sctp_pcb.c revision 243882
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 243882 2012-12-05 08:04:20Z glebius $");
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
98	SCTP_INP_INFO_RUNLOCK();
99}
100
101/*
102 * Addresses are added to VRF's (Virtual Router's). For BSD we
103 * have only the default VRF 0. We maintain a hash list of
104 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
105 * these has a list of addresses. When we add a new address
106 * to a VRF we lookup the ifn/ifn_index, if the ifn does
107 * not exist we create it and add it to the list of IFN's
108 * within the VRF. Once we have the sctp_ifn, we add the
109 * address to the list. So we look something like:
110 *
111 * hash-vrf-table
112 *   vrf-> ifn-> ifn -> ifn
113 *   vrf    |
114 *    ...   +--ifa-> ifa -> ifa
115 *   vrf
116 *
117 * We keep these separate lists since the SCTP subsystem will
118 * point to these from its source address selection nets structure.
119 * When an address is deleted it does not happen right away on
120 * the SCTP side, it gets scheduled. What we do when a
121 * delete happens is immediately remove the address from
122 * the master list and decrement the refcount. As our
123 * addip iterator works through and frees the src address
124 * selection pointing to the sctp_ifa, eventually the refcount
125 * will reach 0 and we will delete it. Note that it is assumed
126 * that any locking on system level ifn/ifa is done at the
127 * caller of these functions and these routines will only
128 * lock the SCTP structures as they add or delete things.
129 *
130 * Other notes on VRF concepts.
131 *  - An endpoint can be in multiple VRF's
132 *  - An association lives within a VRF and only one VRF.
133 *  - Any incoming packet we can deduce the VRF for by
134 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
135 *  - Any downward send call or connect call must supply the
136 *    VRF via ancillary data or via some sort of set default
137 *    VRF socket option call (again for BSD no brainer since
138 *    the VRF is always 0).
139 *  - An endpoint may add multiple VRF's to it.
140 *  - Listening sockets can accept associations in any
141 *    of the VRF's they are in but the assoc will end up
142 *    in only one VRF (gotten from the packet or connect/send).
143 *
144 */
145
146struct sctp_vrf *
147sctp_allocate_vrf(int vrf_id)
148{
149	struct sctp_vrf *vrf = NULL;
150	struct sctp_vrflist *bucket;
151
152	/* First allocate the VRF structure */
153	vrf = sctp_find_vrf(vrf_id);
154	if (vrf) {
155		/* Already allocated */
156		return (vrf);
157	}
158	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
159	    SCTP_M_VRF);
160	if (vrf == NULL) {
161		/* No memory */
162#ifdef INVARIANTS
163		panic("No memory for VRF:%d", vrf_id);
164#endif
165		return (NULL);
166	}
167	/* setup the VRF */
168	memset(vrf, 0, sizeof(struct sctp_vrf));
169	vrf->vrf_id = vrf_id;
170	LIST_INIT(&vrf->ifnlist);
171	vrf->total_ifa_count = 0;
172	vrf->refcount = 0;
173	/* now also setup table ids */
174	SCTP_INIT_VRF_TABLEID(vrf);
175	/* Init the HASH of addresses */
176	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
177	    &vrf->vrf_addr_hashmark);
178	if (vrf->vrf_addr_hash == NULL) {
179		/* No memory */
180#ifdef INVARIANTS
181		panic("No memory for VRF:%d", vrf_id);
182#endif
183		SCTP_FREE(vrf, SCTP_M_VRF);
184		return (NULL);
185	}
186	/* Add it to the hash table */
187	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
188	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
189	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
190	return (vrf);
191}
192
193
194struct sctp_ifn *
195sctp_find_ifn(void *ifn, uint32_t ifn_index)
196{
197	struct sctp_ifn *sctp_ifnp;
198	struct sctp_ifnlist *hash_ifn_head;
199
200	/*
201	 * We assume the lock is held for the addresses if that's wrong
202	 * problems could occur :-)
203	 */
204	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
205	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
206		if (sctp_ifnp->ifn_index == ifn_index) {
207			return (sctp_ifnp);
208		}
209		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
210			return (sctp_ifnp);
211		}
212	}
213	return (NULL);
214}
215
216
217
218struct sctp_vrf *
219sctp_find_vrf(uint32_t vrf_id)
220{
221	struct sctp_vrflist *bucket;
222	struct sctp_vrf *liste;
223
224	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
225	LIST_FOREACH(liste, bucket, next_vrf) {
226		if (vrf_id == liste->vrf_id) {
227			return (liste);
228		}
229	}
230	return (NULL);
231}
232
233void
234sctp_free_vrf(struct sctp_vrf *vrf)
235{
236	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
237		if (vrf->vrf_addr_hash) {
238			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
239			vrf->vrf_addr_hash = NULL;
240		}
241		/* We zero'd the count */
242		LIST_REMOVE(vrf, next_vrf);
243		SCTP_FREE(vrf, SCTP_M_VRF);
244		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
245	}
246}
247
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
261void
262sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
263{
264	struct sctp_ifn *sctp_ifnp;
265
266	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
267	if (sctp_ifnp != NULL) {
268		sctp_ifnp->ifn_mtu = mtu;
269	}
270}
271
272
273void
274sctp_free_ifa(struct sctp_ifa *sctp_ifap)
275{
276	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
277		/* We zero'd the count */
278		if (sctp_ifap->ifn_p) {
279			sctp_free_ifn(sctp_ifap->ifn_p);
280		}
281		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
282		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
283	}
284}
285
286static void
287sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
288{
289	struct sctp_ifn *found;
290
291	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
292	if (found == NULL) {
293		/* Not in the list.. sorry */
294		return;
295	}
296	if (hold_addr_lock == 0)
297		SCTP_IPI_ADDR_WLOCK();
298	LIST_REMOVE(sctp_ifnp, next_bucket);
299	LIST_REMOVE(sctp_ifnp, next_ifn);
300	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
301	    sctp_ifnp->registered_af);
302	if (hold_addr_lock == 0)
303		SCTP_IPI_ADDR_WUNLOCK();
304	/* Take away the reference, and possibly free it */
305	sctp_free_ifn(sctp_ifnp);
306}
307
308void
309sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
310    const char *if_name, uint32_t ifn_index)
311{
312	struct sctp_vrf *vrf;
313	struct sctp_ifa *sctp_ifap = NULL;
314
315	SCTP_IPI_ADDR_RLOCK();
316	vrf = sctp_find_vrf(vrf_id);
317	if (vrf == NULL) {
318		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
319		goto out;
320
321	}
322	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
323	if (sctp_ifap == NULL) {
324		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
325		goto out;
326	}
327	if (sctp_ifap->ifn_p == NULL) {
328		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
329		goto out;
330	}
331	if (if_name) {
332		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
333			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
334			    sctp_ifap->ifn_p->ifn_name, if_name);
335			goto out;
336		}
337	} else {
338		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
339			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
340			    sctp_ifap->ifn_p->ifn_index, ifn_index);
341			goto out;
342		}
343	}
344
345	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
346	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
347out:
348	SCTP_IPI_ADDR_RUNLOCK();
349}
350
351void
352sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
353    const char *if_name, uint32_t ifn_index)
354{
355	struct sctp_vrf *vrf;
356	struct sctp_ifa *sctp_ifap = NULL;
357
358	SCTP_IPI_ADDR_RLOCK();
359	vrf = sctp_find_vrf(vrf_id);
360	if (vrf == NULL) {
361		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
362		goto out;
363
364	}
365	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
366	if (sctp_ifap == NULL) {
367		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
368		goto out;
369	}
370	if (sctp_ifap->ifn_p == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
372		goto out;
373	}
374	if (if_name) {
375		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
376			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
377			    sctp_ifap->ifn_p->ifn_name, if_name);
378			goto out;
379		}
380	} else {
381		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
382			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
383			    sctp_ifap->ifn_p->ifn_index, ifn_index);
384			goto out;
385		}
386	}
387
388	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
389	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
390out:
391	SCTP_IPI_ADDR_RUNLOCK();
392}
393
394/*-
395 * Add an ifa to an ifn.
396 * Register the interface as necessary.
397 * NOTE: ADDR write lock MUST be held.
398 */
399static void
400sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
401{
402	int ifa_af;
403
404	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
405	sctp_ifap->ifn_p = sctp_ifnp;
406	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
407	/* update address counts */
408	sctp_ifnp->ifa_count++;
409	ifa_af = sctp_ifap->address.sa.sa_family;
410	switch (ifa_af) {
411#ifdef INET
412	case AF_INET:
413		sctp_ifnp->num_v4++;
414		break;
415#endif
416#ifdef INET6
417	case AF_INET6:
418		sctp_ifnp->num_v6++;
419		break;
420#endif
421	default:
422		break;
423	}
424	if (sctp_ifnp->ifa_count == 1) {
425		/* register the new interface */
426		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
427		sctp_ifnp->registered_af = ifa_af;
428	}
429}
430
431/*-
432 * Remove an ifa from its ifn.
433 * If no more addresses exist, remove the ifn too. Otherwise, re-register
434 * the interface based on the remaining address families left.
435 * NOTE: ADDR write lock MUST be held.
436 */
437static void
438sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
439{
440	LIST_REMOVE(sctp_ifap, next_ifa);
441	if (sctp_ifap->ifn_p) {
442		/* update address counts */
443		sctp_ifap->ifn_p->ifa_count--;
444		switch (sctp_ifap->address.sa.sa_family) {
445#ifdef INET
446		case AF_INET:
447			sctp_ifap->ifn_p->num_v4--;
448			break;
449#endif
450#ifdef INET6
451		case AF_INET6:
452			sctp_ifap->ifn_p->num_v6--;
453			break;
454#endif
455		default:
456			break;
457		}
458
459		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
460			/* remove the ifn, possibly freeing it */
461			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
462		} else {
463			/* re-register address family type, if needed */
464			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
465			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
466				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
467				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
468				sctp_ifap->ifn_p->registered_af = AF_INET;
469			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
470			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
471				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
472				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				sctp_ifap->ifn_p->registered_af = AF_INET6;
474			}
475			/* free the ifn refcount */
476			sctp_free_ifn(sctp_ifap->ifn_p);
477		}
478		sctp_ifap->ifn_p = NULL;
479	}
480}
481
482struct sctp_ifa *
483sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
484    uint32_t ifn_type, const char *if_name, void *ifa,
485    struct sockaddr *addr, uint32_t ifa_flags,
486    int dynamic_add)
487{
488	struct sctp_vrf *vrf;
489	struct sctp_ifn *sctp_ifnp = NULL;
490	struct sctp_ifa *sctp_ifap = NULL;
491	struct sctp_ifalist *hash_addr_head;
492	struct sctp_ifnlist *hash_ifn_head;
493	uint32_t hash_of_addr;
494	int new_ifn_af = 0;
495
496#ifdef SCTP_DEBUG
497	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
498	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
499#endif
500	SCTP_IPI_ADDR_WLOCK();
501	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
502	if (sctp_ifnp) {
503		vrf = sctp_ifnp->vrf;
504	} else {
505		vrf = sctp_find_vrf(vrf_id);
506		if (vrf == NULL) {
507			vrf = sctp_allocate_vrf(vrf_id);
508			if (vrf == NULL) {
509				SCTP_IPI_ADDR_WUNLOCK();
510				return (NULL);
511			}
512		}
513	}
514	if (sctp_ifnp == NULL) {
515		/*
516		 * build one and add it, can't hold lock until after malloc
517		 * done though.
518		 */
519		SCTP_IPI_ADDR_WUNLOCK();
520		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
521		    sizeof(struct sctp_ifn), SCTP_M_IFN);
522		if (sctp_ifnp == NULL) {
523#ifdef INVARIANTS
524			panic("No memory for IFN");
525#endif
526			return (NULL);
527		}
528		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
529		sctp_ifnp->ifn_index = ifn_index;
530		sctp_ifnp->ifn_p = ifn;
531		sctp_ifnp->ifn_type = ifn_type;
532		sctp_ifnp->refcount = 0;
533		sctp_ifnp->vrf = vrf;
534		atomic_add_int(&vrf->refcount, 1);
535		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
536		if (if_name != NULL) {
537			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
538		} else {
539			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
540		}
541		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
542		LIST_INIT(&sctp_ifnp->ifalist);
543		SCTP_IPI_ADDR_WLOCK();
544		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
545		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
546		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
547		new_ifn_af = 1;
548	}
549	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
550	if (sctp_ifap) {
551		/* Hmm, it already exists? */
552		if ((sctp_ifap->ifn_p) &&
553		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
554			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
555			    sctp_ifap->ifn_p->ifn_name, ifn_index,
556			    (void *)sctp_ifap);
557			if (new_ifn_af) {
558				/* Remove the created one that we don't want */
559				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
560			}
561			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
562				/* easy to solve, just switch back to active */
563				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
564				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
565				sctp_ifap->ifn_p = sctp_ifnp;
566				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
567			}
568	exit_stage_left:
569			SCTP_IPI_ADDR_WUNLOCK();
570			return (sctp_ifap);
571		} else {
572			if (sctp_ifap->ifn_p) {
573				/*
574				 * The last IFN gets the address, remove the
575				 * old one
576				 */
577				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
578				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
579				    sctp_ifap->ifn_p->ifn_index, if_name,
580				    ifn_index);
581				/* remove the address from the old ifn */
582				sctp_remove_ifa_from_ifn(sctp_ifap);
583				/* move the address over to the new ifn */
584				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
585				goto exit_stage_left;
586			} else {
587				/* repair ifnp which was NULL ? */
588				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
589				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
590				    (void *)sctp_ifnp, (void *)sctp_ifap);
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592			}
593			goto exit_stage_left;
594		}
595	}
596	SCTP_IPI_ADDR_WUNLOCK();
597	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
598	if (sctp_ifap == NULL) {
599#ifdef INVARIANTS
600		panic("No memory for IFA");
601#endif
602		return (NULL);
603	}
604	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
605	sctp_ifap->ifn_p = sctp_ifnp;
606	atomic_add_int(&sctp_ifnp->refcount, 1);
607	sctp_ifap->vrf_id = vrf_id;
608	sctp_ifap->ifa = ifa;
609	memcpy(&sctp_ifap->address, addr, addr->sa_len);
610	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
611	sctp_ifap->flags = ifa_flags;
612	/* Set scope */
613	switch (sctp_ifap->address.sa.sa_family) {
614#ifdef INET
615	case AF_INET:
616		{
617			struct sockaddr_in *sin;
618
619			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
620			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
621			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
622				sctp_ifap->src_is_loop = 1;
623			}
624			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
625				sctp_ifap->src_is_priv = 1;
626			}
627			sctp_ifnp->num_v4++;
628			if (new_ifn_af)
629				new_ifn_af = AF_INET;
630			break;
631		}
632#endif
633#ifdef INET6
634	case AF_INET6:
635		{
636			/* ok to use deprecated addresses? */
637			struct sockaddr_in6 *sin6;
638
639			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
640			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
641			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
642				sctp_ifap->src_is_loop = 1;
643			}
644			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
645				sctp_ifap->src_is_priv = 1;
646			}
647			sctp_ifnp->num_v6++;
648			if (new_ifn_af)
649				new_ifn_af = AF_INET6;
650			break;
651		}
652#endif
653	default:
654		new_ifn_af = 0;
655		break;
656	}
657	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
658
659	if ((sctp_ifap->src_is_priv == 0) &&
660	    (sctp_ifap->src_is_loop == 0)) {
661		sctp_ifap->src_is_glob = 1;
662	}
663	SCTP_IPI_ADDR_WLOCK();
664	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
665	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
666	sctp_ifap->refcount = 1;
667	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
668	sctp_ifnp->ifa_count++;
669	vrf->total_ifa_count++;
670	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
671	if (new_ifn_af) {
672		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
673		sctp_ifnp->registered_af = new_ifn_af;
674	}
675	SCTP_IPI_ADDR_WUNLOCK();
676	if (dynamic_add) {
677		/*
678		 * Bump up the refcount so that when the timer completes it
679		 * will drop back down.
680		 */
681		struct sctp_laddr *wi;
682
683		atomic_add_int(&sctp_ifap->refcount, 1);
684		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
685		if (wi == NULL) {
686			/*
687			 * Gak, what can we do? We have lost an address
688			 * change can you say HOSED?
689			 */
690			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
691			/* Opps, must decrement the count */
692			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
693			    if_name);
694			return (NULL);
695		}
696		SCTP_INCR_LADDR_COUNT();
697		bzero(wi, sizeof(*wi));
698		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
699		wi->ifa = sctp_ifap;
700		wi->action = SCTP_ADD_IP_ADDRESS;
701
702		SCTP_WQ_ADDR_LOCK();
703		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
704		SCTP_WQ_ADDR_UNLOCK();
705
706		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
707		    (struct sctp_inpcb *)NULL,
708		    (struct sctp_tcb *)NULL,
709		    (struct sctp_nets *)NULL);
710	} else {
711		/* it's ready for use */
712		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
713	}
714	return (sctp_ifap);
715}
716
717void
718sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
719    uint32_t ifn_index, const char *if_name)
720{
721	struct sctp_vrf *vrf;
722	struct sctp_ifa *sctp_ifap = NULL;
723
724	SCTP_IPI_ADDR_WLOCK();
725	vrf = sctp_find_vrf(vrf_id);
726	if (vrf == NULL) {
727		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
728		goto out_now;
729	}
730#ifdef SCTP_DEBUG
731	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
732	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
733#endif
734	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
735	if (sctp_ifap) {
736		/* Validate the delete */
737		if (sctp_ifap->ifn_p) {
738			int valid = 0;
739
740			/*-
741			 * The name has priority over the ifn_index
742			 * if its given. We do this especially for
743			 * panda who might recycle indexes fast.
744			 */
745			if (if_name) {
746				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
747					/* They match its a correct delete */
748					valid = 1;
749				}
750			}
751			if (!valid) {
752				/* last ditch check ifn_index */
753				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
754					valid = 1;
755				}
756			}
757			if (!valid) {
758				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
759				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
760				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
761				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
762				SCTP_IPI_ADDR_WUNLOCK();
763				return;
764			}
765		}
766		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
767		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
768		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
769		vrf->total_ifa_count--;
770		LIST_REMOVE(sctp_ifap, next_bucket);
771		sctp_remove_ifa_from_ifn(sctp_ifap);
772	}
773#ifdef SCTP_DEBUG
774	else {
775		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
776		    ifn_index);
777		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
778	}
779#endif
780
781out_now:
782	SCTP_IPI_ADDR_WUNLOCK();
783	if (sctp_ifap) {
784		struct sctp_laddr *wi;
785
786		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
787		if (wi == NULL) {
788			/*
789			 * Gak, what can we do? We have lost an address
790			 * change can you say HOSED?
791			 */
792			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
793
794			/* Oops, must decrement the count */
795			sctp_free_ifa(sctp_ifap);
796			return;
797		}
798		SCTP_INCR_LADDR_COUNT();
799		bzero(wi, sizeof(*wi));
800		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
801		wi->ifa = sctp_ifap;
802		wi->action = SCTP_DEL_IP_ADDRESS;
803		SCTP_WQ_ADDR_LOCK();
804		/*
805		 * Should this really be a tailq? As it is we will process
806		 * the newest first :-0
807		 */
808		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
809		SCTP_WQ_ADDR_UNLOCK();
810
811		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
812		    (struct sctp_inpcb *)NULL,
813		    (struct sctp_tcb *)NULL,
814		    (struct sctp_nets *)NULL);
815	}
816	return;
817}
818
819
820static struct sctp_tcb *
821sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
822    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
823{
824	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
825	/*
826	 * If we support the TCP model, then we must now dig through to see
827	 * if we can find our endpoint in the list of tcp ep's.
828	 */
829	uint16_t lport, rport;
830	struct sctppcbhead *ephead;
831	struct sctp_inpcb *inp;
832	struct sctp_laddr *laddr;
833	struct sctp_tcb *stcb;
834	struct sctp_nets *net;
835
836	if ((to == NULL) || (from == NULL)) {
837		return (NULL);
838	}
839	switch (to->sa_family) {
840#ifdef INET
841	case AF_INET:
842		if (from->sa_family == AF_INET) {
843			lport = ((struct sockaddr_in *)to)->sin_port;
844			rport = ((struct sockaddr_in *)from)->sin_port;
845		} else {
846			return (NULL);
847		}
848		break;
849#endif
850#ifdef INET6
851	case AF_INET6:
852		if (from->sa_family == AF_INET6) {
853			lport = ((struct sockaddr_in6 *)to)->sin6_port;
854			rport = ((struct sockaddr_in6 *)from)->sin6_port;
855		} else {
856			return (NULL);
857		}
858		break;
859#endif
860	default:
861		return (NULL);
862	}
863	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
864	/*
865	 * Ok now for each of the guys in this bucket we must look and see:
866	 * - Does the remote port match. - Does there single association's
867	 * addresses match this address (to). If so we update p_ep to point
868	 * to this ep and return the tcb from it.
869	 */
870	LIST_FOREACH(inp, ephead, sctp_hash) {
871		SCTP_INP_RLOCK(inp);
872		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
873			SCTP_INP_RUNLOCK(inp);
874			continue;
875		}
876		if (lport != inp->sctp_lport) {
877			SCTP_INP_RUNLOCK(inp);
878			continue;
879		}
880		if (inp->def_vrf_id != vrf_id) {
881			SCTP_INP_RUNLOCK(inp);
882			continue;
883		}
884		/* check to see if the ep has one of the addresses */
885		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
886			/* We are NOT bound all, so look further */
887			int match = 0;
888
889			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
890
891				if (laddr->ifa == NULL) {
892					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
893					continue;
894				}
895				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
896					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
897					continue;
898				}
899				if (laddr->ifa->address.sa.sa_family ==
900				    to->sa_family) {
901					/* see if it matches */
902
903#ifdef INET
904					if (from->sa_family == AF_INET) {
905						struct sockaddr_in *intf_addr,
906						           *sin;
907
908						intf_addr = &laddr->ifa->address.sin;
909						sin = (struct sockaddr_in *)to;
910						if (sin->sin_addr.s_addr ==
911						    intf_addr->sin_addr.s_addr) {
912							match = 1;
913							break;
914						}
915					}
916#endif
917#ifdef INET6
918					if (from->sa_family == AF_INET6) {
919						struct sockaddr_in6 *intf_addr6;
920						struct sockaddr_in6 *sin6;
921
922						sin6 = (struct sockaddr_in6 *)
923						    to;
924						intf_addr6 = &laddr->ifa->address.sin6;
925
926						if (SCTP6_ARE_ADDR_EQUAL(sin6,
927						    intf_addr6)) {
928							match = 1;
929							break;
930						}
931					}
932#endif
933				}
934			}
935			if (match == 0) {
936				/* This endpoint does not have this address */
937				SCTP_INP_RUNLOCK(inp);
938				continue;
939			}
940		}
941		/*
942		 * Ok if we hit here the ep has the address, does it hold
943		 * the tcb?
