1/* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel reference Implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * The SCTP reference implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 *                 ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING.  If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 *    lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 *    http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 *    La Monte H.P. Yarroll <piggy@acm.org>
44 *    Narasimha Budihal     <narsi@refcode.org>
45 *    Karl Knutson          <karl@athena.chicago.il.us>
46 *    Jon Grimm             <jgrimm@us.ibm.com>
47 *    Xingang Guo           <xingang.guo@intel.com>
48 *    Daisy Chang           <daisyc@us.ibm.com>
49 *    Sridhar Samudrala     <samudrala@us.ibm.com>
50 *    Inaky Perez-Gonzalez  <inaky.gonzalez@intel.com>
51 *    Ardelle Fan	    <ardelle.fan@intel.com>
52 *    Ryan Layer	    <rmlayer@us.ibm.com>
53 *    Anup Pemmaiah         <pemmaiah@cc.usu.edu>
54 *    Kevin Gao             <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60#include <linux/types.h>
61#include <linux/kernel.h>
62#include <linux/wait.h>
63#include <linux/time.h>
64#include <linux/ip.h>
65#include <linux/capability.h>
66#include <linux/fcntl.h>
67#include <linux/poll.h>
68#include <linux/init.h>
69#include <linux/crypto.h>
70
71#include <net/ip.h>
72#include <net/icmp.h>
73#include <net/route.h>
74#include <net/ipv6.h>
75#include <net/inet_common.h>
76
77#include <linux/socket.h> /* for sa_family_t */
78#include <net/sock.h>
79#include <net/sctp/sctp.h>
80#include <net/sctp/sm.h>
81
82/* WARNING:  Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
85 */
86
87/* Forward declarations for internal helper functions. */
88static int sctp_writeable(struct sock *sk);
89static void sctp_wfree(struct sk_buff *skb);
90static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91				size_t msg_len);
92static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94static int sctp_wait_for_accept(struct sock *sk, long timeo);
95static void sctp_wait_for_close(struct sock *sk, long timeo);
96static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97					union sctp_addr *addr, int len);
98static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102static int sctp_send_asconf(struct sctp_association *asoc,
103			    struct sctp_chunk *chunk);
104static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105static int sctp_autobind(struct sock *sk);
106static void sctp_sock_migrate(struct sock *, struct sock *,
107			      struct sctp_association *, sctp_socket_type_t);
108static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
109
110extern struct kmem_cache *sctp_bucket_cachep;
111
112/* Get the sndbuf space available at the time on the association.  */
113static inline int sctp_wspace(struct sctp_association *asoc)
114{
115	struct sock *sk = asoc->base.sk;
116	int amt = 0;
117
118	if (asoc->ep->sndbuf_policy) {
119		/* make sure that no association uses more than sk_sndbuf */
120		amt = sk->sk_sndbuf - asoc->sndbuf_used;
121	} else {
122		/* do socket level accounting */
123		amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
124	}
125
126	if (amt < 0)
127		amt = 0;
128
129	return amt;
130}
131
132/* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
135 *
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
139 * tracking.
140 */
141static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
142{
143	struct sctp_association *asoc = chunk->asoc;
144	struct sock *sk = asoc->base.sk;
145
146	/* The sndbuf space is tracked per association.  */
147	sctp_association_hold(asoc);
148
149	skb_set_owner_w(chunk->skb, sk);
150
151	chunk->skb->destructor = sctp_wfree;
152	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
153	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
154
155	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
156				sizeof(struct sk_buff) +
157				sizeof(struct sctp_chunk);
158
159	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
160}
161
162/* Verify that this is a valid address. */
163static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
164				   int len)
165{
166	struct sctp_af *af;
167
168	/* Verify basic sockaddr. */
169	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
170	if (!af)
171		return -EINVAL;
172
173	/* Is this a valid SCTP address?  */
174	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
175		return -EINVAL;
176
177	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
178		return -EINVAL;
179
180	return 0;
181}
182
183/* Look up the association by its id.  If this is not a UDP-style
184 * socket, the ID field is always ignored.
185 */
186struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
187{
188	struct sctp_association *asoc = NULL;
189
190	/* If this is not a UDP-style socket, assoc id should be ignored. */
191	if (!sctp_style(sk, UDP)) {
192		/* Return NULL if the socket state is not ESTABLISHED. It
193		 * could be a TCP-style listening socket or a socket which
194		 * hasn't yet called connect() to establish an association.
195		 */
196		if (!sctp_sstate(sk, ESTABLISHED))
197			return NULL;
198
199		/* Get the first and the only association from the list. */
200		if (!list_empty(&sctp_sk(sk)->ep->asocs))
201			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
202					  struct sctp_association, asocs);
203		return asoc;
204	}
205
206	/* Otherwise this is a UDP-style socket. */
207	if (!id || (id == (sctp_assoc_t)-1))
208		return NULL;
209
210	spin_lock_bh(&sctp_assocs_id_lock);
211	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
212	spin_unlock_bh(&sctp_assocs_id_lock);
213
214	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
215		return NULL;
216
217	return asoc;
218}
219
220/* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
222 * the same.
223 */
224static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225					      struct sockaddr_storage *addr,
226					      sctp_assoc_t id)
227{
228	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
229	struct sctp_transport *transport;
230	union sctp_addr *laddr = (union sctp_addr *)addr;
231
232	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
233					       laddr,
234					       &transport);
235
236	if (!addr_asoc)
237		return NULL;
238
239	id_asoc = sctp_id2assoc(sk, id);
240	if (id_asoc && (id_asoc != addr_asoc))
241		return NULL;
242
243	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
244						(union sctp_addr *)addr);
245
246	return transport;
247}
248
249/* API 3.1.2 bind() - UDP Style Syntax
250 * The syntax of bind() is,
251 *
252 *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
253 *
254 *   sd      - the socket descriptor returned by socket().
255 *   addr    - the address structure (struct sockaddr_in or struct
256 *             sockaddr_in6 [RFC 2553]),
257 *   addr_len - the size of the address structure.
258 */
259SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
260{
261	int retval = 0;
262
263	sctp_lock_sock(sk);
264
265	SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
266			  sk, addr, addr_len);
267
268	/* Disallow binding twice. */
269	if (!sctp_sk(sk)->ep->base.bind_addr.port)
270		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
271				      addr_len);
272	else
273		retval = -EINVAL;
274
275	sctp_release_sock(sk);
276
277	return retval;
278}
279
280static long sctp_get_port_local(struct sock *, union sctp_addr *);
281
282/* Verify this is a valid sockaddr. */
283static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
284					union sctp_addr *addr, int len)
285{
286	struct sctp_af *af;
287
288	/* Check minimum size.  */
289	if (len < sizeof (struct sockaddr))
290		return NULL;
291
292	/* Does this PF support this AF? */
293	if (!opt->pf->af_supported(addr->sa.sa_family, opt))
294		return NULL;
295
296	/* If we get this far, af is valid. */
297	af = sctp_get_af_specific(addr->sa.sa_family);
298
299	if (len < af->sockaddr_len)
300		return NULL;
301
302	return af;
303}
304
305/* Bind a local address either to an endpoint or to an association.  */
306SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
307{
308	struct sctp_sock *sp = sctp_sk(sk);
309	struct sctp_endpoint *ep = sp->ep;
310	struct sctp_bind_addr *bp = &ep->base.bind_addr;
311	struct sctp_af *af;
312	unsigned short snum;
313	int ret = 0;
314
315	/* Common sockaddr verification. */
316	af = sctp_sockaddr_af(sp, addr, len);
317	if (!af) {
318		SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
319				  sk, addr, len);
320		return -EINVAL;
321	}
322
323	snum = ntohs(addr->v4.sin_port);
324
325	SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
326				 ", port: %d, new port: %d, len: %d)\n",
327				 sk,
328				 addr,
329				 bp->port, snum,
330				 len);
331
332	/* PF specific bind() address verification. */
333	if (!sp->pf->bind_verify(sp, addr))
334		return -EADDRNOTAVAIL;
335
336	/* We must either be unbound, or bind to the same port.
337	 * It's OK to allow 0 ports if we are already bound.
338	 * We'll just inhert an already bound port in this case
339	 */
340	if (bp->port) {
341		if (!snum)
342			snum = bp->port;
343		else if (snum != bp->port) {
344			SCTP_DEBUG_PRINTK("sctp_do_bind:"
345				  " New port %d does not match existing port "
346				  "%d.\n", snum, bp->port);
347			return -EINVAL;
348		}
349	}
350
351	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
352		return -EACCES;
353
354	/* Make sure we are allowed to bind here.
355	 * The function sctp_get_port_local() does duplicate address
356	 * detection.
357	 */
358	if ((ret = sctp_get_port_local(sk, addr))) {
359		if (ret == (long) sk) {
360			/* This endpoint has a conflicting address. */
361			return -EINVAL;
362		} else {
363			return -EADDRINUSE;
364		}
365	}
366
367	/* Refresh ephemeral port.  */
368	if (!bp->port)
369		bp->port = inet_sk(sk)->num;
370
371	/* Add the address to the bind address list.  */
372	sctp_local_bh_disable();
373	sctp_write_lock(&ep->base.addr_lock);
374
375	/* Use GFP_ATOMIC since BHs are disabled.  */
376	ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
377	sctp_write_unlock(&ep->base.addr_lock);
378	sctp_local_bh_enable();
379
380	/* Copy back into socket for getsockname() use. */
381	if (!ret) {
382		inet_sk(sk)->sport = htons(inet_sk(sk)->num);
383		af->to_sk_saddr(addr, sk);
384	}
385
386	return ret;
387}
388
389 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
390 *
391 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
392 * at any one time.  If a sender, after sending an ASCONF chunk, decides
393 * it needs to transfer another ASCONF Chunk, it MUST wait until the
394 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
395 * subsequent ASCONF. Note this restriction binds each side, so at any
396 * time two ASCONF may be in-transit on any given association (one sent
397 * from each endpoint).
398 */
399static int sctp_send_asconf(struct sctp_association *asoc,
400			    struct sctp_chunk *chunk)
401{
402	int		retval = 0;
403
404	/* If there is an outstanding ASCONF chunk, queue it for later
405	 * transmission.
406	 */
407	if (asoc->addip_last_asconf) {
408		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
409		goto out;
410	}
411
412	/* Hold the chunk until an ASCONF_ACK is received. */
413	sctp_chunk_hold(chunk);
414	retval = sctp_primitive_ASCONF(asoc, chunk);
415	if (retval)
416		sctp_chunk_free(chunk);
417	else
418		asoc->addip_last_asconf = chunk;
419
420out:
421	return retval;
422}
423
424/* Add a list of addresses as bind addresses to local endpoint or
425 * association.
426 *
427 * Basically run through each address specified in the addrs/addrcnt
428 * array/length pair, determine if it is IPv6 or IPv4 and call
429 * sctp_do_bind() on it.
430 *
431 * If any of them fails, then the operation will be reversed and the
432 * ones that were added will be removed.
433 *
434 * Only sctp_setsockopt_bindx() is supposed to call this function.
435 */
436int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
437{
438	int cnt;
439	int retval = 0;
440	void *addr_buf;
441	struct sockaddr *sa_addr;
442	struct sctp_af *af;
443
444	SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
445			  sk, addrs, addrcnt);
446
447	addr_buf = addrs;
448	for (cnt = 0; cnt < addrcnt; cnt++) {
449		/* The list may contain either IPv4 or IPv6 address;
450		 * determine the address length for walking thru the list.
451		 */
452		sa_addr = (struct sockaddr *)addr_buf;
453		af = sctp_get_af_specific(sa_addr->sa_family);
454		if (!af) {
455			retval = -EINVAL;
456			goto err_bindx_add;
457		}
458
459		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
460				      af->sockaddr_len);
461
462		addr_buf += af->sockaddr_len;
463
464err_bindx_add:
465		if (retval < 0) {
466			/* Failed. Cleanup the ones that have been added */
467			if (cnt > 0)
468				sctp_bindx_rem(sk, addrs, cnt);
469			return retval;
470		}
471	}
472
473	return retval;
474}
475
476/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
477 * associations that are part of the endpoint indicating that a list of local
478 * addresses are added to the endpoint.
479 *
480 * If any of the addresses is already in the bind address list of the
481 * association, we do not send the chunk for that association.  But it will not
482 * affect other associations.
483 *
484 * Only sctp_setsockopt_bindx() is supposed to call this function.
485 */
486static int sctp_send_asconf_add_ip(struct sock		*sk,
487				   struct sockaddr	*addrs,
488				   int 			addrcnt)
489{
490	struct sctp_sock		*sp;
491	struct sctp_endpoint		*ep;
492	struct sctp_association		*asoc;
493	struct sctp_bind_addr		*bp;
494	struct sctp_chunk		*chunk;
495	struct sctp_sockaddr_entry	*laddr;
496	union sctp_addr			*addr;
497	union sctp_addr			saveaddr;
498	void				*addr_buf;
499	struct sctp_af			*af;
500	struct list_head		*pos;
501	struct list_head		*p;
502	int 				i;
503	int 				retval = 0;
504
505	if (!sctp_addip_enable)
506		return retval;
507
508	sp = sctp_sk(sk);
509	ep = sp->ep;
510
511	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
512			  __FUNCTION__, sk, addrs, addrcnt);
513
514	list_for_each(pos, &ep->asocs) {
515		asoc = list_entry(pos, struct sctp_association, asocs);
516
517		if (!asoc->peer.asconf_capable)
518			continue;
519
520		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
521			continue;
522
523		if (!sctp_state(asoc, ESTABLISHED))
524			continue;
525
526		/* Check if any address in the packed array of addresses is
527		 * in the bind address list of the association. If so,
528		 * do not send the asconf chunk to its peer, but continue with
529		 * other associations.
530		 */
531		addr_buf = addrs;
532		for (i = 0; i < addrcnt; i++) {
533			addr = (union sctp_addr *)addr_buf;
534			af = sctp_get_af_specific(addr->v4.sin_family);
535			if (!af) {
536				retval = -EINVAL;
537				goto out;
538			}
539
540			if (sctp_assoc_lookup_laddr(asoc, addr))
541				break;
542
543			addr_buf += af->sockaddr_len;
544		}
545		if (i < addrcnt)
546			continue;
547
548		/* Use the first address in bind addr list of association as
549		 * Address Parameter of ASCONF CHUNK.
550		 */
551		sctp_read_lock(&asoc->base.addr_lock);
552		bp = &asoc->base.bind_addr;
553		p = bp->address_list.next;
554		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
555		sctp_read_unlock(&asoc->base.addr_lock);
556
557		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
558						   addrcnt, SCTP_PARAM_ADD_IP);
559		if (!chunk) {
560			retval = -ENOMEM;
561			goto out;
562		}
563
564		retval = sctp_send_asconf(asoc, chunk);
565		if (retval)
566			goto out;
567
568		/* Add the new addresses to the bind address list with
569		 * use_as_src set to 0.
570		 */
571		sctp_local_bh_disable();
572		sctp_write_lock(&asoc->base.addr_lock);
573		addr_buf = addrs;
574		for (i = 0; i < addrcnt; i++) {
575			addr = (union sctp_addr *)addr_buf;
576			af = sctp_get_af_specific(addr->v4.sin_family);
577			memcpy(&saveaddr, addr, af->sockaddr_len);
578			retval = sctp_add_bind_addr(bp, &saveaddr, 0,
579						    GFP_ATOMIC);
580			addr_buf += af->sockaddr_len;
581		}
582		sctp_write_unlock(&asoc->base.addr_lock);
583		sctp_local_bh_enable();
584	}
585
586out:
587	return retval;
588}
589
590/* Remove a list of addresses from bind addresses list.  Do not remove the
591 * last address.
592 *
593 * Basically run through each address specified in the addrs/addrcnt
594 * array/length pair, determine if it is IPv6 or IPv4 and call
595 * sctp_del_bind() on it.
596 *
597 * If any of them fails, then the operation will be reversed and the
598 * ones that were removed will be added back.
599 *
600 * At least one address has to be left; if only one address is
601 * available, the operation will return -EBUSY.
602 *
603 * Only sctp_setsockopt_bindx() is supposed to call this function.
604 */
605int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
606{
607	struct sctp_sock *sp = sctp_sk(sk);
608	struct sctp_endpoint *ep = sp->ep;
609	int cnt;
610	struct sctp_bind_addr *bp = &ep->base.bind_addr;
611	int retval = 0;
612	void *addr_buf;
613	union sctp_addr *sa_addr;
614	struct sctp_af *af;
615
616	SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
617			  sk, addrs, addrcnt);
618
619	addr_buf = addrs;
620	for (cnt = 0; cnt < addrcnt; cnt++) {
621		/* If the bind address list is empty or if there is only one
622		 * bind address, there is nothing more to be removed (we need
623		 * at least one address here).
624		 */
625		if (list_empty(&bp->address_list) ||
626		    (sctp_list_single_entry(&bp->address_list))) {
627			retval = -EBUSY;
628			goto err_bindx_rem;
629		}
630
631		sa_addr = (union sctp_addr *)addr_buf;
632		af = sctp_get_af_specific(sa_addr->sa.sa_family);
633		if (!af) {
634			retval = -EINVAL;
635			goto err_bindx_rem;
636		}
637
638		if (!af->addr_valid(sa_addr, sp, NULL)) {
639			retval = -EADDRNOTAVAIL;
640			goto err_bindx_rem;
641		}
642
643		if (sa_addr->v4.sin_port != htons(bp->port)) {
644			retval = -EINVAL;
645			goto err_bindx_rem;
646		}
647
648		sctp_local_bh_disable();
649		sctp_write_lock(&ep->base.addr_lock);
650
651		retval = sctp_del_bind_addr(bp, sa_addr);
652
653		sctp_write_unlock(&ep->base.addr_lock);
654		sctp_local_bh_enable();
655
656		addr_buf += af->sockaddr_len;
657err_bindx_rem:
658		if (retval < 0) {
659			/* Failed. Add the ones that has been removed back */
660			if (cnt > 0)
661				sctp_bindx_add(sk, addrs, cnt);
662			return retval;
663		}
664	}
665
666	return retval;
667}
668
669/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
670 * the associations that are part of the endpoint indicating that a list of
671 * local addresses are removed from the endpoint.
672 *
673 * If any of the addresses is already in the bind address list of the
674 * association, we do not send the chunk for that association.  But it will not
675 * affect other associations.
676 *
677 * Only sctp_setsockopt_bindx() is supposed to call this function.
678 */
679static int sctp_send_asconf_del_ip(struct sock		*sk,
680				   struct sockaddr	*addrs,
681				   int			addrcnt)
682{
683	struct sctp_sock	*sp;
684	struct sctp_endpoint	*ep;
685	struct sctp_association	*asoc;
686	struct sctp_transport	*transport;
687	struct sctp_bind_addr	*bp;
688	struct sctp_chunk	*chunk;
689	union sctp_addr		*laddr;
690	void			*addr_buf;
691	struct sctp_af		*af;
692	struct list_head	*pos, *pos1;
693	struct sctp_sockaddr_entry *saddr;
694	int 			i;
695	int 			retval = 0;
696
697	if (!sctp_addip_enable)
698		return retval;
699
700	sp = sctp_sk(sk);
701	ep = sp->ep;
702
703	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
704			  __FUNCTION__, sk, addrs, addrcnt);
705
706	list_for_each(pos, &ep->asocs) {
707		asoc = list_entry(pos, struct sctp_association, asocs);
708
709		if (!asoc->peer.asconf_capable)
710			continue;
711
712		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
713			continue;
714
715		if (!sctp_state(asoc, ESTABLISHED))
716			continue;
717
718		/* Check if any address in the packed array of addresses is
719		 * not present in the bind address list of the association.
720		 * If so, do not send the asconf chunk to its peer, but
721		 * continue with other associations.
722		 */
723		addr_buf = addrs;
724		for (i = 0; i < addrcnt; i++) {
725			laddr = (union sctp_addr *)addr_buf;
726			af = sctp_get_af_specific(laddr->v4.sin_family);
727			if (!af) {
728				retval = -EINVAL;
729				goto out;
730			}
731
732			if (!sctp_assoc_lookup_laddr(asoc, laddr))
733				break;
734
735			addr_buf += af->sockaddr_len;
736		}
737		if (i < addrcnt)
738			continue;
739
740		/* Find one address in the association's bind address list
741		 * that is not in the packed array of addresses. This is to
742		 * make sure that we do not delete all the addresses in the
743		 * association.
744		 */
745		sctp_read_lock(&asoc->base.addr_lock);
746		bp = &asoc->base.bind_addr;
747		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
748					       addrcnt, sp);
749		sctp_read_unlock(&asoc->base.addr_lock);
750		if (!laddr)
751			continue;
752
753		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
754						   SCTP_PARAM_DEL_IP);
755		if (!chunk) {
756			retval = -ENOMEM;
757			goto out;
758		}
759
760		/* Reset use_as_src flag for the addresses in the bind address
761		 * list that are to be deleted.
762		 */
763		sctp_local_bh_disable();
764		sctp_write_lock(&asoc->base.addr_lock);
765		addr_buf = addrs;
766		for (i = 0; i < addrcnt; i++) {
767			laddr = (union sctp_addr *)addr_buf;
768			af = sctp_get_af_specific(laddr->v4.sin_family);
769			list_for_each(pos1, &bp->address_list) {
770				saddr = list_entry(pos1,
771						   struct sctp_sockaddr_entry,
772						   list);
773				if (sctp_cmp_addr_exact(&saddr->a, laddr))
774					saddr->use_as_src = 0;
775			}
776			addr_buf += af->sockaddr_len;
777		}
778		sctp_write_unlock(&asoc->base.addr_lock);
779		sctp_local_bh_enable();
780
781		/* Update the route and saddr entries for all the transports
782		 * as some of the addresses in the bind address list are
783		 * about to be deleted and cannot be used as source addresses.
784		 */
785		list_for_each(pos1, &asoc->peer.transport_addr_list) {
786			transport = list_entry(pos1, struct sctp_transport,
787					       transports);
788			dst_release(transport->dst);
789			sctp_transport_route(transport, NULL,
790					     sctp_sk(asoc->base.sk));
791		}
792
793		retval = sctp_send_asconf(asoc, chunk);
794	}
795out:
796	return retval;
797}
798
799/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
800 *
801 * API 8.1
802 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
803 *                int flags);
804 *
805 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
806 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
807 * or IPv6 addresses.
808 *
809 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
810 * Section 3.1.2 for this usage.
811 *
812 * addrs is a pointer to an array of one or more socket addresses. Each
813 * address is contained in its appropriate structure (i.e. struct
814 * sockaddr_in or struct sockaddr_in6) the family of the address type
815 * must be used to distinguish the address length (note that this
816 * representation is termed a "packed array" of addresses). The caller
817 * specifies the number of addresses in the array with addrcnt.
818 *
819 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
820 * -1, and sets errno to the appropriate error code.
821 *
822 * For SCTP, the port given in each socket address must be the same, or
823 * sctp_bindx() will fail, setting errno to EINVAL.
824 *
825 * The flags parameter is formed from the bitwise OR of zero or more of
826 * the following currently defined flags:
827 *
828 * SCTP_BINDX_ADD_ADDR
829 *
830 * SCTP_BINDX_REM_ADDR
831 *
832 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
833 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
834 * addresses from the association. The two flags are mutually exclusive;
835 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
836 * not remove all addresses from an association; sctp_bindx() will
837 * reject such an attempt with EINVAL.
