paramhdr *)&p6_buf,
6218				    sizeof(p6_buf));
6219				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6220				    phdr == NULL) {
6221					return (-14);
6222				}
6223				p6 = (struct sctp_ipv6addr_param *)phdr;
6224				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6225				    sizeof(p6->addr));
6226				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6227					/* Skip multi-cast addresses */
6228					goto next_param;
6229				}
6230				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6231					/*
6232					 * Link local make no sense without
6233					 * scope
6234					 */
6235					goto next_param;
6236				}
6237				sa = (struct sockaddr *)&sin6;
6238				inp = stcb->sctp_ep;
6239				atomic_add_int(&stcb->asoc.refcnt, 1);
6240				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6241				    dst, stcb);
6242				atomic_add_int(&stcb->asoc.refcnt, -1);
6243				if (stcb_tmp == NULL &&
6244				    (inp == stcb->sctp_ep || inp == NULL)) {
6245					/*
6246					 * we must validate the state again
6247					 * here
6248					 */
6249			add_it_now6:
6250					if (stcb->asoc.state == 0) {
6251						/* the assoc was freed? */
6252						return (-16);
6253					}
6254					/*
6255					 * we must add the address, no scope
6256					 * set
6257					 */
6258					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6259						return (-17);
6260					}
6261				} else if (stcb_tmp == stcb) {
6262					/*
6263					 * we must validate the state again
6264					 * here
6265					 */
6266					if (stcb->asoc.state == 0) {
6267						/* the assoc was freed? */
6268						return (-19);
6269					}
6270					if (net != NULL) {
6271						/* clear flag */
6272						net->dest_state &=
6273						    ~SCTP_ADDR_NOT_IN_ASSOC;
6274					}
6275				} else {
6276					/*
6277					 * strange, address is in another
6278					 * assoc? straighten out locks.
6279					 */
6280					if (stcb_tmp)
6281						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6282							/*
6283							 * in setup state we
6284							 * abort this guy
6285							 */
6286							sctp_abort_an_association(stcb_tmp->sctp_ep,
6287							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6288							goto add_it_now6;
6289						}
6290					SCTP_TCB_UNLOCK(stcb_tmp);
6291
6292					if (stcb->asoc.state == 0) {
6293						/* the assoc was freed? */
6294						return (-21);
6295					}
6296					return (-22);
6297				}
6298			}
6299		} else
6300#endif
6301		if (ptype == SCTP_ECN_CAPABLE) {
6302			ecn_allowed = 1;
6303		} else if (ptype == SCTP_ULP_ADAPTATION) {
6304			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6305				struct sctp_adaptation_layer_indication ai,
6306				                                *aip;
6307
6308				phdr = sctp_get_next_param(m, offset,
6309				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6310				aip = (struct sctp_adaptation_layer_indication *)phdr;
6311				if (aip) {
6312					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6313					stcb->asoc.adaptation_needed = 1;
6314				}
6315			}
6316		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6317			struct sctp_asconf_addr_param lstore, *fee;
6318			int lptype;
6319			struct sockaddr *lsa = NULL;
6320
6321#ifdef INET
6322			struct sctp_asconf_addrv4_param *fii;
6323
6324#endif
6325
6326			stcb->asoc.peer_supports_asconf = 1;
6327			if (plen > sizeof(lstore)) {
6328				return (-23);
6329			}
6330			phdr = sctp_get_next_param(m, offset,
6331			    (struct sctp_paramhdr *)&lstore,
6332			    min(plen, sizeof(lstore)));
6333			if (phdr == NULL) {
6334				return (-24);
6335			}
6336			fee = (struct sctp_asconf_addr_param *)phdr;
6337			lptype = ntohs(fee->addrp.ph.param_type);
6338			switch (lptype) {
6339#ifdef INET
6340			case SCTP_IPV4_ADDRESS:
6341				if (plen !=
6342				    sizeof(struct sctp_asconf_addrv4_param)) {
6343					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6344					    (int)sizeof(struct sctp_asconf_addrv4_param),
6345					    plen);
6346				} else {
6347					fii = (struct sctp_asconf_addrv4_param *)fee;
6348					sin.sin_addr.s_addr = fii->addrp.addr;
6349					lsa = (struct sockaddr *)&sin;
6350				}
6351				break;
6352#endif
6353#ifdef INET6
6354			case SCTP_IPV6_ADDRESS:
6355				if (plen !=