944		 */
945
946		stcb = LIST_FIRST(&inp->sctp_asoc_list);
947		if (stcb == NULL) {
948			SCTP_INP_RUNLOCK(inp);
949			continue;
950		}
951		SCTP_TCB_LOCK(stcb);
952		if (stcb->rport != rport) {
953			/* remote port does not match. */
954			SCTP_TCB_UNLOCK(stcb);
955			SCTP_INP_RUNLOCK(inp);
956			continue;
957		}
958		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
959			SCTP_TCB_UNLOCK(stcb);
960			SCTP_INP_RUNLOCK(inp);
961			continue;
962		}
963		/* Does this TCB have a matching address? */
964		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
965
966			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
967				/* not the same family, can't be a match */
968				continue;
969			}
970			switch (from->sa_family) {
971#ifdef INET
972			case AF_INET:
973				{
974					struct sockaddr_in *sin, *rsin;
975
976					sin = (struct sockaddr_in *)&net->ro._l_addr;
977					rsin = (struct sockaddr_in *)from;
978					if (sin->sin_addr.s_addr ==
979					    rsin->sin_addr.s_addr) {
980						/* found it */
981						if (netp != NULL) {
982							*netp = net;
983						}
984						/*
985						 * Update the endpoint
986						 * pointer
987						 */
988						*inp_p = inp;
989						SCTP_INP_RUNLOCK(inp);
990						return (stcb);
991					}
992					break;
993				}
994#endif
995#ifdef INET6
996			case AF_INET6:
997				{
998					struct sockaddr_in6 *sin6, *rsin6;
999
1000					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1001					rsin6 = (struct sockaddr_in6 *)from;
1002					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1003					    rsin6)) {
1004						/* found it */
1005						if (netp != NULL) {
1006							*netp = net;
1007						}
1008						/*
1009						 * Update the endpoint
1010						 * pointer
1011						 */
1012						*inp_p = inp;
1013						SCTP_INP_RUNLOCK(inp);
1014						return (stcb);
1015					}
1016					break;
1017				}
1018#endif
1019			default:
1020				/* TSNH */
1021				break;
1022			}
1023		}
1024		SCTP_TCB_UNLOCK(stcb);
1025		SCTP_INP_RUNLOCK(inp);
1026	}
1027	return (NULL);
1028}
1029
1030static int
1031sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1032{
1033	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
1034	int ipv4_addr_legal, ipv6_addr_legal;
1035	struct sctp_vrf *vrf;
1036	struct sctp_ifn *sctp_ifn;
1037	struct sctp_ifa *sctp_ifa;
1038
1039	loopback_scope = stcb->asoc.loopback_scope;
1040	ipv4_local_scope = stcb->asoc.ipv4_local_scope;
1041	local_scope = stcb->asoc.local_scope;
1042	site_scope = stcb->asoc.site_scope;
1043	ipv4_addr_legal = ipv6_addr_legal = 0;
1044	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1045		ipv6_addr_legal = 1;
1046		if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) {
1047			ipv4_addr_legal = 1;
1048		}
1049	} else {
1050		ipv4_addr_legal = 1;
1051	}
1052
1053	SCTP_IPI_ADDR_RLOCK();
1054	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
1055	if (vrf == NULL) {
1056		/* no vrf, no addresses */
1057		SCTP_IPI_ADDR_RUNLOCK();
1058		return (0);
1059	}
1060	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1061		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
1062			if ((loopback_scope == 0) &&
1063			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
1064				continue;
1065			}
1066			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
1067				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
1068				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
1069					/*
1070					 * We allow pending addresses, where
1071					 * we have sent an asconf-add to be
1072					 * considered valid.
1073					 */
1074					continue;
1075				}
1076				switch (sctp_ifa->address.sa.sa_family) {
1077#ifdef INET
1078				case AF_INET:
1079					if (ipv4_addr_legal) {
1080						struct sockaddr_in *sin,
1081						           *rsin;
1082
1083						sin = &sctp_ifa->address.sin;
1084						rsin = (struct sockaddr_in *)to;
1085						if ((ipv4_local_scope == 0) &&
1086						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
1087							continue;
1088						}
1089						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1090							SCTP_IPI_ADDR_RUNLOCK();
1091							return (1);
1092						}
1093					}
1094					break;
1095#endif
1096#ifdef INET6
1097				case AF_INET6:
1098					if (ipv6_addr_legal) {
1099						struct sockaddr_in6 *sin6,
1100						            *rsin6;
1101
1102						sin6 = &sctp_ifa->address.sin6;
1103						rsin6 = (struct sockaddr_in6 *)to;
1104						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
1105							if (local_scope == 0)
1106								continue;
1107							if (sin6->sin6_scope_id == 0) {
1108								if (sa6_recoverscope(sin6) != 0)
1109									continue;
1110							}
1111						}
1112						if ((site_scope == 0) &&
1113						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
1114							continue;
1115						}
1116						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1117							SCTP_IPI_ADDR_RUNLOCK();
1118							return (1);
1119						}
1120					}
1121					break;
1122#endif
1123				default:
1124					/* TSNH */
1125					break;
1126				}
1127			}
1128		}
1129	} else {
1130		struct sctp_laddr *laddr;
1131
1132		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
1133			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
1134			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
1135				/*
1136				 * We allow pending addresses, where we have
1137				 * sent an asconf-add to be considered
1138				 * valid.
1139				 */
1140				continue;
1141			}
1142			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
1143				continue;
1144			}
1145			switch (to->sa_family) {
1146#ifdef INET
1147			case AF_INET:
1148				{
1149					struct sockaddr_in *sin, *rsin;
1150
1151					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
1152					rsin = (struct sockaddr_in *)to;
1153					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1154						SCTP_IPI_ADDR_RUNLOCK();
1155						return (1);
1156					}
1157					break;
1158				}
1159#endif
1160#ifdef INET6
1161			case AF_INET6:
1162				{
1163					struct sockaddr_in6 *sin6, *rsin6;
1164
1165					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1166					rsin6 = (struct sockaddr_in6 *)to;
1167					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1168						SCTP_IPI_ADDR_RUNLOCK();
1169						return (1);
1170					}
1171					break;
1172				}
1173
1174#endif
1175			default:
1176				/* TSNH */
1177				break;
1178			}
1179
1180		}
1181	}
1182	SCTP_IPI_ADDR_RUNLOCK();
1183	return (0);
1184}
1185
1186/*
1187 * rules for use
1188 *
1189 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1190 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1191 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1192 * NULL.
1193 */
1194
1195struct sctp_tcb *
1196sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1197    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1198{
1199	struct sctpasochead *head;
1200	struct sctp_inpcb *inp;
1201	struct sctp_tcb *stcb = NULL;
1202	struct sctp_nets *net;
1203	uint16_t rport;
1204
1205	inp = *inp_p;
1206	switch (remote->sa_family) {
1207#ifdef INET
1208	case AF_INET:
1209		rport = (((struct sockaddr_in *)remote)->sin_port);
1210		break;
1211#endif
1212#ifdef INET6
1213	case AF_INET6:
1214		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1215		break;
1216#endif
1217	default:
1218		return (NULL);
1219	}
1220	if (locked_tcb) {
1221		/*
1222		 * UN-lock so we can do proper locking here this occurs when
1223		 * called from load_addresses_from_init.
1224		 */
1225		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1226		SCTP_TCB_UNLOCK(locked_tcb);
1227	}
1228	SCTP_INP_INFO_RLOCK();
1229	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1230	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1231		/*-
1232		 * Now either this guy is our listener or it's the
1233		 * connector. If it is the one that issued the connect, then
1234		 * it's only chance is to be the first TCB in the list. If
1235		 * it is the acceptor, then do the special_lookup to hash
1236		 * and find the real inp.
1237		 */
1238		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1239			/* to is peer addr, from is my addr */
1240			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1241			    netp, inp->def_vrf_id);
1242			if ((stcb != NULL) && (locked_tcb == NULL)) {
1243				/* we have a locked tcb, lower refcount */
1244				SCTP_INP_DECR_REF(inp);
1245			}
1246			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1247				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1248				SCTP_TCB_LOCK(locked_tcb);
1249				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1250				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1251			}
1252			SCTP_INP_INFO_RUNLOCK();
1253			return (stcb);
1254		} else {
1255			SCTP_INP_WLOCK(inp);
1256			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1257				goto null_return;
1258			}
1259			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1260			if (stcb == NULL) {
1261				goto null_return;
1262			}
1263			SCTP_TCB_LOCK(stcb);
1264
1265			if (stcb->rport != rport) {
1266				/* remote port does not match. */
1267				SCTP_TCB_UNLOCK(stcb);
1268				goto null_return;
1269			}
1270			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1271				SCTP_TCB_UNLOCK(stcb);
1272				goto null_return;
1273			}
1274			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1275				SCTP_TCB_UNLOCK(stcb);
1276				goto null_return;
1277			}
1278			/* now look at the list of remote addresses */
1279			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1280#ifdef INVARIANTS
1281				if (net == (TAILQ_NEXT(net, sctp_next))) {
1282					panic("Corrupt net list");
1283				}
1284#endif
1285				if (net->ro._l_addr.sa.sa_family !=
1286				    remote->sa_family) {
1287					/* not the same family */
1288					continue;
1289				}
1290				switch (remote->sa_family) {
1291#ifdef INET
1292				case AF_INET:
1293					{
1294						struct sockaddr_in *sin,
1295						           *rsin;
1296
1297						sin = (struct sockaddr_in *)
1298						    &net->ro._l_addr;
1299						rsin = (struct sockaddr_in *)remote;
1300						if (sin->sin_addr.s_addr ==
1301						    rsin->sin_addr.s_addr) {
1302							/* found it */
1303							if (netp != NULL) {
1304								*netp = net;
1305							}
1306							if (locked_tcb == NULL) {
1307								SCTP_INP_DECR_REF(inp);
1308							} else if (locked_tcb != stcb) {
1309								SCTP_TCB_LOCK(locked_tcb);
1310							}
1311							if (locked_tcb) {
1312								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1313							}
1314							SCTP_INP_WUNLOCK(inp);
1315							SCTP_INP_INFO_RUNLOCK();
1316							return (stcb);
1317						}
1318						break;
1319					}
1320#endif
1321#ifdef INET6
1322				case AF_INET6:
1323					{
1324						struct sockaddr_in6 *sin6,
1325						            *rsin6;
1326
1327						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1328						rsin6 = (struct sockaddr_in6 *)remote;
1329						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1330						    rsin6)) {
1331							/* found it */
1332							if (netp != NULL) {
1333								*netp = net;
1334							}
1335							if (locked_tcb == NULL) {
1336								SCTP_INP_DECR_REF(inp);
1337							} else if (locked_tcb != stcb) {
1338								SCTP_TCB_LOCK(locked_tcb);
1339							}
1340							if (locked_tcb) {
1341								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1342							}
1343							SCTP_INP_WUNLOCK(inp);
1344							SCTP_INP_INFO_RUNLOCK();
1345							return (stcb);
1346						}
1347						break;
1348					}
1349#endif
1350				default:
1351					/* TSNH */
1352					break;
1353				}
1354			}
1355			SCTP_TCB_UNLOCK(stcb);
1356		}
1357	} else {
1358		SCTP_INP_WLOCK(inp);
1359		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1360			goto null_return;
1361		}
1362		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1363		    inp->sctp_hashmark)];
1364		if (head == NULL) {
1365			goto null_return;
1366		}
1367		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1368			if (stcb->rport != rport) {
1369				/* remote port does not match */
1370				continue;
1371			}
1372			SCTP_TCB_LOCK(stcb);
1373			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1374				SCTP_TCB_UNLOCK(stcb);
1375				continue;
1376			}
1377			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1378				SCTP_TCB_UNLOCK(stcb);
1379				continue;
1380			}
1381			/* now look at the list of remote addresses */
1382			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1383#ifdef INVARIANTS
1384				if (net == (TAILQ_NEXT(net, sctp_next))) {
1385					panic("Corrupt net list");
1386				}
1387#endif
1388				if (net->ro._l_addr.sa.sa_family !=
1389				    remote->sa_family) {
1390					/* not the same family */
1391					continue;
1392				}
1393				switch (remote->sa_family) {
1394#ifdef INET
1395				case AF_INET:
1396					{
1397						struct sockaddr_in *sin,
1398						           *rsin;
1399
1400						sin = (struct sockaddr_in *)
1401						    &net->ro._l_addr;
1402						rsin = (struct sockaddr_in *)remote;
1403						if (sin->sin_addr.s_addr ==
1404						    rsin->sin_addr.s_addr) {
1405							/* found it */
1406							if (netp != NULL) {
1407								*netp = net;
1408							}
1409							if (locked_tcb == NULL) {
1410								SCTP_INP_DECR_REF(inp);
1411							} else if (locked_tcb != stcb) {
1412								SCTP_TCB_LOCK(locked_tcb);
1413							}
1414							if (locked_tcb) {
1415								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1416							}
1417							SCTP_INP_WUNLOCK(inp);
1418							SCTP_INP_INFO_RUNLOCK();
1419							return (stcb);
1420						}
1421						break;
1422					}
1423#endif
1424#ifdef INET6
1425				case AF_INET6:
1426					{
1427						struct sockaddr_in6 *sin6,
1428						            *rsin6;
1429
1430						sin6 = (struct sockaddr_in6 *)
1431						    &net->ro._l_addr;
1432						rsin6 = (struct sockaddr_in6 *)remote;
1433						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1434						    rsin6)) {
1435							/* found it */
1436							if (netp != NULL) {
1437								*netp = net;
1438							}
1439							if (locked_tcb == NULL) {
1440								SCTP_INP_DECR_REF(inp);
1441							} else if (locked_tcb != stcb) {
1442								SCTP_TCB_LOCK(locked_tcb);
1443							}
1444							if (locked_tcb) {
1445								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1446							}
1447							SCTP_INP_WUNLOCK(inp);
1448							SCTP_INP_INFO_RUNLOCK();
1449							return (stcb);
1450						}
1451						break;
1452					}
1453#endif
1454				default:
1455					/* TSNH */
1456					break;
1457				}
1458			}
1459			SCTP_TCB_UNLOCK(stcb);
1460		}
1461	}
1462null_return:
1463	/* clean up for returning null */
1464	if (locked_tcb) {
1465		SCTP_TCB_LOCK(locked_tcb);
1466		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1467	}
1468	SCTP_INP_WUNLOCK(inp);
1469	SCTP_INP_INFO_RUNLOCK();
1470	/* not found */
1471	return (NULL);
1472}
1473
1474/*
1475 * Find an association for a specific endpoint using the association id given
1476 * out in the COMM_UP notification
1477 */
1478
1479struct sctp_tcb *
1480sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1481{
1482	/*
1483	 * Use my the assoc_id to find a endpoint
1484	 */
1485	struct sctpasochead *head;
1486	struct sctp_tcb *stcb;
1487	uint32_t id;
1488
1489	if (inp == NULL) {
1490		SCTP_PRINTF("TSNH ep_associd\n");
1491		return (NULL);
1492	}
1493	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1494		SCTP_PRINTF("TSNH ep_associd0\n");
1495		return (NULL);
1496	}
1497	id = (uint32_t) asoc_id;
1498	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1499	if (head == NULL) {
1500		/* invalid id TSNH */
1501		SCTP_PRINTF("TSNH ep_associd1\n");
1502		return (NULL);
1503	}
1504	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1505		if (stcb->asoc.assoc_id == id) {
1506			if (inp != stcb->sctp_ep) {
1507				/*
1508				 * some other guy has the same id active (id
1509				 * collision ??).
1510				 */
1511				SCTP_PRINTF("TSNH ep_associd2\n");
1512				continue;
1513			}
1514			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1515				continue;
1516			}
1517			if (want_lock) {
1518				SCTP_TCB_LOCK(stcb);
1519			}
1520			return (stcb);
1521		}
1522	}
1523	return (NULL);
1524}
1525
1526
1527struct sctp_tcb *
1528sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1529{
1530	struct sctp_tcb *stcb;
1531
1532	SCTP_INP_RLOCK(inp);
1533	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1534	SCTP_INP_RUNLOCK(inp);
1535	return (stcb);
1536}
1537
1538
1539static struct sctp_inpcb *
1540sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1541    uint16_t lport, uint32_t vrf_id)
1542{
1543	struct sctp_inpcb *inp;
1544	struct sctp_laddr *laddr;
1545
1546#ifdef INET
1547	struct sockaddr_in *sin;
1548
1549#endif
1550#ifdef INET6
1551	struct sockaddr_in6 *sin6;
1552	struct sockaddr_in6 *intf_addr6;
1553
1554#endif
1555
1556	int fnd;
1557
1558	/*
1559	 * Endpoint probe expects that the INP_INFO is locked.
1560	 */
1561#ifdef INET
1562	sin = NULL;
1563#endif
1564#ifdef INET6
1565	sin6 = NULL;
1566#endif
1567	switch (nam->sa_family) {
1568#ifdef INET
1569	case AF_INET:
1570		sin = (struct sockaddr_in *)nam;
1571		break;
1572#endif
1573#ifdef INET6
1574	case AF_INET6:
1575		sin6 = (struct sockaddr_in6 *)nam;
1576		break;
1577#endif
1578	default:
1579		/* unsupported family */
1580		return (NULL);
1581	}
1582
1583	if (head == NULL)
1584		return (NULL);
1585
1586	LIST_FOREACH(inp, head, sctp_hash) {
1587		SCTP_INP_RLOCK(inp);
1588		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1589			SCTP_INP_RUNLOCK(inp);
1590			continue;
1591		}
1592		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1593		    (inp->sctp_lport == lport)) {
1594			/* got it */
1595#ifdef INET
1596			if ((nam->sa_family == AF_INET) &&
1597			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1598			    SCTP_IPV6_V6ONLY(inp)) {
1599				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1600				SCTP_INP_RUNLOCK(inp);
1601				continue;
1602			}
1603#endif
1604#ifdef INET6
1605			/* A V6 address and the endpoint is NOT bound V6 */
1606			if (nam->sa_family == AF_INET6 &&
1607			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1608				SCTP_INP_RUNLOCK(inp);
1609				continue;
1610			}
1611#endif
1612			/* does a VRF id match? */
1613			fnd = 0;
1614			if (inp->def_vrf_id == vrf_id)
1615				fnd = 1;
1616
1617			SCTP_INP_RUNLOCK(inp);
1618			if (!fnd)
1619				continue;
1620			return (inp);
1621		}
1622		SCTP_INP_RUNLOCK(inp);
1623	}
1624	switch (nam->sa_family) {
1625#ifdef INET
1626	case AF_INET:
1627		if (sin->sin_addr.s_addr == INADDR_ANY) {
1628			/* Can't hunt for one that has no address specified */
1629			return (NULL);
1630		}
1631		break;
1632#endif
1633#ifdef INET6
1634	case AF_INET6:
1635		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1636			/* Can't hunt for one that has no address specified */
1637			return (NULL);
1638		}
1639		break;
1640#endif
1641	default:
1642		break;
1643	}
1644	/*
1645	 * ok, not bound to all so see if we can find a EP bound to this
1646	 * address.
1647	 */
1648	LIST_FOREACH(inp, head, sctp_hash) {
1649		SCTP_INP_RLOCK(inp);
1650		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1651			SCTP_INP_RUNLOCK(inp);
1652			continue;
1653		}
1654		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1655			SCTP_INP_RUNLOCK(inp);
1656			continue;
1657		}
1658		/*
1659		 * Ok this could be a likely candidate, look at all of its
1660		 * addresses
1661		 */
1662		if (inp->sctp_lport != lport) {
1663			SCTP_INP_RUNLOCK(inp);
1664			continue;
1665		}
1666		/* does a VRF id match? */
1667		fnd = 0;
1668		if (inp->def_vrf_id == vrf_id)
1669			fnd = 1;
1670
1671		if (!fnd) {
1672			SCTP_INP_RUNLOCK(inp);
1673			continue;
1674		}
1675		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1676			if (laddr->ifa == NULL) {
1677				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1678				    __FUNCTION__);
1679				continue;
1680			}
1681			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1682			    (void *)laddr->ifa);
1683			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1684				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1685				continue;
1686			}
1687			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1688				/* possible, see if it matches */
1689				switch (nam->sa_family) {
1690#ifdef INET
1691				case AF_INET:
1692					if (sin->sin_addr.s_addr ==
1693					    laddr->ifa->address.sin.sin_addr.s_addr) {
1694						SCTP_INP_RUNLOCK(inp);
1695						return (inp);
1696					}
1697					break;
1698#endif
1699#ifdef INET6
1700				case AF_INET6:
1701					intf_addr6 = &laddr->ifa->address.sin6;
1702					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1703					    intf_addr6)) {
1704						SCTP_INP_RUNLOCK(inp);
1705						return (inp);
1706					}
1707					break;
1708#endif
1709				}
1710			}
1711		}
1712		SCTP_INP_RUNLOCK(inp);
1713	}
1714	return (NULL);
1715}
1716
1717
1718static struct sctp_inpcb *
1719sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1720{
1721	struct sctppcbhead *head;
1722	struct sctp_inpcb *t_inp;
1723	int fnd;
1724
1725	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1726	    SCTP_BASE_INFO(hashmark))];
1727	LIST_FOREACH(t_inp, head, sctp_hash) {
1728		if (t_inp->sctp_lport != lport) {
1729			continue;
1730		}
1731		/* is it in the VRF in question */
1732		fnd = 0;
1733		if (t_inp->def_vrf_id == vrf_id)
1734			fnd = 1;
1735		if (!fnd)
1736			continue;
1737
1738		/* This one is in use. */
1739		/* check the v6/v4 binding issue */
1740		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1741		    SCTP_IPV6_V6ONLY(t_inp)) {
1742			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1743				/* collision in V6 space */
1744				return (t_inp);
1745			} else {
1746				/* inp is BOUND_V4 no conflict */
1747				continue;
1748			}
1749		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1750			/* t_inp is bound v4 and v6, conflict always */
1751			return (t_inp);
1752		} else {
1753			/* t_inp is bound only V4 */
1754			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1755			    SCTP_IPV6_V6ONLY(inp)) {
1756				/* no conflict */
1757				continue;
1758			}
1759			/* else fall through to conflict */
1760		}
1761		return (t_inp);
1762	}
1763	return (NULL);
1764}
1765
1766
1767int
1768sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1769{
1770	/* For 1-2-1 with port reuse */
1771	struct sctppcbhead *head;
1772	struct sctp_inpcb *tinp;
1773
1774	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1775		/* only works with port reuse on */
1776		return (-1);
1777	}
1778	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1779		return (0);
1780	}
1781	SCTP_INP_RUNLOCK(inp);
1782	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1783	    SCTP_BASE_INFO(hashmark))];
1784	/* Kick out all non-listeners to the TCP hash */
1785	LIST_FOREACH(tinp, head, sctp_hash) {
1786		if (tinp->sctp_lport != inp->sctp_lport) {
1787			continue;
1788		}
1789		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1790			continue;
1791		}
1792		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1793			continue;
1794		}
1795		if (tinp->sctp_socket->so_qlimit) {
1796			continue;
1797		}
1798		SCTP_INP_WLOCK(tinp);
1799		LIST_REMOVE(tinp, sctp_hash);
1800		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1801		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1802		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1803		SCTP_INP_WUNLOCK(tinp);
1804	}
1805	SCTP_INP_WLOCK(inp);
1806	/* Pull from where he was */
1807	LIST_REMOVE(inp, sctp_hash);
1808	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1809	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1810	LIST_INSERT_HEAD(head, inp, sctp_hash);
1811	SCTP_INP_WUNLOCK(inp);
1812	SCTP_INP_RLOCK(inp);
1813	return (0);
1814}
1815
1816
1817struct sctp_inpcb *
1818sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1819    uint32_t vrf_id)
1820{
1821	/*
1822	 * First we check the hash table to see if someone has this port
1823	 * bound with just the port.
1824	 */
1825	struct sctp_inpcb *inp;
1826	struct sctppcbhead *head;
1827	int lport;
1828	unsigned int i;
1829
1830#ifdef INET
1831	struct sockaddr_in *sin;
1832
1833#endif
1834#ifdef INET6
1835	struct sockaddr_in6 *sin6;
1836
1837#endif
1838
1839	switch (nam->sa_family) {
1840#ifdef INET
1841	case AF_INET:
1842		sin = (struct sockaddr_in *)nam;
1843		lport = sin->sin_port;
1844		break;
1845#endif
1846#ifdef INET6
1847	case AF_INET6:
1848		sin6 = (struct sockaddr_in6 *)nam;
1849		lport = sin6->sin6_port;
1850		break;
1851#endif
1852	default:
1853		return (NULL);
1854	}
1855	/*
1856	 * I could cheat here and just cast to one of the types but we will
1857	 * do it right. It also provides the check against an Unsupported
1858	 * type too.
1859	 */
1860	/* Find the head of the ALLADDR chain */
1861	if (have_lock == 0) {
1862		SCTP_INP_INFO_RLOCK();
1863	}
1864	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1865	    SCTP_BASE_INFO(hashmark))];
1866	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1867
1868	/*
1869	 * If the TCP model exists it could be that the main listening
1870	 * endpoint is gone but there still exists a connected socket for
1871	 * this guy. If so we can return the first one that we find. This
1872	 * may NOT be the correct one so the caller should be wary on the
1873	 * returned INP. Currently the only caller that sets find_tcp_pool
1874	 * is in bindx where we are verifying that a user CAN bind the
1875	 * address. He either has bound it already, or someone else has, or
1876	 * its open to bind, so this is good enough.
1877	 */
1878	if (inp == NULL && find_tcp_pool) {
1879		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1880			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1881			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1882			if (inp) {
1883				break;
1884			}
1885		}
1886	}
1887	if (inp) {
1888		SCTP_INP_INCR_REF(inp);
1889	}
1890	if (have_lock == 0) {
1891		SCTP_INP_INFO_RUNLOCK();
1892	}
1893	return (inp);
1894}
1895
1896/*
1897 * Find an association for an endpoint with the pointer to whom you want to
1898 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1899 * need to change the *to to some other struct like a mbuf...