838 *
839 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
840 * additional addresses with an endpoint after calling bind().  Or use
841 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
842 * socket is associated with so that no new association accepted will be
843 * associated with those addresses. If the endpoint supports dynamic
844 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
845 * endpoint to send the appropriate message to the peer to change the
846 * peers address lists.
847 *
848 * Adding and removing addresses from a connected association is
849 * optional functionality. Implementations that do not support this
850 * functionality should return EOPNOTSUPP.
851 *
852 * Basically do nothing but copying the addresses from user to kernel
853 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
854 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
855 * from userspace.
856 *
857 * We don't use copy_from_user() for optimization: we first do the
858 * sanity checks (buffer size -fast- and access check-healthy
859 * pointer); if all of those succeed, then we can alloc the memory
860 * (expensive operation) needed to copy the data to kernel. Then we do
861 * the copying without checking the user space area
862 * (__copy_from_user()).
863 *
864 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
865 * it.
866 *
867 * sk        The sk of the socket
868 * addrs     The pointer to the addresses in user land
869 * addrssize Size of the addrs buffer
870 * op        Operation to perform (add or remove, see the flags of
871 *           sctp_bindx)
872 *
873 * Returns 0 if ok, <0 errno code on error.
874 */
875SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
876				      struct sockaddr __user *addrs,
877				      int addrs_size, int op)
878{
879	struct sockaddr *kaddrs;
880	int err;
881	int addrcnt = 0;
882	int walk_size = 0;
883	struct sockaddr *sa_addr;
884	void *addr_buf;
885	struct sctp_af *af;
886
887	SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
888			  " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
889
890	if (unlikely(addrs_size <= 0))
891		return -EINVAL;
892
893	/* Check the user passed a healthy pointer.  */
894	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
895		return -EFAULT;
896
897	/* Alloc space for the address array in kernel memory.  */
898	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
899	if (unlikely(!kaddrs))
900		return -ENOMEM;
901
902	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
903		kfree(kaddrs);
904		return -EFAULT;
905	}
906
907	/* Walk through the addrs buffer and count the number of addresses. */
908	addr_buf = kaddrs;
909	while (walk_size < addrs_size) {
910		sa_addr = (struct sockaddr *)addr_buf;
911		af = sctp_get_af_specific(sa_addr->sa_family);
912
913		/* If the address family is not supported or if this address
914		 * causes the address buffer to overflow return EINVAL.
915		 */
916		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
917			kfree(kaddrs);
918			return -EINVAL;
919		}
920		addrcnt++;
921		addr_buf += af->sockaddr_len;
922		walk_size += af->sockaddr_len;
923	}
924
925	/* Do the work. */
926	switch (op) {
927	case SCTP_BINDX_ADD_ADDR:
928		err = sctp_bindx_add(sk, kaddrs, addrcnt);
929		if (err)
930			goto out;
931		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
932		break;
933
934	case SCTP_BINDX_REM_ADDR:
935		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
936		if (err)
937			goto out;
938		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
939		break;
940
941	default:
942		err = -EINVAL;
943		break;
944	}
945
946out:
947	kfree(kaddrs);
948
949	return err;
950}
951
952/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
953 *
954 * Common routine for handling connect() and sctp_connectx().
955 * Connect will come in with just a single address.
956 */
957static int __sctp_connect(struct sock* sk,
958			  struct sockaddr *kaddrs,
959			  int addrs_size)
960{
961	struct sctp_sock *sp;
962	struct sctp_endpoint *ep;
963	struct sctp_association *asoc = NULL;
964	struct sctp_association *asoc2;
965	struct sctp_transport *transport;
966	union sctp_addr to;
967	struct sctp_af *af;
968	sctp_scope_t scope;
969	long timeo;
970	int err = 0;
971	int addrcnt = 0;
972	int walk_size = 0;
973	union sctp_addr *sa_addr;
974	void *addr_buf;
975	unsigned short port;
976	unsigned int f_flags = 0;
977
978	sp = sctp_sk(sk);
979	ep = sp->ep;
980
981	/* connect() cannot be done on a socket that is already in ESTABLISHED
982	 * state - UDP-style peeled off socket or a TCP-style socket that
983	 * is already connected.
984	 * It cannot be done even on a TCP-style listening socket.
985	 */
986	if (sctp_sstate(sk, ESTABLISHED) ||
987	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
988		err = -EISCONN;
989		goto out_free;
990	}
991
992	/* Walk through the addrs buffer and count the number of addresses. */
993	addr_buf = kaddrs;
994	while (walk_size < addrs_size) {
995		sa_addr = (union sctp_addr *)addr_buf;
996		af = sctp_get_af_specific(sa_addr->sa.sa_family);
997		port = ntohs(sa_addr->v4.sin_port);
998
999		/* If the address family is not supported or if this address
1000		 * causes the address buffer to overflow return EINVAL.
1001		 */
1002		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003			err = -EINVAL;
1004			goto out_free;
1005		}
1006
1007		err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
1008		if (err)
1009			goto out_free;
1010
1011		/* Make sure the destination port is correctly set
1012		 * in all addresses.
1013		 */
1014		if (asoc && asoc->peer.port && asoc->peer.port != port)
1015			goto out_free;
1016
1017		memcpy(&to, sa_addr, af->sockaddr_len);
1018
1019		/* Check if there already is a matching association on the
1020		 * endpoint (other than the one created here).
1021		 */
1022		asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1023		if (asoc2 && asoc2 != asoc) {
1024			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1025				err = -EISCONN;
1026			else
1027				err = -EALREADY;
1028			goto out_free;
1029		}
1030
1031		/* If we could not find a matching association on the endpoint,
1032		 * make sure that there is no peeled-off association matching
1033		 * the peer address even on another socket.
1034		 */
1035		if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1036			err = -EADDRNOTAVAIL;
1037			goto out_free;
1038		}
1039
1040		if (!asoc) {
1041			/* If a bind() or sctp_bindx() is not called prior to
1042			 * an sctp_connectx() call, the system picks an
1043			 * ephemeral port and will choose an address set
1044			 * equivalent to binding with a wildcard address.
1045			 */
1046			if (!ep->base.bind_addr.port) {
1047				if (sctp_autobind(sk)) {
1048					err = -EAGAIN;
1049					goto out_free;
1050				}
1051			} else {
1052				/*
1053				 * If an unprivileged user inherits a 1-many
1054				 * style socket with open associations on a
1055				 * privileged port, it MAY be permitted to
1056				 * accept new associations, but it SHOULD NOT
1057				 * be permitted to open new associations.
1058				 */
1059				if (ep->base.bind_addr.port < PROT_SOCK &&
1060				    !capable(CAP_NET_BIND_SERVICE)) {
1061					err = -EACCES;
1062					goto out_free;
1063				}
1064			}
1065
1066			scope = sctp_scope(sa_addr);
1067			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1068			if (!asoc) {
1069				err = -ENOMEM;
1070				goto out_free;
1071			}
1072		}
1073
1074		/* Prime the peer's transport structures.  */
1075		transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1076						SCTP_UNKNOWN);
1077		if (!transport) {
1078			err = -ENOMEM;
1079			goto out_free;
1080		}
1081
1082		addrcnt++;
1083		addr_buf += af->sockaddr_len;
1084		walk_size += af->sockaddr_len;
1085	}
1086
1087	err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1088	if (err < 0) {
1089		goto out_free;
1090	}
1091
1092	err = sctp_primitive_ASSOCIATE(asoc, NULL);
1093	if (err < 0) {
1094		goto out_free;
1095	}
1096
1097	/* Initialize sk's dport and daddr for getpeername() */
1098	inet_sk(sk)->dport = htons(asoc->peer.port);
1099	af = sctp_get_af_specific(to.sa.sa_family);
1100	af->to_sk_daddr(&to, sk);
1101	sk->sk_err = 0;
1102
1103	/* in-kernel sockets don't generally have a file allocated to them
1104	 * if all they do is call sock_create_kern().
1105	 */
1106	if (sk->sk_socket->file)
1107		f_flags = sk->sk_socket->file->f_flags;
1108
1109	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1110
1111	err = sctp_wait_for_connect(asoc, &timeo);
1112
1113	/* Don't free association on exit. */
1114	asoc = NULL;
1115
1116out_free:
1117
1118	SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1119			  " kaddrs: %p err: %d\n",
1120			  asoc, kaddrs, err);
1121	if (asoc)
1122		sctp_association_free(asoc);
1123	return err;
1124}
1125
1126/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1127 *
1128 * API 8.9
1129 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1130 *
1131 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1132 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1133 * or IPv6 addresses.
1134 *
1135 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1136 * Section 3.1.2 for this usage.
1137 *
1138 * addrs is a pointer to an array of one or more socket addresses. Each
1139 * address is contained in its appropriate structure (i.e. struct
1140 * sockaddr_in or struct sockaddr_in6) the family of the address type
1141 * must be used to distengish the address length (note that this
1142 * representation is termed a "packed array" of addresses). The caller
1143 * specifies the number of addresses in the array with addrcnt.
1144 *
1145 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1146 * -1, and sets errno to the appropriate error code.
1147 *
1148 * For SCTP, the port given in each socket address must be the same, or
1149 * sctp_connectx() will fail, setting errno to EINVAL.
1150 *
1151 * An application can use sctp_connectx to initiate an association with
1152 * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1153 * allows a caller to specify multiple addresses at which a peer can be
1154 * reached.  The way the SCTP stack uses the list of addresses to set up
1155 * the association is implementation dependant.  This function only
1156 * specifies that the stack will try to make use of all the addresses in
1157 * the list when needed.
1158 *
1159 * Note that the list of addresses passed in is only used for setting up
1160 * the association.  It does not necessarily equal the set of addresses
1161 * the peer uses for the resulting association.  If the caller wants to
1162 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1163 * retrieve them after the association has been set up.
1164 *
1165 * Basically do nothing but copying the addresses from user to kernel
1166 * land and invoking either sctp_connectx(). This is used for tunneling
1167 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1168 *
1169 * We don't use copy_from_user() for optimization: we first do the
1170 * sanity checks (buffer size -fast- and access check-healthy
1171 * pointer); if all of those succeed, then we can alloc the memory
1172 * (expensive operation) needed to copy the data to kernel. Then we do
1173 * the copying without checking the user space area
1174 * (__copy_from_user()).
1175 *
1176 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1177 * it.
1178 *
1179 * sk        The sk of the socket
1180 * addrs     The pointer to the addresses in user land
1181 * addrssize Size of the addrs buffer
1182 *
1183 * Returns 0 if ok, <0 errno code on error.
1184 */
1185SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1186				      struct sockaddr __user *addrs,
1187				      int addrs_size)
1188{
1189	int err = 0;
1190	struct sockaddr *kaddrs;
1191
1192	SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1193			  __FUNCTION__, sk, addrs, addrs_size);
1194
1195	if (unlikely(addrs_size <= 0))
1196		return -EINVAL;
1197
1198	/* Check the user passed a healthy pointer.  */
1199	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1200		return -EFAULT;
1201
1202	/* Alloc space for the address array in kernel memory.  */
1203	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1204	if (unlikely(!kaddrs))
1205		return -ENOMEM;
1206
1207	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1208		err = -EFAULT;
1209	} else {
1210		err = __sctp_connect(sk, kaddrs, addrs_size);
1211	}
1212
1213	kfree(kaddrs);
1214	return err;
1215}
1216
1217SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1218{
1219	struct sctp_endpoint *ep;
1220	struct sctp_association *asoc;
1221	struct list_head *pos, *temp;
1222
1223	SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1224
1225	sctp_lock_sock(sk);
1226	sk->sk_shutdown = SHUTDOWN_MASK;
1227
1228	ep = sctp_sk(sk)->ep;
1229
1230	/* Walk all associations on an endpoint.  */
1231	list_for_each_safe(pos, temp, &ep->asocs) {
1232		asoc = list_entry(pos, struct sctp_association, asocs);
1233
1234		if (sctp_style(sk, TCP)) {
1235			/* A closed association can still be in the list if
1236			 * it belongs to a TCP-style listening socket that is
1237			 * not yet accepted. If so, free it. If not, send an
1238			 * ABORT or SHUTDOWN based on the linger options.
1239			 */
1240			if (sctp_state(asoc, CLOSED)) {
1241				sctp_unhash_established(asoc);
1242				sctp_association_free(asoc);
1243				continue;
1244			}
1245		}
1246
1247		if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1248			struct sctp_chunk *chunk;
1249
1250			chunk = sctp_make_abort_user(asoc, NULL, 0);
1251			if (chunk)
1252				sctp_primitive_ABORT(asoc, chunk);
1253		} else
1254			sctp_primitive_SHUTDOWN(asoc, NULL);
1255	}
1256
1257	/* Clean up any skbs sitting on the receive queue.  */
1258	sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1259	sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1260
1261	/* On a TCP-style socket, block for at most linger_time if set. */
1262	if (sctp_style(sk, TCP) && timeout)
1263		sctp_wait_for_close(sk, timeout);
1264
1265	/* This will run the backlog queue.  */
1266	sctp_release_sock(sk);
1267
1268	/* Supposedly, no process has access to the socket, but
1269	 * the net layers still may.
1270	 */
1271	sctp_local_bh_disable();
1272	sctp_bh_lock_sock(sk);
1273
1274	/* Hold the sock, since sk_common_release() will put sock_put()
1275	 * and we have just a little more cleanup.
1276	 */
1277	sock_hold(sk);
1278	sk_common_release(sk);
1279
1280	sctp_bh_unlock_sock(sk);
1281	sctp_local_bh_enable();
1282
1283	sock_put(sk);
1284
1285	SCTP_DBG_OBJCNT_DEC(sock);
1286}
1287
1288/* Handle EPIPE error. */
1289static int sctp_error(struct sock *sk, int flags, int err)
1290{
1291	if (err == -EPIPE)
1292		err = sock_error(sk) ? : -EPIPE;
1293	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1294		send_sig(SIGPIPE, current, 0);
1295	return err;
1296}
1297
1298/* API 3.1.3 sendmsg() - UDP Style Syntax
1299 *
1300 * An application uses sendmsg() and recvmsg() calls to transmit data to
1301 * and receive data from its peer.
1302 *
1303 *  ssize_t sendmsg(int socket, const struct msghdr *message,
1304 *                  int flags);
1305 *
1306 *  socket  - the socket descriptor of the endpoint.
1307 *  message - pointer to the msghdr structure which contains a single
1308 *            user message and possibly some ancillary data.
1309 *
1310 *            See Section 5 for complete description of the data
1311 *            structures.
1312 *
1313 *  flags   - flags sent or received with the user message, see Section
1314 *            5 for complete description of the flags.
1315 *
1316 * Note:  This function could use a rewrite especially when explicit
1317 * connect support comes in.
1318 */
1319/* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1320
1321SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1322
1323SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1324			     struct msghdr *msg, size_t msg_len)
1325{
1326	struct sctp_sock *sp;
1327	struct sctp_endpoint *ep;
1328	struct sctp_association *new_asoc=NULL, *asoc=NULL;
1329	struct sctp_transport *transport, *chunk_tp;
1330	struct sctp_chunk *chunk;
1331	union sctp_addr to;
1332	struct sockaddr *msg_name = NULL;
1333	struct sctp_sndrcvinfo default_sinfo = { 0 };
1334	struct sctp_sndrcvinfo *sinfo;
1335	struct sctp_initmsg *sinit;
1336	sctp_assoc_t associd = 0;
1337	sctp_cmsgs_t cmsgs = { NULL };
1338	int err;
1339	sctp_scope_t scope;
1340	long timeo;
1341	__u16 sinfo_flags = 0;
1342	struct sctp_datamsg *datamsg;
1343	struct list_head *pos;
1344	int msg_flags = msg->msg_flags;
1345
1346	SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1347			  sk, msg, msg_len);
1348
1349	err = 0;
1350	sp = sctp_sk(sk);
1351	ep = sp->ep;
1352
1353	SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1354
1355	/* We cannot send a message over a TCP-style listening socket. */
1356	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1357		err = -EPIPE;
1358		goto out_nounlock;
1359	}
1360
1361	/* Parse out the SCTP CMSGs.  */
1362	err = sctp_msghdr_parse(msg, &cmsgs);
1363
1364	if (err) {
1365		SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1366		goto out_nounlock;
1367	}
1368
1369	/* Fetch the destination address for this packet.  This
1370	 * address only selects the association--it is not necessarily
1371	 * the address we will send to.
1372	 * For a peeled-off socket, msg_name is ignored.
1373	 */
1374	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1375		int msg_namelen = msg->msg_namelen;
1376
1377		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1378				       msg_namelen);
1379		if (err)
1380			return err;
1381
1382		if (msg_namelen > sizeof(to))
1383			msg_namelen = sizeof(to);
1384		memcpy(&to, msg->msg_name, msg_namelen);
1385		msg_name = msg->msg_name;
1386	}
1387
1388	sinfo = cmsgs.info;
1389	sinit = cmsgs.init;
1390
1391	/* Did the user specify SNDRCVINFO?  */
1392	if (sinfo) {
1393		sinfo_flags = sinfo->sinfo_flags;
1394		associd = sinfo->sinfo_assoc_id;
1395	}
1396
1397	SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1398			  msg_len, sinfo_flags);
1399
1400	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1401	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1402		err = -EINVAL;
1403		goto out_nounlock;
1404	}
1405
1406	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1407	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1408	 * If SCTP_ABORT is set, the message length could be non zero with
1409	 * the msg_iov set to the user abort reason.
1410	 */
1411	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1412	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1413		err = -EINVAL;
1414		goto out_nounlock;
1415	}
1416
1417	/* If SCTP_ADDR_OVER is set, there must be an address
1418	 * specified in msg_name.
1419	 */
1420	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1421		err = -EINVAL;
1422		goto out_nounlock;
1423	}
1424
1425	transport = NULL;
1426
1427	SCTP_DEBUG_PRINTK("About to look up association.\n");
1428
1429	sctp_lock_sock(sk);
1430
1431	/* If a msg_name has been specified, assume this is to be used.  */
1432	if (msg_name) {
1433		/* Look for a matching association on the endpoint. */
1434		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1435		if (!asoc) {
1436			/* If we could not find a matching association on the
1437			 * endpoint, make sure that it is not a TCP-style
1438			 * socket that already has an association or there is
1439			 * no peeled-off association on another socket.
1440			 */
1441			if ((sctp_style(sk, TCP) &&
1442			     sctp_sstate(sk, ESTABLISHED)) ||
1443			    sctp_endpoint_is_peeled_off(ep, &to)) {
1444				err = -EADDRNOTAVAIL;
1445				goto out_unlock;
1446			}
1447		}
1448	} else {
1449		asoc = sctp_id2assoc(sk, associd);
1450		if (!asoc) {
1451			err = -EPIPE;
1452			goto out_unlock;
1453		}
1454	}
1455
1456	if (asoc) {
1457		SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1458
1459		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1460		 * socket that has an association in CLOSED state. This can
1461		 * happen when an accepted socket has an association that is
1462		 * already CLOSED.
1463		 */
1464		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1465			err = -EPIPE;
1466			goto out_unlock;
1467		}
1468
1469		if (sinfo_flags & SCTP_EOF) {
1470			SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1471					  asoc);
1472			sctp_primitive_SHUTDOWN(asoc, NULL);
1473			err = 0;
1474			goto out_unlock;
1475		}
1476		if (sinfo_flags & SCTP_ABORT) {
1477			struct sctp_chunk *chunk;
1478
1479			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1480			if (!chunk) {
1481				err = -ENOMEM;
1482				goto out_unlock;
1483			}
1484
1485			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1486			sctp_primitive_ABORT(asoc, chunk);
1487			err = 0;
1488			goto out_unlock;
1489		}
1490	}
1491
1492	/* Do we need to create the association?  */
1493	if (!asoc) {
1494		SCTP_DEBUG_PRINTK("There is no association yet.\n");
1495
1496		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1497			err = -EINVAL;
1498			goto out_unlock;
1499		}
1500
1501		/* Check for invalid stream against the stream counts,
1502		 * either the default or the user specified stream counts.
1503		 */
1504		if (sinfo) {
1505			if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1506				/* Check against the defaults. */
1507				if (sinfo->sinfo_stream >=
1508				    sp->initmsg.sinit_num_ostreams) {
1509					err = -EINVAL;
1510					goto out_unlock;
1511				}
1512			} else {
1513				/* Check against the requested.  */
1514				if (sinfo->sinfo_stream >=
1515				    sinit->sinit_num_ostreams) {
1516					err = -EINVAL;
1517					goto out_unlock;
1518				}
1519			}
1520		}
1521
1522		/*
1523		 * API 3.1.2 bind() - UDP Style Syntax
1524		 * If a bind() or sctp_bindx() is not called prior to a
1525		 * sendmsg() call that initiates a new association, the
1526		 * system picks an ephemeral port and will choose an address
1527		 * set equivalent to binding with a wildcard address.
1528		 */
1529		if (!ep->base.bind_addr.port) {
1530			if (sctp_autobind(sk)) {
1531				err = -EAGAIN;
1532				goto out_unlock;
1533			}
1534		} else {
1535			/*
1536			 * If an unprivileged user inherits a one-to-many
1537			 * style socket with open associations on a privileged
1538			 * port, it MAY be permitted to accept new associations,
1539			 * but it SHOULD NOT be permitted to open new
1540			 * associations.
1541			 */
1542			if (ep->base.bind_addr.port < PROT_SOCK &&
1543			    !capable(CAP_NET_BIND_SERVICE)) {
1544				err = -EACCES;
1545				goto out_unlock;
1546			}
1547		}
1548
1549		scope = sctp_scope(&to);
1550		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1551		if (!new_asoc) {
1552			err = -ENOMEM;
1553			goto out_unlock;
1554		}
1555		asoc = new_asoc;
1556
1557		/* If the SCTP_INIT ancillary data is specified, set all
1558		 * the association init values accordingly.
1559		 */
1560		if (sinit) {
1561			if (sinit->sinit_num_ostreams) {
1562				asoc->c.sinit_num_ostreams =
1563					sinit->sinit_num_ostreams;
1564			}
1565			if (sinit->sinit_max_instreams) {
1566				asoc->c.sinit_max_instreams =
1567					sinit->sinit_max_instreams;
1568			}
1569			if (sinit->sinit_max_attempts) {
1570				asoc->max_init_attempts
1571					= sinit->sinit_max_attempts;
1572			}
1573			if (sinit->sinit_max_init_timeo) {
1574				asoc->max_init_timeo =
1575				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1576			}
1577		}
1578
1579		/* Prime the peer's transport structures.  */
1580		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1581		if (!transport) {
1582			err = -ENOMEM;
1583			goto out_free;
1584		}
1585		err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1586		if (err < 0) {
1587			err = -ENOMEM;
1588			goto out_free;
1589		}
1590	}
1591
1592	/* ASSERT: we have a valid association at this point.  */
1593	SCTP_DEBUG_PRINTK("We have a valid association.\n");
1594
1595	if (!sinfo) {
1596		/* If the user didn't specify SNDRCVINFO, make up one with
1597		 * some defaults.
1598		 */
1599		default_sinfo.sinfo_stream = asoc->default_stream;
1600		default_sinfo.sinfo_flags = asoc->default_flags;
1601		default_sinfo.sinfo_ppid = asoc->default_ppid;
1602		default_sinfo.sinfo_context = asoc->default_context;
1603		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1604		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1605		sinfo = &default_sinfo;
1606	}
1607
1608	/* API 7.1.7, the sndbuf size per association bounds the
1609	 * maximum size of data that can be sent in a single send call.