6356				    sizeof(struct sctp_asconf_addr_param)) {
6357					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6358					    (int)sizeof(struct sctp_asconf_addr_param),
6359					    plen);
6360				} else {
6361					memcpy(sin6.sin6_addr.s6_addr,
6362					    fee->addrp.addr,
6363					    sizeof(fee->addrp.addr));
6364					lsa = (struct sockaddr *)&sin6;
6365				}
6366				break;
6367#endif
6368			default:
6369				break;
6370			}
6371			if (lsa) {
6372				(void)sctp_set_primary_addr(stcb, sa, NULL);
6373			}
6374		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6375			stcb->asoc.peer_supports_nat = 1;
6376		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6377			/* Peer supports pr-sctp */
6378			stcb->asoc.peer_supports_prsctp = 1;
6379		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6380			/* A supported extension chunk */
6381			struct sctp_supported_chunk_types_param *pr_supported;
6382			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6383			int num_ent, i;
6384
6385			phdr = sctp_get_next_param(m, offset,
6386			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6387			if (phdr == NULL) {
6388				return (-25);
6389			}
6390			stcb->asoc.peer_supports_asconf = 0;
6391			stcb->asoc.peer_supports_prsctp = 0;
6392			stcb->asoc.peer_supports_pktdrop = 0;
6393			stcb->asoc.peer_supports_strreset = 0;
6394			stcb->asoc.peer_supports_nr_sack = 0;
6395			stcb->asoc.peer_supports_auth = 0;
6396			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6397			num_ent = plen - sizeof(struct sctp_paramhdr);
6398			for (i = 0; i < num_ent; i++) {
6399				switch (pr_supported->chunk_types[i]) {
6400				case SCTP_ASCONF:
6401				case SCTP_ASCONF_ACK:
6402					stcb->asoc.peer_supports_asconf = 1;
6403					break;
6404				case SCTP_FORWARD_CUM_TSN:
6405					stcb->asoc.peer_supports_prsctp = 1;
6406					break;
6407				case SCTP_PACKET_DROPPED:
6408					stcb->asoc.peer_supports_pktdrop = 1;
6409					break;
6410				case SCTP_NR_SELECTIVE_ACK:
6411					stcb->asoc.peer_supports_nr_sack = 1;
6412					break;
6413				case SCTP_STREAM_RESET:
6414					stcb->asoc.peer_supports_strreset = 1;
6415					break;
6416				case SCTP_AUTHENTICATION:
6417					stcb->asoc.peer_supports_auth = 1;
6418					break;
6419				default:
6420					/* one I have not learned yet */
6421					break;
6422
6423				}
6424			}
6425		} else if (ptype == SCTP_RANDOM) {
6426			if (plen > sizeof(random_store))
6427				break;
6428			if (got_random) {
6429				/* already processed a RANDOM */
6430				goto next_param;
6431			}
6432			phdr = sctp_get_next_param(m, offset,
6433			    (struct sctp_paramhdr *)random_store,
6434			    min(sizeof(random_store), plen));
6435			if (phdr == NULL)
6436				return (-26);
6437			p_random = (struct sctp_auth_random *)phdr;
6438			random_len = plen - sizeof(*p_random);
6439			/* enforce the random length */
6440			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6441				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6442				return (-27);
6443			}
6444			got_random = 1;
6445		} else if (ptype == SCTP_HMAC_LIST) {
6446			int num_hmacs;
6447			int i;
6448
6449			if (plen > sizeof(hmacs_store))
6450				break;
6451			if (got_hmacs) {
6452				/* already processed a HMAC list */
6453				goto next_param;
6454			}
6455			phdr = sctp_get_next_param(m, offset,
6456			    (struct sctp_paramhdr *)hmacs_store,
6457			    min(plen, sizeof(hmacs_store)));
6458			if (phdr == NULL)
6459				return (-28);
6460			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6461			hmacs_len = plen - sizeof(*hmacs);
6462			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6463			/* validate the hmac list */
6464			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6465				return (-29);
6466			}
6467			if (stcb->asoc.peer_hmacs != NULL)
6468				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6469			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6470			if (stcb->asoc.peer_hmacs != NULL) {
6471				for (i = 0; i < num_hmacs; i++) {
6472					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6473					    ntohs(hmacs->hmac_ids[i]));
6474				}
6475			}
6476			got_hmacs = 1;
6477		} else if (ptype == SCTP_CHUNK_LIST) {
6478			int i;
6479
6480			if (plen > sizeof(chunks_store))