1900 */
1901struct sctp_tcb *
1902sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1903    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1904    uint32_t vrf_id)
1905{
1906	struct sctp_inpcb *inp = NULL;
1907	struct sctp_tcb *retval;
1908
1909	SCTP_INP_INFO_RLOCK();
1910	if (find_tcp_pool) {
1911		if (inp_p != NULL) {
1912			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1913			    vrf_id);
1914		} else {
1915			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1916			    vrf_id);
1917		}
1918		if (retval != NULL) {
1919			SCTP_INP_INFO_RUNLOCK();
1920			return (retval);
1921		}
1922	}
1923	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1924	if (inp_p != NULL) {
1925		*inp_p = inp;
1926	}
1927	SCTP_INP_INFO_RUNLOCK();
1928
1929	if (inp == NULL) {
1930		return (NULL);
1931	}
1932	/*
1933	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1934	 * we now place the source address or from in the to of the find
1935	 * endpoint call. Since in reality this chain is used from the
1936	 * inbound packet side.
1937	 */
1938	if (inp_p != NULL) {
1939		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1940		    NULL);
1941	} else {
1942		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1943		    NULL);
1944	}
1945	return retval;
1946}
1947
1948
1949/*
1950 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1951 * find all addresses that the sender has specified in any address list. Each
1952 * address will be used to lookup the TCB and see if one exits.
1953 */
1954static struct sctp_tcb *
1955sctp_findassociation_special_addr(struct mbuf *m, int offset,
1956    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1957    struct sockaddr *dst)
1958{
1959	struct sctp_paramhdr *phdr, parm_buf;
1960	struct sctp_tcb *retval;
1961	uint32_t ptype, plen;
1962
1963#ifdef INET
1964	struct sockaddr_in sin4;
1965
1966#endif
1967#ifdef INET6
1968	struct sockaddr_in6 sin6;
1969
1970#endif
1971
1972#ifdef INET
1973	memset(&sin4, 0, sizeof(sin4));
1974	sin4.sin_len = sizeof(sin4);
1975	sin4.sin_family = AF_INET;
1976	sin4.sin_port = sh->src_port;
1977#endif
1978#ifdef INET6
1979	memset(&sin6, 0, sizeof(sin6));
1980	sin6.sin6_len = sizeof(sin6);
1981	sin6.sin6_family = AF_INET6;
1982	sin6.sin6_port = sh->src_port;
1983#endif
1984
1985	retval = NULL;
1986	offset += sizeof(struct sctp_init_chunk);
1987
1988	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1989	while (phdr != NULL) {
1990		/* now we must see if we want the parameter */
1991		ptype = ntohs(phdr->param_type);
1992		plen = ntohs(phdr->param_length);
1993		if (plen == 0) {
1994			break;
1995		}
1996#ifdef INET
1997		if (ptype == SCTP_IPV4_ADDRESS &&
1998		    plen == sizeof(struct sctp_ipv4addr_param)) {
1999			/* Get the rest of the address */
2000			struct sctp_ipv4addr_param ip4_parm, *p4;
2001
2002			phdr = sctp_get_next_param(m, offset,
2003			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2004			if (phdr == NULL) {
2005				return (NULL);
2006			}
2007			p4 = (struct sctp_ipv4addr_param *)phdr;
2008			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2009			/* look it up */
2010			retval = sctp_findassociation_ep_addr(inp_p,
2011			    (struct sockaddr *)&sin4, netp, dst, NULL);
2012			if (retval != NULL) {
2013				return (retval);
2014			}
2015		}
2016#endif
2017#ifdef INET6
2018		if (ptype == SCTP_IPV6_ADDRESS &&
2019		    plen == sizeof(struct sctp_ipv6addr_param)) {
2020			/* Get the rest of the address */
2021			struct sctp_ipv6addr_param ip6_parm, *p6;
2022
2023			phdr = sctp_get_next_param(m, offset,
2024			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2025			if (phdr == NULL) {
2026				return (NULL);
2027			}
2028			p6 = (struct sctp_ipv6addr_param *)phdr;
2029			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2030			/* look it up */
2031			retval = sctp_findassociation_ep_addr(inp_p,
2032			    (struct sockaddr *)&sin6, netp, dst, NULL);
2033			if (retval != NULL) {
2034				return (retval);
2035			}
2036		}
2037#endif
2038		offset += SCTP_SIZE32(plen);
2039		phdr = sctp_get_next_param(m, offset, &parm_buf,
2040		    sizeof(parm_buf));
2041	}
2042	return (NULL);
2043}
2044
2045static struct sctp_tcb *
2046sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2047    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2048    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2049{
2050	/*
2051	 * Use my vtag to hash. If we find it we then verify the source addr
2052	 * is in the assoc. If all goes well we save a bit on rec of a
2053	 * packet.
2054	 */
2055	struct sctpasochead *head;
2056	struct sctp_nets *net;
2057	struct sctp_tcb *stcb;
2058
2059	SCTP_INP_INFO_RLOCK();
2060	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2061	    SCTP_BASE_INFO(hashasocmark))];
2062	if (head == NULL) {
2063		/* invalid vtag */
2064		SCTP_INP_INFO_RUNLOCK();
2065		return (NULL);
2066	}
2067	LIST_FOREACH(stcb, head, sctp_asocs) {
2068		SCTP_INP_RLOCK(stcb->sctp_ep);
2069		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2070			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2071			continue;
2072		}
2073		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2074			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2075			continue;
2076		}
2077		SCTP_TCB_LOCK(stcb);
2078		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2079		if (stcb->asoc.my_vtag == vtag) {
2080			/* candidate */
2081			if (stcb->rport != rport) {
2082				SCTP_TCB_UNLOCK(stcb);
2083				continue;
2084			}
2085			if (stcb->sctp_ep->sctp_lport != lport) {
2086				SCTP_TCB_UNLOCK(stcb);
2087				continue;
2088			}
2089			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2090				SCTP_TCB_UNLOCK(stcb);
2091				continue;
2092			}
2093			/* RRS:Need toaddr check here */
2094			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2095				/* Endpoint does not own this address */
2096				SCTP_TCB_UNLOCK(stcb);
2097				continue;
2098			}
2099			if (remote_tag) {
2100				/*
2101				 * If we have both vtags that's all we match
2102				 * on
2103				 */
2104				if (stcb->asoc.peer_vtag == remote_tag) {
2105					/*
2106					 * If both tags match we consider it
2107					 * conclusive and check NO
2108					 * source/destination addresses
2109					 */
2110					goto conclusive;
2111				}
2112			}
2113			if (skip_src_check) {
2114		conclusive:
2115				if (from) {
2116					*netp = sctp_findnet(stcb, from);
2117				} else {
2118					*netp = NULL;	/* unknown */
2119				}
2120				if (inp_p)
2121					*inp_p = stcb->sctp_ep;
2122				SCTP_INP_INFO_RUNLOCK();
2123				return (stcb);
2124			}
2125			net = sctp_findnet(stcb, from);
2126			if (net) {
2127				/* yep its him. */
2128				*netp = net;
2129				SCTP_STAT_INCR(sctps_vtagexpress);
2130				*inp_p = stcb->sctp_ep;
2131				SCTP_INP_INFO_RUNLOCK();
2132				return (stcb);
2133			} else {
2134				/*
2135				 * not him, this should only happen in rare
2136				 * cases so I peg it.
2137				 */
2138				SCTP_STAT_INCR(sctps_vtagbogus);
2139			}
2140		}
2141		SCTP_TCB_UNLOCK(stcb);
2142	}
2143	SCTP_INP_INFO_RUNLOCK();
2144	return (NULL);
2145}
2146
2147/*
2148 * Find an association with the pointer to the inbound IP packet. This can be
2149 * a IPv4 or IPv6 packet.
2150 */
2151struct sctp_tcb *
2152sctp_findassociation_addr(struct mbuf *m, int offset,
2153    struct sockaddr *src, struct sockaddr *dst,
2154    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2155    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2156{
2157	int find_tcp_pool;
2158	struct sctp_tcb *retval;
2159	struct sctp_inpcb *inp;
2160
2161	if (sh->v_tag) {
2162		/* we only go down this path if vtag is non-zero */
2163		retval = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2164		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2165		if (retval) {
2166			return (retval);
2167		}
2168	}
2169	find_tcp_pool = 0;
2170	if ((ch->chunk_type != SCTP_INITIATION) &&
2171	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2172	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2173	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2174		/* Other chunk types go to the tcp pool. */
2175		find_tcp_pool = 1;
2176	}
2177	if (inp_p) {
2178		retval = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2179		    find_tcp_pool, vrf_id);
2180		inp = *inp_p;
2181	} else {
2182		retval = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2183		    find_tcp_pool, vrf_id);
2184	}
2185	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", (void *)retval, (void *)inp);
2186	if (retval == NULL && inp) {
2187		/* Found a EP but not this address */
2188		if ((ch->chunk_type == SCTP_INITIATION) ||
2189		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2190			/*-
2191			 * special hook, we do NOT return linp or an
2192			 * association that is linked to an existing
2193			 * association that is under the TCP pool (i.e. no
2194			 * listener exists). The endpoint finding routine
2195			 * will always find a listener before examining the
2196			 * TCP pool.
2197			 */
2198			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2199				if (inp_p) {
2200					*inp_p = NULL;
2201				}
2202				return (NULL);
2203			}
2204			retval = sctp_findassociation_special_addr(m,
2205			    offset, sh, &inp, netp, dst);
2206			if (inp_p != NULL) {
2207				*inp_p = inp;
2208			}
2209		}
2210	}
2211	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", (void *)retval);
2212	return (retval);
2213}
2214
2215/*
2216 * lookup an association by an ASCONF lookup address.
2217 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2218 */
2219struct sctp_tcb *
2220sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2221    struct sockaddr *dst, struct sctphdr *sh,
2222    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2223{
2224	struct sctp_tcb *stcb;
2225	struct sockaddr_storage remote_store;
2226	struct sctp_paramhdr parm_buf, *phdr;
2227	int ptype;
2228	int zero_address = 0;
2229
2230#ifdef INET
2231	struct sockaddr_in *sin;
2232
2233#endif
2234#ifdef INET6
2235	struct sockaddr_in6 *sin6;
2236
2237#endif
2238
2239	memset(&remote_store, 0, sizeof(remote_store));
2240	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2241	    &parm_buf, sizeof(struct sctp_paramhdr));
2242	if (phdr == NULL) {
2243		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2244		    __FUNCTION__);
2245		return NULL;
2246	}
2247	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2248	/* get the correlation address */
2249	switch (ptype) {
2250#ifdef INET6
2251	case SCTP_IPV6_ADDRESS:
2252		{
2253			/* ipv6 address param */
2254			struct sctp_ipv6addr_param *p6, p6_buf;
2255
2256			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2257				return NULL;
2258			}
2259			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2260			    offset + sizeof(struct sctp_asconf_chunk),
2261			    &p6_buf.ph, sizeof(*p6));
2262			if (p6 == NULL) {
2263				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2264				    __FUNCTION__);
2265				return (NULL);
2266			}
2267			sin6 = (struct sockaddr_in6 *)&remote_store;
2268			sin6->sin6_family = AF_INET6;
2269			sin6->sin6_len = sizeof(*sin6);
2270			sin6->sin6_port = sh->src_port;
2271			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2272			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2273				zero_address = 1;
2274			break;
2275		}
2276#endif
2277#ifdef INET
2278	case SCTP_IPV4_ADDRESS:
2279		{
2280			/* ipv4 address param */
2281			struct sctp_ipv4addr_param *p4, p4_buf;
2282
2283			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2284				return NULL;
2285			}
2286			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2287			    offset + sizeof(struct sctp_asconf_chunk),
2288			    &p4_buf.ph, sizeof(*p4));
2289			if (p4 == NULL) {
2290				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2291				    __FUNCTION__);
2292				return (NULL);
2293			}
2294			sin = (struct sockaddr_in *)&remote_store;
2295			sin->sin_family = AF_INET;
2296			sin->sin_len = sizeof(*sin);
2297			sin->sin_port = sh->src_port;
2298			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2299			if (sin->sin_addr.s_addr == INADDR_ANY)
2300				zero_address = 1;
2301			break;
2302		}
2303#endif
2304	default:
2305		/* invalid address param type */
2306		return NULL;
2307	}
2308
2309	if (zero_address) {
2310		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2311		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2312		if (stcb != NULL) {
2313			SCTP_INP_DECR_REF(*inp_p);
2314		}
2315	} else {
2316		stcb = sctp_findassociation_ep_addr(inp_p,
2317		    (struct sockaddr *)&remote_store, netp,
2318		    dst, NULL);
2319	}
2320	return (stcb);
2321}
2322
2323
2324/*
2325 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2326 * addresses. This way if we don't get a bind we by default pick a ephemeral
2327 * port with all addresses bound.
2328 */
2329int
2330sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2331{
2332	/*
2333	 * we get called when a new endpoint starts up. We need to allocate
2334	 * the sctp_inpcb structure from the zone and init it. Mark it as
2335	 * unbound and find a port that we can use as an ephemeral with
2336	 * INADDR_ANY. If the user binds later no problem we can then add in
2337	 * the specific addresses. And setup the default parameters for the
2338	 * EP.
2339	 */
2340	int i, error;
2341	struct sctp_inpcb *inp;
2342	struct sctp_pcb *m;
2343	struct timeval time;
2344	sctp_sharedkey_t *null_key;
2345
2346	error = 0;
2347
2348	SCTP_INP_INFO_WLOCK();
2349	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2350	if (inp == NULL) {
2351		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2352		SCTP_INP_INFO_WUNLOCK();
2353		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2354		return (ENOBUFS);
2355	}
2356	/* zap it */
2357	bzero(inp, sizeof(*inp));
2358
2359	/* bump generations */
2360	/* setup socket pointers */
2361	inp->sctp_socket = so;
2362	inp->ip_inp.inp.inp_socket = so;
2363#ifdef INET6
2364	if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2365		inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2366	}
2367#endif
2368	inp->sctp_associd_counter = 1;
2369	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2370	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2371	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2372	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2373	/* init the small hash table we use to track asocid <-> tcb */
2374	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2375	if (inp->sctp_asocidhash == NULL) {
2376		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2377		SCTP_INP_INFO_WUNLOCK();
2378		return (ENOBUFS);
2379	}
2380#ifdef IPSEC
2381	{
2382		struct inpcbpolicy *pcb_sp = NULL;
2383
2384		error = ipsec_init_policy(so, &pcb_sp);
2385		/* Arrange to share the policy */
2386		inp->ip_inp.inp.inp_sp = pcb_sp;
2387		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2388	}
2389	if (error != 0) {
2390		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2391		SCTP_INP_INFO_WUNLOCK();
2392		return error;
2393	}
2394#endif				/* IPSEC */
2395	SCTP_INCR_EP_COUNT();
2396	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2397	SCTP_INP_INFO_WUNLOCK();
2398
2399	so->so_pcb = (caddr_t)inp;
2400
2401	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2402		/* UDP style socket */
2403		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2404		    SCTP_PCB_FLAGS_UNBOUND);
2405		/* Be sure it is NON-BLOCKING IO for UDP */
2406		/* SCTP_SET_SO_NBIO(so); */
2407	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2408		/* TCP style socket */
2409		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2410		    SCTP_PCB_FLAGS_UNBOUND);
2411		/* Be sure we have blocking IO by default */
2412		SCTP_CLEAR_SO_NBIO(so);
2413	} else {
2414		/*
2415		 * unsupported socket type (RAW, etc)- in case we missed it
2416		 * in protosw
2417		 */
2418		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2419		so->so_pcb = NULL;
2420		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2421		return (EOPNOTSUPP);
2422	}
2423	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2424		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2425		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2426	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2427		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2428		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2429	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2430		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2431		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2432	}
2433	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2434	    &inp->sctp_hashmark);
2435	if (inp->sctp_tcbhash == NULL) {
2436		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2437		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2438		so->so_pcb = NULL;
2439		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2440		return (ENOBUFS);
2441	}
2442	inp->def_vrf_id = vrf_id;
2443
2444	SCTP_INP_INFO_WLOCK();
2445	SCTP_INP_LOCK_INIT(inp);
2446	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2447	SCTP_INP_READ_INIT(inp);
2448	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2449	/* lock the new ep */
2450	SCTP_INP_WLOCK(inp);
2451
2452	/* add it to the info area */
2453	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2454	SCTP_INP_INFO_WUNLOCK();
2455
2456	TAILQ_INIT(&inp->read_queue);
2457	LIST_INIT(&inp->sctp_addr_list);
2458
2459	LIST_INIT(&inp->sctp_asoc_list);
2460
2461#ifdef SCTP_TRACK_FREED_ASOCS
2462	/* TEMP CODE */
2463	LIST_INIT(&inp->sctp_asoc_free_list);
2464#endif
2465	/* Init the timer structure for signature change */
2466	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2467	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2468
2469	/* now init the actual endpoint default data */
2470	m = &inp->sctp_ep;
2471
2472	/* setup the base timeout information */
2473	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2474	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2475	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2476	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2477	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2478	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2479	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2480	/* all max/min max are in ms */
2481	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2482	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2483	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2484	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2485	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2486
2487	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2488
2489	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2490	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2491	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2492	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2493	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2494	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2495	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2496	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2497
2498	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2499	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2500	/* number of streams to pre-open on a association */
2501	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2502
2503	/* Add adaptation cookie */
2504	m->adaptation_layer_indicator = 0x504C5253;
2505
2506	/* seed random number generator */
2507	m->random_counter = 1;
2508	m->store_at = SCTP_SIGNATURE_SIZE;
2509	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2510	sctp_fill_random_store(m);
2511
2512	/* Minimum cookie size */
2513	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2514	    sizeof(struct sctp_state_cookie);
2515	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2516
2517	/* Setup the initial secret */
2518	(void)SCTP_GETTIME_TIMEVAL(&time);
2519	m->time_of_secret_change = time.tv_sec;
2520
2521	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2522		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2523	}
2524	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2525
2526	/* How long is a cookie good for ? */
2527	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2528	/*
2529	 * Initialize authentication parameters
2530	 */
2531	m->local_hmacs = sctp_default_supported_hmaclist();
2532	m->local_auth_chunks = sctp_alloc_chunklist();
2533	m->default_dscp = 0;
2534#ifdef INET6
2535	m->default_flowlabel = 0;
2536#endif
2537	m->port = 0;		/* encapsulation disabled by default */
2538	sctp_auth_set_default_chunks(m->local_auth_chunks);
2539	LIST_INIT(&m->shared_keys);
2540	/* add default NULL key as key id 0 */
2541	null_key = sctp_alloc_sharedkey();
2542	sctp_insert_sharedkey(&m->shared_keys, null_key);
2543	SCTP_INP_WUNLOCK(inp);
2544#ifdef SCTP_LOG_CLOSING
2545	sctp_log_closing(inp, NULL, 12);
2546#endif
2547	return (error);
2548}
2549
2550
2551void
2552sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2553    struct sctp_tcb *stcb)
2554{
2555	struct sctp_nets *net;
2556	uint16_t lport, rport;
2557	struct sctppcbhead *head;
2558	struct sctp_laddr *laddr, *oladdr;
2559
2560	atomic_add_int(&stcb->asoc.refcnt, 1);
2561	SCTP_TCB_UNLOCK(stcb);
2562	SCTP_INP_INFO_WLOCK();
2563	SCTP_INP_WLOCK(old_inp);
2564	SCTP_INP_WLOCK(new_inp);
2565	SCTP_TCB_LOCK(stcb);
2566	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2567
2568	new_inp->sctp_ep.time_of_secret_change =
2569	    old_inp->sctp_ep.time_of_secret_change;
2570	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2571	    sizeof(old_inp->sctp_ep.secret_key));
2572	new_inp->sctp_ep.current_secret_number =
2573	    old_inp->sctp_ep.current_secret_number;
2574	new_inp->sctp_ep.last_secret_number =
2575	    old_inp->sctp_ep.last_secret_number;
2576	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2577
2578	/* make it so new data pours into the new socket */
2579	stcb->sctp_socket = new_inp->sctp_socket;
2580	stcb->sctp_ep = new_inp;
2581
2582	/* Copy the port across */
2583	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2584	rport = stcb->rport;
2585	/* Pull the tcb from the old association */
2586	LIST_REMOVE(stcb, sctp_tcbhash);
2587	LIST_REMOVE(stcb, sctp_tcblist);
2588	if (stcb->asoc.in_asocid_hash) {
2589		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2590	}
2591	/* Now insert the new_inp into the TCP connected hash */
2592	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2593
2594	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2595	/* Its safe to access */
2596	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2597
2598	/* Now move the tcb into the endpoint list */
2599	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2600	/*
2601	 * Question, do we even need to worry about the ep-hash since we
2602	 * only have one connection? Probably not :> so lets get rid of it
2603	 * and not suck up any kernel memory in that.
2604	 */
2605	if (stcb->asoc.in_asocid_hash) {
2606		struct sctpasochead *lhd;
2607
2608		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2609		    new_inp->hashasocidmark)];
2610		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2611	}
2612	/* Ok. Let's restart timer. */
2613	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2614		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2615		    stcb, net);
2616	}
2617
2618	SCTP_INP_INFO_WUNLOCK();
2619	if (new_inp->sctp_tcbhash != NULL) {
2620		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2621		new_inp->sctp_tcbhash = NULL;
2622	}
2623	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2624		/* Subset bound, so copy in the laddr list from the old_inp */
2625		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2626			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2627			if (laddr == NULL) {
2628				/*
2629				 * Gak, what can we do? This assoc is really
2630				 * HOSED. We probably should send an abort
2631				 * here.
2632				 */
2633				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2634				continue;
2635			}
2636			SCTP_INCR_LADDR_COUNT();
2637			bzero(laddr, sizeof(*laddr));
2638			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2639			laddr->ifa = oladdr->ifa;
2640			atomic_add_int(&laddr->ifa->refcount, 1);
2641			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2642			    sctp_nxt_addr);
2643			new_inp->laddr_count++;
2644			if (oladdr == stcb->asoc.last_used_address) {
2645				stcb->asoc.last_used_address = laddr;
2646			}
2647		}
2648	}
2649	/*
2650	 * Now any running timers need to be adjusted since we really don't
2651	 * care if they are running or not just blast in the new_inp into
2652	 * all of them.
2653	 */
2654
2655	stcb->asoc.dack_timer.ep = (void *)new_inp;
2656	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2657	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2658	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2659	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2660	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2661	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2662	/* now what about the nets? */
2663	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2664		net->pmtu_timer.ep = (void *)new_inp;
2665		net->hb_timer.ep = (void *)new_inp;
2666		net->rxt_timer.ep = (void *)new_inp;
2667	}
2668	SCTP_INP_WUNLOCK(new_inp);
2669	SCTP_INP_WUNLOCK(old_inp);
2670}
2671
2672
2673
2674
2675/* sctp_ifap is used to bypass normal local address validation checks */
2676int
2677sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2678    struct sctp_ifa *sctp_ifap, struct thread *p)
2679{
2680	/* bind a ep to a socket address */
2681	struct sctppcbhead *head;
2682	struct sctp_inpcb *inp, *inp_tmp;
2683	struct inpcb *ip_inp;
2684	int port_reuse_active = 0;
2685	int bindall;
2686	uint16_t lport;
2687	int error;
2688	uint32_t vrf_id;
2689
2690	lport = 0;
2691	error = 0;
2692	bindall = 1;
2693	inp = (struct sctp_inpcb *)so->so_pcb;
2694	ip_inp = (struct inpcb *)so->so_pcb;
2695#ifdef SCTP_DEBUG
2696	if (addr) {
2697		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2698		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2699		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2700		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2701	}
2702#endif
2703	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2704		/* already did a bind, subsequent binds NOT allowed ! */
2705		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2706		return (EINVAL);
2707	}
2708#ifdef INVARIANTS
2709	if (p == NULL)
2710		panic("null proc/thread");
2711#endif
2712	if (addr != NULL) {
2713		switch (addr->sa_family) {
2714#ifdef INET
2715		case AF_INET:
2716			{
2717				struct sockaddr_in *sin;
2718
2719				/* IPV6_V6ONLY socket? */
2720				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2721					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2722					return (EINVAL);
2723				}
2724				if (addr->sa_len != sizeof(*sin)) {
2725					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2726					return (EINVAL);
2727				}
2728				sin = (struct sockaddr_in *)addr;
2729				lport = sin->sin_port;
2730				/*
2731				 * For LOOPBACK the prison_local_ip4() call
2732				 * will transmute the ip address to the
2733				 * proper value.
2734				 */
2735				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2736					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2737					return (error);
2738				}
2739				if (sin->sin_addr.s_addr != INADDR_ANY) {
2740					bindall = 0;
2741				}
2742				break;
2743			}
2744#endif
2745#ifdef INET6
2746		case AF_INET6:
2747			{
2748				/*
2749				 * Only for pure IPv6 Address. (No IPv4
2750				 * Mapped!)