1610	 */
1611	if (msg_len > sk->sk_sndbuf) {
1612		err = -EMSGSIZE;
1613		goto out_free;
1614	}
1615
1616	if (asoc->pmtu_pending)
1617		sctp_assoc_pending_pmtu(asoc);
1618
1619	/* If fragmentation is disabled and the message length exceeds the
1620	 * association fragmentation point, return EMSGSIZE.  The I-D
1621	 * does not specify what this error is, but this looks like
1622	 * a great fit.
1623	 */
1624	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1625		err = -EMSGSIZE;
1626		goto out_free;
1627	}
1628
1629	if (sinfo) {
1630		/* Check for invalid stream. */
1631		if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1632			err = -EINVAL;
1633			goto out_free;
1634		}
1635	}
1636
1637	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1638	if (!sctp_wspace(asoc)) {
1639		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1640		if (err)
1641			goto out_free;
1642	}
1643
1644	/* If an address is passed with the sendto/sendmsg call, it is used
1645	 * to override the primary destination address in the TCP model, or
1646	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1647	 */
1648	if ((sctp_style(sk, TCP) && msg_name) ||
1649	    (sinfo_flags & SCTP_ADDR_OVER)) {
1650		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1651		if (!chunk_tp) {
1652			err = -EINVAL;
1653			goto out_free;
1654		}
1655	} else
1656		chunk_tp = NULL;
1657
1658	/* Auto-connect, if we aren't connected already. */
1659	if (sctp_state(asoc, CLOSED)) {
1660		err = sctp_primitive_ASSOCIATE(asoc, NULL);
1661		if (err < 0)
1662			goto out_free;
1663		SCTP_DEBUG_PRINTK("We associated primitively.\n");
1664	}
1665
1666	/* Break the message into multiple chunks of maximum size. */
1667	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1668	if (!datamsg) {
1669		err = -ENOMEM;
1670		goto out_free;
1671	}
1672
1673	/* Now send the (possibly) fragmented message. */
1674	list_for_each(pos, &datamsg->chunks) {
1675		chunk = list_entry(pos, struct sctp_chunk, frag_list);
1676		sctp_datamsg_track(chunk);
1677
1678		/* Do accounting for the write space.  */
1679		sctp_set_owner_w(chunk);
1680
1681		chunk->transport = chunk_tp;
1682
1683		/* Send it to the lower layers.  Note:  all chunks
1684		 * must either fail or succeed.   The lower layer
1685		 * works that way today.  Keep it that way or this
1686		 * breaks.
1687		 */
1688		err = sctp_primitive_SEND(asoc, chunk);
1689		/* Did the lower layer accept the chunk? */
1690		if (err)
1691			sctp_chunk_free(chunk);
1692		SCTP_DEBUG_PRINTK("We sent primitively.\n");
1693	}
1694
1695	sctp_datamsg_free(datamsg);
1696	if (err)
1697		goto out_free;
1698	else
1699		err = msg_len;
1700
1701	/* If we are already past ASSOCIATE, the lower
1702	 * layers are responsible for association cleanup.
1703	 */
1704	goto out_unlock;
1705
1706out_free:
1707	if (new_asoc)
1708		sctp_association_free(asoc);
1709out_unlock:
1710	sctp_release_sock(sk);
1711
1712out_nounlock:
1713	return sctp_error(sk, msg_flags, err);
1714
1715}
1716
1717/* This is an extended version of skb_pull() that removes the data from the
1718 * start of a skb even when data is spread across the list of skb's in the
1719 * frag_list. len specifies the total amount of data that needs to be removed.
1720 * when 'len' bytes could be removed from the skb, it returns 0.
1721 * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1722 * could not be removed.
1723 */
1724static int sctp_skb_pull(struct sk_buff *skb, int len)
1725{
1726	struct sk_buff *list;
1727	int skb_len = skb_headlen(skb);
1728	int rlen;
1729
1730	if (len <= skb_len) {
1731		__skb_pull(skb, len);
1732		return 0;
1733	}
1734	len -= skb_len;
1735	__skb_pull(skb, skb_len);
1736
1737	for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1738		rlen = sctp_skb_pull(list, len);
1739		skb->len -= (len-rlen);
1740		skb->data_len -= (len-rlen);
1741
1742		if (!rlen)
1743			return 0;
1744
1745		len = rlen;
1746	}
1747
1748	return len;
1749}
1750
1751/* API 3.1.3  recvmsg() - UDP Style Syntax
1752 *
1753 *  ssize_t recvmsg(int socket, struct msghdr *message,
1754 *                    int flags);
1755 *
1756 *  socket  - the socket descriptor of the endpoint.
1757 *  message - pointer to the msghdr structure which contains a single
1758 *            user message and possibly some ancillary data.
1759 *
1760 *            See Section 5 for complete description of the data
1761 *            structures.
1762 *
1763 *  flags   - flags sent or received with the user message, see Section
1764 *            5 for complete description of the flags.
1765 */
1766static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1767
1768SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1769			     struct msghdr *msg, size_t len, int noblock,
1770			     int flags, int *addr_len)
1771{
1772	struct sctp_ulpevent *event = NULL;
1773	struct sctp_sock *sp = sctp_sk(sk);
1774	struct sk_buff *skb;
1775	int copied;
1776	int err = 0;
1777	int skb_len;
1778
1779	SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1780			  "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1781			  "len", len, "knoblauch", noblock,
1782			  "flags", flags, "addr_len", addr_len);
1783
1784	sctp_lock_sock(sk);
1785
1786	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1787		err = -ENOTCONN;
1788		goto out;
1789	}
1790
1791	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1792	if (!skb)
1793		goto out;
1794
1795	/* Get the total length of the skb including any skb's in the
1796	 * frag_list.
1797	 */
1798	skb_len = skb->len;
1799
1800	copied = skb_len;
1801	if (copied > len)
1802		copied = len;
1803
1804	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1805
1806	event = sctp_skb2event(skb);
1807
1808	if (err)
1809		goto out_free;
1810
1811	sock_recv_timestamp(msg, sk, skb);
1812	if (sctp_ulpevent_is_notification(event)) {
1813		msg->msg_flags |= MSG_NOTIFICATION;
1814		sp->pf->event_msgname(event, msg->msg_name, addr_len);
1815	} else {
1816		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1817	}
1818
1819	/* Check if we allow SCTP_SNDRCVINFO. */
1820	if (sp->subscribe.sctp_data_io_event)
1821		sctp_ulpevent_read_sndrcvinfo(event, msg);
1822
1823	err = copied;
1824
1825	/* If skb's length exceeds the user's buffer, update the skb and
1826	 * push it back to the receive_queue so that the next call to
1827	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1828	 */
1829	if (skb_len > copied) {
1830		msg->msg_flags &= ~MSG_EOR;
1831		if (flags & MSG_PEEK)
1832			goto out_free;
1833		sctp_skb_pull(skb, copied);
1834		skb_queue_head(&sk->sk_receive_queue, skb);
1835
1836		/* When only partial message is copied to the user, increase
1837		 * rwnd by that amount. If all the data in the skb is read,
1838		 * rwnd is updated when the event is freed.
1839		 */
1840		sctp_assoc_rwnd_increase(event->asoc, copied);
1841		goto out;
1842	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
1843		   (event->msg_flags & MSG_EOR))
1844		msg->msg_flags |= MSG_EOR;
1845	else
1846		msg->msg_flags &= ~MSG_EOR;
1847
1848out_free:
1849	if (flags & MSG_PEEK) {
1850		/* Release the skb reference acquired after peeking the skb in
1851		 * sctp_skb_recv_datagram().
1852		 */
1853		kfree_skb(skb);
1854	} else {
1855		/* Free the event which includes releasing the reference to
1856		 * the owner of the skb, freeing the skb and updating the
1857		 * rwnd.
1858		 */
1859		sctp_ulpevent_free(event);
1860	}
1861out:
1862	sctp_release_sock(sk);
1863	return err;
1864}
1865
1866/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1867 *
1868 * This option is a on/off flag.  If enabled no SCTP message
1869 * fragmentation will be performed.  Instead if a message being sent
1870 * exceeds the current PMTU size, the message will NOT be sent and
1871 * instead a error will be indicated to the user.
1872 */
1873static int sctp_setsockopt_disable_fragments(struct sock *sk,
1874					    char __user *optval, int optlen)
1875{
1876	int val;
1877
1878	if (optlen < sizeof(int))
1879		return -EINVAL;
1880
1881	if (get_user(val, (int __user *)optval))
1882		return -EFAULT;
1883
1884	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1885
1886	return 0;
1887}
1888
1889static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1890					int optlen)
1891{
1892	if (optlen != sizeof(struct sctp_event_subscribe))
1893		return -EINVAL;
1894	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1895		return -EFAULT;
1896	return 0;
1897}
1898
1899/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1900 *
1901 * This socket option is applicable to the UDP-style socket only.  When
1902 * set it will cause associations that are idle for more than the
1903 * specified number of seconds to automatically close.  An association
1904 * being idle is defined an association that has NOT sent or received
1905 * user data.  The special value of '0' indicates that no automatic
1906 * close of any associations should be performed.  The option expects an
1907 * integer defining the number of seconds of idle time before an
1908 * association is closed.
1909 */
1910static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1911					    int optlen)
1912{
1913	struct sctp_sock *sp = sctp_sk(sk);
1914
1915	/* Applicable to UDP-style socket only */
1916	if (sctp_style(sk, TCP))
1917		return -EOPNOTSUPP;
1918	if (optlen != sizeof(int))
1919		return -EINVAL;
1920	if (copy_from_user(&sp->autoclose, optval, optlen))
1921		return -EFAULT;
1922
1923	return 0;
1924}
1925
1926/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1927 *
1928 * Applications can enable or disable heartbeats for any peer address of
1929 * an association, modify an address's heartbeat interval, force a
1930 * heartbeat to be sent immediately, and adjust the address's maximum
1931 * number of retransmissions sent before an address is considered
1932 * unreachable.  The following structure is used to access and modify an
1933 * address's parameters:
1934 *
1935 *  struct sctp_paddrparams {
1936 *     sctp_assoc_t            spp_assoc_id;
1937 *     struct sockaddr_storage spp_address;
1938 *     uint32_t                spp_hbinterval;
1939 *     uint16_t                spp_pathmaxrxt;
1940 *     uint32_t                spp_pathmtu;
1941 *     uint32_t                spp_sackdelay;
1942 *     uint32_t                spp_flags;
1943 * };
1944 *
1945 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
1946 *                     application, and identifies the association for
1947 *                     this query.
1948 *   spp_address     - This specifies which address is of interest.
1949 *   spp_hbinterval  - This contains the value of the heartbeat interval,
1950 *                     in milliseconds.  If a  value of zero
1951 *                     is present in this field then no changes are to
1952 *                     be made to this parameter.
1953 *   spp_pathmaxrxt  - This contains the maximum number of
1954 *                     retransmissions before this address shall be
1955 *                     considered unreachable. If a  value of zero
1956 *                     is present in this field then no changes are to
1957 *                     be made to this parameter.
1958 *   spp_pathmtu     - When Path MTU discovery is disabled the value
1959 *                     specified here will be the "fixed" path mtu.
1960 *                     Note that if the spp_address field is empty
1961 *                     then all associations on this address will
1962 *                     have this fixed path mtu set upon them.
1963 *
1964 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
1965 *                     the number of milliseconds that sacks will be delayed
1966 *                     for. This value will apply to all addresses of an
1967 *                     association if the spp_address field is empty. Note
1968 *                     also, that if delayed sack is enabled and this
1969 *                     value is set to 0, no change is made to the last
1970 *                     recorded delayed sack timer value.
1971 *
1972 *   spp_flags       - These flags are used to control various features
1973 *                     on an association. The flag field may contain
1974 *                     zero or more of the following options.
1975 *
1976 *                     SPP_HB_ENABLE  - Enable heartbeats on the
1977 *                     specified address. Note that if the address
1978 *                     field is empty all addresses for the association
1979 *                     have heartbeats enabled upon them.
1980 *
1981 *                     SPP_HB_DISABLE - Disable heartbeats on the
1982 *                     speicifed address. Note that if the address
1983 *                     field is empty all addresses for the association
1984 *                     will have their heartbeats disabled. Note also
1985 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
1986 *                     mutually exclusive, only one of these two should
1987 *                     be specified. Enabling both fields will have
1988 *                     undetermined results.
1989 *
1990 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
1991 *                     to be made immediately.
1992 *
1993 *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
1994 *                     heartbeat delayis to be set to the value of 0
1995 *                     milliseconds.
1996 *
1997 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
1998 *                     discovery upon the specified address. Note that
1999 *                     if the address feild is empty then all addresses
2000 *                     on the association are effected.
2001 *
2002 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2003 *                     discovery upon the specified address. Note that
2004 *                     if the address feild is empty then all addresses
2005 *                     on the association are effected. Not also that
2006 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2007 *                     exclusive. Enabling both will have undetermined
2008 *                     results.
2009 *
2010 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2011 *                     on delayed sack. The time specified in spp_sackdelay
2012 *                     is used to specify the sack delay for this address. Note
2013 *                     that if spp_address is empty then all addresses will
2014 *                     enable delayed sack and take on the sack delay
2015 *                     value specified in spp_sackdelay.
2016 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2017 *                     off delayed sack. If the spp_address field is blank then
2018 *                     delayed sack is disabled for the entire association. Note
2019 *                     also that this field is mutually exclusive to
2020 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2021 *                     results.
2022 */
2023static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2024				       struct sctp_transport   *trans,
2025				       struct sctp_association *asoc,
2026				       struct sctp_sock        *sp,
2027				       int                      hb_change,
2028				       int                      pmtud_change,
2029				       int                      sackdelay_change)
2030{
2031	int error;
2032
2033	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2034		error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2035		if (error)
2036			return error;
2037	}
2038
2039	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2040	 * this field is ignored.  Note also that a value of zero indicates
2041	 * the current setting should be left unchanged.
2042	 */
2043	if (params->spp_flags & SPP_HB_ENABLE) {
2044
2045		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2046		 * set.  This lets us use 0 value when this flag
2047		 * is set.
2048		 */
2049		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2050			params->spp_hbinterval = 0;
2051
2052		if (params->spp_hbinterval ||
2053		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2054			if (trans) {
2055				trans->hbinterval =
2056				    msecs_to_jiffies(params->spp_hbinterval);
2057			} else if (asoc) {
2058				asoc->hbinterval =
2059				    msecs_to_jiffies(params->spp_hbinterval);
2060			} else {
2061				sp->hbinterval = params->spp_hbinterval;
2062			}
2063		}
2064	}
2065
2066	if (hb_change) {
2067		if (trans) {
2068			trans->param_flags =
2069				(trans->param_flags & ~SPP_HB) | hb_change;
2070		} else if (asoc) {
2071			asoc->param_flags =
2072				(asoc->param_flags & ~SPP_HB) | hb_change;
2073		} else {
2074			sp->param_flags =
2075				(sp->param_flags & ~SPP_HB) | hb_change;
2076		}
2077	}
2078
2079	/* When Path MTU discovery is disabled the value specified here will
2080	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2081	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2082	 * effect).
2083	 */
2084	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2085		if (trans) {
2086			trans->pathmtu = params->spp_pathmtu;
2087			sctp_assoc_sync_pmtu(asoc);
2088		} else if (asoc) {
2089			asoc->pathmtu = params->spp_pathmtu;
2090			sctp_frag_point(sp, params->spp_pathmtu);
2091		} else {
2092			sp->pathmtu = params->spp_pathmtu;
2093		}
2094	}
2095
2096	if (pmtud_change) {
2097		if (trans) {
2098			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2099				(params->spp_flags & SPP_PMTUD_ENABLE);
2100			trans->param_flags =
2101				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2102			if (update) {
2103				sctp_transport_pmtu(trans);
2104				sctp_assoc_sync_pmtu(asoc);
2105			}
2106		} else if (asoc) {
2107			asoc->param_flags =
2108				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2109		} else {
2110			sp->param_flags =
2111				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2112		}
2113	}
2114
2115	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2116	 * value of this field is ignored.  Note also that a value of zero
2117	 * indicates the current setting should be left unchanged.
2118	 */
2119	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2120		if (trans) {
2121			trans->sackdelay =
2122				msecs_to_jiffies(params->spp_sackdelay);
2123		} else if (asoc) {
2124			asoc->sackdelay =
2125				msecs_to_jiffies(params->spp_sackdelay);
2126		} else {
2127			sp->sackdelay = params->spp_sackdelay;
2128		}
2129	}
2130
2131	if (sackdelay_change) {
2132		if (trans) {
2133			trans->param_flags =
2134				(trans->param_flags & ~SPP_SACKDELAY) |
2135				sackdelay_change;
2136		} else if (asoc) {
2137			asoc->param_flags =
2138				(asoc->param_flags & ~SPP_SACKDELAY) |
2139				sackdelay_change;
2140		} else {
2141			sp->param_flags =
2142				(sp->param_flags & ~SPP_SACKDELAY) |
2143				sackdelay_change;
2144		}
2145	}
2146
2147	/* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2148	 * of this field is ignored.  Note also that a value of zero
2149	 * indicates the current setting should be left unchanged.
2150	 */
2151	if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2152		if (trans) {
2153			trans->pathmaxrxt = params->spp_pathmaxrxt;
2154		} else if (asoc) {
2155			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2156		} else {
2157			sp->pathmaxrxt = params->spp_pathmaxrxt;
2158		}
2159	}
2160
2161	return 0;
2162}
2163
2164static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2165					    char __user *optval, int optlen)
2166{
2167	struct sctp_paddrparams  params;
2168	struct sctp_transport   *trans = NULL;
2169	struct sctp_association *asoc = NULL;
2170	struct sctp_sock        *sp = sctp_sk(sk);
2171	int error;
2172	int hb_change, pmtud_change, sackdelay_change;
2173
2174	if (optlen != sizeof(struct sctp_paddrparams))
2175		return - EINVAL;
2176
2177	if (copy_from_user(&params, optval, optlen))
2178		return -EFAULT;
2179
2180	/* Validate flags and value parameters. */
2181	hb_change        = params.spp_flags & SPP_HB;
2182	pmtud_change     = params.spp_flags & SPP_PMTUD;
2183	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2184
2185	if (hb_change        == SPP_HB ||
2186	    pmtud_change     == SPP_PMTUD ||
2187	    sackdelay_change == SPP_SACKDELAY ||
2188	    params.spp_sackdelay > 500 ||
2189	    (params.spp_pathmtu
2190	    && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2191		return -EINVAL;
2192
2193	/* If an address other than INADDR_ANY is specified, and
2194	 * no transport is found, then the request is invalid.
2195	 */
2196	if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2197		trans = sctp_addr_id2transport(sk, &params.spp_address,
2198					       params.spp_assoc_id);
2199		if (!trans)
2200			return -EINVAL;
2201	}
2202
2203	/* Get association, if assoc_id != 0 and the socket is a one
2204	 * to many style socket, and an association was not found, then
2205	 * the id was invalid.
2206	 */
2207	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2208	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2209		return -EINVAL;
2210
2211	/* Heartbeat demand can only be sent on a transport or
2212	 * association, but not a socket.
2213	 */
2214	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2215		return -EINVAL;
2216
2217	/* Process parameters. */
2218	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2219					    hb_change, pmtud_change,
2220					    sackdelay_change);
2221
2222	if (error)
2223		return error;
2224
2225	/* If changes are for association, also apply parameters to each
2226	 * transport.
2227	 */
2228	if (!trans && asoc) {
2229		struct list_head *pos;
2230
2231		list_for_each(pos, &asoc->peer.transport_addr_list) {
2232			trans = list_entry(pos, struct sctp_transport,
2233					   transports);
2234			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2235						    hb_change, pmtud_change,
2236						    sackdelay_change);
2237		}
2238	}
2239
2240	return 0;
2241}
2242
2243/* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2244 *
2245 *   This options will get or set the delayed ack timer.  The time is set
2246 *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
2247 *   endpoints default delayed ack timer value.  If the assoc_id field is
2248 *   non-zero, then the set or get effects the specified association.
2249 *
2250 *   struct sctp_assoc_value {
2251 *       sctp_assoc_t            assoc_id;
2252 *       uint32_t                assoc_value;
2253 *   };
2254 *
2255 *     assoc_id    - This parameter, indicates which association the
2256 *                   user is preforming an action upon. Note that if
2257 *                   this field's value is zero then the endpoints
2258 *                   default value is changed (effecting future
2259 *                   associations only).
2260 *
2261 *     assoc_value - This parameter contains the number of milliseconds
2262 *                   that the user is requesting the delayed ACK timer
2263 *                   be set to. Note that this value is defined in
2264 *                   the standard to be between 200 and 500 milliseconds.
2265 *
2266 *                   Note: a value of zero will leave the value alone,
2267 *                   but disable SACK delay. A non-zero value will also
2268 *                   enable SACK delay.
2269 */
2270
2271static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2272					    char __user *optval, int optlen)
2273{
2274	struct sctp_assoc_value  params;
2275	struct sctp_transport   *trans = NULL;
2276	struct sctp_association *asoc = NULL;
2277	struct sctp_sock        *sp = sctp_sk(sk);
2278
2279	if (optlen != sizeof(struct sctp_assoc_value))
2280		return - EINVAL;
2281
2282	if (copy_from_user(&params, optval, optlen))
2283		return -EFAULT;
2284
2285	/* Validate value parameter. */
2286	if (params.assoc_value > 500)
2287		return -EINVAL;
2288
2289	/* Get association, if assoc_id != 0 and the socket is a one
2290	 * to many style socket, and an association was not found, then
2291	 * the id was invalid.
2292	 */
2293	asoc = sctp_id2assoc(sk, params.assoc_id);
2294	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2295		return -EINVAL;
2296
2297	if (params.assoc_value) {
2298		if (asoc) {
2299			asoc->sackdelay =
2300				msecs_to_jiffies(params.assoc_value);
2301			asoc->param_flags =
2302				(asoc->param_flags & ~SPP_SACKDELAY) |
2303				SPP_SACKDELAY_ENABLE;
2304		} else {
2305			sp->sackdelay = params.assoc_value;
2306			sp->param_flags =
2307				(sp->param_flags & ~SPP_SACKDELAY) |
2308				SPP_SACKDELAY_ENABLE;
2309		}
2310	} else {
2311		if (asoc) {
2312			asoc->param_flags =
2313				(asoc->param_flags & ~SPP_SACKDELAY) |
2314				SPP_SACKDELAY_DISABLE;
2315		} else {
2316			sp->param_flags =
2317				(sp->param_flags & ~SPP_SACKDELAY) |
2318				SPP_SACKDELAY_DISABLE;
2319		}
2320	}
2321
2322	/* If change is for association, also apply to each transport. */
2323	if (asoc) {
2324		struct list_head *pos;
2325
2326		list_for_each(pos, &asoc->peer.transport_addr_list) {
2327			trans = list_entry(pos, struct sctp_transport,
2328					   transports);
2329			if (params.assoc_value) {
2330				trans->sackdelay =
2331					msecs_to_jiffies(params.assoc_value);
2332				trans->param_flags =
2333					(trans->param_flags & ~SPP_SACKDELAY) |
2334					SPP_SACKDELAY_ENABLE;
2335			} else {
2336				trans->param_flags =
2337					(trans->param_flags & ~SPP_SACKDELAY) |
2338					SPP_SACKDELAY_DISABLE;
2339			}
2340		}
2341	}
2342
2343	return 0;
2344}
2345
2346/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2347 *
2348 * Applications can specify protocol parameters for the default association
2349 * initialization.  The option name argument to setsockopt() and getsockopt()
2350 * is SCTP_INITMSG.
