xref: /openbmc/linux/net/sctp/socket.c (revision 22246614)
1 /* SCTP kernel 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 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  * This SCTP 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  * This SCTP 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. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 				size_t msg_len);
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 					union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 			    struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 			      struct sctp_association *, sctp_socket_type_t);
108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
109 
110 extern struct kmem_cache *sctp_bucket_cachep;
111 extern int sysctl_sctp_mem[3];
112 extern int sysctl_sctp_rmem[3];
113 extern int sysctl_sctp_wmem[3];
114 
115 static int sctp_memory_pressure;
116 static atomic_t sctp_memory_allocated;
117 static atomic_t sctp_sockets_allocated;
118 
119 static void sctp_enter_memory_pressure(void)
120 {
121 	sctp_memory_pressure = 1;
122 }
123 
124 
125 /* Get the sndbuf space available at the time on the association.  */
126 static inline int sctp_wspace(struct sctp_association *asoc)
127 {
128 	int amt;
129 
130 	if (asoc->ep->sndbuf_policy)
131 		amt = asoc->sndbuf_used;
132 	else
133 		amt = atomic_read(&asoc->base.sk->sk_wmem_alloc);
134 
135 	if (amt >= asoc->base.sk->sk_sndbuf) {
136 		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
137 			amt = 0;
138 		else {
139 			amt = sk_stream_wspace(asoc->base.sk);
140 			if (amt < 0)
141 				amt = 0;
142 		}
143 	} else {
144 		amt = asoc->base.sk->sk_sndbuf - amt;
145 	}
146 	return amt;
147 }
148 
149 /* Increment the used sndbuf space count of the corresponding association by
150  * the size of the outgoing data chunk.
151  * Also, set the skb destructor for sndbuf accounting later.
152  *
153  * Since it is always 1-1 between chunk and skb, and also a new skb is always
154  * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155  * destructor in the data chunk skb for the purpose of the sndbuf space
156  * tracking.
157  */
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
159 {
160 	struct sctp_association *asoc = chunk->asoc;
161 	struct sock *sk = asoc->base.sk;
162 
163 	/* The sndbuf space is tracked per association.  */
164 	sctp_association_hold(asoc);
165 
166 	skb_set_owner_w(chunk->skb, sk);
167 
168 	chunk->skb->destructor = sctp_wfree;
169 	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
170 	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
171 
172 	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 				sizeof(struct sk_buff) +
174 				sizeof(struct sctp_chunk);
175 
176 	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 	sk->sk_wmem_queued += chunk->skb->truesize;
178 	sk_mem_charge(sk, chunk->skb->truesize);
179 }
180 
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
183 				   int len)
184 {
185 	struct sctp_af *af;
186 
187 	/* Verify basic sockaddr. */
188 	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
189 	if (!af)
190 		return -EINVAL;
191 
192 	/* Is this a valid SCTP address?  */
193 	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
194 		return -EINVAL;
195 
196 	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
197 		return -EINVAL;
198 
199 	return 0;
200 }
201 
202 /* Look up the association by its id.  If this is not a UDP-style
203  * socket, the ID field is always ignored.
204  */
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
206 {
207 	struct sctp_association *asoc = NULL;
208 
209 	/* If this is not a UDP-style socket, assoc id should be ignored. */
210 	if (!sctp_style(sk, UDP)) {
211 		/* Return NULL if the socket state is not ESTABLISHED. It
212 		 * could be a TCP-style listening socket or a socket which
213 		 * hasn't yet called connect() to establish an association.
214 		 */
215 		if (!sctp_sstate(sk, ESTABLISHED))
216 			return NULL;
217 
218 		/* Get the first and the only association from the list. */
219 		if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 					  struct sctp_association, asocs);
222 		return asoc;
223 	}
224 
225 	/* Otherwise this is a UDP-style socket. */
226 	if (!id || (id == (sctp_assoc_t)-1))
227 		return NULL;
228 
229 	spin_lock_bh(&sctp_assocs_id_lock);
230 	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 	spin_unlock_bh(&sctp_assocs_id_lock);
232 
233 	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
234 		return NULL;
235 
236 	return asoc;
237 }
238 
239 /* Look up the transport from an address and an assoc id. If both address and
240  * id are specified, the associations matching the address and the id should be
241  * the same.
242  */
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 					      struct sockaddr_storage *addr,
245 					      sctp_assoc_t id)
246 {
247 	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 	struct sctp_transport *transport;
249 	union sctp_addr *laddr = (union sctp_addr *)addr;
250 
251 	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 					       laddr,
253 					       &transport);
254 
255 	if (!addr_asoc)
256 		return NULL;
257 
258 	id_asoc = sctp_id2assoc(sk, id);
259 	if (id_asoc && (id_asoc != addr_asoc))
260 		return NULL;
261 
262 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 						(union sctp_addr *)addr);
264 
265 	return transport;
266 }
267 
268 /* API 3.1.2 bind() - UDP Style Syntax
269  * The syntax of bind() is,
270  *
271  *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
272  *
273  *   sd      - the socket descriptor returned by socket().
274  *   addr    - the address structure (struct sockaddr_in or struct
275  *             sockaddr_in6 [RFC 2553]),
276  *   addr_len - the size of the address structure.
277  */
278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
279 {
280 	int retval = 0;
281 
282 	sctp_lock_sock(sk);
283 
284 	SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
285 			  sk, addr, addr_len);
286 
287 	/* Disallow binding twice. */
288 	if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
290 				      addr_len);
291 	else
292 		retval = -EINVAL;
293 
294 	sctp_release_sock(sk);
295 
296 	return retval;
297 }
298 
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
300 
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 					union sctp_addr *addr, int len)
304 {
305 	struct sctp_af *af;
306 
307 	/* Check minimum size.  */
308 	if (len < sizeof (struct sockaddr))
309 		return NULL;
310 
311 	/* Does this PF support this AF? */
312 	if (!opt->pf->af_supported(addr->sa.sa_family, opt))
313 		return NULL;
314 
315 	/* If we get this far, af is valid. */
316 	af = sctp_get_af_specific(addr->sa.sa_family);
317 
318 	if (len < af->sockaddr_len)
319 		return NULL;
320 
321 	return af;
322 }
323 
324 /* Bind a local address either to an endpoint or to an association.  */
325 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
326 {
327 	struct sctp_sock *sp = sctp_sk(sk);
328 	struct sctp_endpoint *ep = sp->ep;
329 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
330 	struct sctp_af *af;
331 	unsigned short snum;
332 	int ret = 0;
333 
334 	/* Common sockaddr verification. */
335 	af = sctp_sockaddr_af(sp, addr, len);
336 	if (!af) {
337 		SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
338 				  sk, addr, len);
339 		return -EINVAL;
340 	}
341 
342 	snum = ntohs(addr->v4.sin_port);
343 
344 	SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
345 				 ", port: %d, new port: %d, len: %d)\n",
346 				 sk,
347 				 addr,
348 				 bp->port, snum,
349 				 len);
350 
351 	/* PF specific bind() address verification. */
352 	if (!sp->pf->bind_verify(sp, addr))
353 		return -EADDRNOTAVAIL;
354 
355 	/* We must either be unbound, or bind to the same port.
356 	 * It's OK to allow 0 ports if we are already bound.
357 	 * We'll just inhert an already bound port in this case
358 	 */
359 	if (bp->port) {
360 		if (!snum)
361 			snum = bp->port;
362 		else if (snum != bp->port) {
363 			SCTP_DEBUG_PRINTK("sctp_do_bind:"
364 				  " New port %d does not match existing port "
365 				  "%d.\n", snum, bp->port);
366 			return -EINVAL;
367 		}
368 	}
369 
370 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
371 		return -EACCES;
372 
373 	/* Make sure we are allowed to bind here.
374 	 * The function sctp_get_port_local() does duplicate address
375 	 * detection.
376 	 */
377 	addr->v4.sin_port = htons(snum);
378 	if ((ret = sctp_get_port_local(sk, addr))) {
379 		if (ret == (long) sk) {
380 			/* This endpoint has a conflicting address. */
381 			return -EINVAL;
382 		} else {
383 			return -EADDRINUSE;
384 		}
385 	}
386 
387 	/* Refresh ephemeral port.  */
388 	if (!bp->port)
389 		bp->port = inet_sk(sk)->num;
390 
391 	/* Add the address to the bind address list.
392 	 * Use GFP_ATOMIC since BHs will be disabled.
393 	 */
394 	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
395 
396 	/* Copy back into socket for getsockname() use. */
397 	if (!ret) {
398 		inet_sk(sk)->sport = htons(inet_sk(sk)->num);
399 		af->to_sk_saddr(addr, sk);
400 	}
401 
402 	return ret;
403 }
404 
405  /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
406  *
407  * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408  * at any one time.  If a sender, after sending an ASCONF chunk, decides
409  * it needs to transfer another ASCONF Chunk, it MUST wait until the
410  * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411  * subsequent ASCONF. Note this restriction binds each side, so at any
412  * time two ASCONF may be in-transit on any given association (one sent
413  * from each endpoint).
414  */
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 			    struct sctp_chunk *chunk)
417 {
418 	int		retval = 0;
419 
420 	/* If there is an outstanding ASCONF chunk, queue it for later
421 	 * transmission.
422 	 */
423 	if (asoc->addip_last_asconf) {
424 		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 		goto out;
426 	}
427 
428 	/* Hold the chunk until an ASCONF_ACK is received. */
429 	sctp_chunk_hold(chunk);
430 	retval = sctp_primitive_ASCONF(asoc, chunk);
431 	if (retval)
432 		sctp_chunk_free(chunk);
433 	else
434 		asoc->addip_last_asconf = chunk;
435 
436 out:
437 	return retval;
438 }
439 
440 /* Add a list of addresses as bind addresses to local endpoint or
441  * association.
442  *
443  * Basically run through each address specified in the addrs/addrcnt
444  * array/length pair, determine if it is IPv6 or IPv4 and call
445  * sctp_do_bind() on it.
446  *
447  * If any of them fails, then the operation will be reversed and the
448  * ones that were added will be removed.
449  *
450  * Only sctp_setsockopt_bindx() is supposed to call this function.
451  */
452 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
453 {
454 	int cnt;
455 	int retval = 0;
456 	void *addr_buf;
457 	struct sockaddr *sa_addr;
458 	struct sctp_af *af;
459 
460 	SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
461 			  sk, addrs, addrcnt);
462 
463 	addr_buf = addrs;
464 	for (cnt = 0; cnt < addrcnt; cnt++) {
465 		/* The list may contain either IPv4 or IPv6 address;
466 		 * determine the address length for walking thru the list.
467 		 */
468 		sa_addr = (struct sockaddr *)addr_buf;
469 		af = sctp_get_af_specific(sa_addr->sa_family);
470 		if (!af) {
471 			retval = -EINVAL;
472 			goto err_bindx_add;
473 		}
474 
475 		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
476 				      af->sockaddr_len);
477 
478 		addr_buf += af->sockaddr_len;
479 
480 err_bindx_add:
481 		if (retval < 0) {
482 			/* Failed. Cleanup the ones that have been added */
483 			if (cnt > 0)
484 				sctp_bindx_rem(sk, addrs, cnt);
485 			return retval;
486 		}
487 	}
488 
489 	return retval;
490 }
491 
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493  * associations that are part of the endpoint indicating that a list of local
494  * addresses are added to the endpoint.