2751				 */
2752				struct sockaddr_in6 *sin6;
2753
2754				sin6 = (struct sockaddr_in6 *)addr;
2755
2756				if (addr->sa_len != sizeof(*sin6)) {
2757					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2758					return (EINVAL);
2759				}
2760				lport = sin6->sin6_port;
2761				/*
2762				 * For LOOPBACK the prison_local_ip6() call
2763				 * will transmute the ipv6 address to the
2764				 * proper value.
2765				 */
2766				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2767				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2768					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2769					return (error);
2770				}
2771				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2772					bindall = 0;
2773					/* KAME hack: embed scopeid */
2774					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2775						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2776						return (EINVAL);
2777					}
2778				}
2779				/* this must be cleared for ifa_ifwithaddr() */
2780				sin6->sin6_scope_id = 0;
2781				break;
2782			}
2783#endif
2784		default:
2785			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2786			return (EAFNOSUPPORT);
2787		}
2788	}
2789	SCTP_INP_INFO_WLOCK();
2790	SCTP_INP_WLOCK(inp);
2791	/* Setup a vrf_id to be the default for the non-bind-all case. */
2792	vrf_id = inp->def_vrf_id;
2793
2794	/* increase our count due to the unlock we do */
2795	SCTP_INP_INCR_REF(inp);
2796	if (lport) {
2797		/*
2798		 * Did the caller specify a port? if so we must see if an ep
2799		 * already has this one bound.
2800		 */
2801		/* got to be root to get at low ports */
2802		if (ntohs(lport) < IPPORT_RESERVED) {
2803			if (p && (error =
2804			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2805			    )) {
2806				SCTP_INP_DECR_REF(inp);
2807				SCTP_INP_WUNLOCK(inp);
2808				SCTP_INP_INFO_WUNLOCK();
2809				return (error);
2810			}
2811		}
2812		if (p == NULL) {
2813			SCTP_INP_DECR_REF(inp);
2814			SCTP_INP_WUNLOCK(inp);
2815			SCTP_INP_INFO_WUNLOCK();
2816			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2817			return (error);
2818		}
2819		SCTP_INP_WUNLOCK(inp);
2820		if (bindall) {
2821			vrf_id = inp->def_vrf_id;
2822			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2823			if (inp_tmp != NULL) {
2824				/*
2825				 * lock guy returned and lower count note
2826				 * that we are not bound so inp_tmp should
2827				 * NEVER be inp. And it is this inp
2828				 * (inp_tmp) that gets the reference bump,
2829				 * so we must lower it.
2830				 */
2831				SCTP_INP_DECR_REF(inp_tmp);
2832				/* unlock info */
2833				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2834				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2835					/*
2836					 * Ok, must be one-2-one and
2837					 * allowing port re-use
2838					 */
2839					port_reuse_active = 1;
2840					goto continue_anyway;
2841				}
2842				SCTP_INP_DECR_REF(inp);
2843				SCTP_INP_INFO_WUNLOCK();
2844				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2845				return (EADDRINUSE);
2846			}
2847		} else {
2848			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2849			if (inp_tmp != NULL) {
2850				/*
2851				 * lock guy returned and lower count note
2852				 * that we are not bound so inp_tmp should
2853				 * NEVER be inp. And it is this inp
2854				 * (inp_tmp) that gets the reference bump,
2855				 * so we must lower it.
2856				 */
2857				SCTP_INP_DECR_REF(inp_tmp);
2858				/* unlock info */
2859				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2860				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2861					/*
2862					 * Ok, must be one-2-one and
2863					 * allowing port re-use
2864					 */
2865					port_reuse_active = 1;
2866					goto continue_anyway;
2867				}
2868				SCTP_INP_DECR_REF(inp);
2869				SCTP_INP_INFO_WUNLOCK();
2870				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2871				return (EADDRINUSE);
2872			}
2873		}
2874continue_anyway:
2875		SCTP_INP_WLOCK(inp);
2876		if (bindall) {
2877			/* verify that no lport is not used by a singleton */
2878			if ((port_reuse_active == 0) &&
2879			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2880				/* Sorry someone already has this one bound */
2881				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2882				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2883					port_reuse_active = 1;
2884				} else {
2885					SCTP_INP_DECR_REF(inp);
2886					SCTP_INP_WUNLOCK(inp);
2887					SCTP_INP_INFO_WUNLOCK();
2888					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2889					return (EADDRINUSE);
2890				}
2891			}
2892		}
2893	} else {
2894		uint16_t first, last, candidate;
2895		uint16_t count;
2896		int done;
2897
2898		if (ip_inp->inp_flags & INP_HIGHPORT) {
2899			first = MODULE_GLOBAL(ipport_hifirstauto);
2900			last = MODULE_GLOBAL(ipport_hilastauto);
2901		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2902			if (p && (error =
2903			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2904			    )) {
2905				SCTP_INP_DECR_REF(inp);
2906				SCTP_INP_WUNLOCK(inp);
2907				SCTP_INP_INFO_WUNLOCK();
2908				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2909				return (error);
2910			}
2911			first = MODULE_GLOBAL(ipport_lowfirstauto);
2912			last = MODULE_GLOBAL(ipport_lowlastauto);
2913		} else {
2914			first = MODULE_GLOBAL(ipport_firstauto);
2915			last = MODULE_GLOBAL(ipport_lastauto);
2916		}
2917		if (first > last) {
2918			uint16_t temp;
2919
2920			temp = first;
2921			first = last;
2922			last = temp;
2923		}
2924		count = last - first + 1;	/* number of candidates */
2925		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2926
2927		done = 0;
2928		while (!done) {
2929			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2930				done = 1;
2931			}
2932			if (!done) {
2933				if (--count == 0) {
2934					SCTP_INP_DECR_REF(inp);
2935					SCTP_INP_WUNLOCK(inp);
2936					SCTP_INP_INFO_WUNLOCK();
2937					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2938					return (EADDRINUSE);
2939				}
2940				if (candidate == last)
2941					candidate = first;
2942				else
2943					candidate = candidate + 1;
2944			}
2945		}
2946		lport = htons(candidate);
2947	}
2948	SCTP_INP_DECR_REF(inp);
2949	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2950	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2951		/*
2952		 * this really should not happen. The guy did a non-blocking
2953		 * bind and then did a close at the same time.
2954		 */
2955		SCTP_INP_WUNLOCK(inp);
2956		SCTP_INP_INFO_WUNLOCK();
2957		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2958		return (EINVAL);
2959	}
2960	/* ok we look clear to give out this port, so lets setup the binding */
2961	if (bindall) {
2962		/* binding to all addresses, so just set in the proper flags */
2963		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2964		/* set the automatic addr changes from kernel flag */
2965		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2966			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2967			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2968		} else {
2969			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2970			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2971		}
2972		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2973			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2974		} else {
2975			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2976		}
2977		/*
2978		 * set the automatic mobility_base from kernel flag (by
2979		 * micchie)
2980		 */
2981		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
2982			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
2983			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2984		} else {
2985			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
2986			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2987		}
2988		/*
2989		 * set the automatic mobility_fasthandoff from kernel flag
2990		 * (by micchie)
2991		 */
2992		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
2993			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
2994			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2995		} else {
2996			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
2997			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2998		}
2999	} else {
3000		/*
3001		 * bind specific, make sure flags is off and add a new
3002		 * address structure to the sctp_addr_list inside the ep
3003		 * structure.
3004		 *
3005		 * We will need to allocate one and insert it at the head. The
3006		 * socketopt call can just insert new addresses in there as
3007		 * well. It will also have to do the embed scope kame hack
3008		 * too (before adding).
3009		 */
3010		struct sctp_ifa *ifa;
3011		struct sockaddr_storage store_sa;
3012
3013		memset(&store_sa, 0, sizeof(store_sa));
3014		switch (addr->sa_family) {
3015#ifdef INET
3016		case AF_INET:
3017			{
3018				struct sockaddr_in *sin;
3019
3020				sin = (struct sockaddr_in *)&store_sa;
3021				memcpy(sin, addr, sizeof(struct sockaddr_in));
3022				sin->sin_port = 0;
3023				break;
3024			}
3025#endif
3026#ifdef INET6
3027		case AF_INET6:
3028			{
3029				struct sockaddr_in6 *sin6;
3030
3031				sin6 = (struct sockaddr_in6 *)&store_sa;
3032				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3033				sin6->sin6_port = 0;
3034				break;
3035			}
3036#endif
3037		default:
3038			break;
3039		}
3040		/*
3041		 * first find the interface with the bound address need to
3042		 * zero out the port to find the address! yuck! can't do
3043		 * this earlier since need port for sctp_pcb_findep()
3044		 */
3045		if (sctp_ifap != NULL) {
3046			ifa = sctp_ifap;
3047		} else {
3048			/*
3049			 * Note for BSD we hit here always other O/S's will
3050			 * pass things in via the sctp_ifap argument
3051			 * (Panda).
3052			 */
3053			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3054			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3055		}
3056		if (ifa == NULL) {
3057			/* Can't find an interface with that address */
3058			SCTP_INP_WUNLOCK(inp);
3059			SCTP_INP_INFO_WUNLOCK();
3060			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3061			return (EADDRNOTAVAIL);
3062		}
3063#ifdef INET6
3064		if (addr->sa_family == AF_INET6) {
3065			/* GAK, more FIXME IFA lock? */
3066			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3067				/* Can't bind a non-existent addr. */
3068				SCTP_INP_WUNLOCK(inp);
3069				SCTP_INP_INFO_WUNLOCK();
3070				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3071				return (EINVAL);
3072			}
3073		}
3074#endif
3075		/* we're not bound all */
3076		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3077		/* allow bindx() to send ASCONF's for binding changes */
3078		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3079		/* clear automatic addr changes from kernel flag */
3080		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3081
3082		/* add this address to the endpoint list */
3083		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3084		if (error != 0) {
3085			SCTP_INP_WUNLOCK(inp);
3086			SCTP_INP_INFO_WUNLOCK();
3087			return (error);
3088		}
3089		inp->laddr_count++;
3090	}
3091	/* find the bucket */
3092	if (port_reuse_active) {
3093		/* Put it into tcp 1-2-1 hash */
3094		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3095		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3096	} else {
3097		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3098	}
3099	/* put it in the bucket */
3100	LIST_INSERT_HEAD(head, inp, sctp_hash);
3101	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3102	    (void *)head, ntohs(lport), port_reuse_active);
3103	/* set in the port */
3104	inp->sctp_lport = lport;
3105
3106	/* turn off just the unbound flag */
3107	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3108	SCTP_INP_WUNLOCK(inp);
3109	SCTP_INP_INFO_WUNLOCK();
3110	return (0);
3111}
3112
3113
3114static void
3115sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3116{
3117	struct sctp_iterator *it, *nit;
3118
3119	/*
3120	 * We enter with the only the ITERATOR_LOCK in place and a write
3121	 * lock on the inp_info stuff.
3122	 */
3123	it = sctp_it_ctl.cur_it;
3124	if (it && (it->vn != curvnet)) {
3125		/* Its not looking at our VNET */
3126		return;
3127	}
3128	if (it && (it->inp == inp)) {
3129		/*
3130		 * This is tricky and we hold the iterator lock, but when it
3131		 * returns and gets the lock (when we release it) the
3132		 * iterator will try to operate on inp. We need to stop that
3133		 * from happening. But of course the iterator has a
3134		 * reference on the stcb and inp. We can mark it and it will
3135		 * stop.
3136		 *
3137		 * If its a single iterator situation, we set the end iterator
3138		 * flag. Otherwise we set the iterator to go to the next
3139		 * inp.
3140		 *
3141		 */
3142		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3143			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3144		} else {
3145			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3146		}
3147	}
3148	/*
3149	 * Now go through and remove any single reference to our inp that
3150	 * may be still pending on the list
3151	 */
3152	SCTP_IPI_ITERATOR_WQ_LOCK();
3153	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3154		if (it->vn != curvnet) {
3155			continue;
3156		}
3157		if (it->inp == inp) {
3158			/* This one points to me is it inp specific? */
3159			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3160				/* Remove and free this one */
3161				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3162				    it, sctp_nxt_itr);
3163				if (it->function_atend != NULL) {
3164					(*it->function_atend) (it->pointer, it->val);
3165				}
3166				SCTP_FREE(it, SCTP_M_ITER);
3167			} else {
3168				it->inp = LIST_NEXT(it->inp, sctp_list);
3169				if (it->inp) {
3170					SCTP_INP_INCR_REF(it->inp);
3171				}
3172			}
3173			/*
3174			 * When its put in the refcnt is incremented so decr
3175			 * it
3176			 */
3177			SCTP_INP_DECR_REF(inp);
3178		}
3179	}
3180	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3181}
3182
3183/* release sctp_inpcb unbind the port */
3184void
3185sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3186{
3187	/*
3188	 * Here we free a endpoint. We must find it (if it is in the Hash
3189	 * table) and remove it from there. Then we must also find it in the
3190	 * overall list and remove it from there. After all removals are
3191	 * complete then any timer has to be stopped. Then start the actual
3192	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3193	 * all associations. d) finally the ep itself.
3194	 */
3195	struct sctp_tcb *asoc, *nasoc;
3196	struct sctp_laddr *laddr, *nladdr;
3197	struct inpcb *ip_pcb;
3198	struct socket *so;
3199	int being_refed = 0;
3200	struct sctp_queued_to_read *sq, *nsq;
3201	int cnt;
3202	sctp_sharedkey_t *shared_key, *nshared_key;
3203
3204
3205#ifdef SCTP_LOG_CLOSING
3206	sctp_log_closing(inp, NULL, 0);
3207#endif
3208	SCTP_ITERATOR_LOCK();
3209	/* mark any iterators on the list or being processed */
3210	sctp_iterator_inp_being_freed(inp);
3211	SCTP_ITERATOR_UNLOCK();
3212	so = inp->sctp_socket;
3213	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3214		/* been here before.. eeks.. get out of here */
3215		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3216#ifdef SCTP_LOG_CLOSING
3217		sctp_log_closing(inp, NULL, 1);
3218#endif
3219		return;
3220	}
3221	SCTP_ASOC_CREATE_LOCK(inp);
3222	SCTP_INP_INFO_WLOCK();
3223
3224	SCTP_INP_WLOCK(inp);
3225	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3226		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3227		/* socket is gone, so no more wakeups allowed */
3228		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3229		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3230		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3231
3232	}
3233	/* First time through we have the socket lock, after that no more. */
3234	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3235	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3236
3237	if (inp->control) {
3238		sctp_m_freem(inp->control);
3239		inp->control = NULL;
3240	}
3241	if (inp->pkt) {
3242		sctp_m_freem(inp->pkt);
3243		inp->pkt = NULL;
3244	}
3245	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3246					 * here but I will be nice :> (i.e.
3247					 * ip_pcb = ep;) */
3248	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3249		int cnt_in_sd;
3250
3251		cnt_in_sd = 0;
3252		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3253			SCTP_TCB_LOCK(asoc);
3254			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3255				/* Skip guys being freed */
3256				cnt_in_sd++;
3257				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3258					/*
3259					 * Special case - we did not start a
3260					 * kill timer on the asoc due to it
3261					 * was not closed. So go ahead and
3262					 * start it now.
3263					 */
3264					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3265					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3266				}
3267				SCTP_TCB_UNLOCK(asoc);
3268				continue;
3269			}
3270			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3271			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3272			    (asoc->asoc.total_output_queue_size == 0)) {
3273				/*
3274				 * If we have data in queue, we don't want
3275				 * to just free since the app may have done,
3276				 * send()/close or connect/send/close. And
3277				 * it wants the data to get across first.
3278				 */
3279				/* Just abandon things in the front states */
3280				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3281				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3282					cnt_in_sd++;
3283				}
3284				continue;
3285			}
3286			/* Disconnect the socket please */
3287			asoc->sctp_socket = NULL;
3288			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3289			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3290			    (asoc->asoc.control_pdapi) ||
3291			    (asoc->asoc.size_on_all_streams > 0) ||
3292			    (so && (so->so_rcv.sb_cc > 0))) {
3293				/* Left with Data unread */
3294				struct mbuf *op_err;
3295
3296				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3297				    0, M_NOWAIT, 1, MT_DATA);
3298				if (op_err) {
3299					/* Fill in the user initiated abort */
3300					struct sctp_paramhdr *ph;
3301					uint32_t *ippp;
3302
3303					SCTP_BUF_LEN(op_err) =
3304					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3305					ph = mtod(op_err,
3306					    struct sctp_paramhdr *);
3307					ph->param_type = htons(
3308					    SCTP_CAUSE_USER_INITIATED_ABT);
3309					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3310					ippp = (uint32_t *) (ph + 1);
3311					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3312				}
3313				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3314				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3315				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3316				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3317				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3318					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3319				}
3320				if (sctp_free_assoc(inp, asoc,
3321				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3322					cnt_in_sd++;
3323				}
3324				continue;
3325			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3326				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3327			    (asoc->asoc.stream_queue_cnt == 0)) {
3328				if (asoc->asoc.locked_on_sending) {
3329					goto abort_anyway;
3330				}
3331				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3332				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3333					struct sctp_nets *netp;
3334
3335					if (asoc->asoc.alternate) {
3336						netp = asoc->asoc.alternate;
3337					} else {
3338						netp = asoc->asoc.primary_destination;
3339					}
3340					/*
3341					 * there is nothing queued to send,
3342					 * so I send shutdown
3343					 */
3344					sctp_send_shutdown(asoc, netp);
3345					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3346					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3347						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3348					}
3349					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3350					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3351					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3352					    netp);
3353					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3354					    asoc->asoc.primary_destination);
3355					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3356				}
3357			} else {
3358				/* mark into shutdown pending */
3359				struct sctp_stream_queue_pending *sp;
3360
3361				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3362				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3363				    asoc->asoc.primary_destination);
3364				if (asoc->asoc.locked_on_sending) {
3365					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3366					    sctp_streamhead);
3367					if (sp == NULL) {
3368						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3369						    (void *)asoc->asoc.locked_on_sending,
3370						    asoc->asoc.locked_on_sending->stream_no);
3371					} else {
3372						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3373							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3374					}
3375				}
3376				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3377				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3378				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3379					struct mbuf *op_err;
3380
3381			abort_anyway:
3382					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3383					    0, M_NOWAIT, 1, MT_DATA);
3384					if (op_err) {
3385						/*
3386						 * Fill in the user
3387						 * initiated abort
3388						 */
3389						struct sctp_paramhdr *ph;
3390						uint32_t *ippp;
3391
3392						SCTP_BUF_LEN(op_err) =
3393						    (sizeof(struct sctp_paramhdr) +
3394						    sizeof(uint32_t));
3395						ph = mtod(op_err,
3396						    struct sctp_paramhdr *);
3397						ph->param_type = htons(
3398						    SCTP_CAUSE_USER_INITIATED_ABT);
3399						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3400						ippp = (uint32_t *) (ph + 1);
3401						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3402					}
3403					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3404					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3405					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3406					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3407					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3408						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3409					}
3410					if (sctp_free_assoc(inp, asoc,
3411					    SCTP_PCBFREE_NOFORCE,
3412					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3413						cnt_in_sd++;
3414					}
3415					continue;
3416				} else {
3417					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3418				}
3419			}
3420			cnt_in_sd++;
3421			SCTP_TCB_UNLOCK(asoc);
3422		}
3423		/* now is there some left in our SHUTDOWN state? */
3424		if (cnt_in_sd) {
3425#ifdef SCTP_LOG_CLOSING
3426			sctp_log_closing(inp, NULL, 2);
3427#endif
3428			inp->sctp_socket = NULL;
3429			SCTP_INP_WUNLOCK(inp);
3430			SCTP_ASOC_CREATE_UNLOCK(inp);
3431			SCTP_INP_INFO_WUNLOCK();
3432			return;
3433		}
3434	}
3435	inp->sctp_socket = NULL;
3436	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3437	    SCTP_PCB_FLAGS_UNBOUND) {
3438		/*
3439		 * ok, this guy has been bound. It's port is somewhere in
3440		 * the SCTP_BASE_INFO(hash table). Remove it!
3441		 */
3442		LIST_REMOVE(inp, sctp_hash);
3443		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3444	}
3445	/*
3446	 * If there is a timer running to kill us, forget it, since it may
3447	 * have a contest on the INP lock.. which would cause us to die ...
3448	 */
3449	cnt = 0;
3450	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3451		SCTP_TCB_LOCK(asoc);
3452		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3453			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3454				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3455				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3456			}
3457			cnt++;
3458			SCTP_TCB_UNLOCK(asoc);
3459			continue;
3460		}
3461		/* Free associations that are NOT killing us */
3462		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3463		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3464			struct mbuf *op_err;
3465			uint32_t *ippp;
3466
3467			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3468			    0, M_NOWAIT, 1, MT_DATA);
3469			if (op_err) {
3470				/* Fill in the user initiated abort */
3471				struct sctp_paramhdr *ph;
3472
3473				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3474				    sizeof(uint32_t));
3475				ph = mtod(op_err, struct sctp_paramhdr *);
3476				ph->param_type = htons(
3477				    SCTP_CAUSE_USER_INITIATED_ABT);
3478				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3479				ippp = (uint32_t *) (ph + 1);
3480				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3481
3482			}
3483			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3484			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3485			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3486		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3487			cnt++;
3488			SCTP_TCB_UNLOCK(asoc);
3489			continue;
3490		}
3491		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3492		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3493			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3494		}
3495		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3496			cnt++;
3497		}
3498	}
3499	if (cnt) {
3500		/* Ok we have someone out there that will kill us */
3501		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3502#ifdef SCTP_LOG_CLOSING
3503		sctp_log_closing(inp, NULL, 3);
3504#endif
3505		SCTP_INP_WUNLOCK(inp);
3506		SCTP_ASOC_CREATE_UNLOCK(inp);
3507		SCTP_INP_INFO_WUNLOCK();
3508		return;
3509	}
3510	if (SCTP_INP_LOCK_CONTENDED(inp))
3511		being_refed++;
3512	if (SCTP_INP_READ_CONTENDED(inp))
3513		being_refed++;
3514	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3515		being_refed++;
3516
3517	if ((inp->refcount) ||
3518	    (being_refed) ||
3519	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3520		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3521#ifdef SCTP_LOG_CLOSING
3522		sctp_log_closing(inp, NULL, 4);
3523#endif
3524		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3525		SCTP_INP_WUNLOCK(inp);
3526		SCTP_ASOC_CREATE_UNLOCK(inp);
3527		SCTP_INP_INFO_WUNLOCK();
3528		return;
3529	}
3530	inp->sctp_ep.signature_change.type = 0;
3531	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3532	/*
3533	 * Remove it from the list .. last thing we need a lock for.
3534	 */
3535	LIST_REMOVE(inp, sctp_list);
3536	SCTP_INP_WUNLOCK(inp);
3537	SCTP_ASOC_CREATE_UNLOCK(inp);
3538	SCTP_INP_INFO_WUNLOCK();
3539	/*
3540	 * Now we release all locks. Since this INP cannot be found anymore
3541	 * except possibly by the kill timer that might be running. We call
3542	 * the drain function here. It should hit the case were it sees the
3543	 * ACTIVE flag cleared and exit out freeing us to proceed and
3544	 * destroy everything.
3545	 */
3546	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3547		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3548	} else {
3549		/* Probably un-needed */
3550		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3551	}
3552
3553#ifdef SCTP_LOG_CLOSING
3554	sctp_log_closing(inp, NULL, 5);
3555#endif
3556
3557
3558	if ((inp->sctp_asocidhash) != NULL) {
3559		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3560		inp->sctp_asocidhash = NULL;
3561	}
3562	/* sa_ignore FREED_MEMORY */
3563	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3564		/* Its only abandoned if it had data left */
3565		if (sq->length)
3566			SCTP_STAT_INCR(sctps_left_abandon);
3567
3568		TAILQ_REMOVE(&inp->read_queue, sq, next);
3569		sctp_free_remote_addr(sq->whoFrom);
3570		if (so)
3571			so->so_rcv.sb_cc -= sq->length;
3572		if (sq->data) {
3573			sctp_m_freem(sq->data);
3574			sq->data = NULL;
3575		}
3576		/*
3577		 * no need to free the net count, since at this point all
3578		 * assoc's are gone.
3579		 */
3580		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3581		SCTP_DECR_READQ_COUNT();
3582	}
3583	/* Now the sctp_pcb things */
3584	/*
3585	 * free each asoc if it is not already closed/free. we can't use the
3586	 * macro here since le_next will get freed as part of the
3587	 * sctp_free_assoc() call.