2351 *
2352 * Setting initialization parameters is effective only on an unconnected
2353 * socket (for UDP-style sockets only future associations are effected
2354 * by the change).  With TCP-style sockets, this option is inherited by
2355 * sockets derived from a listener socket.
2356 */
2357static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2358{
2359	struct sctp_initmsg sinit;
2360	struct sctp_sock *sp = sctp_sk(sk);
2361
2362	if (optlen != sizeof(struct sctp_initmsg))
2363		return -EINVAL;
2364	if (copy_from_user(&sinit, optval, optlen))
2365		return -EFAULT;
2366
2367	if (sinit.sinit_num_ostreams)
2368		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2369	if (sinit.sinit_max_instreams)
2370		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2371	if (sinit.sinit_max_attempts)
2372		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2373	if (sinit.sinit_max_init_timeo)
2374		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2375
2376	return 0;
2377}
2378
2379/*
2380 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2381 *
2382 *   Applications that wish to use the sendto() system call may wish to
2383 *   specify a default set of parameters that would normally be supplied
2384 *   through the inclusion of ancillary data.  This socket option allows
2385 *   such an application to set the default sctp_sndrcvinfo structure.
2386 *   The application that wishes to use this socket option simply passes
2387 *   in to this call the sctp_sndrcvinfo structure defined in Section
2388 *   5.2.2) The input parameters accepted by this call include
2389 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2390 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2391 *   to this call if the caller is using the UDP model.
2392 */
2393static int sctp_setsockopt_default_send_param(struct sock *sk,
2394						char __user *optval, int optlen)
2395{
2396	struct sctp_sndrcvinfo info;
2397	struct sctp_association *asoc;
2398	struct sctp_sock *sp = sctp_sk(sk);
2399
2400	if (optlen != sizeof(struct sctp_sndrcvinfo))
2401		return -EINVAL;
2402	if (copy_from_user(&info, optval, optlen))
2403		return -EFAULT;
2404
2405	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2406	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2407		return -EINVAL;
2408
2409	if (asoc) {
2410		asoc->default_stream = info.sinfo_stream;
2411		asoc->default_flags = info.sinfo_flags;
2412		asoc->default_ppid = info.sinfo_ppid;
2413		asoc->default_context = info.sinfo_context;
2414		asoc->default_timetolive = info.sinfo_timetolive;
2415	} else {
2416		sp->default_stream = info.sinfo_stream;
2417		sp->default_flags = info.sinfo_flags;
2418		sp->default_ppid = info.sinfo_ppid;
2419		sp->default_context = info.sinfo_context;
2420		sp->default_timetolive = info.sinfo_timetolive;
2421	}
2422
2423	return 0;
2424}
2425
2426/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2427 *
2428 * Requests that the local SCTP stack use the enclosed peer address as
2429 * the association primary.  The enclosed address must be one of the
2430 * association peer's addresses.
2431 */
2432static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2433					int optlen)
2434{
2435	struct sctp_prim prim;
2436	struct sctp_transport *trans;
2437
2438	if (optlen != sizeof(struct sctp_prim))
2439		return -EINVAL;
2440
2441	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2442		return -EFAULT;
2443
2444	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2445	if (!trans)
2446		return -EINVAL;
2447
2448	sctp_assoc_set_primary(trans->asoc, trans);
2449
2450	return 0;
2451}
2452
2453/*
2454 * 7.1.5 SCTP_NODELAY
2455 *
2456 * Turn on/off any Nagle-like algorithm.  This means that packets are
2457 * generally sent as soon as possible and no unnecessary delays are
2458 * introduced, at the cost of more packets in the network.  Expects an
2459 *  integer boolean flag.
2460 */
2461static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2462					int optlen)
2463{
2464	int val;
2465
2466	if (optlen < sizeof(int))
2467		return -EINVAL;
2468	if (get_user(val, (int __user *)optval))
2469		return -EFAULT;
2470
2471	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2472	return 0;
2473}
2474
2475/*
2476 *
2477 * 7.1.1 SCTP_RTOINFO
2478 *
2479 * The protocol parameters used to initialize and bound retransmission
2480 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2481 * and modify these parameters.
2482 * All parameters are time values, in milliseconds.  A value of 0, when
2483 * modifying the parameters, indicates that the current value should not
2484 * be changed.
2485 *
2486 */
2487static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2488	struct sctp_rtoinfo rtoinfo;
2489	struct sctp_association *asoc;
2490
2491	if (optlen != sizeof (struct sctp_rtoinfo))
2492		return -EINVAL;
2493
2494	if (copy_from_user(&rtoinfo, optval, optlen))
2495		return -EFAULT;
2496
2497	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2498
2499	/* Set the values to the specific association */
2500	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2501		return -EINVAL;
2502
2503	if (asoc) {
2504		if (rtoinfo.srto_initial != 0)
2505			asoc->rto_initial =
2506				msecs_to_jiffies(rtoinfo.srto_initial);
2507		if (rtoinfo.srto_max != 0)
2508			asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2509		if (rtoinfo.srto_min != 0)
2510			asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2511	} else {
2512		/* If there is no association or the association-id = 0
2513		 * set the values to the endpoint.
2514		 */
2515		struct sctp_sock *sp = sctp_sk(sk);
2516
2517		if (rtoinfo.srto_initial != 0)
2518			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2519		if (rtoinfo.srto_max != 0)
2520			sp->rtoinfo.srto_max = rtoinfo.srto_max;
2521		if (rtoinfo.srto_min != 0)
2522			sp->rtoinfo.srto_min = rtoinfo.srto_min;
2523	}
2524
2525	return 0;
2526}
2527
2528/*
2529 *
2530 * 7.1.2 SCTP_ASSOCINFO
2531 *
2532 * This option is used to tune the maximum retransmission attempts
2533 * of the association.
2534 * Returns an error if the new association retransmission value is
2535 * greater than the sum of the retransmission value  of the peer.
2536 * See [SCTP] for more information.
2537 *
2538 */
2539static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2540{
2541
2542	struct sctp_assocparams assocparams;
2543	struct sctp_association *asoc;
2544
2545	if (optlen != sizeof(struct sctp_assocparams))
2546		return -EINVAL;
2547	if (copy_from_user(&assocparams, optval, optlen))
2548		return -EFAULT;
2549
2550	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2551
2552	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2553		return -EINVAL;
2554
2555	/* Set the values to the specific association */
2556	if (asoc) {
2557		if (assocparams.sasoc_asocmaxrxt != 0) {
2558			__u32 path_sum = 0;
2559			int   paths = 0;
2560			struct list_head *pos;
2561			struct sctp_transport *peer_addr;
2562
2563			list_for_each(pos, &asoc->peer.transport_addr_list) {
2564				peer_addr = list_entry(pos,
2565						struct sctp_transport,
2566						transports);
2567				path_sum += peer_addr->pathmaxrxt;
2568				paths++;
2569			}
2570
2571			/* Only validate asocmaxrxt if we have more then
2572			 * one path/transport.  We do this because path
2573			 * retransmissions are only counted when we have more
2574			 * then one path.
2575			 */
2576			if (paths > 1 &&
2577			    assocparams.sasoc_asocmaxrxt > path_sum)
2578				return -EINVAL;
2579
2580			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2581		}
2582
2583		if (assocparams.sasoc_cookie_life != 0) {
2584			asoc->cookie_life.tv_sec =
2585					assocparams.sasoc_cookie_life / 1000;
2586			asoc->cookie_life.tv_usec =
2587					(assocparams.sasoc_cookie_life % 1000)
2588					* 1000;
2589		}
2590	} else {
2591		/* Set the values to the endpoint */
2592		struct sctp_sock *sp = sctp_sk(sk);
2593
2594		if (assocparams.sasoc_asocmaxrxt != 0)
2595			sp->assocparams.sasoc_asocmaxrxt =
2596						assocparams.sasoc_asocmaxrxt;
2597		if (assocparams.sasoc_cookie_life != 0)
2598			sp->assocparams.sasoc_cookie_life =
2599						assocparams.sasoc_cookie_life;
2600	}
2601	return 0;
2602}
2603
2604/*
2605 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2606 *
2607 * This socket option is a boolean flag which turns on or off mapped V4
2608 * addresses.  If this option is turned on and the socket is type
2609 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2610 * If this option is turned off, then no mapping will be done of V4
2611 * addresses and a user will receive both PF_INET6 and PF_INET type
2612 * addresses on the socket.
2613 */
2614static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2615{
2616	int val;
2617	struct sctp_sock *sp = sctp_sk(sk);
2618
2619	if (optlen < sizeof(int))
2620		return -EINVAL;
2621	if (get_user(val, (int __user *)optval))
2622		return -EFAULT;
2623	if (val)
2624		sp->v4mapped = 1;
2625	else
2626		sp->v4mapped = 0;
2627
2628	return 0;
2629}
2630
2631/*
2632 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2633 *
2634 * This socket option specifies the maximum size to put in any outgoing
2635 * SCTP chunk.  If a message is larger than this size it will be
2636 * fragmented by SCTP into the specified size.  Note that the underlying
2637 * SCTP implementation may fragment into smaller sized chunks when the
2638 * PMTU of the underlying association is smaller than the value set by
2639 * the user.
2640 */
2641static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2642{
2643	struct sctp_association *asoc;
2644	struct list_head *pos;
2645	struct sctp_sock *sp = sctp_sk(sk);
2646	int val;
2647
2648	if (optlen < sizeof(int))
2649		return -EINVAL;
2650	if (get_user(val, (int __user *)optval))
2651		return -EFAULT;
2652	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2653		return -EINVAL;
2654	sp->user_frag = val;
2655
2656	/* Update the frag_point of the existing associations. */
2657	list_for_each(pos, &(sp->ep->asocs)) {
2658		asoc = list_entry(pos, struct sctp_association, asocs);
2659		asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2660	}
2661
2662	return 0;
2663}
2664
2665
2666/*
2667 *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2668 *
2669 *   Requests that the peer mark the enclosed address as the association
2670 *   primary. The enclosed address must be one of the association's
2671 *   locally bound addresses. The following structure is used to make a
2672 *   set primary request:
2673 */
2674static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2675					     int optlen)
2676{
2677	struct sctp_sock	*sp;
2678	struct sctp_endpoint	*ep;
2679	struct sctp_association	*asoc = NULL;
2680	struct sctp_setpeerprim	prim;
2681	struct sctp_chunk	*chunk;
2682	int 			err;
2683
2684	sp = sctp_sk(sk);
2685	ep = sp->ep;
2686
2687	if (!sctp_addip_enable)
2688		return -EPERM;
2689
2690	if (optlen != sizeof(struct sctp_setpeerprim))
2691		return -EINVAL;
2692
2693	if (copy_from_user(&prim, optval, optlen))
2694		return -EFAULT;
2695
2696	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2697	if (!asoc)
2698		return -EINVAL;
2699
2700	if (!asoc->peer.asconf_capable)
2701		return -EPERM;
2702
2703	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2704		return -EPERM;
2705
2706	if (!sctp_state(asoc, ESTABLISHED))
2707		return -ENOTCONN;
2708
2709	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2710		return -EADDRNOTAVAIL;
2711
2712	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
2713	chunk = sctp_make_asconf_set_prim(asoc,
2714					  (union sctp_addr *)&prim.sspp_addr);
2715	if (!chunk)
2716		return -ENOMEM;
2717
2718	err = sctp_send_asconf(asoc, chunk);
2719
2720	SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2721
2722	return err;
2723}
2724
2725static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2726					  int optlen)
2727{
2728	struct sctp_setadaptation adaptation;
2729
2730	if (optlen != sizeof(struct sctp_setadaptation))
2731		return -EINVAL;
2732	if (copy_from_user(&adaptation, optval, optlen))
2733		return -EFAULT;
2734
2735	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2736
2737	return 0;
2738}
2739
2740/*
2741 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
2742 *
2743 * The context field in the sctp_sndrcvinfo structure is normally only
2744 * used when a failed message is retrieved holding the value that was
2745 * sent down on the actual send call.  This option allows the setting of
2746 * a default context on an association basis that will be received on
2747 * reading messages from the peer.  This is especially helpful in the
2748 * one-2-many model for an application to keep some reference to an
2749 * internal state machine that is processing messages on the
2750 * association.  Note that the setting of this value only effects
2751 * received messages from the peer and does not effect the value that is
2752 * saved with outbound messages.
2753 */
2754static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2755				   int optlen)
2756{
2757	struct sctp_assoc_value params;
2758	struct sctp_sock *sp;
2759	struct sctp_association *asoc;
2760
2761	if (optlen != sizeof(struct sctp_assoc_value))
2762		return -EINVAL;
2763	if (copy_from_user(&params, optval, optlen))
2764		return -EFAULT;
2765
2766	sp = sctp_sk(sk);
2767
2768	if (params.assoc_id != 0) {
2769		asoc = sctp_id2assoc(sk, params.assoc_id);
2770		if (!asoc)
2771			return -EINVAL;
2772		asoc->default_rcv_context = params.assoc_value;
2773	} else {
2774		sp->default_rcv_context = params.assoc_value;
2775	}
2776
2777	return 0;
2778}
2779
2780/*
2781 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2782 *
2783 * This options will at a minimum specify if the implementation is doing
2784 * fragmented interleave.  Fragmented interleave, for a one to many
2785 * socket, is when subsequent calls to receive a message may return
2786 * parts of messages from different associations.  Some implementations
2787 * may allow you to turn this value on or off.  If so, when turned off,
2788 * no fragment interleave will occur (which will cause a head of line
2789 * blocking amongst multiple associations sharing the same one to many
2790 * socket).  When this option is turned on, then each receive call may
2791 * come from a different association (thus the user must receive data
2792 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2793 * association each receive belongs to.
2794 *
2795 * This option takes a boolean value.  A non-zero value indicates that
2796 * fragmented interleave is on.  A value of zero indicates that
2797 * fragmented interleave is off.
2798 *
2799 * Note that it is important that an implementation that allows this
2800 * option to be turned on, have it off by default.  Otherwise an unaware
2801 * application using the one to many model may become confused and act
2802 * incorrectly.
2803 */
2804static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2805					       char __user *optval,
2806					       int optlen)
2807{
2808	int val;
2809
2810	if (optlen != sizeof(int))
2811		return -EINVAL;
2812	if (get_user(val, (int __user *)optval))
2813		return -EFAULT;
2814
2815	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2816
2817	return 0;
2818}
2819
2820/*
2821 * 7.1.25.  Set or Get the sctp partial delivery point
2822 *       (SCTP_PARTIAL_DELIVERY_POINT)
2823 * This option will set or get the SCTP partial delivery point.  This
2824 * point is the size of a message where the partial delivery API will be
2825 * invoked to help free up rwnd space for the peer.  Setting this to a
2826 * lower value will cause partial delivery's to happen more often.  The
2827 * calls argument is an integer that sets or gets the partial delivery
2828 * point.
2829 */
2830static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2831						  char __user *optval,
2832						  int optlen)
2833{
2834	u32 val;
2835
2836	if (optlen != sizeof(u32))
2837		return -EINVAL;
2838	if (get_user(val, (int __user *)optval))
2839		return -EFAULT;
2840
2841	sctp_sk(sk)->pd_point = val;
2842
2843	return 0; /* is this the right error code? */
2844}
2845
2846/*
2847 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
2848 *
2849 * This option will allow a user to change the maximum burst of packets
2850 * that can be emitted by this association.  Note that the default value
2851 * is 4, and some implementations may restrict this setting so that it
2852 * can only be lowered.
2853 *
2854 * NOTE: This text doesn't seem right.  Do this on a socket basis with
2855 * future associations inheriting the socket value.
2856 */
2857static int sctp_setsockopt_maxburst(struct sock *sk,
2858				    char __user *optval,
2859				    int optlen)
2860{
2861	int val;
2862
2863	if (optlen != sizeof(int))
2864		return -EINVAL;
2865	if (get_user(val, (int __user *)optval))
2866		return -EFAULT;
2867
2868	if (val < 0)
2869		return -EINVAL;
2870
2871	sctp_sk(sk)->max_burst = val;
2872
2873	return 0;
2874}
2875
2876/* API 6.2 setsockopt(), getsockopt()
2877 *
2878 * Applications use setsockopt() and getsockopt() to set or retrieve
2879 * socket options.  Socket options are used to change the default
2880 * behavior of sockets calls.  They are described in Section 7.
2881 *
2882 * The syntax is:
2883 *
2884 *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
2885 *                    int __user *optlen);
2886 *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2887 *                    int optlen);
2888 *
2889 *   sd      - the socket descript.
2890 *   level   - set to IPPROTO_SCTP for all SCTP options.
2891 *   optname - the option name.
2892 *   optval  - the buffer to store the value of the option.
2893 *   optlen  - the size of the buffer.
2894 */
2895SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2896				char __user *optval, int optlen)
2897{
2898	int retval = 0;
2899
2900	SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2901			  sk, optname);
2902
2903	/* I can hardly begin to describe how wrong this is.  This is
2904	 * so broken as to be worse than useless.  The API draft
2905	 * REALLY is NOT helpful here...  I am not convinced that the
2906	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2907	 * are at all well-founded.
2908	 */
2909	if (level != SOL_SCTP) {
2910		struct sctp_af *af = sctp_sk(sk)->pf->af;
2911		retval = af->setsockopt(sk, level, optname, optval, optlen);
2912		goto out_nounlock;
2913	}
2914
2915	sctp_lock_sock(sk);
2916
2917	switch (optname) {
2918	case SCTP_SOCKOPT_BINDX_ADD:
2919		/* 'optlen' is the size of the addresses buffer. */
2920		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2921					       optlen, SCTP_BINDX_ADD_ADDR);
2922		break;
2923
2924	case SCTP_SOCKOPT_BINDX_REM:
2925		/* 'optlen' is the size of the addresses buffer. */
2926		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2927					       optlen, SCTP_BINDX_REM_ADDR);
2928		break;
2929
2930	case SCTP_SOCKOPT_CONNECTX:
2931		/* 'optlen' is the size of the addresses buffer. */
2932		retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2933					       optlen);
2934		break;
2935
2936	case SCTP_DISABLE_FRAGMENTS:
2937		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2938		break;
2939
2940	case SCTP_EVENTS:
2941		retval = sctp_setsockopt_events(sk, optval, optlen);
2942		break;
2943
2944	case SCTP_AUTOCLOSE:
2945		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2946		break;
2947
2948	case SCTP_PEER_ADDR_PARAMS:
2949		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2950		break;
2951
2952	case SCTP_DELAYED_ACK_TIME:
2953		retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2954		break;
2955	case SCTP_PARTIAL_DELIVERY_POINT:
2956		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
2957		break;
2958
2959	case SCTP_INITMSG:
2960		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2961		break;
2962	case SCTP_DEFAULT_SEND_PARAM:
2963		retval = sctp_setsockopt_default_send_param(sk, optval,
2964							    optlen);
2965		break;
2966	case SCTP_PRIMARY_ADDR:
2967		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2968		break;
2969	case SCTP_SET_PEER_PRIMARY_ADDR:
2970		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2971		break;
2972	case SCTP_NODELAY:
2973		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2974		break;
2975	case SCTP_RTOINFO:
2976		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2977		break;
2978	case SCTP_ASSOCINFO:
2979		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2980		break;
2981	case SCTP_I_WANT_MAPPED_V4_ADDR:
2982		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2983		break;
2984	case SCTP_MAXSEG:
2985		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2986		break;
2987	case SCTP_ADAPTATION_LAYER:
2988		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
2989		break;
2990	case SCTP_CONTEXT:
2991		retval = sctp_setsockopt_context(sk, optval, optlen);
2992		break;
2993	case SCTP_FRAGMENT_INTERLEAVE:
2994		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
2995		break;
2996	case SCTP_MAX_BURST:
2997		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
2998		break;
2999	default:
3000		retval = -ENOPROTOOPT;
3001		break;
3002	}
3003
3004	sctp_release_sock(sk);
3005
3006out_nounlock:
3007	return retval;
3008}
3009
3010/* API 3.1.6 connect() - UDP Style Syntax
3011 *
3012 * An application may use the connect() call in the UDP model to initiate an
3013 * association without sending data.
3014 *
3015 * The syntax is:
3016 *
3017 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3018 *
3019 * sd: the socket descriptor to have a new association added to.
3020 *
3021 * nam: the address structure (either struct sockaddr_in or struct
3022 *    sockaddr_in6 defined in RFC2553 [7]).
3023 *
3024 * len: the size of the address.
3025 */
3026SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3027			     int addr_len)
3028{
3029	int err = 0;
3030	struct sctp_af *af;
3031
3032	sctp_lock_sock(sk);
3033
3034	SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3035			  __FUNCTION__, sk, addr, addr_len);
3036
3037	/* Validate addr_len before calling common connect/connectx routine. */
3038	af = sctp_get_af_specific(addr->sa_family);
3039	if (!af || addr_len < af->sockaddr_len) {
3040		err = -EINVAL;
3041	} else {
3042		/* Pass correct addr len to common routine (so it knows there
3043		 * is only one address being passed.
3044		 */
3045		err = __sctp_connect(sk, addr, af->sockaddr_len);
3046	}
3047
3048	sctp_release_sock(sk);
3049	return err;
3050}
3051
3052SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3053{
3054	return -EOPNOTSUPP; /* STUB */
3055}
3056
3057/* 4.1.4 accept() - TCP Style Syntax
3058 *
3059 * Applications use accept() call to remove an established SCTP
3060 * association from the accept queue of the endpoint.  A new socket
3061 * descriptor will be returned from accept() to represent the newly
3062 * formed association.
3063 */
3064SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3065{
3066	struct sctp_sock *sp;
3067	struct sctp_endpoint *ep;
3068	struct sock *newsk = NULL;
3069	struct sctp_association *asoc;
3070	long timeo;
3071	int error = 0;
3072
3073	sctp_lock_sock(sk);
3074
3075	sp = sctp_sk(sk);
3076	ep = sp->ep;
3077
3078	if (!sctp_style(sk, TCP)) {
3079		error = -EOPNOTSUPP;
3080		goto out;
3081	}
3082
3083	if (!sctp_sstate(sk, LISTENING)) {
3084		error = -EINVAL;
3085		goto out;
3086	}
3087
3088	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3089
3090	error = sctp_wait_for_accept(sk, timeo);
3091	if (error)
3092		goto out;
3093
3094	/* We treat the list of associations on the endpoint as the accept
3095	 * queue and pick the first association on the list.
3096	 */
3097	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3098
3099	newsk = sp->pf->create_accept_sk(sk, asoc);
3100	if (!newsk) {
3101		error = -ENOMEM;
3102		goto out;
3103	}
3104
3105	/* Populate the fields of the newsk from the oldsk and migrate the
3106	 * asoc to the newsk.
3107	 */
3108	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3109
3110out:
3111	sctp_release_sock(sk);
3112	*err = error;
3113	return newsk;
3114}
3115
3116/* The SCTP ioctl handler. */
3117SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3118{
3119	return -ENOIOCTLCMD;
3120}
3121
3122/* This is the function which gets called during socket creation to
3123 * initialized the SCTP-specific portion of the sock.