495  *
496  * If any of the addresses is already in the bind address list of the
497  * association, we do not send the chunk for that association.  But it will not
498  * affect other associations.
499  *
500  * Only sctp_setsockopt_bindx() is supposed to call this function.
501  */
502 static int sctp_send_asconf_add_ip(struct sock		*sk,
503 				   struct sockaddr	*addrs,
504 				   int 			addrcnt)
505 {
506 	struct sctp_sock		*sp;
507 	struct sctp_endpoint		*ep;
508 	struct sctp_association		*asoc;
509 	struct sctp_bind_addr		*bp;
510 	struct sctp_chunk		*chunk;
511 	struct sctp_sockaddr_entry	*laddr;
512 	union sctp_addr			*addr;
513 	union sctp_addr			saveaddr;
514 	void				*addr_buf;
515 	struct sctp_af			*af;
516 	struct list_head		*p;
517 	int 				i;
518 	int 				retval = 0;
519 
520 	if (!sctp_addip_enable)
521 		return retval;
522 
523 	sp = sctp_sk(sk);
524 	ep = sp->ep;
525 
526 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
527 			  __func__, sk, addrs, addrcnt);
528 
529 	list_for_each_entry(asoc, &ep->asocs, asocs) {
530 
531 		if (!asoc->peer.asconf_capable)
532 			continue;
533 
534 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
535 			continue;
536 
537 		if (!sctp_state(asoc, ESTABLISHED))
538 			continue;
539 
540 		/* Check if any address in the packed array of addresses is
541 		 * in the bind address list of the association. If so,
542 		 * do not send the asconf chunk to its peer, but continue with
543 		 * other associations.
544 		 */
545 		addr_buf = addrs;
546 		for (i = 0; i < addrcnt; i++) {
547 			addr = (union sctp_addr *)addr_buf;
548 			af = sctp_get_af_specific(addr->v4.sin_family);
549 			if (!af) {
550 				retval = -EINVAL;
551 				goto out;
552 			}
553 
554 			if (sctp_assoc_lookup_laddr(asoc, addr))
555 				break;
556 
557 			addr_buf += af->sockaddr_len;
558 		}
559 		if (i < addrcnt)
560 			continue;
561 
562 		/* Use the first valid address in bind addr list of
563 		 * association as Address Parameter of ASCONF CHUNK.
564 		 */
565 		bp = &asoc->base.bind_addr;
566 		p = bp->address_list.next;
567 		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 						   addrcnt, SCTP_PARAM_ADD_IP);
570 		if (!chunk) {
571 			retval = -ENOMEM;
572 			goto out;
573 		}
574 
575 		retval = sctp_send_asconf(asoc, chunk);
576 		if (retval)
577 			goto out;
578 
579 		/* Add the new addresses to the bind address list with
580 		 * use_as_src set to 0.
581 		 */
582 		addr_buf = addrs;
583 		for (i = 0; i < addrcnt; i++) {
584 			addr = (union sctp_addr *)addr_buf;
585 			af = sctp_get_af_specific(addr->v4.sin_family);
586 			memcpy(&saveaddr, addr, af->sockaddr_len);
587 			retval = sctp_add_bind_addr(bp, &saveaddr,
588 						    SCTP_ADDR_NEW, GFP_ATOMIC);
589 			addr_buf += af->sockaddr_len;
590 		}
591 	}
592 
593 out:
594 	return retval;
595 }
596 
597 /* Remove a list of addresses from bind addresses list.  Do not remove the
598  * last address.
599  *
600  * Basically run through each address specified in the addrs/addrcnt
601  * array/length pair, determine if it is IPv6 or IPv4 and call
602  * sctp_del_bind() on it.
603  *
604  * If any of them fails, then the operation will be reversed and the
605  * ones that were removed will be added back.
606  *
607  * At least one address has to be left; if only one address is
608  * available, the operation will return -EBUSY.
609  *
610  * Only sctp_setsockopt_bindx() is supposed to call this function.
611  */
612 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
613 {
614 	struct sctp_sock *sp = sctp_sk(sk);
615 	struct sctp_endpoint *ep = sp->ep;
616 	int cnt;
617 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
618 	int retval = 0;
619 	void *addr_buf;
620 	union sctp_addr *sa_addr;
621 	struct sctp_af *af;
622 
623 	SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
624 			  sk, addrs, addrcnt);
625 
626 	addr_buf = addrs;
627 	for (cnt = 0; cnt < addrcnt; cnt++) {
628 		/* If the bind address list is empty or if there is only one
629 		 * bind address, there is nothing more to be removed (we need
630 		 * at least one address here).
631 		 */
632 		if (list_empty(&bp->address_list) ||
633 		    (sctp_list_single_entry(&bp->address_list))) {
634 			retval = -EBUSY;
635 			goto err_bindx_rem;
636 		}
637 
638 		sa_addr = (union sctp_addr *)addr_buf;
639 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
640 		if (!af) {
641 			retval = -EINVAL;
642 			goto err_bindx_rem;
643 		}
644 
645 		if (!af->addr_valid(sa_addr, sp, NULL)) {
646 			retval = -EADDRNOTAVAIL;
647 			goto err_bindx_rem;
648 		}
649 
650 		if (sa_addr->v4.sin_port != htons(bp->port)) {
651 			retval = -EINVAL;
652 			goto err_bindx_rem;
653 		}
654 
655 		/* FIXME - There is probably a need to check if sk->sk_saddr and
656 		 * sk->sk_rcv_addr are currently set to one of the addresses to
657 		 * be removed. This is something which needs to be looked into
658 		 * when we are fixing the outstanding issues with multi-homing
659 		 * socket routing and failover schemes. Refer to comments in
660 		 * sctp_do_bind(). -daisy
661 		 */
662 		retval = sctp_del_bind_addr(bp, sa_addr);
663 
664 		addr_buf += af->sockaddr_len;
665 err_bindx_rem:
666 		if (retval < 0) {
667 			/* Failed. Add the ones that has been removed back */
668 			if (cnt > 0)
669 				sctp_bindx_add(sk, addrs, cnt);
670 			return retval;
671 		}
672 	}
673 
674 	return retval;
675 }
676 
677 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
678  * the associations that are part of the endpoint indicating that a list of
679  * local addresses are removed from the endpoint.
680  *
681  * If any of the addresses is already in the bind address list of the
682  * association, we do not send the chunk for that association.  But it will not
683  * affect other associations.
684  *
685  * Only sctp_setsockopt_bindx() is supposed to call this function.
686  */
687 static int sctp_send_asconf_del_ip(struct sock		*sk,
688 				   struct sockaddr	*addrs,
689 				   int			addrcnt)
690 {
691 	struct sctp_sock	*sp;
692 	struct sctp_endpoint	*ep;
693 	struct sctp_association	*asoc;
694 	struct sctp_transport	*transport;
695 	struct sctp_bind_addr	*bp;
696 	struct sctp_chunk	*chunk;
697 	union sctp_addr		*laddr;
698 	void			*addr_buf;
699 	struct sctp_af		*af;
700 	struct sctp_sockaddr_entry *saddr;
701 	int 			i;
702 	int 			retval = 0;
703 
704 	if (!sctp_addip_enable)
705 		return retval;
706 
707 	sp = sctp_sk(sk);
708 	ep = sp->ep;
709 
710 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 			  __func__, sk, addrs, addrcnt);
712 
713 	list_for_each_entry(asoc, &ep->asocs, asocs) {
714 
715 		if (!asoc->peer.asconf_capable)
716 			continue;
717 
718 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
719 			continue;
720 
721 		if (!sctp_state(asoc, ESTABLISHED))
722 			continue;
723 
724 		/* Check if any address in the packed array of addresses is
725 		 * not present in the bind address list of the association.
726 		 * If so, do not send the asconf chunk to its peer, but
727 		 * continue with other associations.
728 		 */
729 		addr_buf = addrs;
730 		for (i = 0; i < addrcnt; i++) {
731 			laddr = (union sctp_addr *)addr_buf;
732 			af = sctp_get_af_specific(laddr->v4.sin_family);
733 			if (!af) {
734 				retval = -EINVAL;
735 				goto out;
736 			}
737 
738 			if (!sctp_assoc_lookup_laddr(asoc, laddr))
739 				break;
740 
741 			addr_buf += af->sockaddr_len;
742 		}
743 		if (i < addrcnt)
744 			continue;
745 
746 		/* Find one address in the association's bind address list
747 		 * that is not in the packed array of addresses. This is to
748 		 * make sure that we do not delete all the addresses in the
749 		 * association.
750 		 */
751 		bp = &asoc->base.bind_addr;
752 		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
753 					       addrcnt, sp);
754 		if (!laddr)
755 			continue;
756 
757 		/* We do not need RCU protection throughout this loop
758 		 * because this is done under a socket lock from the
759 		 * setsockopt call.
760 		 */
761 		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
762 						   SCTP_PARAM_DEL_IP);
763 		if (!chunk) {
764 			retval = -ENOMEM;
765 			goto out;
766 		}
767 
768 		/* Reset use_as_src flag for the addresses in the bind address
769 		 * list that are to be deleted.
770 		 */
771 		addr_buf = addrs;
772 		for (i = 0; i < addrcnt; i++) {
773 			laddr = (union sctp_addr *)addr_buf;
774 			af = sctp_get_af_specific(laddr->v4.sin_family);
775 			list_for_each_entry(saddr, &bp->address_list, list) {
776 				if (sctp_cmp_addr_exact(&saddr->a, laddr))
777 					saddr->state = SCTP_ADDR_DEL;
778 			}
779 			addr_buf += af->sockaddr_len;
780 		}
781 
782 		/* Update the route and saddr entries for all the transports
783 		 * as some of the addresses in the bind address list are
784 		 * about to be deleted and cannot be used as source addresses.
785 		 */
786 		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
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 	}
795 out:
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  */
875 SCTP_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 
946 out:
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  */
957 static 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 = NULL;
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 		/* Save current address so we can work with it */
1008 		memcpy(&to, sa_addr, af->sockaddr_len);
1009 
1010 		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1011 		if (err)
1012 			goto out_free;
1013 
1014 		/* Make sure the destination port is correctly set
1015 		 * in all addresses.
1016 		 */
1017 		if (asoc && asoc->peer.port && asoc->peer.port != port)
1018 			goto out_free;
1019 
1020 
1021 		/* Check if there already is a matching association on the
1022 		 * endpoint (other than the one created here).
1023 		 */
1024 		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1025 		if (asoc2 && asoc2 != asoc) {
1026 			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1027 				err = -EISCONN;
1028 			else
1029 				err = -EALREADY;
1030 			goto out_free;
1031 		}
1032 
1033 		/* If we could not find a matching association on the endpoint,
1034 		 * make sure that there is no peeled-off association matching
1035 		 * the peer address even on another socket.
1036 		 */
1037 		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1038 			err = -EADDRNOTAVAIL;
1039 			goto out_free;
1040 		}
1041 
1042 		if (!asoc) {
1043 			/* If a bind() or sctp_bindx() is not called prior to
1044 			 * an sctp_connectx() call, the system picks an
1045 			 * ephemeral port and will choose an address set
1046 			 * equivalent to binding with a wildcard address.
1047 			 */
1048 			if (!ep->base.bind_addr.port) {
1049 				if (sctp_autobind(sk)) {
1050 					err = -EAGAIN;
1051 					goto out_free;
1052 				}
1053 			} else {
1054 				/*
1055 				 * If an unprivileged user inherits a 1-many
1056 				 * style socket with open associations on a
1057 				 * privileged port, it MAY be permitted to
1058 				 * accept new associations, but it SHOULD NOT
1059 				 * be permitted to open new associations.