6481				break;
6482			if (got_chklist) {
6483				/* already processed a Chunks list */
6484				goto next_param;
6485			}
6486			phdr = sctp_get_next_param(m, offset,
6487			    (struct sctp_paramhdr *)chunks_store,
6488			    min(plen, sizeof(chunks_store)));
6489			if (phdr == NULL)
6490				return (-30);
6491			chunks = (struct sctp_auth_chunk_list *)phdr;
6492			num_chunks = plen - sizeof(*chunks);
6493			if (stcb->asoc.peer_auth_chunks != NULL)
6494				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6495			else
6496				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6497			for (i = 0; i < num_chunks; i++) {
6498				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6499				    stcb->asoc.peer_auth_chunks);
6500				/* record asconf/asconf-ack if listed */
6501				if (chunks->chunk_types[i] == SCTP_ASCONF)
6502					saw_asconf = 1;
6503				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6504					saw_asconf_ack = 1;
6505
6506			}
6507			got_chklist = 1;
6508		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6509			    (ptype == SCTP_STATE_COOKIE) ||
6510			    (ptype == SCTP_UNRECOG_PARAM) ||
6511			    (ptype == SCTP_COOKIE_PRESERVE) ||
6512			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6513			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6514			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6515			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6516		    (ptype == SCTP_SUCCESS_REPORT)) {
6517			 /* don't care */ ;
6518		} else {
6519			if ((ptype & 0x8000) == 0x0000) {
6520				/*
6521				 * must stop processing the rest of the
6522				 * param's. Any report bits were handled
6523				 * with the call to
6524				 * sctp_arethere_unrecognized_parameters()
6525				 * when the INIT or INIT-ACK was first seen.
6526				 */
6527				break;
6528			}
6529		}
6530
6531next_param:
6532		offset += SCTP_SIZE32(plen);
6533		if (offset >= limit) {
6534			break;
6535		}
6536		phdr = sctp_get_next_param(m, offset, &parm_buf,
6537		    sizeof(parm_buf));
6538	}
6539	/* Now check to see if we need to purge any addresses */
6540	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6541		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6542		    SCTP_ADDR_NOT_IN_ASSOC) {
6543			/* This address has been removed from the asoc */
6544			/* remove and free it */
6545			stcb->asoc.numnets--;
6546			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6547			sctp_free_remote_addr(net);
6548			if (net == stcb->asoc.primary_destination) {
6549				stcb->asoc.primary_destination = NULL;
6550				sctp_select_primary_destination(stcb);
6551			}
6552		}
6553	}
6554	if (ecn_allowed == 0) {
6555		stcb->asoc.ecn_allowed = 0;
6556	}
6557	/* validate authentication required parameters */
6558	if (got_random && got_hmacs) {
6559		stcb->asoc.peer_supports_auth = 1;
6560	} else {
6561		stcb->asoc.peer_supports_auth = 0;
6562	}
6563	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6564		/* peer does not support auth but sent a chunks list? */
6565		return (-31);
6566	}
6567	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6568	    !stcb->asoc.peer_supports_auth) {
6569		/* peer supports asconf but not auth? */
6570		return (-32);
6571	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6572	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6573		return (-33);
6574	}
6575	/* concatenate the full random key */
6576	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6577	if (chunks != NULL) {
6578		keylen += sizeof(*chunks) + num_chunks;
6579	}
6580	new_key = sctp_alloc_key(keylen);
6581	if (new_key != NULL) {
6582		/* copy in the RANDOM */
6583		if (p_random != NULL) {
6584			keylen = sizeof(*p_random) + random_len;
6585			bcopy(p_random, new_key->key, keylen);
6586		}
6587		/* append in the AUTH chunks */
6588		if (chunks != NULL) {
6589			bcopy(chunks, new_key->key + keylen,
6590			    sizeof(*chunks) + num_chunks);
6591			keylen += sizeof(*chunks) + num_chunks;
6592		}
6593		/* append in the HMACs */
6594		if (hmacs != NULL) {
6595			bcopy(hmacs, new_key->key + keylen,
6596			    sizeof(*hmacs) + hmacs_len);
6597		}
6598	} else {
6599		/* failed to get memory for the key */
6600		return (-34);
6601	}