3124 * The sock structure should already be zero-filled memory.
3125 */
3126SCTP_STATIC int sctp_init_sock(struct sock *sk)
3127{
3128	struct sctp_endpoint *ep;
3129	struct sctp_sock *sp;
3130
3131	SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3132
3133	sp = sctp_sk(sk);
3134
3135	/* Initialize the SCTP per socket area.  */
3136	switch (sk->sk_type) {
3137	case SOCK_SEQPACKET:
3138		sp->type = SCTP_SOCKET_UDP;
3139		break;
3140	case SOCK_STREAM:
3141		sp->type = SCTP_SOCKET_TCP;
3142		break;
3143	default:
3144		return -ESOCKTNOSUPPORT;
3145	}
3146
3147	/* Initialize default send parameters. These parameters can be
3148	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3149	 */
3150	sp->default_stream = 0;
3151	sp->default_ppid = 0;
3152	sp->default_flags = 0;
3153	sp->default_context = 0;
3154	sp->default_timetolive = 0;
3155
3156	sp->default_rcv_context = 0;
3157	sp->max_burst = sctp_max_burst;
3158
3159	/* Initialize default setup parameters. These parameters
3160	 * can be modified with the SCTP_INITMSG socket option or
3161	 * overridden by the SCTP_INIT CMSG.
3162	 */
3163	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3164	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3165	sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3166	sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3167
3168	/* Initialize default RTO related parameters.  These parameters can
3169	 * be modified for with the SCTP_RTOINFO socket option.
3170	 */
3171	sp->rtoinfo.srto_initial = sctp_rto_initial;
3172	sp->rtoinfo.srto_max     = sctp_rto_max;
3173	sp->rtoinfo.srto_min     = sctp_rto_min;
3174
3175	/* Initialize default association related parameters. These parameters
3176	 * can be modified with the SCTP_ASSOCINFO socket option.
3177	 */
3178	sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3179	sp->assocparams.sasoc_number_peer_destinations = 0;
3180	sp->assocparams.sasoc_peer_rwnd = 0;
3181	sp->assocparams.sasoc_local_rwnd = 0;
3182	sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3183
3184	/* Initialize default event subscriptions. By default, all the
3185	 * options are off.
3186	 */
3187	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3188
3189	/* Default Peer Address Parameters.  These defaults can
3190	 * be modified via SCTP_PEER_ADDR_PARAMS
3191	 */
3192	sp->hbinterval  = sctp_hb_interval;
3193	sp->pathmaxrxt  = sctp_max_retrans_path;
3194	sp->pathmtu     = 0; // allow default discovery
3195	sp->sackdelay   = sctp_sack_timeout;
3196	sp->param_flags = SPP_HB_ENABLE |
3197			  SPP_PMTUD_ENABLE |
3198			  SPP_SACKDELAY_ENABLE;
3199
3200	/* If enabled no SCTP message fragmentation will be performed.
3201	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3202	 */
3203	sp->disable_fragments = 0;
3204
3205	/* Enable Nagle algorithm by default.  */
3206	sp->nodelay           = 0;
3207
3208	/* Enable by default. */
3209	sp->v4mapped          = 1;
3210
3211	/* Auto-close idle associations after the configured
3212	 * number of seconds.  A value of 0 disables this
3213	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3214	 * for UDP-style sockets only.
3215	 */
3216	sp->autoclose         = 0;
3217
3218	/* User specified fragmentation limit. */
3219	sp->user_frag         = 0;
3220
3221	sp->adaptation_ind = 0;
3222
3223	sp->pf = sctp_get_pf_specific(sk->sk_family);
3224
3225	/* Control variables for partial data delivery. */
3226	atomic_set(&sp->pd_mode, 0);
3227	skb_queue_head_init(&sp->pd_lobby);
3228	sp->frag_interleave = 0;
3229
3230	/* Create a per socket endpoint structure.  Even if we
3231	 * change the data structure relationships, this may still
3232	 * be useful for storing pre-connect address information.
3233	 */
3234	ep = sctp_endpoint_new(sk, GFP_KERNEL);
3235	if (!ep)
3236		return -ENOMEM;
3237
3238	sp->ep = ep;
3239	sp->hmac = NULL;
3240
3241	SCTP_DBG_OBJCNT_INC(sock);
3242	return 0;
3243}
3244
3245/* Cleanup any SCTP per socket resources.  */
3246SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3247{
3248	struct sctp_endpoint *ep;
3249
3250	SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3251
3252	/* Release our hold on the endpoint. */
3253	ep = sctp_sk(sk)->ep;
3254	sctp_endpoint_free(ep);
3255
3256	return 0;
3257}
3258
3259/* API 4.1.7 shutdown() - TCP Style Syntax
3260 *     int shutdown(int socket, int how);
3261 *
3262 *     sd      - the socket descriptor of the association to be closed.
3263 *     how     - Specifies the type of shutdown.  The  values  are
3264 *               as follows:
3265 *               SHUT_RD
3266 *                     Disables further receive operations. No SCTP
3267 *                     protocol action is taken.
3268 *               SHUT_WR
3269 *                     Disables further send operations, and initiates
3270 *                     the SCTP shutdown sequence.
3271 *               SHUT_RDWR
3272 *                     Disables further send  and  receive  operations
3273 *                     and initiates the SCTP shutdown sequence.
3274 */
3275SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3276{
3277	struct sctp_endpoint *ep;
3278	struct sctp_association *asoc;
3279
3280	if (!sctp_style(sk, TCP))
3281		return;
3282
3283	if (how & SEND_SHUTDOWN) {
3284		ep = sctp_sk(sk)->ep;
3285		if (!list_empty(&ep->asocs)) {
3286			asoc = list_entry(ep->asocs.next,
3287					  struct sctp_association, asocs);
3288			sctp_primitive_SHUTDOWN(asoc, NULL);
3289		}
3290	}
3291}
3292
3293/* 7.2.1 Association Status (SCTP_STATUS)
3294
3295 * Applications can retrieve current status information about an
3296 * association, including association state, peer receiver window size,
3297 * number of unacked data chunks, and number of data chunks pending
3298 * receipt.  This information is read-only.
3299 */
3300static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3301				       char __user *optval,
3302				       int __user *optlen)
3303{
3304	struct sctp_status status;
3305	struct sctp_association *asoc = NULL;
3306	struct sctp_transport *transport;
3307	sctp_assoc_t associd;
3308	int retval = 0;
3309
3310	if (len < sizeof(status)) {
3311		retval = -EINVAL;
3312		goto out;
3313	}
3314
3315	len = sizeof(status);
3316	if (copy_from_user(&status, optval, len)) {
3317		retval = -EFAULT;
3318		goto out;
3319	}
3320
3321	associd = status.sstat_assoc_id;
3322	asoc = sctp_id2assoc(sk, associd);
3323	if (!asoc) {
3324		retval = -EINVAL;
3325		goto out;
3326	}
3327
3328	transport = asoc->peer.primary_path;
3329
3330	status.sstat_assoc_id = sctp_assoc2id(asoc);
3331	status.sstat_state = asoc->state;
3332	status.sstat_rwnd =  asoc->peer.rwnd;
3333	status.sstat_unackdata = asoc->unack_data;
3334
3335	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3336	status.sstat_instrms = asoc->c.sinit_max_instreams;
3337	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3338	status.sstat_fragmentation_point = asoc->frag_point;
3339	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3340	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3341			transport->af_specific->sockaddr_len);
3342	/* Map ipv4 address into v4-mapped-on-v6 address.  */
3343	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3344		(union sctp_addr *)&status.sstat_primary.spinfo_address);
3345	status.sstat_primary.spinfo_state = transport->state;
3346	status.sstat_primary.spinfo_cwnd = transport->cwnd;
3347	status.sstat_primary.spinfo_srtt = transport->srtt;
3348	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3349	status.sstat_primary.spinfo_mtu = transport->pathmtu;
3350
3351	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3352		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3353
3354	if (put_user(len, optlen)) {
3355		retval = -EFAULT;
3356		goto out;
3357	}
3358
3359	SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3360			  len, status.sstat_state, status.sstat_rwnd,
3361			  status.sstat_assoc_id);
3362
3363	if (copy_to_user(optval, &status, len)) {
3364		retval = -EFAULT;
3365		goto out;
3366	}
3367
3368out:
3369	return (retval);
3370}
3371
3372
3373/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3374 *
3375 * Applications can retrieve information about a specific peer address
3376 * of an association, including its reachability state, congestion
3377 * window, and retransmission timer values.  This information is
3378 * read-only.
3379 */
3380static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3381					  char __user *optval,
3382					  int __user *optlen)
3383{
3384	struct sctp_paddrinfo pinfo;
3385	struct sctp_transport *transport;
3386	int retval = 0;
3387
3388	if (len < sizeof(pinfo)) {
3389		retval = -EINVAL;
3390		goto out;
3391	}
3392
3393	len = sizeof(pinfo);
3394	if (copy_from_user(&pinfo, optval, len)) {
3395		retval = -EFAULT;
3396		goto out;
3397	}
3398
3399	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3400					   pinfo.spinfo_assoc_id);
3401	if (!transport)
3402		return -EINVAL;
3403
3404	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3405	pinfo.spinfo_state = transport->state;
3406	pinfo.spinfo_cwnd = transport->cwnd;
3407	pinfo.spinfo_srtt = transport->srtt;
3408	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3409	pinfo.spinfo_mtu = transport->pathmtu;
3410
3411	if (pinfo.spinfo_state == SCTP_UNKNOWN)
3412		pinfo.spinfo_state = SCTP_ACTIVE;
3413
3414	if (put_user(len, optlen)) {
3415		retval = -EFAULT;
3416		goto out;
3417	}
3418
3419	if (copy_to_user(optval, &pinfo, len)) {
3420		retval = -EFAULT;
3421		goto out;
3422	}
3423
3424out:
3425	return (retval);
3426}
3427
3428/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3429 *
3430 * This option is a on/off flag.  If enabled no SCTP message
3431 * fragmentation will be performed.  Instead if a message being sent
3432 * exceeds the current PMTU size, the message will NOT be sent and
3433 * instead a error will be indicated to the user.
3434 */
3435static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3436					char __user *optval, int __user *optlen)
3437{
3438	int val;
3439
3440	if (len < sizeof(int))
3441		return -EINVAL;
3442
3443	len = sizeof(int);
3444	val = (sctp_sk(sk)->disable_fragments == 1);
3445	if (put_user(len, optlen))
3446		return -EFAULT;
3447	if (copy_to_user(optval, &val, len))
3448		return -EFAULT;
3449	return 0;
3450}
3451
3452/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3453 *
3454 * This socket option is used to specify various notifications and
3455 * ancillary data the user wishes to receive.
3456 */
3457static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3458				  int __user *optlen)
3459{
3460	if (len < sizeof(struct sctp_event_subscribe))
3461		return -EINVAL;
3462	len = sizeof(struct sctp_event_subscribe);
3463	if (put_user(len, optlen))
3464		return -EFAULT;
3465	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3466		return -EFAULT;
3467	return 0;
3468}
3469
3470/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3471 *
3472 * This socket option is applicable to the UDP-style socket only.  When
3473 * set it will cause associations that are idle for more than the
3474 * specified number of seconds to automatically close.  An association
3475 * being idle is defined an association that has NOT sent or received
3476 * user data.  The special value of '0' indicates that no automatic
3477 * close of any associations should be performed.  The option expects an
3478 * integer defining the number of seconds of idle time before an
3479 * association is closed.
3480 */
3481static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3482{
3483	/* Applicable to UDP-style socket only */
3484	if (sctp_style(sk, TCP))
3485		return -EOPNOTSUPP;
3486	if (len < sizeof(int))
3487		return -EINVAL;
3488	len = sizeof(int);
3489	if (put_user(len, optlen))
3490		return -EFAULT;
3491	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3492		return -EFAULT;
3493	return 0;
3494}
3495
3496/* Helper routine to branch off an association to a new socket.  */
3497SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3498				struct socket **sockp)
3499{
3500	struct sock *sk = asoc->base.sk;
3501	struct socket *sock;
3502	struct inet_sock *inetsk;
3503	struct sctp_af *af;
3504	int err = 0;
3505
3506	/* An association cannot be branched off from an already peeled-off
3507	 * socket, nor is this supported for tcp style sockets.
3508	 */
3509	if (!sctp_style(sk, UDP))
3510		return -EINVAL;
3511
3512	/* Create a new socket.  */
3513	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3514	if (err < 0)
3515		return err;
3516
3517	/* Populate the fields of the newsk from the oldsk and migrate the
3518	 * asoc to the newsk.
3519	 */
3520	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3521
3522	/* Make peeled-off sockets more like 1-1 accepted sockets.
3523	 * Set the daddr and initialize id to something more random
3524	 */
3525	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3526	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3527	inetsk = inet_sk(sock->sk);
3528	inetsk->id = asoc->next_tsn ^ jiffies;
3529
3530	*sockp = sock;
3531
3532	return err;
3533}
3534
3535static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3536{
3537	sctp_peeloff_arg_t peeloff;
3538	struct socket *newsock;
3539	int retval = 0;
3540	struct sctp_association *asoc;
3541
3542	if (len < sizeof(sctp_peeloff_arg_t))
3543		return -EINVAL;
3544	len = sizeof(sctp_peeloff_arg_t);
3545	if (copy_from_user(&peeloff, optval, len))
3546		return -EFAULT;
3547
3548	asoc = sctp_id2assoc(sk, peeloff.associd);
3549	if (!asoc) {
3550		retval = -EINVAL;
3551		goto out;
3552	}
3553
3554	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3555
3556	retval = sctp_do_peeloff(asoc, &newsock);
3557	if (retval < 0)
3558		goto out;
3559
3560	/* Map the socket to an unused fd that can be returned to the user.  */
3561	retval = sock_map_fd(newsock);
3562	if (retval < 0) {
3563		sock_release(newsock);
3564		goto out;
3565	}
3566
3567	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3568			  __FUNCTION__, sk, asoc, newsock->sk, retval);
3569
3570	/* Return the fd mapped to the new socket.  */
3571	peeloff.sd = retval;
3572	if (put_user(len, optlen))
3573		return -EFAULT;
3574	if (copy_to_user(optval, &peeloff, len))
3575		retval = -EFAULT;
3576
3577out:
3578	return retval;
3579}
3580
3581/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3582 *
3583 * Applications can enable or disable heartbeats for any peer address of
3584 * an association, modify an address's heartbeat interval, force a
3585 * heartbeat to be sent immediately, and adjust the address's maximum
3586 * number of retransmissions sent before an address is considered
3587 * unreachable.  The following structure is used to access and modify an
3588 * address's parameters:
3589 *
3590 *  struct sctp_paddrparams {
3591 *     sctp_assoc_t            spp_assoc_id;
3592 *     struct sockaddr_storage spp_address;
3593 *     uint32_t                spp_hbinterval;
3594 *     uint16_t                spp_pathmaxrxt;
3595 *     uint32_t                spp_pathmtu;
3596 *     uint32_t                spp_sackdelay;
3597 *     uint32_t                spp_flags;
3598 * };
3599 *
3600 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
3601 *                     application, and identifies the association for
3602 *                     this query.
3603 *   spp_address     - This specifies which address is of interest.
3604 *   spp_hbinterval  - This contains the value of the heartbeat interval,
3605 *                     in milliseconds.  If a  value of zero
3606 *                     is present in this field then no changes are to
3607 *                     be made to this parameter.
3608 *   spp_pathmaxrxt  - This contains the maximum number of
3609 *                     retransmissions before this address shall be
3610 *                     considered unreachable. If a  value of zero
3611 *                     is present in this field then no changes are to
3612 *                     be made to this parameter.
3613 *   spp_pathmtu     - When Path MTU discovery is disabled the value
3614 *                     specified here will be the "fixed" path mtu.
3615 *                     Note that if the spp_address field is empty
3616 *                     then all associations on this address will
3617 *                     have this fixed path mtu set upon them.
3618 *
3619 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
3620 *                     the number of milliseconds that sacks will be delayed
3621 *                     for. This value will apply to all addresses of an
3622 *                     association if the spp_address field is empty. Note
3623 *                     also, that if delayed sack is enabled and this
3624 *                     value is set to 0, no change is made to the last
3625 *                     recorded delayed sack timer value.
3626 *
3627 *   spp_flags       - These flags are used to control various features
3628 *                     on an association. The flag field may contain
3629 *                     zero or more of the following options.
3630 *
3631 *                     SPP_HB_ENABLE  - Enable heartbeats on the
3632 *                     specified address. Note that if the address
3633 *                     field is empty all addresses for the association
3634 *                     have heartbeats enabled upon them.
3635 *
3636 *                     SPP_HB_DISABLE - Disable heartbeats on the
3637 *                     speicifed address. Note that if the address
3638 *                     field is empty all addresses for the association
3639 *                     will have their heartbeats disabled. Note also
3640 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
3641 *                     mutually exclusive, only one of these two should
3642 *                     be specified. Enabling both fields will have
3643 *                     undetermined results.
3644 *
3645 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
3646 *                     to be made immediately.
3647 *
3648 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
3649 *                     discovery upon the specified address. Note that
3650 *                     if the address feild is empty then all addresses
3651 *                     on the association are effected.
3652 *
3653 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
3654 *                     discovery upon the specified address. Note that
3655 *                     if the address feild is empty then all addresses
3656 *                     on the association are effected. Not also that
3657 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3658 *                     exclusive. Enabling both will have undetermined
3659 *                     results.
3660 *
3661 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
3662 *                     on delayed sack. The time specified in spp_sackdelay
3663 *                     is used to specify the sack delay for this address. Note
3664 *                     that if spp_address is empty then all addresses will
3665 *                     enable delayed sack and take on the sack delay
3666 *                     value specified in spp_sackdelay.
3667 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
3668 *                     off delayed sack. If the spp_address field is blank then
3669 *                     delayed sack is disabled for the entire association. Note
3670 *                     also that this field is mutually exclusive to
3671 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
3672 *                     results.
3673 */
3674static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3675					    char __user *optval, int __user *optlen)
3676{
3677	struct sctp_paddrparams  params;
3678	struct sctp_transport   *trans = NULL;
3679	struct sctp_association *asoc = NULL;
3680	struct sctp_sock        *sp = sctp_sk(sk);
3681
3682	if (len < sizeof(struct sctp_paddrparams))
3683		return -EINVAL;
3684	len = sizeof(struct sctp_paddrparams);
3685	if (copy_from_user(&params, optval, len))
3686		return -EFAULT;
3687
3688	/* If an address other than INADDR_ANY is specified, and
3689	 * no transport is found, then the request is invalid.
3690	 */
3691	if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3692		trans = sctp_addr_id2transport(sk, &params.spp_address,
3693					       params.spp_assoc_id);
3694		if (!trans) {
3695			SCTP_DEBUG_PRINTK("Failed no transport\n");
3696			return -EINVAL;
3697		}
3698	}
3699
3700	/* Get association, if assoc_id != 0 and the socket is a one
3701	 * to many style socket, and an association was not found, then
3702	 * the id was invalid.
3703	 */
3704	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3705	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3706		SCTP_DEBUG_PRINTK("Failed no association\n");
3707		return -EINVAL;
3708	}
3709
3710	if (trans) {
3711		/* Fetch transport values. */
3712		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3713		params.spp_pathmtu    = trans->pathmtu;
3714		params.spp_pathmaxrxt = trans->pathmaxrxt;
3715		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
3716
3717		/*draft-11 doesn't say what to return in spp_flags*/
3718		params.spp_flags      = trans->param_flags;
3719	} else if (asoc) {
3720		/* Fetch association values. */
3721		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3722		params.spp_pathmtu    = asoc->pathmtu;
3723		params.spp_pathmaxrxt = asoc->pathmaxrxt;
3724		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
3725
3726		/*draft-11 doesn't say what to return in spp_flags*/
3727		params.spp_flags      = asoc->param_flags;
3728	} else {
3729		/* Fetch socket values. */
3730		params.spp_hbinterval = sp->hbinterval;
3731		params.spp_pathmtu    = sp->pathmtu;
3732		params.spp_sackdelay  = sp->sackdelay;
3733		params.spp_pathmaxrxt = sp->pathmaxrxt;
3734
3735		/*draft-11 doesn't say what to return in spp_flags*/
3736		params.spp_flags      = sp->param_flags;
3737	}
3738
3739	if (copy_to_user(optval, &params, len))
3740		return -EFAULT;
3741
3742	if (put_user(len, optlen))
3743		return -EFAULT;
3744
3745	return 0;
3746}
3747
3748/* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3749 *
3750 *   This options will get or set the delayed ack timer.  The time is set
3751 *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
3752 *   endpoints default delayed ack timer value.  If the assoc_id field is
3753 *   non-zero, then the set or get effects the specified association.
3754 *
3755 *   struct sctp_assoc_value {
3756 *       sctp_assoc_t            assoc_id;
3757 *       uint32_t                assoc_value;
3758 *   };
3759 *
3760 *     assoc_id    - This parameter, indicates which association the
3761 *                   user is preforming an action upon. Note that if
3762 *                   this field's value is zero then the endpoints
3763 *                   default value is changed (effecting future
3764 *                   associations only).
3765 *
3766 *     assoc_value - This parameter contains the number of milliseconds
3767 *                   that the user is requesting the delayed ACK timer
3768 *                   be set to. Note that this value is defined in
3769 *                   the standard to be between 200 and 500 milliseconds.
3770 *
3771 *                   Note: a value of zero will leave the value alone,
3772 *                   but disable SACK delay. A non-zero value will also
3773 *                   enable SACK delay.
3774 */
3775static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3776					    char __user *optval,
3777					    int __user *optlen)
3778{
3779	struct sctp_assoc_value  params;
3780	struct sctp_association *asoc = NULL;
3781	struct sctp_sock        *sp = sctp_sk(sk);
3782
3783	if (len < sizeof(struct sctp_assoc_value))
3784		return - EINVAL;
3785
3786	len = sizeof(struct sctp_assoc_value);
3787
3788	if (copy_from_user(&params, optval, len))
3789		return -EFAULT;
3790
3791	/* Get association, if assoc_id != 0 and the socket is a one
3792	 * to many style socket, and an association was not found, then
3793	 * the id was invalid.
3794	 */
3795	asoc = sctp_id2assoc(sk, params.assoc_id);
3796	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3797		return -EINVAL;
3798
3799	if (asoc) {
3800		/* Fetch association values. */
3801		if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3802			params.assoc_value = jiffies_to_msecs(
3803				asoc->sackdelay);
3804		else
3805			params.assoc_value = 0;
3806	} else {
3807		/* Fetch socket values. */
3808		if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3809			params.assoc_value  = sp->sackdelay;
3810		else
3811			params.assoc_value  = 0;
3812	}
3813
3814	if (copy_to_user(optval, &params, len))
3815		return -EFAULT;
3816
3817	if (put_user(len, optlen))
3818		return -EFAULT;
3819
3820	return 0;
3821}
3822
3823/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3824 *
3825 * Applications can specify protocol parameters for the default association
3826 * initialization.  The option name argument to setsockopt() and getsockopt()
3827 * is SCTP_INITMSG.
3828 *
3829 * Setting initialization parameters is effective only on an unconnected
3830 * socket (for UDP-style sockets only future associations are effected
3831 * by the change).  With TCP-style sockets, this option is inherited by
3832 * sockets derived from a listener socket.