1060 				 */
1061 				if (ep->base.bind_addr.port < PROT_SOCK &&
1062 				    !capable(CAP_NET_BIND_SERVICE)) {
1063 					err = -EACCES;
1064 					goto out_free;
1065 				}
1066 			}
1067 
1068 			scope = sctp_scope(&to);
1069 			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1070 			if (!asoc) {
1071 				err = -ENOMEM;
1072 				goto out_free;
1073 			}
1074 		}
1075 
1076 		/* Prime the peer's transport structures.  */
1077 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1078 						SCTP_UNKNOWN);
1079 		if (!transport) {
1080 			err = -ENOMEM;
1081 			goto out_free;
1082 		}
1083 
1084 		addrcnt++;
1085 		addr_buf += af->sockaddr_len;
1086 		walk_size += af->sockaddr_len;
1087 	}
1088 
1089 	err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1090 	if (err < 0) {
1091 		goto out_free;
1092 	}
1093 
1094 	err = sctp_primitive_ASSOCIATE(asoc, NULL);
1095 	if (err < 0) {
1096 		goto out_free;
1097 	}
1098 
1099 	/* Initialize sk's dport and daddr for getpeername() */
1100 	inet_sk(sk)->dport = htons(asoc->peer.port);
1101 	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1102 	af->to_sk_daddr(sa_addr, sk);
1103 	sk->sk_err = 0;
1104 
1105 	/* in-kernel sockets don't generally have a file allocated to them
1106 	 * if all they do is call sock_create_kern().
1107 	 */
1108 	if (sk->sk_socket->file)
1109 		f_flags = sk->sk_socket->file->f_flags;
1110 
1111 	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1112 
1113 	err = sctp_wait_for_connect(asoc, &timeo);
1114 
1115 	/* Don't free association on exit. */
1116 	asoc = NULL;
1117 
1118 out_free:
1119 
1120 	SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1121 			  " kaddrs: %p err: %d\n",
1122 			  asoc, kaddrs, err);
1123 	if (asoc)
1124 		sctp_association_free(asoc);
1125 	return err;
1126 }
1127 
1128 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1129  *
1130  * API 8.9
1131  * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1132  *
1133  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1134  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1135  * or IPv6 addresses.
1136  *
1137  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1138  * Section 3.1.2 for this usage.
1139  *
1140  * addrs is a pointer to an array of one or more socket addresses. Each
1141  * address is contained in its appropriate structure (i.e. struct
1142  * sockaddr_in or struct sockaddr_in6) the family of the address type
1143  * must be used to distengish the address length (note that this
1144  * representation is termed a "packed array" of addresses). The caller
1145  * specifies the number of addresses in the array with addrcnt.
1146  *
1147  * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1148  * -1, and sets errno to the appropriate error code.
1149  *
1150  * For SCTP, the port given in each socket address must be the same, or
1151  * sctp_connectx() will fail, setting errno to EINVAL.
1152  *
1153  * An application can use sctp_connectx to initiate an association with
1154  * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1155  * allows a caller to specify multiple addresses at which a peer can be
1156  * reached.  The way the SCTP stack uses the list of addresses to set up
1157  * the association is implementation dependant.  This function only
1158  * specifies that the stack will try to make use of all the addresses in
1159  * the list when needed.
1160  *
1161  * Note that the list of addresses passed in is only used for setting up
1162  * the association.  It does not necessarily equal the set of addresses
1163  * the peer uses for the resulting association.  If the caller wants to
1164  * find out the set of peer addresses, it must use sctp_getpaddrs() to
1165  * retrieve them after the association has been set up.
1166  *
1167  * Basically do nothing but copying the addresses from user to kernel
1168  * land and invoking either sctp_connectx(). This is used for tunneling
1169  * the sctp_connectx() request through sctp_setsockopt() from userspace.
1170  *
1171  * We don't use copy_from_user() for optimization: we first do the
1172  * sanity checks (buffer size -fast- and access check-healthy
1173  * pointer); if all of those succeed, then we can alloc the memory
1174  * (expensive operation) needed to copy the data to kernel. Then we do
1175  * the copying without checking the user space area
1176  * (__copy_from_user()).
1177  *
1178  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1179  * it.
1180  *
1181  * sk        The sk of the socket
1182  * addrs     The pointer to the addresses in user land
1183  * addrssize Size of the addrs buffer
1184  *
1185  * Returns 0 if ok, <0 errno code on error.
1186  */
1187 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1188 				      struct sockaddr __user *addrs,
1189 				      int addrs_size)
1190 {
1191 	int err = 0;
1192 	struct sockaddr *kaddrs;
1193 
1194 	SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1195 			  __func__, sk, addrs, addrs_size);
1196 
1197 	if (unlikely(addrs_size <= 0))
1198 		return -EINVAL;
1199 
1200 	/* Check the user passed a healthy pointer.  */
1201 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1202 		return -EFAULT;
1203 
1204 	/* Alloc space for the address array in kernel memory.  */
1205 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1206 	if (unlikely(!kaddrs))
1207 		return -ENOMEM;
1208 
1209 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1210 		err = -EFAULT;
1211 	} else {
1212 		err = __sctp_connect(sk, kaddrs, addrs_size);
1213 	}
1214 
1215 	kfree(kaddrs);
1216 	return err;
1217 }
1218 
1219 /* API 3.1.4 close() - UDP Style Syntax
1220  * Applications use close() to perform graceful shutdown (as described in
1221  * Section 10.1 of [SCTP]) on ALL the associations currently represented
1222  * by a UDP-style socket.
1223  *
1224  * The syntax is
1225  *
1226  *   ret = close(int sd);
1227  *
1228  *   sd      - the socket descriptor of the associations to be closed.
1229  *
1230  * To gracefully shutdown a specific association represented by the
1231  * UDP-style socket, an application should use the sendmsg() call,
1232  * passing no user data, but including the appropriate flag in the
1233  * ancillary data (see Section xxxx).
1234  *
1235  * If sd in the close() call is a branched-off socket representing only
1236  * one association, the shutdown is performed on that association only.
1237  *
1238  * 4.1.6 close() - TCP Style Syntax
1239  *
1240  * Applications use close() to gracefully close down an association.
1241  *
1242  * The syntax is:
1243  *
1244  *    int close(int sd);
1245  *
1246  *      sd      - the socket descriptor of the association to be closed.
1247  *
1248  * After an application calls close() on a socket descriptor, no further
1249  * socket operations will succeed on that descriptor.
1250  *
1251  * API 7.1.4 SO_LINGER
1252  *
1253  * An application using the TCP-style socket can use this option to
1254  * perform the SCTP ABORT primitive.  The linger option structure is:
1255  *
1256  *  struct  linger {
1257  *     int     l_onoff;                // option on/off
1258  *     int     l_linger;               // linger time
1259  * };
1260  *
1261  * To enable the option, set l_onoff to 1.  If the l_linger value is set
1262  * to 0, calling close() is the same as the ABORT primitive.  If the
1263  * value is set to a negative value, the setsockopt() call will return
1264  * an error.  If the value is set to a positive value linger_time, the
1265  * close() can be blocked for at most linger_time ms.  If the graceful
1266  * shutdown phase does not finish during this period, close() will
1267  * return but the graceful shutdown phase continues in the system.
1268  */
1269 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1270 {
1271 	struct sctp_endpoint *ep;
1272 	struct sctp_association *asoc;
1273 	struct list_head *pos, *temp;
1274 
1275 	SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1276 
1277 	sctp_lock_sock(sk);
1278 	sk->sk_shutdown = SHUTDOWN_MASK;
1279 
1280 	ep = sctp_sk(sk)->ep;
1281 
1282 	/* Walk all associations on an endpoint.  */
1283 	list_for_each_safe(pos, temp, &ep->asocs) {
1284 		asoc = list_entry(pos, struct sctp_association, asocs);
1285 
1286 		if (sctp_style(sk, TCP)) {
1287 			/* A closed association can still be in the list if
1288 			 * it belongs to a TCP-style listening socket that is
1289 			 * not yet accepted. If so, free it. If not, send an
1290 			 * ABORT or SHUTDOWN based on the linger options.
1291 			 */
1292 			if (sctp_state(asoc, CLOSED)) {
1293 				sctp_unhash_established(asoc);
1294 				sctp_association_free(asoc);
1295 				continue;
1296 			}
1297 		}
1298 
1299 		if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1300 			struct sctp_chunk *chunk;
1301 
1302 			chunk = sctp_make_abort_user(asoc, NULL, 0);
1303 			if (chunk)
1304 				sctp_primitive_ABORT(asoc, chunk);
1305 		} else
1306 			sctp_primitive_SHUTDOWN(asoc, NULL);
1307 	}
1308 
1309 	/* Clean up any skbs sitting on the receive queue.  */
1310 	sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1311 	sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1312 
1313 	/* On a TCP-style socket, block for at most linger_time if set. */
1314 	if (sctp_style(sk, TCP) && timeout)
1315 		sctp_wait_for_close(sk, timeout);
1316 
1317 	/* This will run the backlog queue.  */
1318 	sctp_release_sock(sk);
1319 
1320 	/* Supposedly, no process has access to the socket, but
1321 	 * the net layers still may.
1322 	 */
1323 	sctp_local_bh_disable();
1324 	sctp_bh_lock_sock(sk);
1325 
1326 	/* Hold the sock, since sk_common_release() will put sock_put()
1327 	 * and we have just a little more cleanup.
1328 	 */
1329 	sock_hold(sk);
1330 	sk_common_release(sk);
1331 
1332 	sctp_bh_unlock_sock(sk);
1333 	sctp_local_bh_enable();
1334 
1335 	sock_put(sk);
1336 
1337 	SCTP_DBG_OBJCNT_DEC(sock);
1338 }
1339 
1340 /* Handle EPIPE error. */
1341 static int sctp_error(struct sock *sk, int flags, int err)
1342 {
1343 	if (err == -EPIPE)
1344 		err = sock_error(sk) ? : -EPIPE;
1345 	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1346 		send_sig(SIGPIPE, current, 0);
1347 	return err;
1348 }
1349 
1350 /* API 3.1.3 sendmsg() - UDP Style Syntax
1351  *
1352  * An application uses sendmsg() and recvmsg() calls to transmit data to
1353  * and receive data from its peer.
1354  *
1355  *  ssize_t sendmsg(int socket, const struct msghdr *message,
1356  *                  int flags);
1357  *
1358  *  socket  - the socket descriptor of the endpoint.
1359  *  message - pointer to the msghdr structure which contains a single
1360  *            user message and possibly some ancillary data.
1361  *
1362  *            See Section 5 for complete description of the data
1363  *            structures.
1364  *
1365  *  flags   - flags sent or received with the user message, see Section
1366  *            5 for complete description of the flags.
1367  *
1368  * Note:  This function could use a rewrite especially when explicit
1369  * connect support comes in.