3588	 */
3589	if (so) {
3590#ifdef IPSEC
3591		ipsec_delete_pcbpolicy(ip_pcb);
3592#endif				/* IPSEC */
3593
3594		/* Unlocks not needed since the socket is gone now */
3595	}
3596	if (ip_pcb->inp_options) {
3597		(void)sctp_m_free(ip_pcb->inp_options);
3598		ip_pcb->inp_options = 0;
3599	}
3600#ifdef INET6
3601	if (ip_pcb->inp_vflag & INP_IPV6) {
3602		struct in6pcb *in6p;
3603
3604		in6p = (struct in6pcb *)inp;
3605		ip6_freepcbopts(in6p->in6p_outputopts);
3606	}
3607#endif				/* INET6 */
3608	ip_pcb->inp_vflag = 0;
3609	/* free up authentication fields */
3610	if (inp->sctp_ep.local_auth_chunks != NULL)
3611		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3612	if (inp->sctp_ep.local_hmacs != NULL)
3613		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3614
3615	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3616		LIST_REMOVE(shared_key, next);
3617		sctp_free_sharedkey(shared_key);
3618		/* sa_ignore FREED_MEMORY */
3619	}
3620
3621	/*
3622	 * if we have an address list the following will free the list of
3623	 * ifaddr's that are set into this ep. Again macro limitations here,
3624	 * since the LIST_FOREACH could be a bad idea.
3625	 */
3626	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3627		sctp_remove_laddr(laddr);
3628	}
3629
3630#ifdef SCTP_TRACK_FREED_ASOCS
3631	/* TEMP CODE */
3632	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3633		LIST_REMOVE(asoc, sctp_tcblist);
3634		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3635		SCTP_DECR_ASOC_COUNT();
3636	}
3637	/* *** END TEMP CODE *** */
3638#endif
3639	/* Now lets see about freeing the EP hash table. */
3640	if (inp->sctp_tcbhash != NULL) {
3641		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3642		inp->sctp_tcbhash = NULL;
3643	}
3644	/* Now we must put the ep memory back into the zone pool */
3645	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3646	SCTP_INP_LOCK_DESTROY(inp);
3647	SCTP_INP_READ_DESTROY(inp);
3648	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3649	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3650	SCTP_DECR_EP_COUNT();
3651}
3652
3653
3654struct sctp_nets *
3655sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3656{
3657	struct sctp_nets *net;
3658
3659	/* locate the address */
3660	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3661		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3662			return (net);
3663	}
3664	return (NULL);
3665}
3666
3667
3668int
3669sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3670{
3671	struct sctp_ifa *sctp_ifa;
3672
3673	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3674	if (sctp_ifa) {
3675		return (1);
3676	} else {
3677		return (0);
3678	}
3679}
3680
3681/*
3682 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3683 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3684 * stats of stuff.
3685 */
3686int
3687sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3688    struct sctp_nets **netp, int set_scope, int from)
3689{
3690	/*
3691	 * The following is redundant to the same lines in the
3692	 * sctp_aloc_assoc() but is needed since others call the add address
3693	 * function
3694	 */
3695	struct sctp_nets *net, *netfirst;
3696	int addr_inscope;
3697
3698	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3699	    from);
3700	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3701
3702	netfirst = sctp_findnet(stcb, newaddr);
3703	if (netfirst) {
3704		/*
3705		 * Lie and return ok, we don't want to make the association
3706		 * go away for this behavior. It will happen in the TCP
3707		 * model in a connected socket. It does not reach the hash
3708		 * table until after the association is built so it can't be
3709		 * found. Mark as reachable, since the initial creation will
3710		 * have been cleared and the NOT_IN_ASSOC flag will have
3711		 * been added... and we don't want to end up removing it
3712		 * back out.
3713		 */
3714		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3715			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3716			    SCTP_ADDR_UNCONFIRMED);
3717		} else {
3718			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3719		}
3720
3721		return (0);
3722	}
3723	addr_inscope = 1;
3724	switch (newaddr->sa_family) {
3725#ifdef INET
3726	case AF_INET:
3727		{
3728			struct sockaddr_in *sin;
3729
3730			sin = (struct sockaddr_in *)newaddr;
3731			if (sin->sin_addr.s_addr == 0) {
3732				/* Invalid address */
3733				return (-1);
3734			}
3735			/* zero out the bzero area */
3736			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3737
3738			/* assure len is set */
3739			sin->sin_len = sizeof(struct sockaddr_in);
3740			if (set_scope) {
3741#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3742				stcb->ipv4_local_scope = 1;
3743#else
3744				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3745					stcb->asoc.ipv4_local_scope = 1;
3746				}
3747#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3748			} else {
3749				/* Validate the address is in scope */
3750				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3751				    (stcb->asoc.ipv4_local_scope == 0)) {
3752					addr_inscope = 0;
3753				}
3754			}
3755			break;
3756		}
3757#endif
3758#ifdef INET6
3759	case AF_INET6:
3760		{
3761			struct sockaddr_in6 *sin6;
3762
3763			sin6 = (struct sockaddr_in6 *)newaddr;
3764			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3765				/* Invalid address */
3766				return (-1);
3767			}
3768			/* assure len is set */
3769			sin6->sin6_len = sizeof(struct sockaddr_in6);
3770			if (set_scope) {
3771				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3772					stcb->asoc.loopback_scope = 1;
3773					stcb->asoc.local_scope = 0;
3774					stcb->asoc.ipv4_local_scope = 1;
3775					stcb->asoc.site_scope = 1;
3776				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3777					/*
3778					 * If the new destination is a
3779					 * LINK_LOCAL we must have common
3780					 * site scope. Don't set the local
3781					 * scope since we may not share all
3782					 * links, only loopback can do this.
3783					 * Links on the local network would
3784					 * also be on our private network
3785					 * for v4 too.
3786					 */
3787					stcb->asoc.ipv4_local_scope = 1;
3788					stcb->asoc.site_scope = 1;
3789				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3790					/*
3791					 * If the new destination is
3792					 * SITE_LOCAL then we must have site
3793					 * scope in common.
3794					 */
3795					stcb->asoc.site_scope = 1;
3796				}
3797			} else {
3798				/* Validate the address is in scope */
3799				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3800				    (stcb->asoc.loopback_scope == 0)) {
3801					addr_inscope = 0;
3802				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3803				    (stcb->asoc.local_scope == 0)) {
3804					addr_inscope = 0;
3805				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3806				    (stcb->asoc.site_scope == 0)) {
3807					addr_inscope = 0;
3808				}
3809			}
3810			break;
3811		}
3812#endif
3813	default:
3814		/* not supported family type */
3815		return (-1);
3816	}
3817	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3818	if (net == NULL) {
3819		return (-1);
3820	}
3821	SCTP_INCR_RADDR_COUNT();
3822	bzero(net, sizeof(struct sctp_nets));
3823	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3824	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3825	switch (newaddr->sa_family) {
3826#ifdef INET
3827	case AF_INET:
3828		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3829		break;
3830#endif
3831#ifdef INET6
3832	case AF_INET6:
3833		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3834		break;
3835#endif
3836	default:
3837		break;
3838	}
3839	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3840	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3841		stcb->asoc.loopback_scope = 1;
3842		stcb->asoc.ipv4_local_scope = 1;
3843		stcb->asoc.local_scope = 0;
3844		stcb->asoc.site_scope = 1;
3845		addr_inscope = 1;
3846	}
3847	net->failure_threshold = stcb->asoc.def_net_failure;
3848	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3849	if (addr_inscope == 0) {
3850		net->dest_state = (SCTP_ADDR_REACHABLE |
3851		    SCTP_ADDR_OUT_OF_SCOPE);
3852	} else {
3853		if (from == SCTP_ADDR_IS_CONFIRMED)
3854			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3855			net->dest_state = SCTP_ADDR_REACHABLE;
3856		else
3857			net->dest_state = SCTP_ADDR_REACHABLE |
3858			    SCTP_ADDR_UNCONFIRMED;
3859	}
3860	/*
3861	 * We set this to 0, the timer code knows that this means its an
3862	 * initial value
3863	 */
3864	net->rto_needed = 1;
3865	net->RTO = 0;
3866	net->RTO_measured = 0;
3867	stcb->asoc.numnets++;
3868	net->ref_count = 1;
3869	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3870	net->port = stcb->asoc.port;
3871	net->dscp = stcb->asoc.default_dscp;
3872#ifdef INET6
3873	net->flowlabel = stcb->asoc.default_flowlabel;
3874#endif
3875	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3876		net->dest_state |= SCTP_ADDR_NOHB;
3877	} else {
3878		net->dest_state &= ~SCTP_ADDR_NOHB;
3879	}
3880	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3881		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3882	} else {
3883		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3884	}
3885	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3886	/* Init the timer structure */
3887	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3888	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3889	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3890
3891	/* Now generate a route for this guy */
3892#ifdef INET6
3893	/* KAME hack: embed scopeid */
3894	if (newaddr->sa_family == AF_INET6) {
3895		struct sockaddr_in6 *sin6;
3896
3897		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3898		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3899		sin6->sin6_scope_id = 0;
3900	}
3901#endif
3902	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3903
3904	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3905		/* Get source address */
3906		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3907		    stcb,
3908		    (sctp_route_t *) & net->ro,
3909		    net,
3910		    0,
3911		    stcb->asoc.vrf_id);
3912		/* Now get the interface MTU */
3913		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3914			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3915		}
3916		if (net->mtu > 0) {
3917			uint32_t rmtu;
3918
3919			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3920			if (rmtu == 0) {
3921				/*
3922				 * Start things off to match mtu of
3923				 * interface please.
3924				 */
3925				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3926				    net->ro.ro_rt, net->mtu);
3927			} else {
3928				/*
3929				 * we take the route mtu over the interface,
3930				 * since the route may be leading out the
3931				 * loopback, or a different interface.
3932				 */
3933				net->mtu = rmtu;
3934			}
3935		}
3936	}
3937	if (net->mtu == 0) {
3938		switch (newaddr->sa_family) {
3939#ifdef INET
3940		case AF_INET:
3941			net->mtu = SCTP_DEFAULT_MTU;
3942			break;
3943#endif
3944#ifdef INET6
3945		case AF_INET6:
3946			net->mtu = 1280;
3947			break;
3948#endif
3949		default:
3950			break;
3951		}
3952	}
3953	if (net->port) {
3954		net->mtu -= (uint32_t) sizeof(struct udphdr);
3955	}
3956	if (from == SCTP_ALLOC_ASOC) {
3957		stcb->asoc.smallest_mtu = net->mtu;
3958	}
3959	if (stcb->asoc.smallest_mtu > net->mtu) {
3960		stcb->asoc.smallest_mtu = net->mtu;
3961	}
3962#ifdef INET6
3963	if (newaddr->sa_family == AF_INET6) {
3964		struct sockaddr_in6 *sin6;
3965
3966		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3967		(void)sa6_recoverscope(sin6);
3968	}
3969#endif
3970
3971	/* JRS - Use the congestion control given in the CC module */
3972	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
3973		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
3974
3975	/*
3976	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3977	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3978	 */
3979	net->find_pseudo_cumack = 1;
3980	net->find_rtx_pseudo_cumack = 1;
3981	net->src_addr_selected = 0;
3982	/* Choose an initial flowid. */
3983	net->flowid = stcb->asoc.my_vtag ^
3984	    ntohs(stcb->rport) ^
3985	    ntohs(stcb->sctp_ep->sctp_lport);
3986#ifdef INVARIANTS
3987	net->flowidset = 1;
3988#endif
3989	if (netp) {
3990		*netp = net;
3991	}
3992	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3993	if (net->ro.ro_rt == NULL) {
3994		/* Since we have no route put it at the back */
3995		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3996	} else if (netfirst == NULL) {
3997		/* We are the first one in the pool. */
3998		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3999	} else if (netfirst->ro.ro_rt == NULL) {
4000		/*
4001		 * First one has NO route. Place this one ahead of the first
4002		 * one.
4003		 */
4004		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4005	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
4006		/*
4007		 * This one has a different interface than the one at the
4008		 * top of the list. Place it ahead.
4009		 */
4010		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4011	} else {
4012		/*
4013		 * Ok we have the same interface as the first one. Move
4014		 * forward until we find either a) one with a NULL route...
4015		 * insert ahead of that b) one with a different ifp.. insert
4016		 * after that. c) end of the list.. insert at the tail.
4017		 */
4018		struct sctp_nets *netlook;
4019
4020		do {
4021			netlook = TAILQ_NEXT(netfirst, sctp_next);
4022			if (netlook == NULL) {
4023				/* End of the list */
4024				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4025				break;
4026			} else if (netlook->ro.ro_rt == NULL) {
4027				/* next one has NO route */
4028				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4029				break;
4030			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4031				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4032				    net, sctp_next);
4033				break;
4034			}
4035			/* Shift forward */
4036			netfirst = netlook;
4037		} while (netlook != NULL);
4038	}
4039
4040	/* got to have a primary set */
4041	if (stcb->asoc.primary_destination == 0) {
4042		stcb->asoc.primary_destination = net;
4043	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4044		    (net->ro.ro_rt) &&
4045	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4046		/* No route to current primary adopt new primary */
4047		stcb->asoc.primary_destination = net;
4048	}
4049	/* Validate primary is first */
4050	net = TAILQ_FIRST(&stcb->asoc.nets);
4051	if ((net != stcb->asoc.primary_destination) &&
4052	    (stcb->asoc.primary_destination)) {
4053		/*
4054		 * first one on the list is NOT the primary sctp_cmpaddr()
4055		 * is much more efficient if the primary is the first on the
4056		 * list, make it so.
4057		 */
4058		TAILQ_REMOVE(&stcb->asoc.nets,
4059		    stcb->asoc.primary_destination, sctp_next);
4060		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4061		    stcb->asoc.primary_destination, sctp_next);
4062	}
4063	return (0);
4064}
4065
4066
4067static uint32_t
4068sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4069{
4070	uint32_t id;
4071	struct sctpasochead *head;
4072	struct sctp_tcb *lstcb;
4073
4074	SCTP_INP_WLOCK(inp);
4075try_again:
4076	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4077		/* TSNH */
4078		SCTP_INP_WUNLOCK(inp);
4079		return (0);
4080	}
4081	/*
4082	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4083	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4084	 */
4085	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4086		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4087	}
4088	id = inp->sctp_associd_counter;
4089	inp->sctp_associd_counter++;
4090	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4091	if (lstcb) {
4092		goto try_again;
4093	}
4094	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4095	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4096	stcb->asoc.in_asocid_hash = 1;
4097	SCTP_INP_WUNLOCK(inp);
4098	return id;
4099}
4100
4101/*
4102 * allocate an association and add it to the endpoint. The caller must be
4103 * careful to add all additional addresses once they are know right away or
4104 * else the assoc will be may experience a blackout scenario.
4105 */
4106struct sctp_tcb *
4107sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4108    int *error, uint32_t override_tag, uint32_t vrf_id,
4109    struct thread *p
4110)
4111{
4112	/* note the p argument is only valid in unbound sockets */
4113
4114	struct sctp_tcb *stcb;
4115	struct sctp_association *asoc;
4116	struct sctpasochead *head;
4117	uint16_t rport;
4118	int err;
4119
4120	/*
4121	 * Assumption made here: Caller has done a
4122	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4123	 * address does not exist already.
4124	 */
4125	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4126		/* Hit max assoc, sorry no more */
4127		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4128		*error = ENOBUFS;
4129		return (NULL);
4130	}
4131	if (firstaddr == NULL) {
4132		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4133		*error = EINVAL;
4134		return (NULL);
4135	}
4136	SCTP_INP_RLOCK(inp);
4137	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4138	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4139	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4140		/*
4141		 * If its in the TCP pool, its NOT allowed to create an
4142		 * association. The parent listener needs to call
4143		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4144		 * off, or connected one does this.. its an error.
4145		 */
4146		SCTP_INP_RUNLOCK(inp);
4147		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4148		*error = EINVAL;
4149		return (NULL);
4150	}
4151	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4152	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4153		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4154		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4155			SCTP_INP_RUNLOCK(inp);
4156			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4157			*error = EINVAL;
4158			return (NULL);
4159		}
4160	}
4161	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4162#ifdef SCTP_DEBUG
4163	if (firstaddr) {
4164		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4165		switch (firstaddr->sa_family) {
4166#ifdef INET
4167		case AF_INET:
4168			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4169			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4170			break;
4171#endif
4172#ifdef INET6
4173		case AF_INET6:
4174			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4175			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4176			break;
4177#endif
4178		default:
4179			break;
4180		}
4181	} else {
4182		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4183	}
4184#endif				/* SCTP_DEBUG */
4185	switch (firstaddr->sa_family) {
4186#ifdef INET
4187	case AF_INET:
4188		{
4189			struct sockaddr_in *sin;
4190
4191			sin = (struct sockaddr_in *)firstaddr;
4192			if ((ntohs(sin->sin_port) == 0) ||
4193			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4194			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4195			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4196				/* Invalid address */
4197				SCTP_INP_RUNLOCK(inp);
4198				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4199				*error = EINVAL;
4200				return (NULL);
4201			}
4202			rport = sin->sin_port;
4203			break;
4204		}
4205#endif
4206#ifdef INET6
4207	case AF_INET6:
4208		{
4209			struct sockaddr_in6 *sin6;
4210
4211			sin6 = (struct sockaddr_in6 *)firstaddr;
4212			if ((ntohs(sin6->sin6_port) == 0) ||
4213			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4214			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4215				/* Invalid address */
4216				SCTP_INP_RUNLOCK(inp);
4217				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4218				*error = EINVAL;
4219				return (NULL);
4220			}
4221			rport = sin6->sin6_port;
4222			break;
4223		}
4224#endif
4225	default:
4226		/* not supported family type */
4227		SCTP_INP_RUNLOCK(inp);
4228		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4229		*error = EINVAL;
4230		return (NULL);
4231	}
4232	SCTP_INP_RUNLOCK(inp);
4233	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4234		/*
4235		 * If you have not performed a bind, then we need to do the
4236		 * ephemeral bind for you.
4237		 */
4238		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4239		    (struct sockaddr *)NULL,
4240		    (struct sctp_ifa *)NULL,
4241		    p
4242		    ))) {
4243			/* bind error, probably perm */
4244			*error = err;
4245			return (NULL);
4246		}
4247	}
4248	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4249	if (stcb == NULL) {
4250		/* out of memory? */
4251		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4252		*error = ENOMEM;
4253		return (NULL);
4254	}
4255	SCTP_INCR_ASOC_COUNT();
4256
4257	bzero(stcb, sizeof(*stcb));
4258	asoc = &stcb->asoc;
4259
4260	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4261	SCTP_TCB_LOCK_INIT(stcb);
4262	SCTP_TCB_SEND_LOCK_INIT(stcb);
4263	stcb->rport = rport;
4264	/* setup back pointer's */
4265	stcb->sctp_ep = inp;
4266	stcb->sctp_socket = inp->sctp_socket;
4267	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4268		/* failed */
4269		SCTP_TCB_LOCK_DESTROY(stcb);
4270		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4271		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4272		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4273		SCTP_DECR_ASOC_COUNT();
4274		*error = err;
4275		return (NULL);
4276	}
4277	/* and the port */
4278	SCTP_INP_INFO_WLOCK();
4279	SCTP_INP_WLOCK(inp);
4280	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4281		/* inpcb freed while alloc going on */
4282		SCTP_TCB_LOCK_DESTROY(stcb);
4283		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4284		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4285		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4286		SCTP_INP_WUNLOCK(inp);
4287		SCTP_INP_INFO_WUNLOCK();
4288		SCTP_DECR_ASOC_COUNT();
4289		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4290		*error = EINVAL;
4291		return (NULL);
4292	}
4293	SCTP_TCB_LOCK(stcb);
4294
4295	/* now that my_vtag is set, add it to the hash */
4296	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4297	/* put it in the bucket in the vtag hash of assoc's for the system */
4298	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4299	SCTP_INP_INFO_WUNLOCK();
4300
4301	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4302		/* failure.. memory error? */
4303		if (asoc->strmout) {
4304			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4305			asoc->strmout = NULL;
4306		}
4307		if (asoc->mapping_array) {
4308			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4309			asoc->mapping_array = NULL;
4310		}
4311		if (asoc->nr_mapping_array) {
4312			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4313			asoc->nr_mapping_array = NULL;
4314		}
4315		SCTP_DECR_ASOC_COUNT();
4316		SCTP_TCB_LOCK_DESTROY(stcb);
4317		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4318		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4319		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4320		SCTP_INP_WUNLOCK(inp);
4321		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4322		*error = ENOBUFS;
4323		return (NULL);
4324	}
4325	/* Init all the timers */
4326	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4327	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4328	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4329	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4330	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4331	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4332	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4333
4334	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4335	/* now file the port under the hash as well */
4336	if (inp->sctp_tcbhash != NULL) {
4337		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4338		    inp->sctp_hashmark)];
4339		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4340	}
4341	SCTP_INP_WUNLOCK(inp);
4342	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4343	return (stcb);
4344}
4345
4346
4347void
4348sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4349{
4350	struct sctp_association *asoc;
4351
4352	asoc = &stcb->asoc;
4353	asoc->numnets--;
4354	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4355	if (net == asoc->primary_destination) {
4356		/* Reset primary */
4357		struct sctp_nets *lnet;
4358
4359		lnet = TAILQ_FIRST(&asoc->nets);
4360		/*
4361		 * Mobility adaptation Ideally, if deleted destination is
4362		 * the primary, it becomes a fast retransmission trigger by
4363		 * the subsequent SET PRIMARY. (by micchie)
4364		 */
4365		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4366		    SCTP_MOBILITY_BASE) ||
4367		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4368		    SCTP_MOBILITY_FASTHANDOFF)) {
4369			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4370			if (asoc->deleted_primary != NULL) {
4371				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4372				goto out;
4373			}
4374			asoc->deleted_primary = net;
4375			atomic_add_int(&net->ref_count, 1);
4376			memset(&net->lastsa, 0, sizeof(net->lastsa));
4377			memset(&net->lastsv, 0, sizeof(net->lastsv));
4378			sctp_mobility_feature_on(stcb->sctp_ep,
4379			    SCTP_MOBILITY_PRIM_DELETED);
4380			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4381			    stcb->sctp_ep, stcb, NULL);
4382		}
4383out:
4384		/* Try to find a confirmed primary */
4385		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4386	}
4387	if (net == asoc->last_data_chunk_from) {
4388		/* Reset primary */
4389		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4390	}
4391	if (net == asoc->last_control_chunk_from) {
4392		/* Clear net */
4393		asoc->last_control_chunk_from = NULL;
4394	}
4395	if (net == stcb->asoc.alternate) {
4396		sctp_free_remote_addr(stcb->asoc.alternate);
4397		stcb->asoc.alternate = NULL;
4398	}
4399	sctp_free_remote_addr(net);
4400}
4401
4402/*
4403 * remove a remote endpoint address from an association, it will fail if the
4404 * address does not exist.
4405 */
4406int
4407sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4408{
4409	/*
4410	 * Here we need to remove a remote address. This is quite simple, we
4411	 * first find it in the list of address for the association
4412	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4413	 * on that item. Note we do not allow it to be removed if there are
4414	 * no other addresses.