3833 */
3834static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3835{
3836	if (len < sizeof(struct sctp_initmsg))
3837		return -EINVAL;
3838	len = sizeof(struct sctp_initmsg);
3839	if (put_user(len, optlen))
3840		return -EFAULT;
3841	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3842		return -EFAULT;
3843	return 0;
3844}
3845
3846static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3847					      char __user *optval,
3848					      int __user *optlen)
3849{
3850	sctp_assoc_t id;
3851	struct sctp_association *asoc;
3852	struct list_head *pos;
3853	int cnt = 0;
3854
3855	if (len < sizeof(sctp_assoc_t))
3856		return -EINVAL;
3857
3858	if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3859		return -EFAULT;
3860
3861	/* For UDP-style sockets, id specifies the association to query.  */
3862	asoc = sctp_id2assoc(sk, id);
3863	if (!asoc)
3864		return -EINVAL;
3865
3866	list_for_each(pos, &asoc->peer.transport_addr_list) {
3867		cnt ++;
3868	}
3869
3870	return cnt;
3871}
3872
3873/*
3874 * Old API for getting list of peer addresses. Does not work for 32-bit
3875 * programs running on a 64-bit kernel
3876 */
3877static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3878					  char __user *optval,
3879					  int __user *optlen)
3880{
3881	struct sctp_association *asoc;
3882	struct list_head *pos;
3883	int cnt = 0;
3884	struct sctp_getaddrs_old getaddrs;
3885	struct sctp_transport *from;
3886	void __user *to;
3887	union sctp_addr temp;
3888	struct sctp_sock *sp = sctp_sk(sk);
3889	int addrlen;
3890
3891	if (len < sizeof(struct sctp_getaddrs_old))
3892		return -EINVAL;
3893
3894	len = sizeof(struct sctp_getaddrs_old);
3895
3896	if (copy_from_user(&getaddrs, optval, len))
3897		return -EFAULT;
3898
3899	if (getaddrs.addr_num <= 0) return -EINVAL;
3900
3901	/* For UDP-style sockets, id specifies the association to query.  */
3902	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3903	if (!asoc)
3904		return -EINVAL;
3905
3906	to = (void __user *)getaddrs.addrs;
3907	list_for_each(pos, &asoc->peer.transport_addr_list) {
3908		from = list_entry(pos, struct sctp_transport, transports);
3909		memcpy(&temp, &from->ipaddr, sizeof(temp));
3910		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3911		addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3912		if (copy_to_user(to, &temp, addrlen))
3913			return -EFAULT;
3914		to += addrlen ;
3915		cnt ++;
3916		if (cnt >= getaddrs.addr_num) break;
3917	}
3918	getaddrs.addr_num = cnt;
3919	if (put_user(len, optlen))
3920		return -EFAULT;
3921	if (copy_to_user(optval, &getaddrs, len))
3922		return -EFAULT;
3923
3924	return 0;
3925}
3926
3927static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3928				      char __user *optval, int __user *optlen)
3929{
3930	struct sctp_association *asoc;
3931	struct list_head *pos;
3932	int cnt = 0;
3933	struct sctp_getaddrs getaddrs;
3934	struct sctp_transport *from;
3935	void __user *to;
3936	union sctp_addr temp;
3937	struct sctp_sock *sp = sctp_sk(sk);
3938	int addrlen;
3939	size_t space_left;
3940	int bytes_copied;
3941
3942	if (len < sizeof(struct sctp_getaddrs))
3943		return -EINVAL;
3944
3945	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3946		return -EFAULT;
3947
3948	/* For UDP-style sockets, id specifies the association to query.  */
3949	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3950	if (!asoc)
3951		return -EINVAL;
3952
3953	to = optval + offsetof(struct sctp_getaddrs,addrs);
3954	space_left = len - offsetof(struct sctp_getaddrs,addrs);
3955
3956	list_for_each(pos, &asoc->peer.transport_addr_list) {
3957		from = list_entry(pos, struct sctp_transport, transports);
3958		memcpy(&temp, &from->ipaddr, sizeof(temp));
3959		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3960		addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3961		if (space_left < addrlen)
3962			return -ENOMEM;
3963		if (copy_to_user(to, &temp, addrlen))
3964			return -EFAULT;
3965		to += addrlen;
3966		cnt++;
3967		space_left -= addrlen;
3968	}
3969
3970	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3971		return -EFAULT;
3972	bytes_copied = ((char __user *)to) - optval;
3973	if (put_user(bytes_copied, optlen))
3974		return -EFAULT;
3975
3976	return 0;
3977}
3978
3979static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3980					       char __user *optval,
3981					       int __user *optlen)
3982{
3983	sctp_assoc_t id;
3984	struct sctp_bind_addr *bp;
3985	struct sctp_association *asoc;
3986	struct list_head *pos, *temp;
3987	struct sctp_sockaddr_entry *addr;
3988	rwlock_t *addr_lock;
3989	int cnt = 0;
3990
3991	if (len < sizeof(sctp_assoc_t))
3992		return -EINVAL;
3993
3994	if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3995		return -EFAULT;
3996
3997	/*
3998	 *  For UDP-style sockets, id specifies the association to query.
3999	 *  If the id field is set to the value '0' then the locally bound
4000	 *  addresses are returned without regard to any particular
4001	 *  association.
4002	 */
4003	if (0 == id) {
4004		bp = &sctp_sk(sk)->ep->base.bind_addr;
4005		addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4006	} else {
4007		asoc = sctp_id2assoc(sk, id);
4008		if (!asoc)
4009			return -EINVAL;
4010		bp = &asoc->base.bind_addr;
4011		addr_lock = &asoc->base.addr_lock;
4012	}
4013
4014	sctp_read_lock(addr_lock);
4015
4016	/* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4017	 * addresses from the global local address list.
4018	 */
4019	if (sctp_list_single_entry(&bp->address_list)) {
4020		addr = list_entry(bp->address_list.next,
4021				  struct sctp_sockaddr_entry, list);
4022		if (sctp_is_any(&addr->a)) {
4023			list_for_each_safe(pos, temp, &sctp_local_addr_list) {
4024				addr = list_entry(pos,
4025						  struct sctp_sockaddr_entry,
4026						  list);
4027				if ((PF_INET == sk->sk_family) &&
4028				    (AF_INET6 == addr->a.sa.sa_family))
4029					continue;
4030				cnt++;
4031			}
4032		} else {
4033			cnt = 1;
4034		}
4035		goto done;
4036	}
4037
4038	list_for_each(pos, &bp->address_list) {
4039		cnt ++;
4040	}
4041
4042done:
4043	sctp_read_unlock(addr_lock);
4044	return cnt;
4045}
4046
4047/* Helper function that copies local addresses to user and returns the number
4048 * of addresses copied.
4049 */
4050static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4051					int max_addrs, void *to,
4052					int *bytes_copied)
4053{
4054	struct list_head *pos, *next;
4055	struct sctp_sockaddr_entry *addr;
4056	union sctp_addr temp;
4057	int cnt = 0;
4058	int addrlen;
4059
4060	list_for_each_safe(pos, next, &sctp_local_addr_list) {
4061		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4062		if ((PF_INET == sk->sk_family) &&
4063		    (AF_INET6 == addr->a.sa.sa_family))
4064			continue;
4065		memcpy(&temp, &addr->a, sizeof(temp));
4066		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4067								&temp);
4068		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4069		memcpy(to, &temp, addrlen);
4070
4071		to += addrlen;
4072		*bytes_copied += addrlen;
4073		cnt ++;
4074		if (cnt >= max_addrs) break;
4075	}
4076
4077	return cnt;
4078}
4079
4080static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4081			    size_t space_left, int *bytes_copied)
4082{
4083	struct list_head *pos, *next;
4084	struct sctp_sockaddr_entry *addr;
4085	union sctp_addr temp;
4086	int cnt = 0;
4087	int addrlen;
4088
4089	list_for_each_safe(pos, next, &sctp_local_addr_list) {
4090		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4091		if ((PF_INET == sk->sk_family) &&
4092		    (AF_INET6 == addr->a.sa.sa_family))
4093			continue;
4094		memcpy(&temp, &addr->a, sizeof(temp));
4095		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4096								&temp);
4097		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4098		if (space_left < addrlen)
4099			return -ENOMEM;
4100		memcpy(to, &temp, addrlen);
4101
4102		to += addrlen;
4103		cnt ++;
4104		space_left -= addrlen;
4105		*bytes_copied += addrlen;
4106	}
4107
4108	return cnt;
4109}
4110
4111/* Old API for getting list of local addresses. Does not work for 32-bit
4112 * programs running on a 64-bit kernel
4113 */
4114static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4115					   char __user *optval, int __user *optlen)
4116{
4117	struct sctp_bind_addr *bp;
4118	struct sctp_association *asoc;
4119	struct list_head *pos;
4120	int cnt = 0;
4121	struct sctp_getaddrs_old getaddrs;
4122	struct sctp_sockaddr_entry *addr;
4123	void __user *to;
4124	union sctp_addr temp;
4125	struct sctp_sock *sp = sctp_sk(sk);
4126	int addrlen;
4127	rwlock_t *addr_lock;
4128	int err = 0;
4129	void *addrs;
4130	void *buf;
4131	int bytes_copied = 0;
4132
4133	if (len < sizeof(struct sctp_getaddrs_old))
4134		return -EINVAL;
4135
4136	len = sizeof(struct sctp_getaddrs_old);
4137	if (copy_from_user(&getaddrs, optval, len))
4138		return -EFAULT;
4139
4140	if (getaddrs.addr_num <= 0) return -EINVAL;
4141	/*
4142	 *  For UDP-style sockets, id specifies the association to query.
4143	 *  If the id field is set to the value '0' then the locally bound
4144	 *  addresses are returned without regard to any particular
4145	 *  association.
4146	 */
4147	if (0 == getaddrs.assoc_id) {
4148		bp = &sctp_sk(sk)->ep->base.bind_addr;
4149		addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4150	} else {
4151		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4152		if (!asoc)
4153			return -EINVAL;
4154		bp = &asoc->base.bind_addr;
4155		addr_lock = &asoc->base.addr_lock;
4156	}
4157
4158	to = getaddrs.addrs;
4159
4160	/* Allocate space for a local instance of packed array to hold all
4161	 * the data.  We store addresses here first and then put write them
4162	 * to the user in one shot.
4163	 */
4164	addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4165			GFP_KERNEL);
4166	if (!addrs)
4167		return -ENOMEM;
4168
4169	sctp_read_lock(addr_lock);
4170
4171	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4172	 * addresses from the global local address list.
4173	 */
4174	if (sctp_list_single_entry(&bp->address_list)) {
4175		addr = list_entry(bp->address_list.next,
4176				  struct sctp_sockaddr_entry, list);
4177		if (sctp_is_any(&addr->a)) {
4178			cnt = sctp_copy_laddrs_old(sk, bp->port,
4179						   getaddrs.addr_num,
4180						   addrs, &bytes_copied);
4181			goto copy_getaddrs;
4182		}
4183	}
4184
4185	buf = addrs;
4186	list_for_each(pos, &bp->address_list) {
4187		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4188		memcpy(&temp, &addr->a, sizeof(temp));
4189		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4190		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4191		memcpy(buf, &temp, addrlen);
4192		buf += addrlen;
4193		bytes_copied += addrlen;
4194		cnt ++;
4195		if (cnt >= getaddrs.addr_num) break;
4196	}
4197
4198copy_getaddrs:
4199	sctp_read_unlock(addr_lock);
4200
4201	/* copy the entire address list into the user provided space */
4202	if (copy_to_user(to, addrs, bytes_copied)) {
4203		err = -EFAULT;
4204		goto error;
4205	}
4206
4207	/* copy the leading structure back to user */
4208	getaddrs.addr_num = cnt;
4209	if (copy_to_user(optval, &getaddrs, len))
4210		err = -EFAULT;
4211
4212error:
4213	kfree(addrs);
4214	return err;
4215}
4216
4217static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4218				       char __user *optval, int __user *optlen)
4219{
4220	struct sctp_bind_addr *bp;
4221	struct sctp_association *asoc;
4222	struct list_head *pos;
4223	int cnt = 0;
4224	struct sctp_getaddrs getaddrs;
4225	struct sctp_sockaddr_entry *addr;
4226	void __user *to;
4227	union sctp_addr temp;
4228	struct sctp_sock *sp = sctp_sk(sk);
4229	int addrlen;
4230	rwlock_t *addr_lock;
4231	int err = 0;
4232	size_t space_left;
4233	int bytes_copied = 0;
4234	void *addrs;
4235	void *buf;
4236
4237	if (len < sizeof(struct sctp_getaddrs))
4238		return -EINVAL;
4239
4240	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4241		return -EFAULT;
4242
4243	/*
4244	 *  For UDP-style sockets, id specifies the association to query.
4245	 *  If the id field is set to the value '0' then the locally bound
4246	 *  addresses are returned without regard to any particular
4247	 *  association.
4248	 */
4249	if (0 == getaddrs.assoc_id) {
4250		bp = &sctp_sk(sk)->ep->base.bind_addr;
4251		addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4252	} else {
4253		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4254		if (!asoc)
4255			return -EINVAL;
4256		bp = &asoc->base.bind_addr;
4257		addr_lock = &asoc->base.addr_lock;
4258	}
4259
4260	to = optval + offsetof(struct sctp_getaddrs,addrs);
4261	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4262
4263	addrs = kmalloc(space_left, GFP_KERNEL);
4264	if (!addrs)
4265		return -ENOMEM;
4266
4267	sctp_read_lock(addr_lock);
4268
4269	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4270	 * addresses from the global local address list.
4271	 */
4272	if (sctp_list_single_entry(&bp->address_list)) {
4273		addr = list_entry(bp->address_list.next,
4274				  struct sctp_sockaddr_entry, list);
4275		if (sctp_is_any(&addr->a)) {
4276			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4277						space_left, &bytes_copied);
4278			if (cnt < 0) {
4279				err = cnt;
4280				goto error;
4281			}
4282			goto copy_getaddrs;
4283		}
4284	}
4285
4286	buf = addrs;
4287	list_for_each(pos, &bp->address_list) {
4288		addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4289		memcpy(&temp, &addr->a, sizeof(temp));
4290		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4291		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4292		if (space_left < addrlen) {
4293			err =  -ENOMEM;
4294			goto error;
4295		}
4296		memcpy(buf, &temp, addrlen);
4297		buf += addrlen;
4298		bytes_copied += addrlen;
4299		cnt ++;
4300		space_left -= addrlen;
4301	}
4302
4303copy_getaddrs:
4304	sctp_read_unlock(addr_lock);
4305
4306	if (copy_to_user(to, addrs, bytes_copied)) {
4307		err = -EFAULT;
4308		goto error;
4309	}
4310	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4311		err = -EFAULT;
4312		goto error;
4313	}
4314	if (put_user(bytes_copied, optlen))
4315		err = -EFAULT;
4316error:
4317	kfree(addrs);
4318	return err;
4319}
4320
4321/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4322 *
4323 * Requests that the local SCTP stack use the enclosed peer address as
4324 * the association primary.  The enclosed address must be one of the
4325 * association peer's addresses.
4326 */
4327static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4328					char __user *optval, int __user *optlen)
4329{
4330	struct sctp_prim prim;
4331	struct sctp_association *asoc;
4332	struct sctp_sock *sp = sctp_sk(sk);
4333
4334	if (len < sizeof(struct sctp_prim))
4335		return -EINVAL;
4336
4337	len = sizeof(struct sctp_prim);
4338
4339	if (copy_from_user(&prim, optval, len))
4340		return -EFAULT;
4341
4342	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4343	if (!asoc)
4344		return -EINVAL;
4345
4346	if (!asoc->peer.primary_path)
4347		return -ENOTCONN;
4348
4349	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4350		asoc->peer.primary_path->af_specific->sockaddr_len);
4351
4352	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4353			(union sctp_addr *)&prim.ssp_addr);
4354
4355	if (put_user(len, optlen))
4356		return -EFAULT;
4357	if (copy_to_user(optval, &prim, len))
4358		return -EFAULT;
4359
4360	return 0;
4361}
4362
4363/*
4364 * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4365 *
4366 * Requests that the local endpoint set the specified Adaptation Layer
4367 * Indication parameter for all future INIT and INIT-ACK exchanges.
4368 */
4369static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4370				  char __user *optval, int __user *optlen)
4371{
4372	struct sctp_setadaptation adaptation;
4373
4374	if (len < sizeof(struct sctp_setadaptation))
4375		return -EINVAL;
4376
4377	len = sizeof(struct sctp_setadaptation);
4378
4379	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4380
4381	if (put_user(len, optlen))
4382		return -EFAULT;
4383	if (copy_to_user(optval, &adaptation, len))
4384		return -EFAULT;
4385
4386	return 0;
4387}
4388
4389/*
4390 *
4391 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4392 *
4393 *   Applications that wish to use the sendto() system call may wish to
4394 *   specify a default set of parameters that would normally be supplied
4395 *   through the inclusion of ancillary data.  This socket option allows
4396 *   such an application to set the default sctp_sndrcvinfo structure.
4397
4398
4399 *   The application that wishes to use this socket option simply passes
4400 *   in to this call the sctp_sndrcvinfo structure defined in Section
4401 *   5.2.2) The input parameters accepted by this call include
4402 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4403 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4404 *   to this call if the caller is using the UDP model.
4405 *
4406 *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4407 */
4408static int sctp_getsockopt_default_send_param(struct sock *sk,
4409					int len, char __user *optval,
4410					int __user *optlen)
4411{
4412	struct sctp_sndrcvinfo info;
4413	struct sctp_association *asoc;
4414	struct sctp_sock *sp = sctp_sk(sk);
4415
4416	if (len < sizeof(struct sctp_sndrcvinfo))
4417		return -EINVAL;
4418
4419	len = sizeof(struct sctp_sndrcvinfo);
4420
4421	if (copy_from_user(&info, optval, len))
4422		return -EFAULT;
4423
4424	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4425	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4426		return -EINVAL;
4427
4428	if (asoc) {
4429		info.sinfo_stream = asoc->default_stream;
4430		info.sinfo_flags = asoc->default_flags;
4431		info.sinfo_ppid = asoc->default_ppid;
4432		info.sinfo_context = asoc->default_context;
4433		info.sinfo_timetolive = asoc->default_timetolive;
4434	} else {
4435		info.sinfo_stream = sp->default_stream;
4436		info.sinfo_flags = sp->default_flags;
4437		info.sinfo_ppid = sp->default_ppid;
4438		info.sinfo_context = sp->default_context;
4439		info.sinfo_timetolive = sp->default_timetolive;
4440	}
4441
4442	if (put_user(len, optlen))
4443		return -EFAULT;
4444	if (copy_to_user(optval, &info, len))
4445		return -EFAULT;
4446
4447	return 0;
4448}
4449
4450/*
4451 *
4452 * 7.1.5 SCTP_NODELAY
4453 *
4454 * Turn on/off any Nagle-like algorithm.  This means that packets are
4455 * generally sent as soon as possible and no unnecessary delays are
4456 * introduced, at the cost of more packets in the network.  Expects an
4457 * integer boolean flag.
4458 */
4459
4460static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4461				   char __user *optval, int __user *optlen)
4462{
4463	int val;
4464
4465	if (len < sizeof(int))
4466		return -EINVAL;
4467
4468	len = sizeof(int);
4469	val = (sctp_sk(sk)->nodelay == 1);
4470	if (put_user(len, optlen))
4471		return -EFAULT;
4472	if (copy_to_user(optval, &val, len))
4473		return -EFAULT;
4474	return 0;
4475}
4476
4477/*
4478 *
4479 * 7.1.1 SCTP_RTOINFO
4480 *
4481 * The protocol parameters used to initialize and bound retransmission
4482 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4483 * and modify these parameters.
4484 * All parameters are time values, in milliseconds.  A value of 0, when
4485 * modifying the parameters, indicates that the current value should not
4486 * be changed.
4487 *
4488 */
4489static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4490				char __user *optval,
4491				int __user *optlen) {
4492	struct sctp_rtoinfo rtoinfo;
4493	struct sctp_association *asoc;
4494
4495	if (len < sizeof (struct sctp_rtoinfo))
4496		return -EINVAL;
4497
4498	len = sizeof(struct sctp_rtoinfo);
4499
4500	if (copy_from_user(&rtoinfo, optval, len))
4501		return -EFAULT;
4502
4503	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4504
4505	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4506		return -EINVAL;
4507
4508	/* Values corresponding to the specific association. */
4509	if (asoc) {
4510		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4511		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4512		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4513	} else {
4514		/* Values corresponding to the endpoint. */
4515		struct sctp_sock *sp = sctp_sk(sk);
4516
4517		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4518		rtoinfo.srto_max = sp->rtoinfo.srto_max;
4519		rtoinfo.srto_min = sp->rtoinfo.srto_min;
4520	}
4521
4522	if (put_user(len, optlen))
4523		return -EFAULT;
4524
4525	if (copy_to_user(optval, &rtoinfo, len))
4526		return -EFAULT;
4527
4528	return 0;
4529}
4530
4531/*
4532 *
4533 * 7.1.2 SCTP_ASSOCINFO
4534 *
4535 * This option is used to tune the maximum retransmission attempts
4536 * of the association.
4537 * Returns an error if the new association retransmission value is
4538 * greater than the sum of the retransmission value  of the peer.
4539 * See [SCTP] for more information.
4540 *
4541 */
4542static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4543				     char __user *optval,
4544				     int __user *optlen)
4545{
4546
4547	struct sctp_assocparams assocparams;
4548	struct sctp_association *asoc;
4549	struct list_head *pos;
4550	int cnt = 0;
4551
4552	if (len < sizeof (struct sctp_assocparams))
4553		return -EINVAL;
4554
4555	len = sizeof(struct sctp_assocparams);
4556
4557	if (copy_from_user(&assocparams, optval, len))
4558		return -EFAULT;
4559
4560	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4561
4562	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4563		return -EINVAL;
4564
4565	/* Values correspoinding to the specific association */
4566	if (asoc) {
4567		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4568		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4569		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4570		assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4571						* 1000) +
4572						(asoc->cookie_life.tv_usec
4573						/ 1000);
4574
4575		list_for_each(pos, &asoc->peer.transport_addr_list) {
4576			cnt ++;
4577		}
4578
4579		assocparams.sasoc_number_peer_destinations = cnt;
4580	} else {
4581		/* Values corresponding to the endpoint */
4582		struct sctp_sock *sp = sctp_sk(sk);
4583
4584		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4585		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4586		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4587		assocparams.sasoc_cookie_life =
4588					sp->assocparams.sasoc_cookie_life;
4589		assocparams.sasoc_number_peer_destinations =
4590					sp->assocparams.
4591					sasoc_number_peer_destinations;
4592	}
4593
4594	if (put_user(len, optlen))
4595		return -EFAULT;
4596
4597	if (copy_to_user(optval, &assocparams, len))
4598		return -EFAULT;
4599
4600	return 0;
4601}
4602
4603/*
4604 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4605 *
4606 * This socket option is a boolean flag which turns on or off mapped V4
4607 * addresses.  If this option is turned on and the socket is type
4608 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4609 * If this option is turned off, then no mapping will be done of V4
4610 * addresses and a user will receive both PF_INET6 and PF_INET type
4611 * addresses on the socket.