1370  */
1371 /* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1372 
1373 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1374 
1375 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1376 			     struct msghdr *msg, size_t msg_len)
1377 {
1378 	struct sctp_sock *sp;
1379 	struct sctp_endpoint *ep;
1380 	struct sctp_association *new_asoc=NULL, *asoc=NULL;
1381 	struct sctp_transport *transport, *chunk_tp;
1382 	struct sctp_chunk *chunk;
1383 	union sctp_addr to;
1384 	struct sockaddr *msg_name = NULL;
1385 	struct sctp_sndrcvinfo default_sinfo = { 0 };
1386 	struct sctp_sndrcvinfo *sinfo;
1387 	struct sctp_initmsg *sinit;
1388 	sctp_assoc_t associd = 0;
1389 	sctp_cmsgs_t cmsgs = { NULL };
1390 	int err;
1391 	sctp_scope_t scope;
1392 	long timeo;
1393 	__u16 sinfo_flags = 0;
1394 	struct sctp_datamsg *datamsg;
1395 	int msg_flags = msg->msg_flags;
1396 
1397 	SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1398 			  sk, msg, msg_len);
1399 
1400 	err = 0;
1401 	sp = sctp_sk(sk);
1402 	ep = sp->ep;
1403 
1404 	SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1405 
1406 	/* We cannot send a message over a TCP-style listening socket. */
1407 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1408 		err = -EPIPE;
1409 		goto out_nounlock;
1410 	}
1411 
1412 	/* Parse out the SCTP CMSGs.  */
1413 	err = sctp_msghdr_parse(msg, &cmsgs);
1414 
1415 	if (err) {
1416 		SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1417 		goto out_nounlock;
1418 	}
1419 
1420 	/* Fetch the destination address for this packet.  This
1421 	 * address only selects the association--it is not necessarily
1422 	 * the address we will send to.
1423 	 * For a peeled-off socket, msg_name is ignored.
1424 	 */
1425 	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1426 		int msg_namelen = msg->msg_namelen;
1427 
1428 		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1429 				       msg_namelen);
1430 		if (err)
1431 			return err;
1432 
1433 		if (msg_namelen > sizeof(to))
1434 			msg_namelen = sizeof(to);
1435 		memcpy(&to, msg->msg_name, msg_namelen);
1436 		msg_name = msg->msg_name;
1437 	}
1438 
1439 	sinfo = cmsgs.info;
1440 	sinit = cmsgs.init;
1441 
1442 	/* Did the user specify SNDRCVINFO?  */
1443 	if (sinfo) {
1444 		sinfo_flags = sinfo->sinfo_flags;
1445 		associd = sinfo->sinfo_assoc_id;
1446 	}
1447 
1448 	SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1449 			  msg_len, sinfo_flags);
1450 
1451 	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1452 	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1453 		err = -EINVAL;
1454 		goto out_nounlock;
1455 	}
1456 
1457 	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1458 	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1459 	 * If SCTP_ABORT is set, the message length could be non zero with
1460 	 * the msg_iov set to the user abort reason.
1461 	 */
1462 	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1463 	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1464 		err = -EINVAL;
1465 		goto out_nounlock;
1466 	}
1467 
1468 	/* If SCTP_ADDR_OVER is set, there must be an address
1469 	 * specified in msg_name.
1470 	 */
1471 	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1472 		err = -EINVAL;
1473 		goto out_nounlock;
1474 	}
1475 
1476 	transport = NULL;
1477 
1478 	SCTP_DEBUG_PRINTK("About to look up association.\n");
1479 
1480 	sctp_lock_sock(sk);
1481 
1482 	/* If a msg_name has been specified, assume this is to be used.  */
1483 	if (msg_name) {
1484 		/* Look for a matching association on the endpoint. */
1485 		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1486 		if (!asoc) {
1487 			/* If we could not find a matching association on the
1488 			 * endpoint, make sure that it is not a TCP-style
1489 			 * socket that already has an association or there is
1490 			 * no peeled-off association on another socket.
1491 			 */
1492 			if ((sctp_style(sk, TCP) &&
1493 			     sctp_sstate(sk, ESTABLISHED)) ||
1494 			    sctp_endpoint_is_peeled_off(ep, &to)) {
1495 				err = -EADDRNOTAVAIL;
1496 				goto out_unlock;
1497 			}
1498 		}
1499 	} else {
1500 		asoc = sctp_id2assoc(sk, associd);
1501 		if (!asoc) {
1502 			err = -EPIPE;
1503 			goto out_unlock;
1504 		}
1505 	}
1506 
1507 	if (asoc) {
1508 		SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1509 
1510 		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1511 		 * socket that has an association in CLOSED state. This can
1512 		 * happen when an accepted socket has an association that is
1513 		 * already CLOSED.
1514 		 */
1515 		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1516 			err = -EPIPE;
1517 			goto out_unlock;
1518 		}
1519 
1520 		if (sinfo_flags & SCTP_EOF) {
1521 			SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1522 					  asoc);
1523 			sctp_primitive_SHUTDOWN(asoc, NULL);
1524 			err = 0;
1525 			goto out_unlock;
1526 		}
1527 		if (sinfo_flags & SCTP_ABORT) {
1528 
1529 			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1530 			if (!chunk) {
1531 				err = -ENOMEM;
1532 				goto out_unlock;
1533 			}
1534 
1535 			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1536 			sctp_primitive_ABORT(asoc, chunk);
1537 			err = 0;
1538 			goto out_unlock;
1539 		}
1540 	}
1541 
1542 	/* Do we need to create the association?  */
1543 	if (!asoc) {
1544 		SCTP_DEBUG_PRINTK("There is no association yet.\n");
1545 
1546 		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1547 			err = -EINVAL;
1548 			goto out_unlock;
1549 		}
1550 
1551 		/* Check for invalid stream against the stream counts,
1552 		 * either the default or the user specified stream counts.
1553 		 */
1554 		if (sinfo) {
1555 			if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1556 				/* Check against the defaults. */
1557 				if (sinfo->sinfo_stream >=
1558 				    sp->initmsg.sinit_num_ostreams) {
1559 					err = -EINVAL;
1560 					goto out_unlock;
1561 				}
1562 			} else {
1563 				/* Check against the requested.  */
1564 				if (sinfo->sinfo_stream >=
1565 				    sinit->sinit_num_ostreams) {
1566 					err = -EINVAL;
1567 					goto out_unlock;
1568 				}
1569 			}
1570 		}
1571 
1572 		/*
1573 		 * API 3.1.2 bind() - UDP Style Syntax
1574 		 * If a bind() or sctp_bindx() is not called prior to a
1575 		 * sendmsg() call that initiates a new association, the
1576 		 * system picks an ephemeral port and will choose an address
1577 		 * set equivalent to binding with a wildcard address.
1578 		 */
1579 		if (!ep->base.bind_addr.port) {
1580 			if (sctp_autobind(sk)) {
1581 				err = -EAGAIN;
1582 				goto out_unlock;
1583 			}
1584 		} else {
1585 			/*
1586 			 * If an unprivileged user inherits a one-to-many
1587 			 * style socket with open associations on a privileged
1588 			 * port, it MAY be permitted to accept new associations,
1589 			 * but it SHOULD NOT be permitted to open new
1590 			 * associations.
1591 			 */
1592 			if (ep->base.bind_addr.port < PROT_SOCK &&
1593 			    !capable(CAP_NET_BIND_SERVICE)) {
1594 				err = -EACCES;
1595 				goto out_unlock;
1596 			}
1597 		}
1598 
1599 		scope = sctp_scope(&to);
1600 		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1601 		if (!new_asoc) {
1602 			err = -ENOMEM;
1603 			goto out_unlock;
1604 		}
1605 		asoc = new_asoc;
1606 
1607 		/* If the SCTP_INIT ancillary data is specified, set all
1608 		 * the association init values accordingly.
1609 		 */
1610 		if (sinit) {
1611 			if (sinit->sinit_num_ostreams) {
1612 				asoc->c.sinit_num_ostreams =
1613 					sinit->sinit_num_ostreams;
1614 			}
1615 			if (sinit->sinit_max_instreams) {
1616 				asoc->c.sinit_max_instreams =
1617 					sinit->sinit_max_instreams;
1618 			}
1619 			if (sinit->sinit_max_attempts) {
1620 				asoc->max_init_attempts
1621 					= sinit->sinit_max_attempts;
1622 			}
1623 			if (sinit->sinit_max_init_timeo) {
1624 				asoc->max_init_timeo =
1625 				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1626 			}
1627 		}
1628 
1629 		/* Prime the peer's transport structures.  */
1630 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1631 		if (!transport) {
1632 			err = -ENOMEM;
1633 			goto out_free;
1634 		}
1635 		err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1636 		if (err < 0) {
1637 			err = -ENOMEM;
1638 			goto out_free;
1639 		}
1640 	}
1641 
1642 	/* ASSERT: we have a valid association at this point.  */
1643 	SCTP_DEBUG_PRINTK("We have a valid association.\n");
1644 
1645 	if (!sinfo) {
1646 		/* If the user didn't specify SNDRCVINFO, make up one with
1647 		 * some defaults.
1648 		 */
1649 		default_sinfo.sinfo_stream = asoc->default_stream;
1650 		default_sinfo.sinfo_flags = asoc->default_flags;
1651 		default_sinfo.sinfo_ppid = asoc->default_ppid;
1652 		default_sinfo.sinfo_context = asoc->default_context;
1653 		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1654 		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1655 		sinfo = &default_sinfo;
1656 	}
1657 
1658 	/* API 7.1.7, the sndbuf size per association bounds the
1659 	 * maximum size of data that can be sent in a single send call.
1660 	 */
1661 	if (msg_len > sk->sk_sndbuf) {
1662 		err = -EMSGSIZE;
1663 		goto out_free;
1664 	}
1665 
1666 	if (asoc->pmtu_pending)
1667 		sctp_assoc_pending_pmtu(asoc);
1668 
1669 	/* If fragmentation is disabled and the message length exceeds the
1670 	 * association fragmentation point, return EMSGSIZE.  The I-D
1671 	 * does not specify what this error is, but this looks like
1672 	 * a great fit.
1673 	 */
1674 	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1675 		err = -EMSGSIZE;
1676 		goto out_free;
1677 	}
1678 
1679 	if (sinfo) {
1680 		/* Check for invalid stream. */
1681 		if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1682 			err = -EINVAL;
1683 			goto out_free;
1684 		}
1685 	}
1686 
1687 	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1688 	if (!sctp_wspace(asoc)) {
1689 		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1690 		if (err)
1691 			goto out_free;
1692 	}
1693 
1694 	/* If an address is passed with the sendto/sendmsg call, it is used
1695 	 * to override the primary destination address in the TCP model, or
1696 	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1697 	 */
1698 	if ((sctp_style(sk, TCP) && msg_name) ||
1699 	    (sinfo_flags & SCTP_ADDR_OVER)) {
1700 		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1701 		if (!chunk_tp) {
1702 			err = -EINVAL;
1703 			goto out_free;
1704 		}
1705 	} else
1706 		chunk_tp = NULL;
1707 
1708 	/* Auto-connect, if we aren't connected already. */
1709 	if (sctp_state(asoc, CLOSED)) {
1710 		err = sctp_primitive_ASSOCIATE(asoc, NULL);
1711 		if (err < 0)
1712 			goto out_free;
1713 		SCTP_DEBUG_PRINTK("We associated primitively.\n");
1714 	}
1715 
1716 	/* Break the message into multiple chunks of maximum size. */
1717 	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1718 	if (!datamsg) {
1719 		err = -ENOMEM;
1720 		goto out_free;
1721 	}
1722 
1723 	/* Now send the (possibly) fragmented message. */
1724 	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1725 		sctp_chunk_hold(chunk);
1726 
1727 		/* Do accounting for the write space.  */
1728 		sctp_set_owner_w(chunk);
1729 
1730 		chunk->transport = chunk_tp;
1731 
1732 		/* Send it to the lower layers.  Note:  all chunks
1733 		 * must either fail or succeed.   The lower layer
1734 		 * works that way today.  Keep it that way or this
1735 		 * breaks.