6602	if (stcb->asoc.authinfo.peer_random != NULL)
6603		sctp_free_key(stcb->asoc.authinfo.peer_random);
6604	stcb->asoc.authinfo.peer_random = new_key;
6605	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6606	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6607
6608	return (0);
6609}
6610
6611int
6612sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6613    struct sctp_nets *net)
6614{
6615	/* make sure the requested primary address exists in the assoc */
6616	if (net == NULL && sa)
6617		net = sctp_findnet(stcb, sa);
6618
6619	if (net == NULL) {
6620		/* didn't find the requested primary address! */
6621		return (-1);
6622	} else {
6623		/* set the primary address */
6624		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6625			/* Must be confirmed, so queue to set */
6626			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6627			return (0);
6628		}
6629		stcb->asoc.primary_destination = net;
6630		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6631			sctp_free_remote_addr(stcb->asoc.alternate);
6632			stcb->asoc.alternate = NULL;
6633		}
6634		net = TAILQ_FIRST(&stcb->asoc.nets);
6635		if (net != stcb->asoc.primary_destination) {
6636			/*
6637			 * first one on the list is NOT the primary
6638			 * sctp_cmpaddr() is much more efficient if the
6639			 * primary is the first on the list, make it so.
6640			 */
6641			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6642			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6643		}
6644		return (0);
6645	}
6646}
6647
6648int
6649sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6650{
6651	/*
6652	 * This function serves two purposes. It will see if a TAG can be
6653	 * re-used and return 1 for yes it is ok and 0 for don't use that
6654	 * tag. A secondary function it will do is purge out old tags that
6655	 * can be removed.
6656	 */
6657	struct sctpvtaghead *chain;
6658	struct sctp_tagblock *twait_block;
6659	struct sctpasochead *head;
6660	struct sctp_tcb *stcb;
6661	int i;
6662
6663	SCTP_INP_INFO_RLOCK();
6664	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6665	    SCTP_BASE_INFO(hashasocmark))];
6666	if (head == NULL) {
6667		/* invalid vtag */
6668		goto skip_vtag_check;
6669	}
6670	LIST_FOREACH(stcb, head, sctp_asocs) {
6671		/*
6672		 * We choose not to lock anything here. TCB's can't be
6673		 * removed since we have the read lock, so they can't be
6674		 * freed on us, same thing for the INP. I may be wrong with
6675		 * this assumption, but we will go with it for now :-)
6676		 */
6677		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6678			continue;
6679		}
6680		if (stcb->asoc.my_vtag == tag) {
6681			/* candidate */
6682			if (stcb->rport != rport) {
6683				continue;
6684			}
6685			if (stcb->sctp_ep->sctp_lport != lport) {
6686				continue;
6687			}
6688			/* Its a used tag set */
6689			SCTP_INP_INFO_RUNLOCK();
6690			return (0);
6691		}
6692	}
6693skip_vtag_check:
6694
6695	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6696	/* Now what about timed wait ? */
6697	if (!LIST_EMPTY(chain)) {
6698		/*
6699		 * Block(s) are present, lets see if we have this tag in the
6700		 * list
6701		 */
6702		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6703			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6704				if (twait_block->vtag_block[i].v_tag == 0) {
6705					/* not used */
6706					continue;
6707				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6708				    now->tv_sec) {
6709					/* Audit expires this guy */
6710					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6711					twait_block->vtag_block[i].v_tag = 0;
6712					twait_block->vtag_block[i].lport = 0;
6713					twait_block->vtag_block[i].rport = 0;
6714				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6715					    (twait_block->vtag_block[i].lport == lport) &&
6716				    (twait_block->vtag_block[i].rport == rport)) {
6717					/* Bad tag, sorry :< */
6718					SCTP_INP_INFO_RUNLOCK();
6719					return (0);
6720				}
6721			}
6722		}
6723	}
6724	SCTP_INP_INFO_RUNLOCK();
6725	return (1);
6726}
6727
6728static void
6729sctp_drain_mbufs(struct sctp_tcb *stcb)
6730{
6731	/*
6732	 * We must hunt this association for MBUF's past the cumack (i.e.