4612 */
4613static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4614				    char __user *optval, int __user *optlen)
4615{
4616	int val;
4617	struct sctp_sock *sp = sctp_sk(sk);
4618
4619	if (len < sizeof(int))
4620		return -EINVAL;
4621
4622	len = sizeof(int);
4623	val = sp->v4mapped;
4624	if (put_user(len, optlen))
4625		return -EFAULT;
4626	if (copy_to_user(optval, &val, len))
4627		return -EFAULT;
4628
4629	return 0;
4630}
4631
4632/*
4633 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
4634 * (chapter and verse is quoted at sctp_setsockopt_context())
4635 */
4636static int sctp_getsockopt_context(struct sock *sk, int len,
4637				   char __user *optval, int __user *optlen)
4638{
4639	struct sctp_assoc_value params;
4640	struct sctp_sock *sp;
4641	struct sctp_association *asoc;
4642
4643	if (len < sizeof(struct sctp_assoc_value))
4644		return -EINVAL;
4645
4646	len = sizeof(struct sctp_assoc_value);
4647
4648	if (copy_from_user(&params, optval, len))
4649		return -EFAULT;
4650
4651	sp = sctp_sk(sk);
4652
4653	if (params.assoc_id != 0) {
4654		asoc = sctp_id2assoc(sk, params.assoc_id);
4655		if (!asoc)
4656			return -EINVAL;
4657		params.assoc_value = asoc->default_rcv_context;
4658	} else {
4659		params.assoc_value = sp->default_rcv_context;
4660	}
4661
4662	if (put_user(len, optlen))
4663		return -EFAULT;
4664	if (copy_to_user(optval, &params, len))
4665		return -EFAULT;
4666
4667	return 0;
4668}
4669
4670/*
4671 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4672 *
4673 * This socket option specifies the maximum size to put in any outgoing
4674 * SCTP chunk.  If a message is larger than this size it will be
4675 * fragmented by SCTP into the specified size.  Note that the underlying
4676 * SCTP implementation may fragment into smaller sized chunks when the
4677 * PMTU of the underlying association is smaller than the value set by
4678 * the user.
4679 */
4680static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4681				  char __user *optval, int __user *optlen)
4682{
4683	int val;
4684
4685	if (len < sizeof(int))
4686		return -EINVAL;
4687
4688	len = sizeof(int);
4689
4690	val = sctp_sk(sk)->user_frag;
4691	if (put_user(len, optlen))
4692		return -EFAULT;
4693	if (copy_to_user(optval, &val, len))
4694		return -EFAULT;
4695
4696	return 0;
4697}
4698
4699/*
4700 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4701 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4702 */
4703static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4704					       char __user *optval, int __user *optlen)
4705{
4706	int val;
4707
4708	if (len < sizeof(int))
4709		return -EINVAL;
4710
4711	len = sizeof(int);
4712
4713	val = sctp_sk(sk)->frag_interleave;
4714	if (put_user(len, optlen))
4715		return -EFAULT;
4716	if (copy_to_user(optval, &val, len))
4717		return -EFAULT;
4718
4719	return 0;
4720}
4721
4722/*
4723 * 7.1.25.  Set or Get the sctp partial delivery point
4724 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4725 */
4726static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4727						  char __user *optval,
4728						  int __user *optlen)
4729{
4730        u32 val;
4731
4732	if (len < sizeof(u32))
4733		return -EINVAL;
4734
4735	len = sizeof(u32);
4736
4737	val = sctp_sk(sk)->pd_point;
4738	if (put_user(len, optlen))
4739		return -EFAULT;
4740	if (copy_to_user(optval, &val, len))
4741		return -EFAULT;
4742
4743	return -ENOTSUPP;
4744}
4745
4746/*
4747 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
4748 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
4749 */
4750static int sctp_getsockopt_maxburst(struct sock *sk, int len,
4751				    char __user *optval,
4752				    int __user *optlen)
4753{
4754        int val;
4755
4756	if (len < sizeof(int))
4757		return -EINVAL;
4758
4759	len = sizeof(int);
4760
4761	val = sctp_sk(sk)->max_burst;
4762	if (put_user(len, optlen))
4763		return -EFAULT;
4764	if (copy_to_user(optval, &val, len))
4765		return -EFAULT;
4766
4767	return -ENOTSUPP;
4768}
4769
4770SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4771				char __user *optval, int __user *optlen)
4772{
4773	int retval = 0;
4774	int len;
4775
4776	SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4777			  sk, optname);
4778
4779	/* I can hardly begin to describe how wrong this is.  This is
4780	 * so broken as to be worse than useless.  The API draft
4781	 * REALLY is NOT helpful here...  I am not convinced that the
4782	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4783	 * are at all well-founded.
4784	 */
4785	if (level != SOL_SCTP) {
4786		struct sctp_af *af = sctp_sk(sk)->pf->af;
4787
4788		retval = af->getsockopt(sk, level, optname, optval, optlen);
4789		return retval;
4790	}
4791
4792	if (get_user(len, optlen))
4793		return -EFAULT;
4794
4795	sctp_lock_sock(sk);
4796
4797	switch (optname) {
4798	case SCTP_STATUS:
4799		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4800		break;
4801	case SCTP_DISABLE_FRAGMENTS:
4802		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4803							   optlen);
4804		break;
4805	case SCTP_EVENTS:
4806		retval = sctp_getsockopt_events(sk, len, optval, optlen);
4807		break;
4808	case SCTP_AUTOCLOSE:
4809		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4810		break;
4811	case SCTP_SOCKOPT_PEELOFF:
4812		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4813		break;
4814	case SCTP_PEER_ADDR_PARAMS:
4815		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4816							  optlen);
4817		break;
4818	case SCTP_DELAYED_ACK_TIME:
4819		retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4820							  optlen);
4821		break;
4822	case SCTP_INITMSG:
4823		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4824		break;
4825	case SCTP_GET_PEER_ADDRS_NUM_OLD:
4826		retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4827							    optlen);
4828		break;
4829	case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4830		retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4831							     optlen);
4832		break;
4833	case SCTP_GET_PEER_ADDRS_OLD:
4834		retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4835							optlen);
4836		break;
4837	case SCTP_GET_LOCAL_ADDRS_OLD:
4838		retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4839							 optlen);
4840		break;
4841	case SCTP_GET_PEER_ADDRS:
4842		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4843						    optlen);
4844		break;
4845	case SCTP_GET_LOCAL_ADDRS:
4846		retval = sctp_getsockopt_local_addrs(sk, len, optval,
4847						     optlen);
4848		break;
4849	case SCTP_DEFAULT_SEND_PARAM:
4850		retval = sctp_getsockopt_default_send_param(sk, len,
4851							    optval, optlen);
4852		break;
4853	case SCTP_PRIMARY_ADDR:
4854		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4855		break;
4856	case SCTP_NODELAY:
4857		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4858		break;
4859	case SCTP_RTOINFO:
4860		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4861		break;
4862	case SCTP_ASSOCINFO:
4863		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4864		break;
4865	case SCTP_I_WANT_MAPPED_V4_ADDR:
4866		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4867		break;
4868	case SCTP_MAXSEG:
4869		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4870		break;
4871	case SCTP_GET_PEER_ADDR_INFO:
4872		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4873							optlen);
4874		break;
4875	case SCTP_ADAPTATION_LAYER:
4876		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4877							optlen);
4878		break;
4879	case SCTP_CONTEXT:
4880		retval = sctp_getsockopt_context(sk, len, optval, optlen);
4881		break;
4882	case SCTP_FRAGMENT_INTERLEAVE:
4883		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4884							     optlen);
4885		break;
4886	case SCTP_PARTIAL_DELIVERY_POINT:
4887		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
4888								optlen);
4889		break;
4890	case SCTP_MAX_BURST:
4891		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
4892		break;
4893	default:
4894		retval = -ENOPROTOOPT;
4895		break;
4896	}
4897
4898	sctp_release_sock(sk);
4899	return retval;
4900}
4901
4902static void sctp_hash(struct sock *sk)
4903{
4904	/* STUB */
4905}
4906
4907static void sctp_unhash(struct sock *sk)
4908{
4909	/* STUB */
4910}
4911
4912static struct sctp_bind_bucket *sctp_bucket_create(
4913	struct sctp_bind_hashbucket *head, unsigned short snum);
4914
4915static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4916{
4917	struct sctp_bind_hashbucket *head; /* hash list */
4918	struct sctp_bind_bucket *pp; /* hash list port iterator */
4919	unsigned short snum;
4920	int ret;
4921
4922	snum = ntohs(addr->v4.sin_port);
4923
4924	SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4925	sctp_local_bh_disable();
4926
4927	if (snum == 0) {
4928		/* Search for an available port.
4929		 *
4930		 * 'sctp_port_rover' was the last port assigned, so
4931		 * we start to search from 'sctp_port_rover +
4932		 * 1'. What we do is first check if port 'rover' is
4933		 * already in the hash table; if not, we use that; if
4934		 * it is, we try next.
4935		 */
4936		int low = sysctl_local_port_range[0];
4937		int high = sysctl_local_port_range[1];
4938		int remaining = (high - low) + 1;
4939		int rover;
4940		int index;
4941
4942		sctp_spin_lock(&sctp_port_alloc_lock);
4943		rover = sctp_port_rover;
4944		do {
4945			rover++;
4946			if ((rover < low) || (rover > high))
4947				rover = low;
4948			index = sctp_phashfn(rover);
4949			head = &sctp_port_hashtable[index];
4950			sctp_spin_lock(&head->lock);
4951			for (pp = head->chain; pp; pp = pp->next)
4952				if (pp->port == rover)
4953					goto next;
4954			break;
4955		next:
4956			sctp_spin_unlock(&head->lock);
4957		} while (--remaining > 0);
4958		sctp_port_rover = rover;
4959		sctp_spin_unlock(&sctp_port_alloc_lock);
4960
4961		/* Exhausted local port range during search? */
4962		ret = 1;
4963		if (remaining <= 0)
4964			goto fail;
4965
4966		/* OK, here is the one we will use.  HEAD (the port
4967		 * hash table list entry) is non-NULL and we hold it's
4968		 * mutex.
4969		 */
4970		snum = rover;
4971	} else {
4972		/* We are given an specific port number; we verify
4973		 * that it is not being used. If it is used, we will
4974		 * exahust the search in the hash list corresponding
4975		 * to the port number (snum) - we detect that with the
4976		 * port iterator, pp being NULL.
4977		 */
4978		head = &sctp_port_hashtable[sctp_phashfn(snum)];
4979		sctp_spin_lock(&head->lock);
4980		for (pp = head->chain; pp; pp = pp->next) {
4981			if (pp->port == snum)
4982				goto pp_found;
4983		}
4984	}
4985	pp = NULL;
4986	goto pp_not_found;
4987pp_found:
4988	if (!hlist_empty(&pp->owner)) {
4989		/* We had a port hash table hit - there is an
4990		 * available port (pp != NULL) and it is being
4991		 * used by other socket (pp->owner not empty); that other
4992		 * socket is going to be sk2.
4993		 */
4994		int reuse = sk->sk_reuse;
4995		struct sock *sk2;
4996		struct hlist_node *node;
4997
4998		SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4999		if (pp->fastreuse && sk->sk_reuse &&
5000			sk->sk_state != SCTP_SS_LISTENING)
5001			goto success;
5002
5003		/* Run through the list of sockets bound to the port
5004		 * (pp->port) [via the pointers bind_next and
5005		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5006		 * we get the endpoint they describe and run through
5007		 * the endpoint's list of IP (v4 or v6) addresses,
5008		 * comparing each of the addresses with the address of
5009		 * the socket sk. If we find a match, then that means
5010		 * that this port/socket (sk) combination are already
5011		 * in an endpoint.
5012		 */
5013		sk_for_each_bound(sk2, node, &pp->owner) {
5014			struct sctp_endpoint *ep2;
5015			ep2 = sctp_sk(sk2)->ep;
5016
5017			if (reuse && sk2->sk_reuse &&
5018			    sk2->sk_state != SCTP_SS_LISTENING)
5019				continue;
5020
5021			if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5022						 sctp_sk(sk))) {
5023				ret = (long)sk2;
5024				goto fail_unlock;
5025			}
5026		}
5027		SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5028	}
5029pp_not_found:
5030	/* If there was a hash table miss, create a new port.  */
5031	ret = 1;
5032	if (!pp && !(pp = sctp_bucket_create(head, snum)))
5033		goto fail_unlock;
5034
5035	/* In either case (hit or miss), make sure fastreuse is 1 only
5036	 * if sk->sk_reuse is too (that is, if the caller requested
5037	 * SO_REUSEADDR on this socket -sk-).
5038	 */
5039	if (hlist_empty(&pp->owner)) {
5040		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5041			pp->fastreuse = 1;
5042		else
5043			pp->fastreuse = 0;
5044	} else if (pp->fastreuse &&
5045		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5046		pp->fastreuse = 0;
5047
5048success:
5049	if (!sctp_sk(sk)->bind_hash) {
5050		inet_sk(sk)->num = snum;
5051		sk_add_bind_node(sk, &pp->owner);
5052		sctp_sk(sk)->bind_hash = pp;
5053	}
5054	ret = 0;
5055
5056fail_unlock:
5057	sctp_spin_unlock(&head->lock);
5058
5059fail:
5060	sctp_local_bh_enable();
5061	return ret;
5062}
5063
5064/* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5065 * port is requested.
5066 */
5067static int sctp_get_port(struct sock *sk, unsigned short snum)
5068{
5069	long ret;
5070	union sctp_addr addr;
5071	struct sctp_af *af = sctp_sk(sk)->pf->af;
5072
5073	/* Set up a dummy address struct from the sk. */
5074	af->from_sk(&addr, sk);
5075	addr.v4.sin_port = htons(snum);
5076
5077	/* Note: sk->sk_num gets filled in if ephemeral port request. */
5078	ret = sctp_get_port_local(sk, &addr);
5079
5080	return (ret ? 1 : 0);
5081}
5082
5083/*
5084 * 3.1.3 listen() - UDP Style Syntax
5085 *
5086 *   By default, new associations are not accepted for UDP style sockets.
5087 *   An application uses listen() to mark a socket as being able to
5088 *   accept new associations.
5089 */
5090SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5091{
5092	struct sctp_sock *sp = sctp_sk(sk);
5093	struct sctp_endpoint *ep = sp->ep;
5094
5095	/* Only UDP style sockets that are not peeled off are allowed to
5096	 * listen().
5097	 */
5098	if (!sctp_style(sk, UDP))
5099		return -EINVAL;
5100
5101	/* If backlog is zero, disable listening. */
5102	if (!backlog) {
5103		if (sctp_sstate(sk, CLOSED))
5104			return 0;
5105
5106		sctp_unhash_endpoint(ep);
5107		sk->sk_state = SCTP_SS_CLOSED;
5108	}
5109
5110	/* Return if we are already listening. */
5111	if (sctp_sstate(sk, LISTENING))
5112		return 0;
5113
5114	/*
5115	 * If a bind() or sctp_bindx() is not called prior to a listen()
5116	 * call that allows new associations to be accepted, the system
5117	 * picks an ephemeral port and will choose an address set equivalent
5118	 * to binding with a wildcard address.
5119	 *
5120	 * This is not currently spelled out in the SCTP sockets
5121	 * extensions draft, but follows the practice as seen in TCP
5122	 * sockets.
5123	 *
5124	 * Additionally, turn off fastreuse flag since we are not listening
5125	 */
5126	sk->sk_state = SCTP_SS_LISTENING;
5127	if (!ep->base.bind_addr.port) {
5128		if (sctp_autobind(sk))
5129			return -EAGAIN;
5130	} else
5131		sctp_sk(sk)->bind_hash->fastreuse = 0;
5132
5133	sctp_hash_endpoint(ep);
5134	return 0;
5135}
5136
5137/*
5138 * 4.1.3 listen() - TCP Style Syntax
5139 *
5140 *   Applications uses listen() to ready the SCTP endpoint for accepting
5141 *   inbound associations.
5142 */
5143SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5144{
5145	struct sctp_sock *sp = sctp_sk(sk);
5146	struct sctp_endpoint *ep = sp->ep;
5147
5148	/* If backlog is zero, disable listening. */
5149	if (!backlog) {
5150		if (sctp_sstate(sk, CLOSED))
5151			return 0;
5152
5153		sctp_unhash_endpoint(ep);
5154		sk->sk_state = SCTP_SS_CLOSED;
5155	}
5156
5157	if (sctp_sstate(sk, LISTENING))
5158		return 0;
5159
5160	/*
5161	 * If a bind() or sctp_bindx() is not called prior to a listen()
5162	 * call that allows new associations to be accepted, the system
5163	 * picks an ephemeral port and will choose an address set equivalent
5164	 * to binding with a wildcard address.
5165	 *
5166	 * This is not currently spelled out in the SCTP sockets
5167	 * extensions draft, but follows the practice as seen in TCP
5168	 * sockets.
5169	 */
5170	sk->sk_state = SCTP_SS_LISTENING;
5171	if (!ep->base.bind_addr.port) {
5172		if (sctp_autobind(sk))
5173			return -EAGAIN;
5174	} else
5175		sctp_sk(sk)->bind_hash->fastreuse = 0;
5176
5177	sk->sk_max_ack_backlog = backlog;
5178	sctp_hash_endpoint(ep);
5179	return 0;
5180}
5181
5182/*
5183 *  Move a socket to LISTENING state.
5184 */
5185int sctp_inet_listen(struct socket *sock, int backlog)
5186{
5187	struct sock *sk = sock->sk;
5188	struct crypto_hash *tfm = NULL;
5189	int err = -EINVAL;
5190
5191	if (unlikely(backlog < 0))
5192		goto out;
5193
5194	sctp_lock_sock(sk);
5195
5196	if (sock->state != SS_UNCONNECTED)
5197		goto out;
5198
5199	/* Allocate HMAC for generating cookie. */
5200	if (sctp_hmac_alg) {
5201		tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5202		if (IS_ERR(tfm)) {
5203			if (net_ratelimit()) {
5204				printk(KERN_INFO
5205				       "SCTP: failed to load transform for %s: %ld\n",
5206					sctp_hmac_alg, PTR_ERR(tfm));
5207			}
5208			err = -ENOSYS;
5209			goto out;
5210		}
5211	}
5212
5213	switch (sock->type) {
5214	case SOCK_SEQPACKET:
5215		err = sctp_seqpacket_listen(sk, backlog);
5216		break;
5217	case SOCK_STREAM:
5218		err = sctp_stream_listen(sk, backlog);
5219		break;
5220	default:
5221		break;
5222	}
5223
5224	if (err)
5225		goto cleanup;
5226
5227	/* Store away the transform reference. */
5228	sctp_sk(sk)->hmac = tfm;
5229out:
5230	sctp_release_sock(sk);
5231	return err;
5232cleanup:
5233	crypto_free_hash(tfm);
5234	goto out;
5235}
5236
5237/*
5238 * This function is done by modeling the current datagram_poll() and the
5239 * tcp_poll().  Note that, based on these implementations, we don't
5240 * lock the socket in this function, even though it seems that,
5241 * ideally, locking or some other mechanisms can be used to ensure
5242 * the integrity of the counters (sndbuf and wmem_alloc) used
5243 * in this place.  We assume that we don't need locks either until proven
5244 * otherwise.
5245 *
5246 * Another thing to note is that we include the Async I/O support
5247 * here, again, by modeling the current TCP/UDP code.  We don't have
5248 * a good way to test with it yet.
5249 */
5250unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5251{
5252	struct sock *sk = sock->sk;
5253	struct sctp_sock *sp = sctp_sk(sk);
5254	unsigned int mask;
5255
5256	poll_wait(file, sk->sk_sleep, wait);
5257
5258	/* A TCP-style listening socket becomes readable when the accept queue
5259	 * is not empty.
5260	 */
5261	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5262		return (!list_empty(&sp->ep->asocs)) ?
5263			(POLLIN | POLLRDNORM) : 0;
5264
5265	mask = 0;
5266
5267	/* Is there any exceptional events?  */
5268	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5269		mask |= POLLERR;
5270	if (sk->sk_shutdown & RCV_SHUTDOWN)
5271		mask |= POLLRDHUP;
5272	if (sk->sk_shutdown == SHUTDOWN_MASK)
5273		mask |= POLLHUP;
5274
5275	/* Is it readable?  Reconsider this code with TCP-style support.  */
5276	if (!skb_queue_empty(&sk->sk_receive_queue) ||
5277	    (sk->sk_shutdown & RCV_SHUTDOWN))
5278		mask |= POLLIN | POLLRDNORM;
5279
5280	/* The association is either gone or not ready.  */
5281	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5282		return mask;
5283
5284	/* Is it writable?  */
5285	if (sctp_writeable(sk)) {
5286		mask |= POLLOUT | POLLWRNORM;
5287	} else {
5288		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5289		/*
5290		 * Since the socket is not locked, the buffer
5291		 * might be made available after the writeable check and
5292		 * before the bit is set.  This could cause a lost I/O
5293		 * signal.  tcp_poll() has a race breaker for this race
5294		 * condition.  Based on their implementation, we put
5295		 * in the following code to cover it as well.
5296		 */
5297		if (sctp_writeable(sk))
5298			mask |= POLLOUT | POLLWRNORM;
5299	}
5300	return mask;
5301}
5302
5303/********************************************************************
5304 * 2nd Level Abstractions
5305 ********************************************************************/
5306
5307static struct sctp_bind_bucket *sctp_bucket_create(
5308	struct sctp_bind_hashbucket *head, unsigned short snum)
5309{
5310	struct sctp_bind_bucket *pp;
5311
5312	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5313	SCTP_DBG_OBJCNT_INC(bind_bucket);
5314	if (pp) {
5315		pp->port = snum;
5316		pp->fastreuse = 0;
5317		INIT_HLIST_HEAD(&pp->owner);
5318		if ((pp->next = head->chain) != NULL)
5319			pp->next->pprev = &pp->next;
5320		head->chain = pp;
5321		pp->pprev = &head->chain;
5322	}
5323	return pp;
5324}
5325
5326/* Caller must hold hashbucket lock for this tb with local BH disabled */
5327static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5328{
5329	if (pp && hlist_empty(&pp->owner)) {
5330		if (pp->next)
5331			pp->next->pprev = pp->pprev;
5332		*(pp->pprev) = pp->next;
5333		kmem_cache_free(sctp_bucket_cachep, pp);
5334		SCTP_DBG_OBJCNT_DEC(bind_bucket);
5335	}
5336}
5337
5338/* Release this socket's reference to a local port.  */
5339static inline void __sctp_put_port(struct sock *sk)
5340{
5341	struct sctp_bind_hashbucket *head =
5342		&sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5343	struct sctp_bind_bucket *pp;
5344
5345	sctp_spin_lock(&head->lock);
5346	pp = sctp_sk(sk)->bind_hash;
5347	__sk_del_bind_node(sk);
5348	sctp_sk(sk)->bind_hash = NULL;
5349	inet_sk(sk)->num = 0;
5350	sctp_bucket_destroy(pp);
5351	sctp_spin_unlock(&head->lock);
5352}
5353
5354void sctp_put_port(struct sock *sk)
5355{
5356	sctp_local_bh_disable();
5357	__sctp_put_port(sk);
5358	sctp_local_bh_enable();
5359}
5360
5361/*
5362 * The system picks an ephemeral port and choose an address set equivalent
5363 * to binding with a wildcard address.
5364 * One of those addresses will be the primary address for the association.
5365 * This automatically enables the multihoming capability of SCTP.