1736 		 */
1737 		err = sctp_primitive_SEND(asoc, chunk);
1738 		/* Did the lower layer accept the chunk? */
1739 		if (err)
1740 			sctp_chunk_free(chunk);
1741 		SCTP_DEBUG_PRINTK("We sent primitively.\n");
1742 	}
1743 
1744 	sctp_datamsg_put(datamsg);
1745 	if (err)
1746 		goto out_free;
1747 	else
1748 		err = msg_len;
1749 
1750 	/* If we are already past ASSOCIATE, the lower
1751 	 * layers are responsible for association cleanup.
1752 	 */
1753 	goto out_unlock;
1754 
1755 out_free:
1756 	if (new_asoc)
1757 		sctp_association_free(asoc);
1758 out_unlock:
1759 	sctp_release_sock(sk);
1760 
1761 out_nounlock:
1762 	return sctp_error(sk, msg_flags, err);
1763 
1764 #if 0
1765 do_sock_err:
1766 	if (msg_len)
1767 		err = msg_len;
1768 	else
1769 		err = sock_error(sk);
1770 	goto out;
1771 
1772 do_interrupted:
1773 	if (msg_len)
1774 		err = msg_len;
1775 	goto out;
1776 #endif /* 0 */
1777 }
1778 
1779 /* This is an extended version of skb_pull() that removes the data from the
1780  * start of a skb even when data is spread across the list of skb's in the
1781  * frag_list. len specifies the total amount of data that needs to be removed.
1782  * when 'len' bytes could be removed from the skb, it returns 0.
1783  * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1784  * could not be removed.
1785  */
1786 static int sctp_skb_pull(struct sk_buff *skb, int len)
1787 {
1788 	struct sk_buff *list;
1789 	int skb_len = skb_headlen(skb);
1790 	int rlen;
1791 
1792 	if (len <= skb_len) {
1793 		__skb_pull(skb, len);
1794 		return 0;
1795 	}
1796 	len -= skb_len;
1797 	__skb_pull(skb, skb_len);
1798 
1799 	for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1800 		rlen = sctp_skb_pull(list, len);
1801 		skb->len -= (len-rlen);
1802 		skb->data_len -= (len-rlen);
1803 
1804 		if (!rlen)
1805 			return 0;
1806 
1807 		len = rlen;
1808 	}
1809 
1810 	return len;
1811 }
1812 
1813 /* API 3.1.3  recvmsg() - UDP Style Syntax
1814  *
1815  *  ssize_t recvmsg(int socket, struct msghdr *message,
1816  *                    int flags);
1817  *
1818  *  socket  - the socket descriptor of the endpoint.
1819  *  message - pointer to the msghdr structure which contains a single
1820  *            user message and possibly some ancillary data.
1821  *
1822  *            See Section 5 for complete description of the data
1823  *            structures.
1824  *
1825  *  flags   - flags sent or received with the user message, see Section
1826  *            5 for complete description of the flags.
1827  */
1828 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1829 
1830 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1831 			     struct msghdr *msg, size_t len, int noblock,
1832 			     int flags, int *addr_len)
1833 {
1834 	struct sctp_ulpevent *event = NULL;
1835 	struct sctp_sock *sp = sctp_sk(sk);
1836 	struct sk_buff *skb;
1837 	int copied;
1838 	int err = 0;
1839 	int skb_len;
1840 
1841 	SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1842 			  "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1843 			  "len", len, "knoblauch", noblock,
1844 			  "flags", flags, "addr_len", addr_len);
1845 
1846 	sctp_lock_sock(sk);
1847 
1848 	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1849 		err = -ENOTCONN;
1850 		goto out;
1851 	}
1852 
1853 	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1854 	if (!skb)
1855 		goto out;
1856 
1857 	/* Get the total length of the skb including any skb's in the
1858 	 * frag_list.
1859 	 */
1860 	skb_len = skb->len;
1861 
1862 	copied = skb_len;
1863 	if (copied > len)
1864 		copied = len;
1865 
1866 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1867 
1868 	event = sctp_skb2event(skb);
1869 
1870 	if (err)
1871 		goto out_free;
1872 
1873 	sock_recv_timestamp(msg, sk, skb);
1874 	if (sctp_ulpevent_is_notification(event)) {
1875 		msg->msg_flags |= MSG_NOTIFICATION;
1876 		sp->pf->event_msgname(event, msg->msg_name, addr_len);
1877 	} else {
1878 		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1879 	}
1880 
1881 	/* Check if we allow SCTP_SNDRCVINFO. */
1882 	if (sp->subscribe.sctp_data_io_event)
1883 		sctp_ulpevent_read_sndrcvinfo(event, msg);
1884 #if 0
1885 	/* FIXME: we should be calling IP/IPv6 layers.  */
1886 	if (sk->sk_protinfo.af_inet.cmsg_flags)
1887 		ip_cmsg_recv(msg, skb);
1888 #endif
1889 
1890 	err = copied;
1891 
1892 	/* If skb's length exceeds the user's buffer, update the skb and
1893 	 * push it back to the receive_queue so that the next call to
1894 	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1895 	 */
1896 	if (skb_len > copied) {
1897 		msg->msg_flags &= ~MSG_EOR;
1898 		if (flags & MSG_PEEK)
1899 			goto out_free;
1900 		sctp_skb_pull(skb, copied);
1901 		skb_queue_head(&sk->sk_receive_queue, skb);
1902 
1903 		/* When only partial message is copied to the user, increase
1904 		 * rwnd by that amount. If all the data in the skb is read,
1905 		 * rwnd is updated when the event is freed.
1906 		 */
1907 		if (!sctp_ulpevent_is_notification(event))
1908 			sctp_assoc_rwnd_increase(event->asoc, copied);
1909 		goto out;
1910 	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
1911 		   (event->msg_flags & MSG_EOR))
1912 		msg->msg_flags |= MSG_EOR;
1913 	else
1914 		msg->msg_flags &= ~MSG_EOR;
1915 
1916 out_free:
1917 	if (flags & MSG_PEEK) {
1918 		/* Release the skb reference acquired after peeking the skb in
1919 		 * sctp_skb_recv_datagram().
1920 		 */
1921 		kfree_skb(skb);
1922 	} else {
1923 		/* Free the event which includes releasing the reference to
1924 		 * the owner of the skb, freeing the skb and updating the
1925 		 * rwnd.
1926 		 */
1927 		sctp_ulpevent_free(event);
1928 	}
1929 out:
1930 	sctp_release_sock(sk);
1931 	return err;
1932 }
1933 
1934 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1935  *
1936  * This option is a on/off flag.  If enabled no SCTP message
1937  * fragmentation will be performed.  Instead if a message being sent
1938  * exceeds the current PMTU size, the message will NOT be sent and
1939  * instead a error will be indicated to the user.
1940  */
1941 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1942 					    char __user *optval, int optlen)
1943 {
1944 	int val;
1945 
1946 	if (optlen < sizeof(int))
1947 		return -EINVAL;
1948 
1949 	if (get_user(val, (int __user *)optval))
1950 		return -EFAULT;
1951 
1952 	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1953 
1954 	return 0;
1955 }
1956 
1957 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1958 					int optlen)
1959 {
1960 	if (optlen > sizeof(struct sctp_event_subscribe))
1961 		return -EINVAL;
1962 	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1963 		return -EFAULT;
1964 	return 0;
1965 }
1966 
1967 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1968  *
1969  * This socket option is applicable to the UDP-style socket only.  When
1970  * set it will cause associations that are idle for more than the
1971  * specified number of seconds to automatically close.  An association
1972  * being idle is defined an association that has NOT sent or received
1973  * user data.  The special value of '0' indicates that no automatic
1974  * close of any associations should be performed.  The option expects an
1975  * integer defining the number of seconds of idle time before an
1976  * association is closed.
1977  */
1978 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1979 					    int optlen)
1980 {
1981 	struct sctp_sock *sp = sctp_sk(sk);
1982 
1983 	/* Applicable to UDP-style socket only */
1984 	if (sctp_style(sk, TCP))
1985 		return -EOPNOTSUPP;
1986 	if (optlen != sizeof(int))
1987 		return -EINVAL;
1988 	if (copy_from_user(&sp->autoclose, optval, optlen))
1989 		return -EFAULT;
1990 
1991 	return 0;
1992 }
1993 
1994 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1995  *
1996  * Applications can enable or disable heartbeats for any peer address of
1997  * an association, modify an address's heartbeat interval, force a
1998  * heartbeat to be sent immediately, and adjust the address's maximum
1999  * number of retransmissions sent before an address is considered
2000  * unreachable.  The following structure is used to access and modify an
2001  * address's parameters:
2002  *
2003  *  struct sctp_paddrparams {
2004  *     sctp_assoc_t            spp_assoc_id;
2005  *     struct sockaddr_storage spp_address;
2006  *     uint32_t                spp_hbinterval;
2007  *     uint16_t                spp_pathmaxrxt;
2008  *     uint32_t                spp_pathmtu;
2009  *     uint32_t                spp_sackdelay;
2010  *     uint32_t                spp_flags;
2011  * };
2012  *
2013  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2014  *                     application, and identifies the association for
2015  *                     this query.
2016  *   spp_address     - This specifies which address is of interest.
2017  *   spp_hbinterval  - This contains the value of the heartbeat interval,
2018  *                     in milliseconds.  If a  value of zero
2019  *                     is present in this field then no changes are to
2020  *                     be made to this parameter.
2021  *   spp_pathmaxrxt  - This contains the maximum number of
2022  *                     retransmissions before this address shall be
2023  *                     considered unreachable. If a  value of zero
2024  *                     is present in this field then no changes are to
2025  *                     be made to this parameter.
2026  *   spp_pathmtu     - When Path MTU discovery is disabled the value
2027  *                     specified here will be the "fixed" path mtu.
2028  *                     Note that if the spp_address field is empty
2029  *                     then all associations on this address will
2030  *                     have this fixed path mtu set upon them.
2031  *
2032  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2033  *                     the number of milliseconds that sacks will be delayed
2034  *                     for. This value will apply to all addresses of an
2035  *                     association if the spp_address field is empty. Note
2036  *                     also, that if delayed sack is enabled and this
2037  *                     value is set to 0, no change is made to the last
2038  *                     recorded delayed sack timer value.
2039  *
2040  *   spp_flags       - These flags are used to control various features
2041  *                     on an association. The flag field may contain
2042  *                     zero or more of the following options.
2043  *
2044  *                     SPP_HB_ENABLE  - Enable heartbeats on the
2045  *                     specified address. Note that if the address
2046  *                     field is empty all addresses for the association
2047  *                     have heartbeats enabled upon them.
2048  *
2049  *                     SPP_HB_DISABLE - Disable heartbeats on the
2050  *                     speicifed address. Note that if the address
2051  *                     field is empty all addresses for the association
2052  *                     will have their heartbeats disabled. Note also
2053  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2054  *                     mutually exclusive, only one of these two should
2055  *                     be specified. Enabling both fields will have
2056  *                     undetermined results.
2057  *
2058  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2059  *                     to be made immediately.
2060  *
2061  *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2062  *                     heartbeat delayis to be set to the value of 0
2063  *                     milliseconds.