6733	 * out of order data that we can renege on).
6734	 */
6735	struct sctp_association *asoc;
6736	struct sctp_tmit_chunk *chk, *nchk;
6737	uint32_t cumulative_tsn_p1;
6738	struct sctp_queued_to_read *ctl, *nctl;
6739	int cnt, strmat;
6740	uint32_t gap, i;
6741	int fnd = 0;
6742
6743	/* We look for anything larger than the cum-ack + 1 */
6744
6745	asoc = &stcb->asoc;
6746	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6747		/* none we can reneg on. */
6748		return;
6749	}
6750	SCTP_STAT_INCR(sctps_protocol_drains_done);
6751	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6752	cnt = 0;
6753	/* First look in the re-assembly queue */
6754	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6755		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6756			/* Yep it is above cum-ack */
6757			cnt++;
6758			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6759			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6760			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6761			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6762			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6763			if (chk->data) {
6764				sctp_m_freem(chk->data);
6765				chk->data = NULL;
6766			}
6767			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6768		}
6769	}
6770	/* Ok that was fun, now we will drain all the inbound streams? */
6771	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6772		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6773			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6774				/* Yep it is above cum-ack */
6775				cnt++;
6776				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6777				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6778				sctp_ucount_decr(asoc->cnt_on_all_streams);
6779				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6780				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6781				if (ctl->data) {
6782					sctp_m_freem(ctl->data);
6783					ctl->data = NULL;
6784				}
6785				sctp_free_remote_addr(ctl->whoFrom);
6786				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6787				SCTP_DECR_READQ_COUNT();
6788			}
6789		}
6790	}
6791	if (cnt) {
6792		/* We must back down to see what the new highest is */
6793		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6794			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6795			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6796				asoc->highest_tsn_inside_map = i;
6797				fnd = 1;
6798				break;
6799			}
6800		}
6801		if (!fnd) {
6802			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6803		}
6804		/*
6805		 * Question, should we go through the delivery queue? The
6806		 * only reason things are on here is the app not reading OR
6807		 * a p-d-api up. An attacker COULD send enough in to
6808		 * initiate the PD-API and then send a bunch of stuff to
6809		 * other streams... these would wind up on the delivery
6810		 * queue.. and then we would not get to them. But in order
6811		 * to do this I then have to back-track and un-deliver
6812		 * sequence numbers in streams.. el-yucko. I think for now
6813		 * we will NOT look at the delivery queue and leave it to be
6814		 * something to consider later. An alternative would be to
6815		 * abort the P-D-API with a notification and then deliver
6816		 * the data.... Or another method might be to keep track of
6817		 * how many times the situation occurs and if we see a
6818		 * possible attack underway just abort the association.
6819		 */
6820#ifdef SCTP_DEBUG
6821		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6822#endif
6823		/*
6824		 * Now do we need to find a new
6825		 * asoc->highest_tsn_inside_map?