5366 */
5367static int sctp_autobind(struct sock *sk)
5368{
5369	union sctp_addr autoaddr;
5370	struct sctp_af *af;
5371	__be16 port;
5372
5373	/* Initialize a local sockaddr structure to INADDR_ANY. */
5374	af = sctp_sk(sk)->pf->af;
5375
5376	port = htons(inet_sk(sk)->num);
5377	af->inaddr_any(&autoaddr, port);
5378
5379	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5380}
5381
5382/* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
5383 *
5384 * From RFC 2292
5385 * 4.2 The cmsghdr Structure *
5386 *
5387 * When ancillary data is sent or received, any number of ancillary data
5388 * objects can be specified by the msg_control and msg_controllen members of
5389 * the msghdr structure, because each object is preceded by
5390 * a cmsghdr structure defining the object's length (the cmsg_len member).
5391 * Historically Berkeley-derived implementations have passed only one object
5392 * at a time, but this API allows multiple objects to be
5393 * passed in a single call to sendmsg() or recvmsg(). The following example
5394 * shows two ancillary data objects in a control buffer.
5395 *
5396 *   |<--------------------------- msg_controllen -------------------------->|
5397 *   |                                                                       |
5398 *
5399 *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
5400 *
5401 *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5402 *   |                                   |                                   |
5403 *
5404 *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
5405 *
5406 *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
5407 *   |                                |  |                                |  |
5408 *
5409 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5410 *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
5411 *
5412 *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
5413 *
5414 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5415 *    ^
5416 *    |
5417 *
5418 * msg_control
5419 * points here
5420 */
5421SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5422				  sctp_cmsgs_t *cmsgs)
5423{
5424	struct cmsghdr *cmsg;
5425
5426	for (cmsg = CMSG_FIRSTHDR(msg);
5427	     cmsg != NULL;
5428	     cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5429		if (!CMSG_OK(msg, cmsg))
5430			return -EINVAL;
5431
5432		/* Should we parse this header or ignore?  */
5433		if (cmsg->cmsg_level != IPPROTO_SCTP)
5434			continue;
5435
5436		/* Strictly check lengths following example in SCM code.  */
5437		switch (cmsg->cmsg_type) {
5438		case SCTP_INIT:
5439			/* SCTP Socket API Extension
5440			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5441			 *
5442			 * This cmsghdr structure provides information for
5443			 * initializing new SCTP associations with sendmsg().
5444			 * The SCTP_INITMSG socket option uses this same data
5445			 * structure.  This structure is not used for
5446			 * recvmsg().
5447			 *
5448			 * cmsg_level    cmsg_type      cmsg_data[]
5449			 * ------------  ------------   ----------------------
5450			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
5451			 */
5452			if (cmsg->cmsg_len !=
5453			    CMSG_LEN(sizeof(struct sctp_initmsg)))
5454				return -EINVAL;
5455			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5456			break;
5457
5458		case SCTP_SNDRCV:
5459			/* SCTP Socket API Extension
5460			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5461			 *
5462			 * This cmsghdr structure specifies SCTP options for
5463			 * sendmsg() and describes SCTP header information
5464			 * about a received message through recvmsg().
5465			 *
5466			 * cmsg_level    cmsg_type      cmsg_data[]
5467			 * ------------  ------------   ----------------------
5468			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
5469			 */
5470			if (cmsg->cmsg_len !=
5471			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5472				return -EINVAL;
5473
5474			cmsgs->info =
5475				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5476
5477			/* Minimally, validate the sinfo_flags. */
5478			if (cmsgs->info->sinfo_flags &
5479			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5480			      SCTP_ABORT | SCTP_EOF))
5481				return -EINVAL;
5482			break;
5483
5484		default:
5485			return -EINVAL;
5486		}
5487	}
5488	return 0;
5489}
5490
5491/*
5492 * Wait for a packet..
5493 * Note: This function is the same function as in core/datagram.c
5494 * with a few modifications to make lksctp work.
5495 */
5496static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5497{
5498	int error;
5499	DEFINE_WAIT(wait);
5500
5501	prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5502
5503	/* Socket errors? */
5504	error = sock_error(sk);
5505	if (error)
5506		goto out;
5507
5508	if (!skb_queue_empty(&sk->sk_receive_queue))
5509		goto ready;
5510
5511	/* Socket shut down?  */
5512	if (sk->sk_shutdown & RCV_SHUTDOWN)
5513		goto out;
5514
5515	/* Sequenced packets can come disconnected.  If so we report the
5516	 * problem.
5517	 */
5518	error = -ENOTCONN;
5519
5520	/* Is there a good reason to think that we may receive some data?  */
5521	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5522		goto out;
5523
5524	/* Handle signals.  */
5525	if (signal_pending(current))
5526		goto interrupted;
5527
5528	/* Let another process have a go.  Since we are going to sleep
5529	 * anyway.  Note: This may cause odd behaviors if the message
5530	 * does not fit in the user's buffer, but this seems to be the
5531	 * only way to honor MSG_DONTWAIT realistically.
5532	 */
5533	sctp_release_sock(sk);
5534	*timeo_p = schedule_timeout(*timeo_p);
5535	sctp_lock_sock(sk);
5536
5537ready:
5538	finish_wait(sk->sk_sleep, &wait);
5539	return 0;
5540
5541interrupted:
5542	error = sock_intr_errno(*timeo_p);
5543
5544out:
5545	finish_wait(sk->sk_sleep, &wait);
5546	*err = error;
5547	return error;
5548}
5549
5550/* Receive a datagram.
5551 * Note: This is pretty much the same routine as in core/datagram.c
5552 * with a few changes to make lksctp work.
5553 */
5554static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5555					      int noblock, int *err)
5556{
5557	int error;
5558	struct sk_buff *skb;
5559	long timeo;
5560
5561	timeo = sock_rcvtimeo(sk, noblock);
5562
5563	SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5564			  timeo, MAX_SCHEDULE_TIMEOUT);
5565
5566	do {
5567		/* Again only user level code calls this function,
5568		 * so nothing interrupt level
5569		 * will suddenly eat the receive_queue.
5570		 *
5571		 *  Look at current nfs client by the way...
5572		 *  However, this function was corrent in any case. 8)
5573		 */
5574		if (flags & MSG_PEEK) {
5575			spin_lock_bh(&sk->sk_receive_queue.lock);
5576			skb = skb_peek(&sk->sk_receive_queue);
5577			if (skb)
5578				atomic_inc(&skb->users);
5579			spin_unlock_bh(&sk->sk_receive_queue.lock);
5580		} else {
5581			skb = skb_dequeue(&sk->sk_receive_queue);
5582		}
5583
5584		if (skb)
5585			return skb;
5586
5587		/* Caller is allowed not to check sk->sk_err before calling. */
5588		error = sock_error(sk);
5589		if (error)
5590			goto no_packet;
5591
5592		if (sk->sk_shutdown & RCV_SHUTDOWN)
5593			break;
5594
5595		/* User doesn't want to wait.  */
5596		error = -EAGAIN;
5597		if (!timeo)
5598			goto no_packet;
5599	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5600
5601	return NULL;
5602
5603no_packet:
5604	*err = error;
5605	return NULL;
5606}
5607
5608/* If sndbuf has changed, wake up per association sndbuf waiters.  */
5609static void __sctp_write_space(struct sctp_association *asoc)
5610{
5611	struct sock *sk = asoc->base.sk;
5612	struct socket *sock = sk->sk_socket;
5613
5614	if ((sctp_wspace(asoc) > 0) && sock) {
5615		if (waitqueue_active(&asoc->wait))
5616			wake_up_interruptible(&asoc->wait);
5617
5618		if (sctp_writeable(sk)) {
5619			if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5620				wake_up_interruptible(sk->sk_sleep);
5621
5622			/* Note that we try to include the Async I/O support
5623			 * here by modeling from the current TCP/UDP code.
5624			 * We have not tested with it yet.
5625			 */
5626			if (sock->fasync_list &&
5627			    !(sk->sk_shutdown & SEND_SHUTDOWN))
5628				sock_wake_async(sock, 2, POLL_OUT);
5629		}
5630	}
5631}
5632
5633/* Do accounting for the sndbuf space.
5634 * Decrement the used sndbuf space of the corresponding association by the
5635 * data size which was just transmitted(freed).
5636 */
5637static void sctp_wfree(struct sk_buff *skb)
5638{
5639	struct sctp_association *asoc;
5640	struct sctp_chunk *chunk;
5641	struct sock *sk;
5642
5643	/* Get the saved chunk pointer.  */
5644	chunk = *((struct sctp_chunk **)(skb->cb));
5645	asoc = chunk->asoc;
5646	sk = asoc->base.sk;
5647	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5648				sizeof(struct sk_buff) +
5649				sizeof(struct sctp_chunk);
5650
5651	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5652
5653	sock_wfree(skb);
5654	__sctp_write_space(asoc);
5655
5656	sctp_association_put(asoc);
5657}
5658
5659/* Do accounting for the receive space on the socket.
5660 * Accounting for the association is done in ulpevent.c
5661 * We set this as a destructor for the cloned data skbs so that
5662 * accounting is done at the correct time.
5663 */
5664void sctp_sock_rfree(struct sk_buff *skb)
5665{
5666	struct sock *sk = skb->sk;
5667	struct sctp_ulpevent *event = sctp_skb2event(skb);
5668
5669	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5670}
5671
5672
5673/* Helper function to wait for space in the sndbuf.  */
5674static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5675				size_t msg_len)
5676{
5677	struct sock *sk = asoc->base.sk;
5678	int err = 0;
5679	long current_timeo = *timeo_p;
5680	DEFINE_WAIT(wait);
5681
5682	SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5683			  asoc, (long)(*timeo_p), msg_len);
5684
5685	/* Increment the association's refcnt.  */
5686	sctp_association_hold(asoc);
5687
5688	/* Wait on the association specific sndbuf space. */
5689	for (;;) {
5690		prepare_to_wait_exclusive(&asoc->wait, &wait,
5691					  TASK_INTERRUPTIBLE);
5692		if (!*timeo_p)
5693			goto do_nonblock;
5694		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5695		    asoc->base.dead)
5696			goto do_error;
5697		if (signal_pending(current))
5698			goto do_interrupted;
5699		if (msg_len <= sctp_wspace(asoc))
5700			break;
5701
5702		/* Let another process have a go.  Since we are going
5703		 * to sleep anyway.
5704		 */
5705		sctp_release_sock(sk);
5706		current_timeo = schedule_timeout(current_timeo);
5707		BUG_ON(sk != asoc->base.sk);
5708		sctp_lock_sock(sk);
5709
5710		*timeo_p = current_timeo;
5711	}
5712
5713out:
5714	finish_wait(&asoc->wait, &wait);
5715
5716	/* Release the association's refcnt.  */
5717	sctp_association_put(asoc);
5718
5719	return err;
5720
5721do_error:
5722	err = -EPIPE;
5723	goto out;
5724
5725do_interrupted:
5726	err = sock_intr_errno(*timeo_p);
5727	goto out;
5728
5729do_nonblock:
5730	err = -EAGAIN;
5731	goto out;
5732}
5733
5734/* If socket sndbuf has changed, wake up all per association waiters.  */
5735void sctp_write_space(struct sock *sk)
5736{
5737	struct sctp_association *asoc;
5738	struct list_head *pos;
5739
5740	/* Wake up the tasks in each wait queue.  */
5741	list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5742		asoc = list_entry(pos, struct sctp_association, asocs);
5743		__sctp_write_space(asoc);
5744	}
5745}
5746
5747/* Is there any sndbuf space available on the socket?
5748 *
5749 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5750 * associations on the same socket.  For a UDP-style socket with
5751 * multiple associations, it is possible for it to be "unwriteable"
5752 * prematurely.  I assume that this is acceptable because
5753 * a premature "unwriteable" is better than an accidental "writeable" which
5754 * would cause an unwanted block under certain circumstances.  For the 1-1
5755 * UDP-style sockets or TCP-style sockets, this code should work.
5756 *  - Daisy
5757 */
5758static int sctp_writeable(struct sock *sk)
5759{
5760	int amt = 0;
5761
5762	amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5763	if (amt < 0)
5764		amt = 0;
5765	return amt;
5766}
5767
5768/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5769 * returns immediately with EINPROGRESS.
5770 */
5771static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5772{
5773	struct sock *sk = asoc->base.sk;
5774	int err = 0;
5775	long current_timeo = *timeo_p;
5776	DEFINE_WAIT(wait);
5777
5778	SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5779			  (long)(*timeo_p));
5780
5781	/* Increment the association's refcnt.  */
5782	sctp_association_hold(asoc);
5783
5784	for (;;) {
5785		prepare_to_wait_exclusive(&asoc->wait, &wait,
5786					  TASK_INTERRUPTIBLE);
5787		if (!*timeo_p)
5788			goto do_nonblock;
5789		if (sk->sk_shutdown & RCV_SHUTDOWN)
5790			break;
5791		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5792		    asoc->base.dead)
5793			goto do_error;
5794		if (signal_pending(current))
5795			goto do_interrupted;
5796
5797		if (sctp_state(asoc, ESTABLISHED))
5798			break;
5799
5800		/* Let another process have a go.  Since we are going
5801		 * to sleep anyway.
5802		 */
5803		sctp_release_sock(sk);
5804		current_timeo = schedule_timeout(current_timeo);
5805		sctp_lock_sock(sk);
5806
5807		*timeo_p = current_timeo;
5808	}
5809
5810out:
5811	finish_wait(&asoc->wait, &wait);
5812
5813	/* Release the association's refcnt.  */
5814	sctp_association_put(asoc);
5815
5816	return err;
5817
5818do_error:
5819	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5820		err = -ETIMEDOUT;
5821	else
5822		err = -ECONNREFUSED;
5823	goto out;
5824
5825do_interrupted:
5826	err = sock_intr_errno(*timeo_p);
5827	goto out;
5828
5829do_nonblock:
5830	err = -EINPROGRESS;
5831	goto out;
5832}
5833
5834static int sctp_wait_for_accept(struct sock *sk, long timeo)
5835{
5836	struct sctp_endpoint *ep;
5837	int err = 0;
5838	DEFINE_WAIT(wait);
5839
5840	ep = sctp_sk(sk)->ep;
5841
5842
5843	for (;;) {
5844		prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5845					  TASK_INTERRUPTIBLE);
5846
5847		if (list_empty(&ep->asocs)) {
5848			sctp_release_sock(sk);
5849			timeo = schedule_timeout(timeo);
5850			sctp_lock_sock(sk);
5851		}
5852
5853		err = -EINVAL;
5854		if (!sctp_sstate(sk, LISTENING))
5855			break;
5856
5857		err = 0;
5858		if (!list_empty(&ep->asocs))
5859			break;
5860
5861		err = sock_intr_errno(timeo);
5862		if (signal_pending(current))
5863			break;
5864
5865		err = -EAGAIN;
5866		if (!timeo)
5867			break;
5868	}
5869
5870	finish_wait(sk->sk_sleep, &wait);
5871
5872	return err;
5873}
5874
5875void sctp_wait_for_close(struct sock *sk, long timeout)
5876{
5877	DEFINE_WAIT(wait);
5878
5879	do {
5880		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5881		if (list_empty(&sctp_sk(sk)->ep->asocs))
5882			break;
5883		sctp_release_sock(sk);
5884		timeout = schedule_timeout(timeout);
5885		sctp_lock_sock(sk);
5886	} while (!signal_pending(current) && timeout);
5887
5888	finish_wait(sk->sk_sleep, &wait);
5889}
5890
5891static void sctp_sock_rfree_frag(struct sk_buff *skb)
5892{
5893	struct sk_buff *frag;
5894
5895	if (!skb->data_len)
5896		goto done;
5897
5898	/* Don't forget the fragments. */
5899	for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5900		sctp_sock_rfree_frag(frag);
5901
5902done:
5903	sctp_sock_rfree(skb);
5904}
5905
5906static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
5907{
5908	struct sk_buff *frag;
5909
5910	if (!skb->data_len)
5911		goto done;
5912
5913	/* Don't forget the fragments. */
5914	for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5915		sctp_skb_set_owner_r_frag(frag, sk);
5916
5917done:
5918	sctp_skb_set_owner_r(skb, sk);
5919}
5920
5921/* Populate the fields of the newsk from the oldsk and migrate the assoc
5922 * and its messages to the newsk.
5923 */
5924static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5925			      struct sctp_association *assoc,
5926			      sctp_socket_type_t type)
5927{
5928	struct sctp_sock *oldsp = sctp_sk(oldsk);
5929	struct sctp_sock *newsp = sctp_sk(newsk);
5930	struct sctp_bind_bucket *pp; /* hash list port iterator */
5931	struct sctp_endpoint *newep = newsp->ep;
5932	struct sk_buff *skb, *tmp;
5933	struct sctp_ulpevent *event;
5934	int flags = 0;
5935
5936	/* Migrate socket buffer sizes and all the socket level options to the
5937	 * new socket.
5938	 */
5939	newsk->sk_sndbuf = oldsk->sk_sndbuf;
5940	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5941	/* Brute force copy old sctp opt. */
5942	inet_sk_copy_descendant(newsk, oldsk);
5943
5944	/* Restore the ep value that was overwritten with the above structure
5945	 * copy.
5946	 */
5947	newsp->ep = newep;
5948	newsp->hmac = NULL;
5949
5950	/* Hook this new socket in to the bind_hash list. */
5951	pp = sctp_sk(oldsk)->bind_hash;
5952	sk_add_bind_node(newsk, &pp->owner);
5953	sctp_sk(newsk)->bind_hash = pp;
5954	inet_sk(newsk)->num = inet_sk(oldsk)->num;
5955
5956	/* Copy the bind_addr list from the original endpoint to the new
5957	 * endpoint so that we can handle restarts properly
5958	 */
5959	if (PF_INET6 == assoc->base.sk->sk_family)
5960		flags = SCTP_ADDR6_ALLOWED;
5961	if (assoc->peer.ipv4_address)
5962		flags |= SCTP_ADDR4_PEERSUPP;
5963	if (assoc->peer.ipv6_address)
5964		flags |= SCTP_ADDR6_PEERSUPP;
5965	sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5966			     &oldsp->ep->base.bind_addr,
5967			     SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5968
5969	/* Move any messages in the old socket's receive queue that are for the
5970	 * peeled off association to the new socket's receive queue.
5971	 */
5972	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5973		event = sctp_skb2event(skb);
5974		if (event->asoc == assoc) {
5975			sctp_sock_rfree_frag(skb);
5976			__skb_unlink(skb, &oldsk->sk_receive_queue);
5977			__skb_queue_tail(&newsk->sk_receive_queue, skb);
5978			sctp_skb_set_owner_r_frag(skb, newsk);
5979		}
5980	}
5981
5982	/* Clean up any messages pending delivery due to partial
5983	 * delivery.   Three cases:
5984	 * 1) No partial deliver;  no work.
5985	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5986	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5987	 */
5988	skb_queue_head_init(&newsp->pd_lobby);
5989	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
5990
5991	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
5992		struct sk_buff_head *queue;
5993
5994		/* Decide which queue to move pd_lobby skbs to. */
5995		if (assoc->ulpq.pd_mode) {
5996			queue = &newsp->pd_lobby;
5997		} else
5998			queue = &newsk->sk_receive_queue;
5999
6000		/* Walk through the pd_lobby, looking for skbs that
6001		 * need moved to the new socket.
6002		 */
6003		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6004			event = sctp_skb2event(skb);
6005			if (event->asoc == assoc) {
6006				sctp_sock_rfree_frag(skb);
6007				__skb_unlink(skb, &oldsp->pd_lobby);
6008				__skb_queue_tail(queue, skb);
6009				sctp_skb_set_owner_r_frag(skb, newsk);
6010			}
6011		}
6012
6013		/* Clear up any skbs waiting for the partial
6014		 * delivery to finish.
6015		 */
6016		if (assoc->ulpq.pd_mode)
6017			sctp_clear_pd(oldsk, NULL);
6018
6019	}
6020
6021	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6022		sctp_sock_rfree_frag(skb);
6023		sctp_skb_set_owner_r_frag(skb, newsk);
6024	}
6025
6026	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6027		sctp_sock_rfree_frag(skb);
6028		sctp_skb_set_owner_r_frag(skb, newsk);
6029	}
6030
6031	/* Set the type of socket to indicate that it is peeled off from the
6032	 * original UDP-style socket or created with the accept() call on a
6033	 * TCP-style socket..
6034	 */
6035	newsp->type = type;
6036
6037	/* Mark the new socket "in-use" by the user so that any packets
6038	 * that may arrive on the association after we've moved it are
6039	 * queued to the backlog.  This prevents a potential race between
6040	 * backlog processing on the old socket and new-packet processing
6041	 * on the new socket.
6042	 *
6043	 * The caller has just allocated newsk so we can guarantee that other
6044	 * paths won't try to lock it and then oldsk.
6045	 */
6046	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6047	sctp_assoc_migrate(assoc, newsk);
6048
6049	/* If the association on the newsk is already closed before accept()
6050	 * is called, set RCV_SHUTDOWN flag.
6051	 */
6052	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6053		newsk->sk_shutdown |= RCV_SHUTDOWN;
6054
6055	newsk->sk_state = SCTP_SS_ESTABLISHED;
6056	sctp_release_sock(newsk);
6057}
6058
6059/* This proto struct describes the ULP interface for SCTP.  */
6060struct proto sctp_prot = {
6061	.name        =	"SCTP",
6062	.owner       =	THIS_MODULE,
6063	.close       =	sctp_close,
6064	.connect     =	sctp_connect,
6065	.disconnect  =	sctp_disconnect,
6066	.accept      =	sctp_accept,
6067	.ioctl       =	sctp_ioctl,
6068	.init        =	sctp_init_sock,
6069	.destroy     =	sctp_destroy_sock,
6070	.shutdown    =	sctp_shutdown,
6071	.setsockopt  =	sctp_setsockopt,
6072	.getsockopt  =	sctp_getsockopt,
6073	.sendmsg     =	sctp_sendmsg,
6074	.recvmsg     =	sctp_recvmsg,
6075	.bind        =	sctp_bind,
6076	.backlog_rcv =	sctp_backlog_rcv,
6077	.hash        =	sctp_hash,
6078	.unhash      =	sctp_unhash,
6079	.get_port    =	sctp_get_port,
6080	.obj_size    =  sizeof(struct sctp_sock),
6081};
6082
6083#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6084struct proto sctpv6_prot = {
6085	.name		= "SCTPv6",
6086	.owner		= THIS_MODULE,
6087	.close		= sctp_close,
6088	.connect	= sctp_connect,
6089	.disconnect	= sctp_disconnect,
6090	.accept		= sctp_accept,
6091	.ioctl		= sctp_ioctl,
6092	.init		= sctp_init_sock,
6093	.destroy	= sctp_destroy_sock,
6094	.shutdown	= sctp_shutdown,
6095	.setsockopt	= sctp_setsockopt,
6096	.getsockopt	= sctp_getsockopt,
6097	.sendmsg	= sctp_sendmsg,
6098	.recvmsg	= sctp_recvmsg,
6099	.bind		= sctp_bind,
6100	.backlog_rcv	= sctp_backlog_rcv,
6101	.hash		= sctp_hash,
6102	.unhash		= sctp_unhash,
6103	.get_port	= sctp_get_port,
6104	.obj_size	= sizeof(struct sctp6_sock),
6105};
6106#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
6107