2064  *
2065  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2066  *                     discovery upon the specified address. Note that
2067  *                     if the address feild is empty then all addresses
2068  *                     on the association are effected.
2069  *
2070  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2071  *                     discovery upon the specified address. Note that
2072  *                     if the address feild is empty then all addresses
2073  *                     on the association are effected. Not also that
2074  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2075  *                     exclusive. Enabling both will have undetermined
2076  *                     results.
2077  *
2078  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2079  *                     on delayed sack. The time specified in spp_sackdelay
2080  *                     is used to specify the sack delay for this address. Note
2081  *                     that if spp_address is empty then all addresses will
2082  *                     enable delayed sack and take on the sack delay
2083  *                     value specified in spp_sackdelay.
2084  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2085  *                     off delayed sack. If the spp_address field is blank then
2086  *                     delayed sack is disabled for the entire association. Note
2087  *                     also that this field is mutually exclusive to
2088  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2089  *                     results.
2090  */
2091 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2092 				       struct sctp_transport   *trans,
2093 				       struct sctp_association *asoc,
2094 				       struct sctp_sock        *sp,
2095 				       int                      hb_change,
2096 				       int                      pmtud_change,
2097 				       int                      sackdelay_change)
2098 {
2099 	int error;
2100 
2101 	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2102 		error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2103 		if (error)
2104 			return error;
2105 	}
2106 
2107 	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2108 	 * this field is ignored.  Note also that a value of zero indicates
2109 	 * the current setting should be left unchanged.
2110 	 */
2111 	if (params->spp_flags & SPP_HB_ENABLE) {
2112 
2113 		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2114 		 * set.  This lets us use 0 value when this flag
2115 		 * is set.
2116 		 */
2117 		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2118 			params->spp_hbinterval = 0;
2119 
2120 		if (params->spp_hbinterval ||
2121 		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2122 			if (trans) {
2123 				trans->hbinterval =
2124 				    msecs_to_jiffies(params->spp_hbinterval);
2125 			} else if (asoc) {
2126 				asoc->hbinterval =
2127 				    msecs_to_jiffies(params->spp_hbinterval);
2128 			} else {
2129 				sp->hbinterval = params->spp_hbinterval;
2130 			}
2131 		}
2132 	}
2133 
2134 	if (hb_change) {
2135 		if (trans) {
2136 			trans->param_flags =
2137 				(trans->param_flags & ~SPP_HB) | hb_change;
2138 		} else if (asoc) {
2139 			asoc->param_flags =
2140 				(asoc->param_flags & ~SPP_HB) | hb_change;
2141 		} else {
2142 			sp->param_flags =
2143 				(sp->param_flags & ~SPP_HB) | hb_change;
2144 		}
2145 	}
2146 
2147 	/* When Path MTU discovery is disabled the value specified here will
2148 	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2149 	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2150 	 * effect).
2151 	 */
2152 	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2153 		if (trans) {
2154 			trans->pathmtu = params->spp_pathmtu;
2155 			sctp_assoc_sync_pmtu(asoc);
2156 		} else if (asoc) {
2157 			asoc->pathmtu = params->spp_pathmtu;
2158 			sctp_frag_point(sp, params->spp_pathmtu);
2159 		} else {
2160 			sp->pathmtu = params->spp_pathmtu;
2161 		}
2162 	}
2163 
2164 	if (pmtud_change) {
2165 		if (trans) {
2166 			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2167 				(params->spp_flags & SPP_PMTUD_ENABLE);
2168 			trans->param_flags =
2169 				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2170 			if (update) {
2171 				sctp_transport_pmtu(trans);
2172 				sctp_assoc_sync_pmtu(asoc);
2173 			}
2174 		} else if (asoc) {
2175 			asoc->param_flags =
2176 				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2177 		} else {
2178 			sp->param_flags =
2179 				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2180 		}
2181 	}
2182 
2183 	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2184 	 * value of this field is ignored.  Note also that a value of zero
2185 	 * indicates the current setting should be left unchanged.
2186 	 */
2187 	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2188 		if (trans) {
2189 			trans->sackdelay =
2190 				msecs_to_jiffies(params->spp_sackdelay);
2191 		} else if (asoc) {
2192 			asoc->sackdelay =
2193 				msecs_to_jiffies(params->spp_sackdelay);
2194 		} else {
2195 			sp->sackdelay = params->spp_sackdelay;
2196 		}
2197 	}
2198 
2199 	if (sackdelay_change) {
2200 		if (trans) {
2201 			trans->param_flags =
2202 				(trans->param_flags & ~SPP_SACKDELAY) |
2203 				sackdelay_change;
2204 		} else if (asoc) {
2205 			asoc->param_flags =
2206 				(asoc->param_flags & ~SPP_SACKDELAY) |
2207 				sackdelay_change;
2208 		} else {
2209 			sp->param_flags =
2210 				(sp->param_flags & ~SPP_SACKDELAY) |
2211 				sackdelay_change;
2212 		}
2213 	}
2214 
2215 	/* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2216 	 * of this field is ignored.  Note also that a value of zero
2217 	 * indicates the current setting should be left unchanged.
2218 	 */
2219 	if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2220 		if (trans) {
2221 			trans->pathmaxrxt = params->spp_pathmaxrxt;
2222 		} else if (asoc) {
2223 			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2224 		} else {
2225 			sp->pathmaxrxt = params->spp_pathmaxrxt;
2226 		}
2227 	}
2228 
2229 	return 0;
2230 }
2231 
2232 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2233 					    char __user *optval, int optlen)
2234 {
2235 	struct sctp_paddrparams  params;
2236 	struct sctp_transport   *trans = NULL;
2237 	struct sctp_association *asoc = NULL;
2238 	struct sctp_sock        *sp = sctp_sk(sk);
2239 	int error;
2240 	int hb_change, pmtud_change, sackdelay_change;
2241 
2242 	if (optlen != sizeof(struct sctp_paddrparams))
2243 		return - EINVAL;
2244 
2245 	if (copy_from_user(&params, optval, optlen))
2246 		return -EFAULT;
2247 
2248 	/* Validate flags and value parameters. */
2249 	hb_change        = params.spp_flags & SPP_HB;
2250 	pmtud_change     = params.spp_flags & SPP_PMTUD;
2251 	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2252 
2253 	if (hb_change        == SPP_HB ||
2254 	    pmtud_change     == SPP_PMTUD ||
2255 	    sackdelay_change == SPP_SACKDELAY ||
2256 	    params.spp_sackdelay > 500 ||
2257 	    (params.spp_pathmtu
2258 	    && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2259 		return -EINVAL;
2260 
2261 	/* If an address other than INADDR_ANY is specified, and
2262 	 * no transport is found, then the request is invalid.
2263 	 */
2264 	if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2265 		trans = sctp_addr_id2transport(sk, &params.spp_address,
2266 					       params.spp_assoc_id);
2267 		if (!trans)
2268 			return -EINVAL;
2269 	}
2270 
2271 	/* Get association, if assoc_id != 0 and the socket is a one
2272 	 * to many style socket, and an association was not found, then
2273 	 * the id was invalid.
2274 	 */
2275 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2276 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2277 		return -EINVAL;
2278 
2279 	/* Heartbeat demand can only be sent on a transport or
2280 	 * association, but not a socket.
2281 	 */
2282 	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2283 		return -EINVAL;
2284 
2285 	/* Process parameters. */
2286 	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2287 					    hb_change, pmtud_change,
2288 					    sackdelay_change);
2289 
2290 	if (error)
2291 		return error;
2292 
2293 	/* If changes are for association, also apply parameters to each
2294 	 * transport.
2295 	 */
2296 	if (!trans && asoc) {
2297 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2298 				transports) {
2299 			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2300 						    hb_change, pmtud_change,
2301 						    sackdelay_change);
2302 		}
2303 	}
2304 
2305 	return 0;
2306 }
2307 
2308 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2309  *
2310  *   This options will get or set the delayed ack timer.  The time is set
2311  *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
2312  *   endpoints default delayed ack timer value.  If the assoc_id field is
2313  *   non-zero, then the set or get effects the specified association.
2314  *
2315  *   struct sctp_assoc_value {
2316  *       sctp_assoc_t            assoc_id;
2317  *       uint32_t                assoc_value;
2318  *   };
2319  *
2320  *     assoc_id    - This parameter, indicates which association the
2321  *                   user is preforming an action upon. Note that if
2322  *                   this field's value is zero then the endpoints
2323  *                   default value is changed (effecting future
2324  *                   associations only).
2325  *
2326  *     assoc_value - This parameter contains the number of milliseconds
2327  *                   that the user is requesting the delayed ACK timer
2328  *                   be set to. Note that this value is defined in
2329  *                   the standard to be between 200 and 500 milliseconds.
2330  *
2331  *                   Note: a value of zero will leave the value alone,
2332  *                   but disable SACK delay. A non-zero value will also
2333  *                   enable SACK delay.
2334  */
2335 
2336 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2337 					    char __user *optval, int optlen)
2338 {
2339 	struct sctp_assoc_value  params;
2340 	struct sctp_transport   *trans = NULL;
2341 	struct sctp_association *asoc = NULL;
2342 	struct sctp_sock        *sp = sctp_sk(sk);
2343 
2344 	if (optlen != sizeof(struct sctp_assoc_value))
2345 		return - EINVAL;
2346 
2347 	if (copy_from_user(&params, optval, optlen))
2348 		return -EFAULT;
2349 
2350 	/* Validate value parameter. */
2351 	if (params.assoc_value > 500)
2352 		return -EINVAL;
2353 
2354 	/* Get association, if assoc_id != 0 and the socket is a one
2355 	 * to many style socket, and an association was not found, then
2356 	 * the id was invalid.
2357 	 */
2358 	asoc = sctp_id2assoc(sk, params.assoc_id);
2359 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2360 		return -EINVAL;
2361 
2362 	if (params.assoc_value) {
2363 		if (asoc) {
2364 			asoc->sackdelay =
2365 				msecs_to_jiffies(params.assoc_value);
2366 			asoc->param_flags =
2367 				(asoc->param_flags & ~SPP_SACKDELAY) |
2368 				SPP_SACKDELAY_ENABLE;
2369 		} else {
2370 			sp->sackdelay = params.assoc_value;
2371 			sp->param_flags =
2372 				(sp->param_flags & ~SPP_SACKDELAY) |
2373 				SPP_SACKDELAY_ENABLE;
2374 		}
2375 	} else {
2376 		if (asoc) {
2377 			asoc->param_flags =
2378 				(asoc->param_flags & ~SPP_SACKDELAY) |
2379 				SPP_SACKDELAY_DISABLE;
2380 		} else {
2381 			sp->param_flags =
2382 				(sp->param_flags & ~SPP_SACKDELAY) |
2383 				SPP_SACKDELAY_DISABLE;
2384 		}
2385 	}
2386 
2387 	/* If change is for association, also apply to each transport. */
2388 	if (asoc) {
2389 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2390 				transports) {
2391 			if (params.assoc_value) {
2392 				trans->sackdelay =
2393 					msecs_to_jiffies(params.assoc_value);
2394 				trans->param_flags =
2395 					(trans->param_flags & ~SPP_SACKDELAY) |
2396 					SPP_SACKDELAY_ENABLE;
2397 			} else {
2398 				trans->param_flags =
2399 					(trans->param_flags & ~SPP_SACKDELAY) |
2400 					SPP_SACKDELAY_DISABLE;
2401 			}
2402 		}
2403 	}
2404 
2405 	return 0;
2406 }
2407 
2408 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2409  *
2410  * Applications can specify protocol parameters for the default association
2411  * initialization.  The option name argument to setsockopt() and getsockopt()
2412  * is SCTP_INITMSG.