6826		 */
6827		asoc->last_revoke_count = cnt;
6828		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6829		/* sa_ignore NO_NULL_CHK */
6830		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6831		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6832	}
6833	/*
6834	 * Another issue, in un-setting the TSN's in the mapping array we
6835	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6836	 * two things to occur. It may cause us to do extra work in checking
6837	 * for our mapping array movement. More importantly it may cause us
6838	 * to SACK every datagram. This may not be a bad thing though since
6839	 * we will recover once we get our cum-ack above and all this stuff
6840	 * we dumped recovered.
6841	 */
6842}
6843
6844void
6845sctp_drain()
6846{
6847	/*
6848	 * We must walk the PCB lists for ALL associations here. The system
6849	 * is LOW on MBUF's and needs help. This is where reneging will
6850	 * occur. We really hope this does NOT happen!
6851	 */
6852	VNET_ITERATOR_DECL(vnet_iter);
6853	VNET_LIST_RLOCK_NOSLEEP();
6854	VNET_FOREACH(vnet_iter) {
6855		CURVNET_SET(vnet_iter);
6856		struct sctp_inpcb *inp;
6857		struct sctp_tcb *stcb;
6858
6859		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6860		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6861#ifdef VIMAGE
6862			continue;
6863#else
6864			return;
6865#endif
6866		}
6867		SCTP_INP_INFO_RLOCK();
6868		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6869			/* For each endpoint */
6870			SCTP_INP_RLOCK(inp);
6871			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6872				/* For each association */
6873				SCTP_TCB_LOCK(stcb);
6874				sctp_drain_mbufs(stcb);
6875				SCTP_TCB_UNLOCK(stcb);
6876			}
6877			SCTP_INP_RUNLOCK(inp);
6878		}
6879		SCTP_INP_INFO_RUNLOCK();
6880		CURVNET_RESTORE();
6881	}
6882	VNET_LIST_RUNLOCK_NOSLEEP();
6883}
6884
6885/*
6886 * start a new iterator
6887 * iterates through all endpoints and associations based on the pcb_state
6888 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6889 * assocs and "ef" (optional) is executed when the iterator completes.
6890 * "inpf" (optional) is executed for each new endpoint as it is being
6891 * iterated through. inpe (optional) is called when the inp completes
6892 * its way through all the stcbs.
6893 */
6894int
6895sctp_initiate_iterator(inp_func inpf,
6896    asoc_func af,
6897    inp_func inpe,
6898    uint32_t pcb_state,
6899    uint32_t pcb_features,
6900    uint32_t asoc_state,
6901    void *argp,
6902    uint32_t argi,
6903    end_func ef,
6904    struct sctp_inpcb *s_inp,
6905    uint8_t chunk_output_off)
6906{
6907	struct sctp_iterator *it = NULL;
6908
6909	if (af == NULL) {
6910		return (-1);
6911	}
6912	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6913	    SCTP_M_ITER);
6914	if (it == NULL) {
6915		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6916		return (ENOMEM);
6917	}
6918	memset(it, 0, sizeof(*it));
6919	it->function_assoc = af;
6920	it->function_inp = inpf;
6921	if (inpf)
6922		it->done_current_ep = 0;
6923	else
6924		it->done_current_ep = 1;
6925	it->function_atend = ef;
6926	it->pointer = argp;
6927	it->val = argi;
6928	it->pcb_flags = pcb_state;
6929	it->pcb_features = pcb_features;
6930	it->asoc_state = asoc_state;
6931	it->function_inp_end = inpe;
6932	it->no_chunk_output = chunk_output_off;
6933	it->vn = curvnet;
6934	if (s_inp) {
6935		/* Assume lock is held here */
6936		it->inp = s_inp;
6937		SCTP_INP_INCR_REF(it->inp);
6938		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6939	} else {
6940		SCTP_INP_INFO_RLOCK();
6941		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6942		if (it->inp) {
6943			SCTP_INP_INCR_REF(it->inp);
6944		}
6945		SCTP_INP_INFO_RUNLOCK();
6946		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6947
6948	}
6949	SCTP_IPI_ITERATOR_WQ_LOCK();
6950
6951	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6952	if (sctp_it_ctl.iterator_running == 0) {
6953		sctp_wakeup_iterator();
6954	}
6955	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6956	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6957	return (0);
6958}
6959