4415	 */
4416	struct sctp_association *asoc;
4417	struct sctp_nets *net, *nnet;
4418
4419	asoc = &stcb->asoc;
4420
4421	/* locate the address */
4422	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4423		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4424			continue;
4425		}
4426		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4427		    remaddr)) {
4428			/* we found the guy */
4429			if (asoc->numnets < 2) {
4430				/* Must have at LEAST two remote addresses */
4431				return (-1);
4432			} else {
4433				sctp_remove_net(stcb, net);
4434				return (0);
4435			}
4436		}
4437	}
4438	/* not found. */
4439	return (-2);
4440}
4441
4442void
4443sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4444{
4445	struct sctpvtaghead *chain;
4446	struct sctp_tagblock *twait_block;
4447	int found = 0;
4448	int i;
4449
4450	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4451	if (!LIST_EMPTY(chain)) {
4452		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4453			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4454				if ((twait_block->vtag_block[i].v_tag == tag) &&
4455				    (twait_block->vtag_block[i].lport == lport) &&
4456				    (twait_block->vtag_block[i].rport == rport)) {
4457					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4458					twait_block->vtag_block[i].v_tag = 0;
4459					twait_block->vtag_block[i].lport = 0;
4460					twait_block->vtag_block[i].rport = 0;
4461					found = 1;
4462					break;
4463				}
4464			}
4465			if (found)
4466				break;
4467		}
4468	}
4469}
4470
4471int
4472sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4473{
4474	struct sctpvtaghead *chain;
4475	struct sctp_tagblock *twait_block;
4476	int found = 0;
4477	int i;
4478
4479	SCTP_INP_INFO_WLOCK();
4480	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4481	if (!LIST_EMPTY(chain)) {
4482		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4483			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4484				if ((twait_block->vtag_block[i].v_tag == tag) &&
4485				    (twait_block->vtag_block[i].lport == lport) &&
4486				    (twait_block->vtag_block[i].rport == rport)) {
4487					found = 1;
4488					break;
4489				}
4490			}
4491			if (found)
4492				break;
4493		}
4494	}
4495	SCTP_INP_INFO_WUNLOCK();
4496	return (found);
4497}
4498
4499
4500void
4501sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4502{
4503	struct sctpvtaghead *chain;
4504	struct sctp_tagblock *twait_block;
4505	struct timeval now;
4506	int set, i;
4507
4508	if (time == 0) {
4509		/* Its disabled */
4510		return;
4511	}
4512	(void)SCTP_GETTIME_TIMEVAL(&now);
4513	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4514	set = 0;
4515	if (!LIST_EMPTY(chain)) {
4516		/* Block(s) present, lets find space, and expire on the fly */
4517		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4518			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4519				if ((twait_block->vtag_block[i].v_tag == 0) &&
4520				    !set) {
4521					twait_block->vtag_block[i].tv_sec_at_expire =
4522					    now.tv_sec + time;
4523					twait_block->vtag_block[i].v_tag = tag;
4524					twait_block->vtag_block[i].lport = lport;
4525					twait_block->vtag_block[i].rport = rport;
4526					set = 1;
4527				} else if ((twait_block->vtag_block[i].v_tag) &&
4528				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4529					/* Audit expires this guy */
4530					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4531					twait_block->vtag_block[i].v_tag = 0;
4532					twait_block->vtag_block[i].lport = 0;
4533					twait_block->vtag_block[i].rport = 0;
4534					if (set == 0) {
4535						/* Reuse it for my new tag */
4536						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4537						twait_block->vtag_block[i].v_tag = tag;
4538						twait_block->vtag_block[i].lport = lport;
4539						twait_block->vtag_block[i].rport = rport;
4540						set = 1;
4541					}
4542				}
4543			}
4544			if (set) {
4545				/*
4546				 * We only do up to the block where we can
4547				 * place our tag for audits
4548				 */
4549				break;
4550			}
4551		}
4552	}
4553	/* Need to add a new block to chain */
4554	if (!set) {
4555		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4556		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4557		if (twait_block == NULL) {
4558#ifdef INVARIANTS
4559			panic("Can not alloc tagblock");
4560#endif
4561			return;
4562		}
4563		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4564		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4565		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4566		twait_block->vtag_block[0].v_tag = tag;
4567		twait_block->vtag_block[0].lport = lport;
4568		twait_block->vtag_block[0].rport = rport;
4569	}
4570}
4571
4572
4573
4574/*-
4575 * Free the association after un-hashing the remote port. This
4576 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4577 * expect that the input to this function IS a locked TCB.
4578 * It will return 0, if it did NOT destroy the association (instead
4579 * it unlocks it. It will return NON-zero if it either destroyed the
4580 * association OR the association is already destroyed.
4581 */
4582int
4583sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4584{
4585	int i;
4586	struct sctp_association *asoc;
4587	struct sctp_nets *net, *nnet;
4588	struct sctp_laddr *laddr, *naddr;
4589	struct sctp_tmit_chunk *chk, *nchk;
4590	struct sctp_asconf_addr *aparam, *naparam;
4591	struct sctp_asconf_ack *aack, *naack;
4592	struct sctp_stream_reset_list *strrst, *nstrrst;
4593	struct sctp_queued_to_read *sq, *nsq;
4594	struct sctp_stream_queue_pending *sp, *nsp;
4595	sctp_sharedkey_t *shared_key, *nshared_key;
4596	struct socket *so;
4597
4598	/* first, lets purge the entry from the hash table. */
4599
4600#ifdef SCTP_LOG_CLOSING
4601	sctp_log_closing(inp, stcb, 6);
4602#endif
4603	if (stcb->asoc.state == 0) {
4604#ifdef SCTP_LOG_CLOSING
4605		sctp_log_closing(inp, NULL, 7);
4606#endif
4607		/* there is no asoc, really TSNH :-0 */
4608		return (1);
4609	}
4610	if (stcb->asoc.alternate) {
4611		sctp_free_remote_addr(stcb->asoc.alternate);
4612		stcb->asoc.alternate = NULL;
4613	}
4614	/* TEMP CODE */
4615	if (stcb->freed_from_where == 0) {
4616		/* Only record the first place free happened from */
4617		stcb->freed_from_where = from_location;
4618	}
4619	/* TEMP CODE */
4620
4621	asoc = &stcb->asoc;
4622	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4623	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4624		/* nothing around */
4625		so = NULL;
4626	else
4627		so = inp->sctp_socket;
4628
4629	/*
4630	 * We used timer based freeing if a reader or writer is in the way.
4631	 * So we first check if we are actually being called from a timer,
4632	 * if so we abort early if a reader or writer is still in the way.
4633	 */
4634	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4635	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4636		/*
4637		 * is it the timer driving us? if so are the reader/writers
4638		 * gone?
4639		 */
4640		if (stcb->asoc.refcnt) {
4641			/* nope, reader or writer in the way */
4642			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4643			/* no asoc destroyed */
4644			SCTP_TCB_UNLOCK(stcb);
4645#ifdef SCTP_LOG_CLOSING
4646			sctp_log_closing(inp, stcb, 8);
4647#endif
4648			return (0);
4649		}
4650	}
4651	/* now clean up any other timers */
4652	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4653	asoc->dack_timer.self = NULL;
4654	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4655	/*-
4656	 * For stream reset we don't blast this unless
4657	 * it is a str-reset timer, it might be the
4658	 * free-asoc timer which we DON'T want to
4659	 * disturb.
4660	 */
4661	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4662		asoc->strreset_timer.self = NULL;
4663	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4664	asoc->asconf_timer.self = NULL;
4665	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4666	asoc->autoclose_timer.self = NULL;
4667	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4668	asoc->shut_guard_timer.self = NULL;
4669	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4670	asoc->delayed_event_timer.self = NULL;
4671	/* Mobility adaptation */
4672	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4673	asoc->delete_prim_timer.self = NULL;
4674	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4675		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4676		net->rxt_timer.self = NULL;
4677		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4678		net->pmtu_timer.self = NULL;
4679		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4680		net->hb_timer.self = NULL;
4681	}
4682	/* Now the read queue needs to be cleaned up (only once) */
4683	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4684		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4685		SCTP_INP_READ_LOCK(inp);
4686		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4687			if (sq->stcb == stcb) {
4688				sq->do_not_ref_stcb = 1;
4689				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4690				/*
4691				 * If there is no end, there never will be
4692				 * now.
4693				 */
4694				if (sq->end_added == 0) {
4695					/* Held for PD-API clear that. */
4696					sq->pdapi_aborted = 1;
4697					sq->held_length = 0;
4698					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4699						/*
4700						 * Need to add a PD-API
4701						 * aborted indication.
4702						 * Setting the control_pdapi
4703						 * assures that it will be
4704						 * added right after this
4705						 * msg.
4706						 */
4707						uint32_t strseq;
4708
4709						stcb->asoc.control_pdapi = sq;
4710						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4711						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4712						    stcb,
4713						    SCTP_PARTIAL_DELIVERY_ABORTED,
4714						    (void *)&strseq,
4715						    SCTP_SO_LOCKED);
4716						stcb->asoc.control_pdapi = NULL;
4717					}
4718				}
4719				/* Add an end to wake them */
4720				sq->end_added = 1;
4721			}
4722		}
4723		SCTP_INP_READ_UNLOCK(inp);
4724		if (stcb->block_entry) {
4725			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4726			stcb->block_entry->error = ECONNRESET;
4727			stcb->block_entry = NULL;
4728		}
4729	}
4730	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4731		/*
4732		 * Someone holds a reference OR the socket is unaccepted
4733		 * yet.
4734		 */
4735		if ((stcb->asoc.refcnt) ||
4736		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4737		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4738			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4739			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4740		}
4741		SCTP_TCB_UNLOCK(stcb);
4742		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4743		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4744			/* nothing around */
4745			so = NULL;
4746		if (so) {
4747			/* Wake any reader/writers */
4748			sctp_sorwakeup(inp, so);
4749			sctp_sowwakeup(inp, so);
4750		}
4751#ifdef SCTP_LOG_CLOSING
4752		sctp_log_closing(inp, stcb, 9);
4753#endif
4754		/* no asoc destroyed */
4755		return (0);
4756	}
4757#ifdef SCTP_LOG_CLOSING
4758	sctp_log_closing(inp, stcb, 10);
4759#endif
4760	/*
4761	 * When I reach here, no others want to kill the assoc yet.. and I
4762	 * own the lock. Now its possible an abort comes in when I do the
4763	 * lock exchange below to grab all the locks to do the final take
4764	 * out. to prevent this we increment the count, which will start a
4765	 * timer and blow out above thus assuring us that we hold exclusive
4766	 * killing of the asoc. Note that after getting back the TCB lock we
4767	 * will go ahead and increment the counter back up and stop any
4768	 * timer a passing stranger may have started :-S
4769	 */
4770	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4771		atomic_add_int(&stcb->asoc.refcnt, 1);
4772
4773		SCTP_TCB_UNLOCK(stcb);
4774		SCTP_INP_INFO_WLOCK();
4775		SCTP_INP_WLOCK(inp);
4776		SCTP_TCB_LOCK(stcb);
4777	}
4778	/* Double check the GONE flag */
4779	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4780	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4781		/* nothing around */
4782		so = NULL;
4783
4784	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4785	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4786		/*
4787		 * For TCP type we need special handling when we are
4788		 * connected. We also include the peel'ed off ones to.
4789		 */
4790		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4791			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4792			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4793			if (so) {
4794				SOCK_LOCK(so);
4795				if (so->so_rcv.sb_cc == 0) {
4796					so->so_state &= ~(SS_ISCONNECTING |
4797					    SS_ISDISCONNECTING |
4798					    SS_ISCONFIRMING |
4799					    SS_ISCONNECTED);
4800				}
4801				socantrcvmore_locked(so);
4802				sctp_sowwakeup(inp, so);
4803				sctp_sorwakeup(inp, so);
4804				SCTP_SOWAKEUP(so);
4805			}
4806		}
4807	}
4808	/*
4809	 * Make it invalid too, that way if its about to run it will abort
4810	 * and return.
4811	 */
4812	/* re-increment the lock */
4813	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4814		atomic_add_int(&stcb->asoc.refcnt, -1);
4815	}
4816	if (stcb->asoc.refcnt) {
4817		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4818		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4819		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4820			SCTP_INP_INFO_WUNLOCK();
4821			SCTP_INP_WUNLOCK(inp);
4822		}
4823		SCTP_TCB_UNLOCK(stcb);
4824		return (0);
4825	}
4826	asoc->state = 0;
4827	if (inp->sctp_tcbhash) {
4828		LIST_REMOVE(stcb, sctp_tcbhash);
4829	}
4830	if (stcb->asoc.in_asocid_hash) {
4831		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4832	}
4833	/* Now lets remove it from the list of ALL associations in the EP */
4834	LIST_REMOVE(stcb, sctp_tcblist);
4835	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4836		SCTP_INP_INCR_REF(inp);
4837		SCTP_INP_WUNLOCK(inp);
4838	}
4839	/* pull from vtag hash */
4840	LIST_REMOVE(stcb, sctp_asocs);
4841	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4842	    inp->sctp_lport, stcb->rport);
4843
4844	/*
4845	 * Now restop the timers to be sure this is paranoia at is finest!
4846	 */
4847	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4848	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4849	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4850	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4851	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4852	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4853	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4854	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4855		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4856		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4857		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4858	}
4859
4860	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4861	/*
4862	 * The chunk lists and such SHOULD be empty but we check them just
4863	 * in case.
4864	 */
4865	/* anything on the wheel needs to be removed */
4866	for (i = 0; i < asoc->streamoutcnt; i++) {
4867		struct sctp_stream_out *outs;
4868
4869		outs = &asoc->strmout[i];
4870		/* now clean up any chunks here */
4871		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4872			TAILQ_REMOVE(&outs->outqueue, sp, next);
4873			sctp_free_spbufspace(stcb, asoc, sp);
4874			if (sp->data) {
4875				if (so) {
4876					/* Still an open socket - report */
4877					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4878					    0, (void *)sp, SCTP_SO_LOCKED);
4879				}
4880				if (sp->data) {
4881					sctp_m_freem(sp->data);
4882					sp->data = NULL;
4883					sp->tail_mbuf = NULL;
4884					sp->length = 0;
4885				}
4886			}
4887			if (sp->net) {
4888				sctp_free_remote_addr(sp->net);
4889				sp->net = NULL;
4890			}
4891			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4892		}
4893	}
4894	/* sa_ignore FREED_MEMORY */
4895	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4896		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4897		SCTP_FREE(strrst, SCTP_M_STRESET);
4898	}
4899	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4900		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4901		if (sq->data) {
4902			sctp_m_freem(sq->data);
4903			sq->data = NULL;
4904		}
4905		sctp_free_remote_addr(sq->whoFrom);
4906		sq->whoFrom = NULL;
4907		sq->stcb = NULL;
4908		/* Free the ctl entry */
4909		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4910		SCTP_DECR_READQ_COUNT();
4911		/* sa_ignore FREED_MEMORY */
4912	}
4913	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4914		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4915		if (chk->data) {
4916			sctp_m_freem(chk->data);
4917			chk->data = NULL;
4918		}
4919		if (chk->holds_key_ref)
4920			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4921		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4922		SCTP_DECR_CHK_COUNT();
4923		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4924		asoc->free_chunk_cnt--;
4925		/* sa_ignore FREED_MEMORY */
4926	}
4927	/* pending send queue SHOULD be empty */
4928	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4929		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4930			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4931#ifdef INVARIANTS
4932		} else {
4933			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4934#endif
4935		}
4936		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4937		if (chk->data) {
4938			if (so) {
4939				/* Still a socket? */
4940				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4941				    0, chk, SCTP_SO_LOCKED);
4942			}
4943			if (chk->data) {
4944				sctp_m_freem(chk->data);
4945				chk->data = NULL;
4946			}
4947		}
4948		if (chk->holds_key_ref)
4949			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4950		if (chk->whoTo) {
4951			sctp_free_remote_addr(chk->whoTo);
4952			chk->whoTo = NULL;
4953		}
4954		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4955		SCTP_DECR_CHK_COUNT();
4956		/* sa_ignore FREED_MEMORY */
4957	}
4958	/* sent queue SHOULD be empty */
4959	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4960		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
4961			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4962				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4963#ifdef INVARIANTS
4964			} else {
4965				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4966#endif
4967			}
4968		}
4969		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4970		if (chk->data) {
4971			if (so) {
4972				/* Still a socket? */
4973				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
4974				    0, chk, SCTP_SO_LOCKED);
4975			}
4976			if (chk->data) {
4977				sctp_m_freem(chk->data);
4978				chk->data = NULL;
4979			}
4980		}
4981		if (chk->holds_key_ref)
4982			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4983		sctp_free_remote_addr(chk->whoTo);
4984		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4985		SCTP_DECR_CHK_COUNT();
4986		/* sa_ignore FREED_MEMORY */
4987	}
4988#ifdef INVARIANTS
4989	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
4990		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
4991			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
4992		}
4993	}
4994#endif
4995	/* control queue MAY not be empty */
4996	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
4997		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4998		if (chk->data) {
4999			sctp_m_freem(chk->data);
5000			chk->data = NULL;
5001		}
5002		if (chk->holds_key_ref)
5003			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5004		sctp_free_remote_addr(chk->whoTo);
5005		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5006		SCTP_DECR_CHK_COUNT();
5007		/* sa_ignore FREED_MEMORY */
5008	}
5009	/* ASCONF queue MAY not be empty */
5010	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
5011		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
5012		if (chk->data) {
5013			sctp_m_freem(chk->data);
5014			chk->data = NULL;
5015		}
5016		if (chk->holds_key_ref)
5017			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5018		sctp_free_remote_addr(chk->whoTo);
5019		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5020		SCTP_DECR_CHK_COUNT();
5021		/* sa_ignore FREED_MEMORY */
5022	}
5023	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5024		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5025		if (chk->data) {
5026			sctp_m_freem(chk->data);
5027			chk->data = NULL;
5028		}
5029		if (chk->holds_key_ref)
5030			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5031		sctp_free_remote_addr(chk->whoTo);
5032		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5033		SCTP_DECR_CHK_COUNT();
5034		/* sa_ignore FREED_MEMORY */
5035	}
5036
5037	if (asoc->mapping_array) {
5038		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5039		asoc->mapping_array = NULL;
5040	}
5041	if (asoc->nr_mapping_array) {
5042		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5043		asoc->nr_mapping_array = NULL;
5044	}
5045	/* the stream outs */
5046	if (asoc->strmout) {
5047		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5048		asoc->strmout = NULL;
5049	}
5050	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5051	if (asoc->strmin) {
5052		struct sctp_queued_to_read *ctl, *nctl;
5053
5054		for (i = 0; i < asoc->streamincnt; i++) {
5055			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5056				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5057				sctp_free_remote_addr(ctl->whoFrom);
5058				if (ctl->data) {
5059					sctp_m_freem(ctl->data);
5060					ctl->data = NULL;
5061				}
5062				/*
5063				 * We don't free the address here since all
5064				 * the net's were freed above.
5065				 */
5066				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5067				SCTP_DECR_READQ_COUNT();
5068			}
5069		}
5070		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5071		asoc->strmin = NULL;
5072	}
5073	asoc->streamincnt = 0;
5074	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5075#ifdef INVARIANTS
5076		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5077			panic("no net's left alloc'ed, or list points to itself");
5078		}
5079#endif
5080		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5081		sctp_free_remote_addr(net);
5082	}
5083	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5084		/* sa_ignore FREED_MEMORY */
5085		sctp_remove_laddr(laddr);
5086	}
5087
5088	/* pending asconf (address) parameters */
5089	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5090		/* sa_ignore FREED_MEMORY */
5091		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5092		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5093	}
5094	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5095		/* sa_ignore FREED_MEMORY */
5096		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5097		if (aack->data != NULL) {
5098			sctp_m_freem(aack->data);
5099		}
5100		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5101	}
5102	/* clean up auth stuff */
5103	if (asoc->local_hmacs)
5104		sctp_free_hmaclist(asoc->local_hmacs);
5105	if (asoc->peer_hmacs)
5106		sctp_free_hmaclist(asoc->peer_hmacs);
5107
5108	if (asoc->local_auth_chunks)
5109		sctp_free_chunklist(asoc->local_auth_chunks);
5110	if (asoc->peer_auth_chunks)
5111		sctp_free_chunklist(asoc->peer_auth_chunks);
5112
5113	sctp_free_authinfo(&asoc->authinfo);
5114
5115	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5116		LIST_REMOVE(shared_key, next);
5117		sctp_free_sharedkey(shared_key);
5118		/* sa_ignore FREED_MEMORY */
5119	}
5120
5121	/* Insert new items here :> */
5122
5123	/* Get rid of LOCK */
5124	SCTP_TCB_LOCK_DESTROY(stcb);
5125	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5126	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5127		SCTP_INP_INFO_WUNLOCK();
5128		SCTP_INP_RLOCK(inp);
5129	}
5130#ifdef SCTP_TRACK_FREED_ASOCS
5131	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5132		/* now clean up the tasoc itself */
5133		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5134		SCTP_DECR_ASOC_COUNT();
5135	} else {
5136		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5137	}
5138#else
5139	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5140	SCTP_DECR_ASOC_COUNT();
5141#endif
5142	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5143		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5144			/*
5145			 * If its NOT the inp_free calling us AND sctp_close
5146			 * as been called, we call back...
5147			 */
5148			SCTP_INP_RUNLOCK(inp);
5149			/*
5150			 * This will start the kill timer (if we are the
5151			 * last one) since we hold an increment yet. But
5152			 * this is the only safe way to do this since
5153			 * otherwise if the socket closes at the same time
5154			 * we are here we might collide in the cleanup.
5155			 */
5156			sctp_inpcb_free(inp,
5157			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5158			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5159			SCTP_INP_DECR_REF(inp);
5160			goto out_of;
5161		} else {
5162			/* The socket is still open. */
5163			SCTP_INP_DECR_REF(inp);
5164		}
5165	}
5166	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5167		SCTP_INP_RUNLOCK(inp);
5168	}
5169out_of:
5170	/* destroyed the asoc */
5171#ifdef SCTP_LOG_CLOSING
5172	sctp_log_closing(inp, NULL, 11);
5173#endif
5174	return (1);
5175}
5176
5177
5178
5179/*
5180 * determine if a destination is "reachable" based upon the addresses bound
5181 * to the current endpoint (e.g. only v4 or v6 currently bound)
5182 */
5183/*
5184 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5185 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5186 * types bound as its endpoint
5187 */
5188int
5189sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5190{
5191	struct sctp_inpcb *inp;
5192	int answer;
5193
5194	/*
5195	 * No locks here, the TCB, in all cases is already locked and an
5196	 * assoc is up. There is either a INP lock by the caller applied (in
5197	 * asconf case when deleting an address) or NOT in the HB case,
5198	 * however if HB then the INP increment is up and the INP will not
5199	 * be removed (on top of the fact that we have a TCB lock). So we
5200	 * only want to read the sctp_flags, which is either bound-all or
5201	 * not.. no protection needed since once an assoc is up you can't be
5202	 * changing your binding.
5203	 */
5204	inp = stcb->sctp_ep;
5205	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5206		/* if bound all, destination is not restricted */
5207		/*
5208		 * RRS: Question during lock work: Is this correct? If you
5209		 * are bound-all you still might need to obey the V4--V6
5210		 * flags??? IMO this bound-all stuff needs to be removed!