2413  *
2414  * Setting initialization parameters is effective only on an unconnected
2415  * socket (for UDP-style sockets only future associations are effected
2416  * by the change).  With TCP-style sockets, this option is inherited by
2417  * sockets derived from a listener socket.
2418  */
2419 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2420 {
2421 	struct sctp_initmsg sinit;
2422 	struct sctp_sock *sp = sctp_sk(sk);
2423 
2424 	if (optlen != sizeof(struct sctp_initmsg))
2425 		return -EINVAL;
2426 	if (copy_from_user(&sinit, optval, optlen))
2427 		return -EFAULT;
2428 
2429 	if (sinit.sinit_num_ostreams)
2430 		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2431 	if (sinit.sinit_max_instreams)
2432 		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2433 	if (sinit.sinit_max_attempts)
2434 		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2435 	if (sinit.sinit_max_init_timeo)
2436 		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2437 
2438 	return 0;
2439 }
2440 
2441 /*
2442  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2443  *
2444  *   Applications that wish to use the sendto() system call may wish to
2445  *   specify a default set of parameters that would normally be supplied
2446  *   through the inclusion of ancillary data.  This socket option allows
2447  *   such an application to set the default sctp_sndrcvinfo structure.
2448  *   The application that wishes to use this socket option simply passes
2449  *   in to this call the sctp_sndrcvinfo structure defined in Section
2450  *   5.2.2) The input parameters accepted by this call include
2451  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2452  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2453  *   to this call if the caller is using the UDP model.
2454  */
2455 static int sctp_setsockopt_default_send_param(struct sock *sk,
2456 						char __user *optval, int optlen)
2457 {
2458 	struct sctp_sndrcvinfo info;
2459 	struct sctp_association *asoc;
2460 	struct sctp_sock *sp = sctp_sk(sk);
2461 
2462 	if (optlen != sizeof(struct sctp_sndrcvinfo))
2463 		return -EINVAL;
2464 	if (copy_from_user(&info, optval, optlen))
2465 		return -EFAULT;
2466 
2467 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2468 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2469 		return -EINVAL;
2470 
2471 	if (asoc) {
2472 		asoc->default_stream = info.sinfo_stream;
2473 		asoc->default_flags = info.sinfo_flags;
2474 		asoc->default_ppid = info.sinfo_ppid;
2475 		asoc->default_context = info.sinfo_context;
2476 		asoc->default_timetolive = info.sinfo_timetolive;
2477 	} else {
2478 		sp->default_stream = info.sinfo_stream;
2479 		sp->default_flags = info.sinfo_flags;
2480 		sp->default_ppid = info.sinfo_ppid;
2481 		sp->default_context = info.sinfo_context;
2482 		sp->default_timetolive = info.sinfo_timetolive;
2483 	}
2484 
2485 	return 0;
2486 }
2487 
2488 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2489  *
2490  * Requests that the local SCTP stack use the enclosed peer address as
2491  * the association primary.  The enclosed address must be one of the
2492  * association peer's addresses.
2493  */
2494 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2495 					int optlen)
2496 {
2497 	struct sctp_prim prim;
2498 	struct sctp_transport *trans;
2499 
2500 	if (optlen != sizeof(struct sctp_prim))
2501 		return -EINVAL;
2502 
2503 	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2504 		return -EFAULT;
2505 
2506 	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2507 	if (!trans)
2508 		return -EINVAL;
2509 
2510 	sctp_assoc_set_primary(trans->asoc, trans);
2511 
2512 	return 0;
2513 }
2514 
2515 /*
2516  * 7.1.5 SCTP_NODELAY
2517  *
2518  * Turn on/off any Nagle-like algorithm.  This means that packets are
2519  * generally sent as soon as possible and no unnecessary delays are
2520  * introduced, at the cost of more packets in the network.  Expects an
2521  *  integer boolean flag.
2522  */
2523 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2524 					int optlen)
2525 {
2526 	int val;
2527 
2528 	if (optlen < sizeof(int))
2529 		return -EINVAL;
2530 	if (get_user(val, (int __user *)optval))
2531 		return -EFAULT;
2532 
2533 	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2534 	return 0;
2535 }
2536 
2537 /*
2538  *
2539  * 7.1.1 SCTP_RTOINFO
2540  *
2541  * The protocol parameters used to initialize and bound retransmission
2542  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2543  * and modify these parameters.
2544  * All parameters are time values, in milliseconds.  A value of 0, when
2545  * modifying the parameters, indicates that the current value should not
2546  * be changed.
2547  *
2548  */
2549 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2550 	struct sctp_rtoinfo rtoinfo;
2551 	struct sctp_association *asoc;
2552 
2553 	if (optlen != sizeof (struct sctp_rtoinfo))
2554 		return -EINVAL;
2555 
2556 	if (copy_from_user(&rtoinfo, optval, optlen))
2557 		return -EFAULT;
2558 
2559 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2560 
2561 	/* Set the values to the specific association */
2562 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2563 		return -EINVAL;
2564 
2565 	if (asoc) {
2566 		if (rtoinfo.srto_initial != 0)
2567 			asoc->rto_initial =
2568 				msecs_to_jiffies(rtoinfo.srto_initial);
2569 		if (rtoinfo.srto_max != 0)
2570 			asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2571 		if (rtoinfo.srto_min != 0)
2572 			asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2573 	} else {
2574 		/* If there is no association or the association-id = 0
2575 		 * set the values to the endpoint.
2576 		 */
2577 		struct sctp_sock *sp = sctp_sk(sk);
2578 
2579 		if (rtoinfo.srto_initial != 0)
2580 			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2581 		if (rtoinfo.srto_max != 0)
2582 			sp->rtoinfo.srto_max = rtoinfo.srto_max;
2583 		if (rtoinfo.srto_min != 0)
2584 			sp->rtoinfo.srto_min = rtoinfo.srto_min;
2585 	}
2586 
2587 	return 0;
2588 }
2589 
2590 /*
2591  *
2592  * 7.1.2 SCTP_ASSOCINFO
2593  *
2594  * This option is used to tune the maximum retransmission attempts
2595  * of the association.
2596  * Returns an error if the new association retransmission value is
2597  * greater than the sum of the retransmission value  of the peer.
2598  * See [SCTP] for more information.
2599  *
2600  */
2601 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2602 {
2603 
2604 	struct sctp_assocparams assocparams;
2605 	struct sctp_association *asoc;
2606 
2607 	if (optlen != sizeof(struct sctp_assocparams))
2608 		return -EINVAL;
2609 	if (copy_from_user(&assocparams, optval, optlen))
2610 		return -EFAULT;
2611 
2612 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2613 
2614 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2615 		return -EINVAL;
2616 
2617 	/* Set the values to the specific association */
2618 	if (asoc) {
2619 		if (assocparams.sasoc_asocmaxrxt != 0) {
2620 			__u32 path_sum = 0;
2621 			int   paths = 0;
2622 			struct sctp_transport *peer_addr;
2623 
2624 			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2625 					transports) {
2626 				path_sum += peer_addr->pathmaxrxt;
2627 				paths++;
2628 			}
2629 
2630 			/* Only validate asocmaxrxt if we have more then
2631 			 * one path/transport.  We do this because path
2632 			 * retransmissions are only counted when we have more
2633 			 * then one path.
2634 			 */
2635 			if (paths > 1 &&
2636 			    assocparams.sasoc_asocmaxrxt > path_sum)
2637 				return -EINVAL;
2638 
2639 			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2640 		}
2641 
2642 		if (assocparams.sasoc_cookie_life != 0) {
2643 			asoc->cookie_life.tv_sec =
2644 					assocparams.sasoc_cookie_life / 1000;
2645 			asoc->cookie_life.tv_usec =
2646 					(assocparams.sasoc_cookie_life % 1000)
2647 					* 1000;
2648 		}
2649 	} else {
2650 		/* Set the values to the endpoint */
2651 		struct sctp_sock *sp = sctp_sk(sk);
2652 
2653 		if (assocparams.sasoc_asocmaxrxt != 0)
2654 			sp->assocparams.sasoc_asocmaxrxt =
2655 						assocparams.sasoc_asocmaxrxt;
2656 		if (assocparams.sasoc_cookie_life != 0)
2657 			sp->assocparams.sasoc_cookie_life =
2658 						assocparams.sasoc_cookie_life;
2659 	}
2660 	return 0;
2661 }
2662 
2663 /*
2664  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2665  *
2666  * This socket option is a boolean flag which turns on or off mapped V4
2667  * addresses.  If this option is turned on and the socket is type
2668  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2669  * If this option is turned off, then no mapping will be done of V4
2670  * addresses and a user will receive both PF_INET6 and PF_INET type
2671  * addresses on the socket.
2672  */
2673 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2674 {
2675 	int val;
2676 	struct sctp_sock *sp = sctp_sk(sk);
2677 
2678 	if (optlen < sizeof(int))
2679 		return -EINVAL;
2680 	if (get_user(val, (int __user *)optval))
2681 		return -EFAULT;
2682 	if (val)
2683 		sp->v4mapped = 1;
2684 	else
2685 		sp->v4mapped = 0;
2686 
2687 	return 0;
2688 }
2689 
2690 /*
2691  * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2692  *
2693  * This socket option specifies the maximum size to put in any outgoing
2694  * SCTP chunk.  If a message is larger than this size it will be
2695  * fragmented by SCTP into the specified size.  Note that the underlying
2696  * SCTP implementation may fragment into smaller sized chunks when the
2697  * PMTU of the underlying association is smaller than the value set by
2698  * the user.
2699  */
2700 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2701 {
2702 	struct sctp_association *asoc;
2703 	struct sctp_sock *sp = sctp_sk(sk);
2704 	int val;
2705 
2706 	if (optlen < sizeof(int))
2707 		return -EINVAL;
2708 	if (get_user(val, (int __user *)optval))
2709 		return -EFAULT;
2710 	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2711 		return -EINVAL;
2712 	sp->user_frag = val;
2713 
2714 	/* Update the frag_point of the existing associations. */
2715 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2716 		asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2717 	}
2718 
2719 	return 0;
2720 }
2721 
2722 
2723 /*
2724  *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2725  *
2726  *   Requests that the peer mark the enclosed address as the association
2727  *   primary. The enclosed address must be one of the association's
2728  *   locally bound addresses. The following structure is used to make a
2729  *   set primary request:
2730  */
2731 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2732 					     int optlen)
2733 {
2734 	struct sctp_sock	*sp;
2735 	struct sctp_endpoint	*ep;
2736 	struct sctp_association	*asoc = NULL;
2737 	struct sctp_setpeerprim	prim;
2738 	struct sctp_chunk	*chunk;
2739 	int 			err;
2740 
2741 	sp = sctp_sk(sk);
2742 	ep = sp->ep;
2743 
2744 	if (!sctp_addip_enable)
2745 		return -EPERM;
2746 
2747 	if (optlen != sizeof(struct sctp_setpeerprim))
2748 		return -EINVAL;
2749 
2750 	if (copy_from_user(&prim, optval, optlen))
2751 		return -EFAULT;
2752 
2753 	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2754 	if (!asoc)
2755 		return -EINVAL;
2756 
2757 	if (!asoc->peer.asconf_capable)
2758 		return -EPERM;
2759 
2760 	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2761 		return -EPERM;
2762 
2763 	if (!sctp_state(asoc, ESTABLISHED))
2764 		return -ENOTCONN;
2765 
2766 	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2767 		return -EADDRNOTAVAIL;
2768 
2769 	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
2770 	chunk = sctp_make_asconf_set_prim(asoc,
2771 					  (union sctp_addr *)&prim.sspp_addr);
2772 	if (!chunk)
2773 		return -ENOMEM;
2774 
2775 	err = sctp_send_asconf(asoc, chunk);
2776 
2777 	SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2778 
2779 	return err;
2780 }
2781 
2782 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2783 					  int optlen)
2784 {
2785 	struct sctp_setadaptation adaptation;
2786 
2787 	if (optlen != sizeof(struct sctp_setadaptation))
2788 		return -EINVAL;
2789 	if (copy_from_user(&adaptation, optval, optlen))
2790 		return -EFAULT;
2791 
2792 	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2793 
2794 	return 0;
2795 }
2796 
2797 /*
2798  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
2799  *
2800  * The context field in the sctp_sndrcvinfo structure is normally only
2801  * used when a failed message is retrieved holding the value that was
2802  * sent down on the actual send call.  This option allows the setting of
2803  * a default context on an association basis that will be received on
2804  * reading messages from the peer.  This is especially helpful in the
2805  * one-2-many model for an application to keep some reference to an
2806  * internal state machine that is processing messages on the
2807  * association.  Note that the setting of this value only effects
2808  * received messages from the peer and does not effect the value that is
2809  * saved with outbound messages.