5211		 */
5212		return (1);
5213	}
5214	/* NOTE: all "scope" checks are done when local addresses are added */
5215	switch (destaddr->sa_family) {
5216#ifdef INET6
5217	case AF_INET6:
5218		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5219		break;
5220#endif
5221#ifdef INET
5222	case AF_INET:
5223		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5224		break;
5225#endif
5226	default:
5227		/* invalid family, so it's unreachable */
5228		answer = 0;
5229		break;
5230	}
5231	return (answer);
5232}
5233
5234/*
5235 * update the inp_vflags on an endpoint
5236 */
5237static void
5238sctp_update_ep_vflag(struct sctp_inpcb *inp)
5239{
5240	struct sctp_laddr *laddr;
5241
5242	/* first clear the flag */
5243	inp->ip_inp.inp.inp_vflag = 0;
5244	/* set the flag based on addresses on the ep list */
5245	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5246		if (laddr->ifa == NULL) {
5247			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5248			    __FUNCTION__);
5249			continue;
5250		}
5251		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5252			continue;
5253		}
5254		switch (laddr->ifa->address.sa.sa_family) {
5255#ifdef INET6
5256		case AF_INET6:
5257			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5258			break;
5259#endif
5260#ifdef INET
5261		case AF_INET:
5262			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5263			break;
5264#endif
5265		default:
5266			break;
5267		}
5268	}
5269}
5270
5271/*
5272 * Add the address to the endpoint local address list There is nothing to be
5273 * done if we are bound to all addresses
5274 */
5275void
5276sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5277{
5278	struct sctp_laddr *laddr;
5279	int fnd, error = 0;
5280
5281	fnd = 0;
5282
5283	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5284		/* You are already bound to all. You have it already */
5285		return;
5286	}
5287#ifdef INET6
5288	if (ifa->address.sa.sa_family == AF_INET6) {
5289		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5290			/* Can't bind a non-useable addr. */
5291			return;
5292		}
5293	}
5294#endif
5295	/* first, is it already present? */
5296	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5297		if (laddr->ifa == ifa) {
5298			fnd = 1;
5299			break;
5300		}
5301	}
5302
5303	if (fnd == 0) {
5304		/* Not in the ep list */
5305		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5306		if (error != 0)
5307			return;
5308		inp->laddr_count++;
5309		/* update inp_vflag flags */
5310		switch (ifa->address.sa.sa_family) {
5311#ifdef INET6
5312		case AF_INET6:
5313			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5314			break;
5315#endif
5316#ifdef INET
5317		case AF_INET:
5318			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5319			break;
5320#endif
5321		default:
5322			break;
5323		}
5324	}
5325	return;
5326}
5327
5328
5329/*
5330 * select a new (hopefully reachable) destination net (should only be used
5331 * when we deleted an ep addr that is the only usable source address to reach
5332 * the destination net)
5333 */
5334static void
5335sctp_select_primary_destination(struct sctp_tcb *stcb)
5336{
5337	struct sctp_nets *net;
5338
5339	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5340		/* for now, we'll just pick the first reachable one we find */
5341		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5342			continue;
5343		if (sctp_destination_is_reachable(stcb,
5344		    (struct sockaddr *)&net->ro._l_addr)) {
5345			/* found a reachable destination */
5346			stcb->asoc.primary_destination = net;
5347		}
5348	}
5349	/* I can't there from here! ...we're gonna die shortly... */
5350}
5351
5352
5353/*
5354 * Delete the address from the endpoint local address list There is nothing
5355 * to be done if we are bound to all addresses
5356 */
5357void
5358sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5359{
5360	struct sctp_laddr *laddr;
5361	int fnd;
5362
5363	fnd = 0;
5364	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5365		/* You are already bound to all. You have it already */
5366		return;
5367	}
5368	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5369		if (laddr->ifa == ifa) {
5370			fnd = 1;
5371			break;
5372		}
5373	}
5374	if (fnd && (inp->laddr_count < 2)) {
5375		/* can't delete unless there are at LEAST 2 addresses */
5376		return;
5377	}
5378	if (fnd) {
5379		/*
5380		 * clean up any use of this address go through our
5381		 * associations and clear any last_used_address that match
5382		 * this one for each assoc, see if a new primary_destination
5383		 * is needed
5384		 */
5385		struct sctp_tcb *stcb;
5386
5387		/* clean up "next_addr_touse" */
5388		if (inp->next_addr_touse == laddr)
5389			/* delete this address */
5390			inp->next_addr_touse = NULL;
5391
5392		/* clean up "last_used_address" */
5393		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5394			struct sctp_nets *net;
5395
5396			SCTP_TCB_LOCK(stcb);
5397			if (stcb->asoc.last_used_address == laddr)
5398				/* delete this address */
5399				stcb->asoc.last_used_address = NULL;
5400			/*
5401			 * Now spin through all the nets and purge any ref
5402			 * to laddr
5403			 */
5404			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5405				if (net->ro._s_addr &&
5406				    (net->ro._s_addr->ifa == laddr->ifa)) {
5407					/* Yep, purge src address selected */
5408					sctp_rtentry_t *rt;
5409
5410					/* delete this address if cached */
5411					rt = net->ro.ro_rt;
5412					if (rt != NULL) {
5413						RTFREE(rt);
5414						net->ro.ro_rt = NULL;
5415					}
5416					sctp_free_ifa(net->ro._s_addr);
5417					net->ro._s_addr = NULL;
5418					net->src_addr_selected = 0;
5419				}
5420			}
5421			SCTP_TCB_UNLOCK(stcb);
5422		}		/* for each tcb */
5423		/* remove it from the ep list */
5424		sctp_remove_laddr(laddr);
5425		inp->laddr_count--;
5426		/* update inp_vflag flags */
5427		sctp_update_ep_vflag(inp);
5428	}
5429	return;
5430}
5431
5432/*
5433 * Add the address to the TCB local address restricted list.
5434 * This is a "pending" address list (eg. addresses waiting for an
5435 * ASCONF-ACK response) and cannot be used as a valid source address.
5436 */
5437void
5438sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5439{
5440	struct sctp_laddr *laddr;
5441	struct sctpladdr *list;
5442
5443	/*
5444	 * Assumes TCB is locked.. and possibly the INP. May need to
5445	 * confirm/fix that if we need it and is not the case.
5446	 */
5447	list = &stcb->asoc.sctp_restricted_addrs;
5448
5449#ifdef INET6
5450	if (ifa->address.sa.sa_family == AF_INET6) {
5451		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5452			/* Can't bind a non-existent addr. */
5453			return;
5454		}
5455	}
5456#endif
5457	/* does the address already exist? */
5458	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5459		if (laddr->ifa == ifa) {
5460			return;
5461		}
5462	}
5463
5464	/* add to the list */
5465	(void)sctp_insert_laddr(list, ifa, 0);
5466	return;
5467}
5468
5469/*
5470 * insert an laddr entry with the given ifa for the desired list
5471 */
5472int
5473sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5474{
5475	struct sctp_laddr *laddr;
5476
5477	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5478	if (laddr == NULL) {
5479		/* out of memory? */
5480		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5481		return (EINVAL);
5482	}
5483	SCTP_INCR_LADDR_COUNT();
5484	bzero(laddr, sizeof(*laddr));
5485	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5486	laddr->ifa = ifa;
5487	laddr->action = act;
5488	atomic_add_int(&ifa->refcount, 1);
5489	/* insert it */
5490	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5491
5492	return (0);
5493}
5494
5495/*
5496 * Remove an laddr entry from the local address list (on an assoc)
5497 */
5498void
5499sctp_remove_laddr(struct sctp_laddr *laddr)
5500{
5501
5502	/* remove from the list */
5503	LIST_REMOVE(laddr, sctp_nxt_addr);
5504	sctp_free_ifa(laddr->ifa);
5505	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5506	SCTP_DECR_LADDR_COUNT();
5507}
5508
5509/*
5510 * Remove a local address from the TCB local address restricted list
5511 */
5512void
5513sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5514{
5515	struct sctp_inpcb *inp;
5516	struct sctp_laddr *laddr;
5517
5518	/*
5519	 * This is called by asconf work. It is assumed that a) The TCB is
5520	 * locked and b) The INP is locked. This is true in as much as I can
5521	 * trace through the entry asconf code where I did these locks.
5522	 * Again, the ASCONF code is a bit different in that it does lock
5523	 * the INP during its work often times. This must be since we don't
5524	 * want other proc's looking up things while what they are looking
5525	 * up is changing :-D
5526	 */
5527
5528	inp = stcb->sctp_ep;
5529	/* if subset bound and don't allow ASCONF's, can't delete last */
5530	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5531	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5532		if (stcb->sctp_ep->laddr_count < 2) {
5533			/* can't delete last address */
5534			return;
5535		}
5536	}
5537	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5538		/* remove the address if it exists */
5539		if (laddr->ifa == NULL)
5540			continue;
5541		if (laddr->ifa == ifa) {
5542			sctp_remove_laddr(laddr);
5543			return;
5544		}
5545	}
5546
5547	/* address not found! */
5548	return;
5549}
5550
5551/*
5552 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5553 */
5554/* sysctl */
5555static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5556static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5557
5558
5559
5560#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5561struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5562int *sctp_cpuarry = NULL;
5563void
5564sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5565{
5566	/* Queue a packet to a processor for the specified core */
5567	struct sctp_mcore_queue *qent;
5568	struct sctp_mcore_ctrl *wkq;
5569	int need_wake = 0;
5570
5571	if (sctp_mcore_workers == NULL) {
5572		/* Something went way bad during setup */
5573		sctp_input_with_port(m, off, 0);
5574		return;
5575	}
5576	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5577	    (sizeof(struct sctp_mcore_queue)),
5578	    SCTP_M_MCORE);
5579	if (qent == NULL) {
5580		/* This is trouble  */
5581		sctp_input_with_port(m, off, 0);
5582		return;
5583	}
5584	qent->vn = curvnet;
5585	qent->m = m;
5586	qent->off = off;
5587	qent->v6 = 0;
5588	wkq = &sctp_mcore_workers[cpu_to_use];
5589	SCTP_MCORE_QLOCK(wkq);
5590
5591	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5592	if (wkq->running == 0) {
5593		need_wake = 1;
5594	}
5595	SCTP_MCORE_QUNLOCK(wkq);
5596	if (need_wake) {
5597		wakeup(&wkq->running);
5598	}
5599}
5600
5601static void
5602sctp_mcore_thread(void *arg)
5603{
5604
5605	struct sctp_mcore_ctrl *wkq;
5606	struct sctp_mcore_queue *qent;
5607
5608	wkq = (struct sctp_mcore_ctrl *)arg;
5609	struct mbuf *m;
5610	int off, v6;
5611
5612	/* Wait for first tickle */
5613	SCTP_MCORE_LOCK(wkq);
5614	wkq->running = 0;
5615	msleep(&wkq->running,
5616	    &wkq->core_mtx,
5617	    0, "wait for pkt", 0);
5618	SCTP_MCORE_UNLOCK(wkq);
5619
5620	/* Bind to our cpu */
5621	thread_lock(curthread);
5622	sched_bind(curthread, wkq->cpuid);
5623	thread_unlock(curthread);
5624
5625	/* Now lets start working */
5626	SCTP_MCORE_LOCK(wkq);
5627	/* Now grab lock and go */
5628	for (;;) {
5629		SCTP_MCORE_QLOCK(wkq);
5630skip_sleep:
5631		wkq->running = 1;
5632		qent = TAILQ_FIRST(&wkq->que);
5633		if (qent) {
5634			TAILQ_REMOVE(&wkq->que, qent, next);
5635			SCTP_MCORE_QUNLOCK(wkq);
5636			CURVNET_SET(qent->vn);
5637			m = qent->m;
5638			off = qent->off;
5639			v6 = qent->v6;
5640			SCTP_FREE(qent, SCTP_M_MCORE);
5641			if (v6 == 0) {
5642				sctp_input_with_port(m, off, 0);
5643			} else {
5644				SCTP_PRINTF("V6 not yet supported\n");
5645				sctp_m_freem(m);
5646			}
5647			CURVNET_RESTORE();
5648			SCTP_MCORE_QLOCK(wkq);
5649		}
5650		wkq->running = 0;
5651		if (!TAILQ_EMPTY(&wkq->que)) {
5652			goto skip_sleep;
5653		}
5654		SCTP_MCORE_QUNLOCK(wkq);
5655		msleep(&wkq->running,
5656		    &wkq->core_mtx,
5657		    0, "wait for pkt", 0);
5658	}
5659}
5660
5661static void
5662sctp_startup_mcore_threads(void)
5663{
5664	int i, cpu;
5665
5666	if (mp_ncpus == 1)
5667		return;
5668
5669	if (sctp_mcore_workers != NULL) {
5670		/*
5671		 * Already been here in some previous vnet?
5672		 */
5673		return;
5674	}
5675	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5676	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5677	    SCTP_M_MCORE);
5678	if (sctp_mcore_workers == NULL) {
5679		/* TSNH I hope */
5680		return;
5681	}
5682	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5683	    sizeof(struct sctp_mcore_ctrl)));
5684	/* Init the structures */
5685	for (i = 0; i <= mp_maxid; i++) {
5686		TAILQ_INIT(&sctp_mcore_workers[i].que);
5687		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5688		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5689		sctp_mcore_workers[i].cpuid = i;
5690	}
5691	if (sctp_cpuarry == NULL) {
5692		SCTP_MALLOC(sctp_cpuarry, int *,
5693		    (mp_ncpus * sizeof(int)),
5694		    SCTP_M_MCORE);
5695		i = 0;
5696		CPU_FOREACH(cpu) {
5697			sctp_cpuarry[i] = cpu;
5698			i++;
5699		}
5700	}
5701	/* Now start them all */
5702	CPU_FOREACH(cpu) {
5703		(void)kproc_create(sctp_mcore_thread,
5704		    (void *)&sctp_mcore_workers[cpu],
5705		    &sctp_mcore_workers[cpu].thread_proc,
5706		    RFPROC,
5707		    SCTP_KTHREAD_PAGES,
5708		    SCTP_MCORE_NAME);
5709
5710	}
5711}
5712
5713#endif
5714
5715void
5716sctp_pcb_init()
5717{
5718	/*
5719	 * SCTP initialization for the PCB structures should be called by
5720	 * the sctp_init() funciton.
5721	 */
5722	int i;
5723	struct timeval tv;
5724
5725	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5726		/* error I was called twice */
5727		return;
5728	}
5729	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5730
5731#if defined(SCTP_LOCAL_TRACE_BUF)
5732	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5733#endif
5734#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5735	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5736	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5737	    SCTP_M_MCORE);
5738#endif
5739	(void)SCTP_GETTIME_TIMEVAL(&tv);
5740#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5741	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5742	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5743	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5744#else
5745	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5746	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5747	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5748#endif
5749	/* init the empty list of (All) Endpoints */
5750	LIST_INIT(&SCTP_BASE_INFO(listhead));
5751
5752
5753	/* init the hash table of endpoints */
5754	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5755	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5756	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5757	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5758	    &SCTP_BASE_INFO(hashasocmark));
5759	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5760	    &SCTP_BASE_INFO(hashmark));
5761	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5762	    &SCTP_BASE_INFO(hashtcpmark));
5763	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5764
5765
5766	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5767	    &SCTP_BASE_INFO(hashvrfmark));
5768
5769	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5770	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5771	/* init the zones */
5772	/*
5773	 * FIX ME: Should check for NULL returns, but if it does fail we are
5774	 * doomed to panic anyways... add later maybe.
5775	 */
5776	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5777	    sizeof(struct sctp_inpcb), maxsockets);
5778
5779	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5780	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5781
5782	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5783	    sizeof(struct sctp_laddr),
5784	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5785
5786	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5787	    sizeof(struct sctp_nets),
5788	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5789
5790	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5791	    sizeof(struct sctp_tmit_chunk),
5792	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5793
5794	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5795	    sizeof(struct sctp_queued_to_read),
5796	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5797
5798	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5799	    sizeof(struct sctp_stream_queue_pending),
5800	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5801
5802	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5803	    sizeof(struct sctp_asconf),
5804	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5805
5806	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5807	    sizeof(struct sctp_asconf_ack),
5808	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5809
5810
5811	/* Master Lock INIT for info structure */
5812	SCTP_INP_INFO_LOCK_INIT();
5813	SCTP_STATLOG_INIT_LOCK();
5814
5815	SCTP_IPI_COUNT_INIT();
5816	SCTP_IPI_ADDR_INIT();
5817#ifdef SCTP_PACKET_LOGGING
5818	SCTP_IP_PKTLOG_INIT();
5819#endif
5820	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5821
5822	SCTP_WQ_ADDR_INIT();
5823	/* not sure if we need all the counts */
5824	SCTP_BASE_INFO(ipi_count_ep) = 0;
5825	/* assoc/tcb zone info */
5826	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5827	/* local addrlist zone info */
5828	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5829	/* remote addrlist zone info */
5830	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5831	/* chunk info */
5832	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5833
5834	/* socket queue zone info */
5835	SCTP_BASE_INFO(ipi_count_readq) = 0;
5836
5837	/* stream out queue cont */
5838	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5839
5840	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5841	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5842
5843	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5844
5845	/* Init the TIMEWAIT list */
5846	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5847		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5848	}
5849
5850	sctp_startup_iterator();
5851
5852#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5853	sctp_startup_mcore_threads();
5854#endif
5855
5856	/*
5857	 * INIT the default VRF which for BSD is the only one, other O/S's
5858	 * may have more. But initially they must start with one and then
5859	 * add the VRF's as addresses are added.
5860	 */
5861	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5862}
5863
5864/*
5865 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5866 */
5867void
5868sctp_pcb_finish(void)
5869{
5870	struct sctp_vrflist *vrf_bucket;
5871	struct sctp_vrf *vrf, *nvrf;
5872	struct sctp_ifn *ifn, *nifn;
5873	struct sctp_ifa *ifa, *nifa;
5874	struct sctpvtaghead *chain;
5875	struct sctp_tagblock *twait_block, *prev_twait_block;
5876	struct sctp_laddr *wi, *nwi;
5877	int i;
5878
5879	/*
5880	 * Free BSD the it thread never exits but we do clean up. The only
5881	 * way freebsd reaches here if we have VRF's but we still add the
5882	 * ifdef to make it compile on old versions.
5883	 */
5884	{
5885		struct sctp_iterator *it, *nit;
5886
5887		SCTP_IPI_ITERATOR_WQ_LOCK();
5888		TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5889			if (it->vn != curvnet) {
5890				continue;
5891			}
5892			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5893			if (it->function_atend != NULL) {
5894				(*it->function_atend) (it->pointer, it->val);
5895			}
5896			SCTP_FREE(it, SCTP_M_ITER);
5897		}
5898		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5899		SCTP_ITERATOR_LOCK();
5900		if ((sctp_it_ctl.cur_it) &&
5901		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5902			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5903		}
5904		SCTP_ITERATOR_UNLOCK();
5905	}
5906
5907	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5908	SCTP_WQ_ADDR_LOCK();
5909	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5910		LIST_REMOVE(wi, sctp_nxt_addr);
5911		SCTP_DECR_LADDR_COUNT();
5912		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5913	}
5914	SCTP_WQ_ADDR_UNLOCK();
5915
5916	/*
5917	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5918	 * destroyed first).
5919	 */
5920	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5921	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5922		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5923			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5924				/* free the ifa */
5925				LIST_REMOVE(ifa, next_bucket);
5926				LIST_REMOVE(ifa, next_ifa);
5927				SCTP_FREE(ifa, SCTP_M_IFA);
5928			}
5929			/* free the ifn */
5930			LIST_REMOVE(ifn, next_bucket);
5931			LIST_REMOVE(ifn, next_ifn);
5932			SCTP_FREE(ifn, SCTP_M_IFN);
5933		}
5934		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5935		/* free the vrf */
5936		LIST_REMOVE(vrf, next_vrf);
5937		SCTP_FREE(vrf, SCTP_M_VRF);
5938	}
5939	/* free the vrf hashes */
5940	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5941	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5942
5943	/*
5944	 * free the TIMEWAIT list elements malloc'd in the function
5945	 * sctp_add_vtag_to_timewait()...
5946	 */
5947	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5948		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5949		if (!LIST_EMPTY(chain)) {
5950			prev_twait_block = NULL;
5951			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5952				if (prev_twait_block) {
5953					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5954				}
5955				prev_twait_block = twait_block;
5956			}
5957			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5958		}
5959	}
5960
5961	/* free the locks and mutexes */
5962#ifdef SCTP_PACKET_LOGGING
5963	SCTP_IP_PKTLOG_DESTROY();
5964#endif
5965	SCTP_IPI_ADDR_DESTROY();
5966	SCTP_STATLOG_DESTROY();
5967	SCTP_INP_INFO_LOCK_DESTROY();
5968
5969	SCTP_WQ_ADDR_DESTROY();
5970
5971	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5972	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5973	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5974	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5975	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5976	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5977	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5978	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5979	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5980	/* Get rid of other stuff to */
5981	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5982		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5983	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5984		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5985	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5986		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5987#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5988	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5989#endif
5990}
5991
5992
5993int
5994sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5995    int offset, int limit,
5996    struct sockaddr *src, struct sockaddr *dst,
5997    struct sockaddr *altsa)
5998{
5999	/*
6000	 * grub through the INIT pulling addresses and loading them to the
6001	 * nets structure in the asoc. The from address in the mbuf should
6002	 * also be loaded (if it is not already). This routine can be called
6003	 * with either INIT or INIT-ACK's as long as the m points to the IP
6004	 * packet and the offset points to the beginning of the parameters.
6005	 */
6006	struct sctp_inpcb *inp;
6007	struct sctp_nets *net, *nnet, *net_tmp;
6008	struct sctp_paramhdr *phdr, parm_buf;
6009	struct sctp_tcb *stcb_tmp;
6010	uint16_t ptype, plen;
6011	struct sockaddr *sa;
6012	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
6013	struct sctp_auth_random *p_random = NULL;
6014	uint16_t random_len = 0;
6015	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
6016	struct sctp_auth_hmac_algo *hmacs = NULL;
6017	uint16_t hmacs_len = 0;
6018	uint8_t saw_asconf = 0;
6019	uint8_t saw_asconf_ack = 0;
6020	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6021	struct sctp_auth_chunk_list *chunks = NULL;
6022	uint16_t num_chunks = 0;
6023	sctp_key_t *new_key;
6024	uint32_t keylen;
6025	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6026	uint8_t ecn_allowed;
6027
6028#ifdef INET
6029	struct sockaddr_in sin;
6030
6031#endif
6032#ifdef INET6
6033	struct sockaddr_in6 sin6;
6034
6035#endif
6036
6037	/* First get the destination address setup too. */
6038#ifdef INET
6039	memset(&sin, 0, sizeof(sin));
6040	sin.sin_family = AF_INET;
6041	sin.sin_len = sizeof(sin);
6042	sin.sin_port = stcb->rport;
6043#endif
6044#ifdef INET6
6045	memset(&sin6, 0, sizeof(sin6));
6046	sin6.sin6_family = AF_INET6;
6047	sin6.sin6_len = sizeof(struct sockaddr_in6);
6048	sin6.sin6_port = stcb->rport;
6049#endif
6050	if (altsa) {
6051		sa = altsa;
6052	} else {
6053		sa = src;
6054	}
6055	/* Turn off ECN until we get through all params */
6056	ecn_allowed = 0;
6057	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6058		/* mark all addresses that we have currently on the list */
6059		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6060	}
6061	/* does the source address already exist? if so skip it */
6062	inp = stcb->sctp_ep;
6063	atomic_add_int(&stcb->asoc.refcnt, 1);
6064	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6065	atomic_add_int(&stcb->asoc.refcnt, -1);
6066
6067	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6068		/* we must add the source address */
6069		/* no scope set here since we have a tcb already. */
6070		switch (sa->sa_family) {
6071#ifdef INET
6072		case AF_INET:
6073			if (stcb->asoc.ipv4_addr_legal) {
6074				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6075					return (-1);
6076				}
6077			}
6078			break;
6079#endif
6080#ifdef INET6
6081		case AF_INET6:
6082			if (stcb->asoc.ipv6_addr_legal) {
6083				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6084					return (-2);
6085				}
6086			}
6087			break;
6088#endif
6089		default:
6090			break;
6091		}
6092	} else {
6093		if (net_tmp != NULL && stcb_tmp == stcb) {
6094			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6095		} else if (stcb_tmp != stcb) {
6096			/* It belongs to another association? */
6097			if (stcb_tmp)
6098				SCTP_TCB_UNLOCK(stcb_tmp);
6099			return (-3);
6100		}
6101	}
6102	if (stcb->asoc.state == 0) {
6103		/* the assoc was freed? */
6104		return (-4);
6105	}
6106	/*
6107	 * peer must explicitly turn this on. This may have been initialized
6108	 * to be "on" in order to allow local addr changes while INIT's are
6109	 * in flight.
6110	 */
6111	stcb->asoc.peer_supports_asconf = 0;
6112	/* now we must go through each of the params. */
6113	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6114	while (phdr) {
6115		ptype = ntohs(phdr->param_type);
6116		plen = ntohs(phdr->param_length);
6117		/*
6118		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6119		 * (uint32_t)ptype, (int)plen);
6120		 */
6121		if (offset + plen > limit) {
6122			break;
6123		}
6124		if (plen == 0) {
6125			break;
6126		}
6127#ifdef INET
6128		if (ptype == SCTP_IPV4_ADDRESS) {
6129			if (stcb->asoc.ipv4_addr_legal) {
6130				struct sctp_ipv4addr_param *p4, p4_buf;
6131
6132				/* ok get the v4 address and check/add */
6133				phdr = sctp_get_next_param(m, offset,
6134				    (struct sctp_paramhdr *)&p4_buf,
6135				    sizeof(p4_buf));
6136				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6137				    phdr == NULL) {
6138					return (-5);
6139				}
6140				p4 = (struct sctp_ipv4addr_param *)phdr;
6141				sin.sin_addr.s_addr = p4->addr;
6142				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6143					/* Skip multi-cast addresses */
6144					goto next_param;
6145				}
6146				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6147				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6148					goto next_param;
6149				}
6150				sa = (struct sockaddr *)&sin;
6151				inp = stcb->sctp_ep;
6152				atomic_add_int(&stcb->asoc.refcnt, 1);
6153				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6154				    dst, stcb);
6155				atomic_add_int(&stcb->asoc.refcnt, -1);
6156
6157				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6158				    inp == NULL) {
6159					/* we must add the source address */
6160					/*
6161					 * no scope set since we have a tcb
6162					 * already
6163					 */
6164
6165					/*
6166					 * we must validate the state again
6167					 * here
6168					 */
6169			add_it_now:
6170					if (stcb->asoc.state == 0) {
6171						/* the assoc was freed? */
6172						return (-7);
6173					}
6174					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6175						return (-8);
6176					}
6177				} else if (stcb_tmp == stcb) {
6178					if (stcb->asoc.state == 0) {
6179						/* the assoc was freed? */
6180						return (-10);
6181					}
6182					if (net != NULL) {
6183						/* clear flag */
6184						net->dest_state &=
6185						    ~SCTP_ADDR_NOT_IN_ASSOC;
6186					}
6187				} else {
6188					/*
6189					 * strange, address is in another
6190					 * assoc? straighten out locks.
6191					 */
6192					if (stcb_tmp) {
6193						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6194							/*
6195							 * in setup state we
6196							 * abort this guy
6197							 */
6198							sctp_abort_an_association(stcb_tmp->sctp_ep,
6199							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6200							goto add_it_now;
6201						}
6202						SCTP_TCB_UNLOCK(stcb_tmp);
6203					}
6204					if (stcb->asoc.state == 0) {
6205						/* the assoc was freed? */
6206						return (-12);
6207					}
6208					return (-13);
6209				}
6210			}
6211		} else
6212#endif
6213#ifdef INET6
6214		if (ptype == SCTP_IPV6_ADDRESS) {
6215			if (stcb->asoc.ipv6_addr_legal) {
6216				/* ok get the v6 address and check/add */
6217				struct sctp_ipv6addr_param *p6, p6_buf;
6218
6219				phdr = sctp_get_next_param(m, offset,
6220				    (struct sctp_paramhdr *)&p6_buf,
6221				    sizeof(p6_buf));
6222				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6223				    phdr == NULL) {
6224					return (-14);
6225				}
6226				p6 = (struct sctp_ipv6addr_param *)phdr;
6227				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6228				    sizeof(p6->addr));
6229				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6230					/* Skip multi-cast addresses */
6231					goto next_param;
6232				}
6233				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6234					/*
6235					 * Link local make no sense without
6236					 * scope
6237					 */
6238					goto next_param;
6239				}
6240				sa = (struct sockaddr *)&sin6;
6241				inp = stcb->sctp_ep;
6242				atomic_add_int(&stcb->asoc.refcnt, 1);
6243				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6244				    dst, stcb);
6245				atomic_add_int(&stcb->asoc.refcnt, -1);
6246				if (stcb_tmp == NULL &&
6247				    (inp == stcb->sctp_ep || inp == NULL)) {
6248					/*
6249					 * we must validate the state again
6250					 * here
6251					 */
6252			add_it_now6:
6253					if (stcb->asoc.state == 0) {
6254						/* the assoc was freed? */
6255						return (-16);
6256					}
6257					/*
6258					 * we must add the address, no scope
6259					 * set
6260					 */
6261					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6262						return (-17);
6263					}
6264				} else if (stcb_tmp == stcb) {
6265					/*
6266					 * we must validate the state again
6267					 * here
6268					 */
6269					if (stcb->asoc.state == 0) {
6270						/* the assoc was freed? */
6271						return (-19);
6272					}
6273					if (net != NULL) {
6274						/* clear flag */
6275						net->dest_state &=
6276						    ~SCTP_ADDR_NOT_IN_ASSOC;
6277					}
6278				} else {
6279					/*
6280					 * strange, address is in another
6281					 * assoc? straighten out locks.
6282					 */
6283					if (stcb_tmp)
6284						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6285							/*
6286							 * in setup state we
6287							 * abort this guy
6288							 */
6289							sctp_abort_an_association(stcb_tmp->sctp_ep,
6290							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6291							goto add_it_now6;
6292						}
6293					SCTP_TCB_UNLOCK(stcb_tmp);
6294
6295					if (stcb->asoc.state == 0) {
6296						/* the assoc was freed? */
6297						return (-21);
6298					}
6299					return (-22);
6300				}
6301			}
6302		} else
6303#endif
6304		if (ptype == SCTP_ECN_CAPABLE) {
6305			ecn_allowed = 1;
6306		} else if (ptype == SCTP_ULP_ADAPTATION) {
6307			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6308				struct sctp_adaptation_layer_indication ai,
6309				                                *aip;
6310
6311				phdr = sctp_get_next_param(m, offset,
6312				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6313				aip = (struct sctp_adaptation_layer_indication *)phdr;
6314				if (aip) {
6315					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6316					stcb->asoc.adaptation_needed = 1;
6317				}
6318			}
6319		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6320			struct sctp_asconf_addr_param lstore, *fee;
6321			int lptype;
6322			struct sockaddr *lsa = NULL;
6323
6324#ifdef INET
6325			struct sctp_asconf_addrv4_param *fii;
6326
6327#endif
6328
6329			stcb->asoc.peer_supports_asconf = 1;
6330			if (plen > sizeof(lstore)) {
6331				return (-23);
6332			}
6333			phdr = sctp_get_next_param(m, offset,
6334			    (struct sctp_paramhdr *)&lstore,
6335			    min(plen, sizeof(lstore)));
6336			if (phdr == NULL) {
6337				return (-24);
6338			}
6339			fee = (struct sctp_asconf_addr_param *)phdr;
6340			lptype = ntohs(fee->addrp.ph.param_type);
6341			switch (lptype) {
6342#ifdef INET
6343			case SCTP_IPV4_ADDRESS:
6344				if (plen !=
6345				    sizeof(struct sctp_asconf_addrv4_param)) {
6346					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6347					    (int)sizeof(struct sctp_asconf_addrv4_param),
6348					    plen);
6349				} else {
6350					fii = (struct sctp_asconf_addrv4_param *)fee;
6351					sin.sin_addr.s_addr = fii->addrp.addr;
6352					lsa = (struct sockaddr *)&sin;
6353				}
6354				break;
6355#endif
6356#ifdef INET6
6357			case SCTP_IPV6_ADDRESS:
6358				if (plen !=
6359				    sizeof(struct sctp_asconf_addr_param)) {
6360					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6361					    (int)sizeof(struct sctp_asconf_addr_param),
6362					    plen);
6363				} else {
6364					memcpy(sin6.sin6_addr.s6_addr,
6365					    fee->addrp.addr,
6366					    sizeof(fee->addrp.addr));
6367					lsa = (struct sockaddr *)&sin6;
6368				}
6369				break;
6370#endif
6371			default:
6372				break;
6373			}
6374			if (lsa) {
6375				(void)sctp_set_primary_addr(stcb, sa, NULL);
6376			}
6377		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6378			stcb->asoc.peer_supports_nat = 1;
6379		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6380			/* Peer supports pr-sctp */
6381			stcb->asoc.peer_supports_prsctp = 1;
6382		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6383			/* A supported extension chunk */
6384			struct sctp_supported_chunk_types_param *pr_supported;
6385			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6386			int num_ent, i;
6387
6388			phdr = sctp_get_next_param(m, offset,
6389			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6390			if (phdr == NULL) {
6391				return (-25);
6392			}
6393			stcb->asoc.peer_supports_asconf = 0;
6394			stcb->asoc.peer_supports_prsctp = 0;
6395			stcb->asoc.peer_supports_pktdrop = 0;
6396			stcb->asoc.peer_supports_strreset = 0;
6397			stcb->asoc.peer_supports_nr_sack = 0;
6398			stcb->asoc.peer_supports_auth = 0;
6399			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6400			num_ent = plen - sizeof(struct sctp_paramhdr);
6401			for (i = 0; i < num_ent; i++) {
6402				switch (pr_supported->chunk_types[i]) {
6403				case SCTP_ASCONF:
6404				case SCTP_ASCONF_ACK:
6405					stcb->asoc.peer_supports_asconf = 1;
6406					break;
6407				case SCTP_FORWARD_CUM_TSN:
6408					stcb->asoc.peer_supports_prsctp = 1;
6409					break;
6410				case SCTP_PACKET_DROPPED:
6411					stcb->asoc.peer_supports_pktdrop = 1;
6412					break;
6413				case SCTP_NR_SELECTIVE_ACK:
6414					stcb->asoc.peer_supports_nr_sack = 1;
6415					break;
6416				case SCTP_STREAM_RESET:
6417					stcb->asoc.peer_supports_strreset = 1;
6418					break;
6419				case SCTP_AUTHENTICATION:
6420					stcb->asoc.peer_supports_auth = 1;
6421					break;
6422				default:
6423					/* one I have not learned yet */
6424					break;
6425
6426				}
6427			}
6428		} else if (ptype == SCTP_RANDOM) {
6429			if (plen > sizeof(random_store))
6430				break;
6431			if (got_random) {
6432				/* already processed a RANDOM */
6433				goto next_param;
6434			}
6435			phdr = sctp_get_next_param(m, offset,
6436			    (struct sctp_paramhdr *)random_store,
6437			    min(sizeof(random_store), plen));
6438			if (phdr == NULL)
6439				return (-26);
6440			p_random = (struct sctp_auth_random *)phdr;
6441			random_len = plen - sizeof(*p_random);
6442			/* enforce the random length */
6443			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6444				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6445				return (-27);
6446			}
6447			got_random = 1;
6448		} else if (ptype == SCTP_HMAC_LIST) {
6449			int num_hmacs;
6450			int i;
6451
6452			if (plen > sizeof(hmacs_store))
6453				break;
6454			if (got_hmacs) {
6455				/* already processed a HMAC list */
6456				goto next_param;
6457			}
6458			phdr = sctp_get_next_param(m, offset,
6459			    (struct sctp_paramhdr *)hmacs_store,
6460			    min(plen, sizeof(hmacs_store)));
6461			if (phdr == NULL)
6462				return (-28);
6463			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6464			hmacs_len = plen - sizeof(*hmacs);
6465			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6466			/* validate the hmac list */
6467			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6468				return (-29);
6469			}
6470			if (stcb->asoc.peer_hmacs != NULL)
6471				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6472			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6473			if (stcb->asoc.peer_hmacs != NULL) {
6474				for (i = 0; i < num_hmacs; i++) {
6475					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6476					    ntohs(hmacs->hmac_ids[i]));
6477				}
6478			}
6479			got_hmacs = 1;
6480		} else if (ptype == SCTP_CHUNK_LIST) {
6481			int i;
6482
6483			if (plen > sizeof(chunks_store))
6484				break;
6485			if (got_chklist) {
6486				/* already processed a Chunks list */
6487				goto next_param;
6488			}
6489			phdr = sctp_get_next_param(m, offset,
6490			    (struct sctp_paramhdr *)chunks_store,
6491			    min(plen, sizeof(chunks_store)));
6492			if (phdr == NULL)
6493				return (-30);
6494			chunks = (struct sctp_auth_chunk_list *)phdr;
6495			num_chunks = plen - sizeof(*chunks);
6496			if (stcb->asoc.peer_auth_chunks != NULL)
6497				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6498			else
6499				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6500			for (i = 0; i < num_chunks; i++) {
6501				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6502				    stcb->asoc.peer_auth_chunks);
6503				/* record asconf/asconf-ack if listed */
6504				if (chunks->chunk_types[i] == SCTP_ASCONF)
6505					saw_asconf = 1;
6506				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6507					saw_asconf_ack = 1;
6508
6509			}
6510			got_chklist = 1;
6511		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6512			    (ptype == SCTP_STATE_COOKIE) ||
6513			    (ptype == SCTP_UNRECOG_PARAM) ||
6514			    (ptype == SCTP_COOKIE_PRESERVE) ||
6515			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6516			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6517			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6518			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6519		    (ptype == SCTP_SUCCESS_REPORT)) {
6520			 /* don't care */ ;
6521		} else {
6522			if ((ptype & 0x8000) == 0x0000) {
6523				/*
6524				 * must stop processing the rest of the
6525				 * param's. Any report bits were handled
6526				 * with the call to
6527				 * sctp_arethere_unrecognized_parameters()
6528				 * when the INIT or INIT-ACK was first seen.
6529				 */
6530				break;
6531			}
6532		}
6533
6534next_param:
6535		offset += SCTP_SIZE32(plen);
6536		if (offset >= limit) {
6537			break;
6538		}
6539		phdr = sctp_get_next_param(m, offset, &parm_buf,
6540		    sizeof(parm_buf));
6541	}
6542	/* Now check to see if we need to purge any addresses */
6543	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6544		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6545		    SCTP_ADDR_NOT_IN_ASSOC) {
6546			/* This address has been removed from the asoc */
6547			/* remove and free it */
6548			stcb->asoc.numnets--;
6549			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6550			sctp_free_remote_addr(net);
6551			if (net == stcb->asoc.primary_destination) {
6552				stcb->asoc.primary_destination = NULL;
6553				sctp_select_primary_destination(stcb);
6554			}
6555		}
6556	}
6557	if (ecn_allowed == 0) {
6558		stcb->asoc.ecn_allowed = 0;
6559	}
6560	/* validate authentication required parameters */
6561	if (got_random && got_hmacs) {
6562		stcb->asoc.peer_supports_auth = 1;
6563	} else {
6564		stcb->asoc.peer_supports_auth = 0;
6565	}
6566	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6567		/* peer does not support auth but sent a chunks list? */
6568		return (-31);
6569	}
6570	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6571	    !stcb->asoc.peer_supports_auth) {
6572		/* peer supports asconf but not auth? */
6573		return (-32);
6574	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6575	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6576		return (-33);
6577	}
6578	/* concatenate the full random key */
6579	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6580	if (chunks != NULL) {
6581		keylen += sizeof(*chunks) + num_chunks;
6582	}
6583	new_key = sctp_alloc_key(keylen);
6584	if (new_key != NULL) {
6585		/* copy in the RANDOM */
6586		if (p_random != NULL) {
6587			keylen = sizeof(*p_random) + random_len;
6588			bcopy(p_random, new_key->key, keylen);
6589		}
6590		/* append in the AUTH chunks */
6591		if (chunks != NULL) {
6592			bcopy(chunks, new_key->key + keylen,
6593			    sizeof(*chunks) + num_chunks);
6594			keylen += sizeof(*chunks) + num_chunks;
6595		}
6596		/* append in the HMACs */
6597		if (hmacs != NULL) {
6598			bcopy(hmacs, new_key->key + keylen,
6599			    sizeof(*hmacs) + hmacs_len);
6600		}
6601	} else {
6602		/* failed to get memory for the key */
6603		return (-34);
6604	}
6605	if (stcb->asoc.authinfo.peer_random != NULL)
6606		sctp_free_key(stcb->asoc.authinfo.peer_random);
6607	stcb->asoc.authinfo.peer_random = new_key;
6608	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6609	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6610
6611	return (0);
6612}
6613
6614int
6615sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6616    struct sctp_nets *net)
6617{
6618	/* make sure the requested primary address exists in the assoc */
6619	if (net == NULL && sa)
6620		net = sctp_findnet(stcb, sa);
6621
6622	if (net == NULL) {
6623		/* didn't find the requested primary address! */
6624		return (-1);
6625	} else {
6626		/* set the primary address */
6627		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6628			/* Must be confirmed, so queue to set */
6629			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6630			return (0);
6631		}
6632		stcb->asoc.primary_destination = net;
6633		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6634			sctp_free_remote_addr(stcb->asoc.alternate);
6635			stcb->asoc.alternate = NULL;
6636		}
6637		net = TAILQ_FIRST(&stcb->asoc.nets);
6638		if (net != stcb->asoc.primary_destination) {
6639			/*
6640			 * first one on the list is NOT the primary
6641			 * sctp_cmpaddr() is much more efficient if the
6642			 * primary is the first on the list, make it so.
6643			 */
6644			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6645			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6646		}
6647		return (0);
6648	}
6649}
6650
6651int
6652sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6653{
6654	/*
6655	 * This function serves two purposes. It will see if a TAG can be
6656	 * re-used and return 1 for yes it is ok and 0 for don't use that
6657	 * tag. A secondary function it will do is purge out old tags that
6658	 * can be removed.
6659	 */
6660	struct sctpvtaghead *chain;
6661	struct sctp_tagblock *twait_block;
6662	struct sctpasochead *head;
6663	struct sctp_tcb *stcb;
6664	int i;
6665
6666	SCTP_INP_INFO_RLOCK();
6667	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6668	    SCTP_BASE_INFO(hashasocmark))];
6669	if (head == NULL) {
6670		/* invalid vtag */
6671		goto skip_vtag_check;
6672	}
6673	LIST_FOREACH(stcb, head, sctp_asocs) {
6674		/*
6675		 * We choose not to lock anything here. TCB's can't be
6676		 * removed since we have the read lock, so they can't be
6677		 * freed on us, same thing for the INP. I may be wrong with
6678		 * this assumption, but we will go with it for now :-)
6679		 */
6680		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6681			continue;
6682		}
6683		if (stcb->asoc.my_vtag == tag) {
6684			/* candidate */
6685			if (stcb->rport != rport) {
6686				continue;
6687			}
6688			if (stcb->sctp_ep->sctp_lport != lport) {
6689				continue;
6690			}
6691			/* Its a used tag set */
6692			SCTP_INP_INFO_RUNLOCK();
6693			return (0);
6694		}
6695	}
6696skip_vtag_check:
6697
6698	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6699	/* Now what about timed wait ? */
6700	if (!LIST_EMPTY(chain)) {
6701		/*
6702		 * Block(s) are present, lets see if we have this tag in the
6703		 * list
6704		 */
6705		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6706			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6707				if (twait_block->vtag_block[i].v_tag == 0) {
6708					/* not used */
6709					continue;
6710				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6711				    now->tv_sec) {
6712					/* Audit expires this guy */
6713					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6714					twait_block->vtag_block[i].v_tag = 0;
6715					twait_block->vtag_block[i].lport = 0;
6716					twait_block->vtag_block[i].rport = 0;
6717				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6718					    (twait_block->vtag_block[i].lport == lport) &&
6719				    (twait_block->vtag_block[i].rport == rport)) {
6720					/* Bad tag, sorry :< */
6721					SCTP_INP_INFO_RUNLOCK();
6722					return (0);
6723				}
6724			}
6725		}
6726	}
6727	SCTP_INP_INFO_RUNLOCK();
6728	return (1);
6729}
6730
6731static void
6732sctp_drain_mbufs(struct sctp_tcb *stcb)
6733{
6734	/*
6735	 * We must hunt this association for MBUF's past the cumack (i.e.
6736	 * out of order data that we can renege on).
6737	 */
6738	struct sctp_association *asoc;
6739	struct sctp_tmit_chunk *chk, *nchk;
6740	uint32_t cumulative_tsn_p1;
6741	struct sctp_queued_to_read *ctl, *nctl;
6742	int cnt, strmat;
6743	uint32_t gap, i;
6744	int fnd = 0;
6745
6746	/* We look for anything larger than the cum-ack + 1 */
6747
6748	asoc = &stcb->asoc;
6749	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6750		/* none we can reneg on. */
6751		return;
6752	}
6753	SCTP_STAT_INCR(sctps_protocol_drains_done);
6754	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6755	cnt = 0;
6756	/* First look in the re-assembly queue */
6757	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6758		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6759			/* Yep it is above cum-ack */
6760			cnt++;
6761			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6762			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6763			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6764			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6765			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6766			if (chk->data) {
6767				sctp_m_freem(chk->data);
6768				chk->data = NULL;
6769			}
6770			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6771		}
6772	}
6773	/* Ok that was fun, now we will drain all the inbound streams? */
6774	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6775		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6776			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6777				/* Yep it is above cum-ack */
6778				cnt++;
6779				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6780				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6781				sctp_ucount_decr(asoc->cnt_on_all_streams);
6782				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6783				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6784				if (ctl->data) {
6785					sctp_m_freem(ctl->data);
6786					ctl->data = NULL;
6787				}
6788				sctp_free_remote_addr(ctl->whoFrom);
6789				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6790				SCTP_DECR_READQ_COUNT();
6791			}
6792		}
6793	}
6794	if (cnt) {
6795		/* We must back down to see what the new highest is */
6796		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6797			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6798			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6799				asoc->highest_tsn_inside_map = i;
6800				fnd = 1;
6801				break;
6802			}
6803		}
6804		if (!fnd) {
6805			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6806		}
6807		/*
6808		 * Question, should we go through the delivery queue? The
6809		 * only reason things are on here is the app not reading OR
6810		 * a p-d-api up. An attacker COULD send enough in to
6811		 * initiate the PD-API and then send a bunch of stuff to
6812		 * other streams... these would wind up on the delivery
6813		 * queue.. and then we would not get to them. But in order
6814		 * to do this I then have to back-track and un-deliver
6815		 * sequence numbers in streams.. el-yucko. I think for now
6816		 * we will NOT look at the delivery queue and leave it to be
6817		 * something to consider later. An alternative would be to
6818		 * abort the P-D-API with a notification and then deliver
6819		 * the data.... Or another method might be to keep track of
6820		 * how many times the situation occurs and if we see a
6821		 * possible attack underway just abort the association.
6822		 */
6823#ifdef SCTP_DEBUG
6824		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6825#endif
6826		/*
6827		 * Now do we need to find a new
6828		 * asoc->highest_tsn_inside_map?
6829		 */
6830		asoc->last_revoke_count = cnt;
6831		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6832		/* sa_ignore NO_NULL_CHK */
6833		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6834		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6835	}
6836	/*
6837	 * Another issue, in un-setting the TSN's in the mapping array we
6838	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6839	 * two things to occur. It may cause us to do extra work in checking
6840	 * for our mapping array movement. More importantly it may cause us
6841	 * to SACK every datagram. This may not be a bad thing though since
6842	 * we will recover once we get our cum-ack above and all this stuff
6843	 * we dumped recovered.
6844	 */
6845}
6846
6847void
6848sctp_drain()
6849{
6850	/*
6851	 * We must walk the PCB lists for ALL associations here. The system
6852	 * is LOW on MBUF's and needs help. This is where reneging will
6853	 * occur. We really hope this does NOT happen!
6854	 */
6855	VNET_ITERATOR_DECL(vnet_iter);
6856	VNET_LIST_RLOCK_NOSLEEP();
6857	VNET_FOREACH(vnet_iter) {
6858		CURVNET_SET(vnet_iter);
6859		struct sctp_inpcb *inp;
6860		struct sctp_tcb *stcb;
6861
6862		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6863		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6864#ifdef VIMAGE
6865			continue;
6866#else
6867			return;
6868#endif
6869		}
6870		SCTP_INP_INFO_RLOCK();
6871		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6872			/* For each endpoint */
6873			SCTP_INP_RLOCK(inp);
6874			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6875				/* For each association */
6876				SCTP_TCB_LOCK(stcb);
6877				sctp_drain_mbufs(stcb);
6878				SCTP_TCB_UNLOCK(stcb);
6879			}
6880			SCTP_INP_RUNLOCK(inp);
6881		}
6882		SCTP_INP_INFO_RUNLOCK();
6883		CURVNET_RESTORE();
6884	}
6885	VNET_LIST_RUNLOCK_NOSLEEP();
6886}
6887
6888/*
6889 * start a new iterator
6890 * iterates through all endpoints and associations based on the pcb_state
6891 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6892 * assocs and "ef" (optional) is executed when the iterator completes.
6893 * "inpf" (optional) is executed for each new endpoint as it is being
6894 * iterated through. inpe (optional) is called when the inp completes
6895 * its way through all the stcbs.
6896 */
6897int
6898sctp_initiate_iterator(inp_func inpf,
6899    asoc_func af,
6900    inp_func inpe,
6901    uint32_t pcb_state,
6902    uint32_t pcb_features,
6903    uint32_t asoc_state,
6904    void *argp,
6905    uint32_t argi,
6906    end_func ef,
6907    struct sctp_inpcb *s_inp,
6908    uint8_t chunk_output_off)
6909{
6910	struct sctp_iterator *it = NULL;
6911
6912	if (af == NULL) {
6913		return (-1);
6914	}
6915	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6916	    SCTP_M_ITER);
6917	if (it == NULL) {
6918		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6919		return (ENOMEM);
6920	}
6921	memset(it, 0, sizeof(*it));
6922	it->function_assoc = af;
6923	it->function_inp = inpf;
6924	if (inpf)
6925		it->done_current_ep = 0;
6926	else
6927		it->done_current_ep = 1;
6928	it->function_atend = ef;
6929	it->pointer = argp;
6930	it->val = argi;
6931	it->pcb_flags = pcb_state;
6932	it->pcb_features = pcb_features;
6933	it->asoc_state = asoc_state;
6934	it->function_inp_end = inpe;
6935	it->no_chunk_output = chunk_output_off;
6936	it->vn = curvnet;
6937	if (s_inp) {
6938		/* Assume lock is held here */
6939		it->inp = s_inp;
6940		SCTP_INP_INCR_REF(it->inp);
6941		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6942	} else {
6943		SCTP_INP_INFO_RLOCK();
6944		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6945		if (it->inp) {
6946			SCTP_INP_INCR_REF(it->inp);
6947		}
6948		SCTP_INP_INFO_RUNLOCK();
6949		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6950
6951	}
6952	SCTP_IPI_ITERATOR_WQ_LOCK();
6953
6954	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6955	if (sctp_it_ctl.iterator_running == 0) {
6956		sctp_wakeup_iterator();
6957	}
6958	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6959	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6960	return (0);
6961}
6962