2810  */
2811 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2812 				   int optlen)
2813 {
2814 	struct sctp_assoc_value params;
2815 	struct sctp_sock *sp;
2816 	struct sctp_association *asoc;
2817 
2818 	if (optlen != sizeof(struct sctp_assoc_value))
2819 		return -EINVAL;
2820 	if (copy_from_user(&params, optval, optlen))
2821 		return -EFAULT;
2822 
2823 	sp = sctp_sk(sk);
2824 
2825 	if (params.assoc_id != 0) {
2826 		asoc = sctp_id2assoc(sk, params.assoc_id);
2827 		if (!asoc)
2828 			return -EINVAL;
2829 		asoc->default_rcv_context = params.assoc_value;
2830 	} else {
2831 		sp->default_rcv_context = params.assoc_value;
2832 	}
2833 
2834 	return 0;
2835 }
2836 
2837 /*
2838  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2839  *
2840  * This options will at a minimum specify if the implementation is doing
2841  * fragmented interleave.  Fragmented interleave, for a one to many
2842  * socket, is when subsequent calls to receive a message may return
2843  * parts of messages from different associations.  Some implementations
2844  * may allow you to turn this value on or off.  If so, when turned off,
2845  * no fragment interleave will occur (which will cause a head of line
2846  * blocking amongst multiple associations sharing the same one to many
2847  * socket).  When this option is turned on, then each receive call may
2848  * come from a different association (thus the user must receive data
2849  * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2850  * association each receive belongs to.
2851  *
2852  * This option takes a boolean value.  A non-zero value indicates that
2853  * fragmented interleave is on.  A value of zero indicates that
2854  * fragmented interleave is off.
2855  *
2856  * Note that it is important that an implementation that allows this
2857  * option to be turned on, have it off by default.  Otherwise an unaware
2858  * application using the one to many model may become confused and act
2859  * incorrectly.
2860  */
2861 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2862 					       char __user *optval,
2863 					       int optlen)
2864 {
2865 	int val;
2866 
2867 	if (optlen != sizeof(int))
2868 		return -EINVAL;
2869 	if (get_user(val, (int __user *)optval))
2870 		return -EFAULT;
2871 
2872 	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2873 
2874 	return 0;
2875 }
2876 
2877 /*
2878  * 7.1.25.  Set or Get the sctp partial delivery point
2879  *       (SCTP_PARTIAL_DELIVERY_POINT)
2880  * This option will set or get the SCTP partial delivery point.  This
2881  * point is the size of a message where the partial delivery API will be
2882  * invoked to help free up rwnd space for the peer.  Setting this to a
2883  * lower value will cause partial delivery's to happen more often.  The
2884  * calls argument is an integer that sets or gets the partial delivery
2885  * point.
2886  */
2887 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2888 						  char __user *optval,
2889 						  int optlen)
2890 {
2891 	u32 val;
2892 
2893 	if (optlen != sizeof(u32))
2894 		return -EINVAL;
2895 	if (get_user(val, (int __user *)optval))
2896 		return -EFAULT;
2897 
2898 	sctp_sk(sk)->pd_point = val;
2899 
2900 	return 0; /* is this the right error code? */
2901 }
2902 
2903 /*
2904  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
2905  *
2906  * This option will allow a user to change the maximum burst of packets
2907  * that can be emitted by this association.  Note that the default value
2908  * is 4, and some implementations may restrict this setting so that it
2909  * can only be lowered.
2910  *
2911  * NOTE: This text doesn't seem right.  Do this on a socket basis with
2912  * future associations inheriting the socket value.
2913  */
2914 static int sctp_setsockopt_maxburst(struct sock *sk,
2915 				    char __user *optval,
2916 				    int optlen)
2917 {
2918 	struct sctp_assoc_value params;
2919 	struct sctp_sock *sp;
2920 	struct sctp_association *asoc;
2921 	int val;
2922 	int assoc_id = 0;
2923 
2924 	if (optlen < sizeof(int))
2925 		return -EINVAL;
2926 
2927 	if (optlen == sizeof(int)) {
2928 		printk(KERN_WARNING
2929 		   "SCTP: Use of int in max_burst socket option deprecated\n");
2930 		printk(KERN_WARNING
2931 		   "SCTP: Use struct sctp_assoc_value instead\n");
2932 		if (copy_from_user(&val, optval, optlen))
2933 			return -EFAULT;
2934 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2935 		if (copy_from_user(&params, optval, optlen))
2936 			return -EFAULT;
2937 		val = params.assoc_value;
2938 		assoc_id = params.assoc_id;
2939 	} else
2940 		return -EINVAL;
2941 
2942 	sp = sctp_sk(sk);
2943 
2944 	if (assoc_id != 0) {
2945 		asoc = sctp_id2assoc(sk, assoc_id);
2946 		if (!asoc)
2947 			return -EINVAL;
2948 		asoc->max_burst = val;
2949 	} else
2950 		sp->max_burst = val;
2951 
2952 	return 0;
2953 }
2954 
2955 /*
2956  * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
2957  *
2958  * This set option adds a chunk type that the user is requesting to be
2959  * received only in an authenticated way.  Changes to the list of chunks
2960  * will only effect future associations on the socket.
2961  */
2962 static int sctp_setsockopt_auth_chunk(struct sock *sk,
2963 				    char __user *optval,
2964 				    int optlen)
2965 {
2966 	struct sctp_authchunk val;
2967 
2968 	if (optlen != sizeof(struct sctp_authchunk))
2969 		return -EINVAL;
2970 	if (copy_from_user(&val, optval, optlen))
2971 		return -EFAULT;
2972 
2973 	switch (val.sauth_chunk) {
2974 		case SCTP_CID_INIT:
2975 		case SCTP_CID_INIT_ACK:
2976 		case SCTP_CID_SHUTDOWN_COMPLETE:
2977 		case SCTP_CID_AUTH:
2978 			return -EINVAL;
2979 	}
2980 
2981 	/* add this chunk id to the endpoint */
2982 	return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
2983 }
2984 
2985 /*
2986  * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
2987  *
2988  * This option gets or sets the list of HMAC algorithms that the local
2989  * endpoint requires the peer to use.
2990  */
2991 static int sctp_setsockopt_hmac_ident(struct sock *sk,
2992 				    char __user *optval,
2993 				    int optlen)
2994 {
2995 	struct sctp_hmacalgo *hmacs;
2996 	int err;
2997 
2998 	if (optlen < sizeof(struct sctp_hmacalgo))
2999 		return -EINVAL;
3000 
3001 	hmacs = kmalloc(optlen, GFP_KERNEL);
3002 	if (!hmacs)
3003 		return -ENOMEM;
3004 
3005 	if (copy_from_user(hmacs, optval, optlen)) {
3006 		err = -EFAULT;
3007 		goto out;
3008 	}
3009 
3010 	if (hmacs->shmac_num_idents == 0 ||
3011 	    hmacs->shmac_num_idents > SCTP_AUTH_NUM_HMACS) {
3012 		err = -EINVAL;
3013 		goto out;
3014 	}
3015 
3016 	err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3017 out:
3018 	kfree(hmacs);
3019 	return err;
3020 }
3021 
3022 /*
3023  * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3024  *
3025  * This option will set a shared secret key which is used to build an
3026  * association shared key.
3027  */
3028 static int sctp_setsockopt_auth_key(struct sock *sk,
3029 				    char __user *optval,
3030 				    int optlen)
3031 {
3032 	struct sctp_authkey *authkey;
3033 	struct sctp_association *asoc;
3034 	int ret;
3035 
3036 	if (optlen <= sizeof(struct sctp_authkey))
3037 		return -EINVAL;
3038 
3039 	authkey = kmalloc(optlen, GFP_KERNEL);
3040 	if (!authkey)
3041 		return -ENOMEM;
3042 
3043 	if (copy_from_user(authkey, optval, optlen)) {
3044 		ret = -EFAULT;
3045 		goto out;
3046 	}
3047 
3048 	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3049 	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3050 		ret = -EINVAL;
3051 		goto out;
3052 	}
3053 
3054 	ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3055 out:
3056 	kfree(authkey);
3057 	return ret;
3058 }
3059 
3060 /*
3061  * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3062  *
3063  * This option will get or set the active shared key to be used to build
3064  * the association shared key.
3065  */
3066 static int sctp_setsockopt_active_key(struct sock *sk,
3067 					char __user *optval,
3068 					int optlen)
3069 {
3070 	struct sctp_authkeyid val;
3071 	struct sctp_association *asoc;
3072 
3073 	if (optlen != sizeof(struct sctp_authkeyid))
3074 		return -EINVAL;
3075 	if (copy_from_user(&val, optval, optlen))
3076 		return -EFAULT;
3077 
3078 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3079 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3080 		return -EINVAL;
3081 
3082 	return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3083 					val.scact_keynumber);
3084 }
3085 
3086 /*
3087  * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3088  *
3089  * This set option will delete a shared secret key from use.
3090  */
3091 static int sctp_setsockopt_del_key(struct sock *sk,
3092 					char __user *optval,
3093 					int optlen)
3094 {
3095 	struct sctp_authkeyid val;
3096 	struct sctp_association *asoc;
3097 
3098 	if (optlen != sizeof(struct sctp_authkeyid))
3099 		return -EINVAL;
3100 	if (copy_from_user(&val, optval, optlen))
3101 		return -EFAULT;
3102 
3103 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3104 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3105 		return -EINVAL;
3106 
3107 	return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3108 				    val.scact_keynumber);
3109 
3110 }
3111 
3112 
3113 /* API 6.2 setsockopt(), getsockopt()
3114  *
3115  * Applications use setsockopt() and getsockopt() to set or retrieve
3116  * socket options.  Socket options are used to change the default
3117  * behavior of sockets calls.  They are described in Section 7.
3118  *
3119  * The syntax is:
3120  *
3121  *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3122  *                    int __user *optlen);
3123  *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3124  *                    int optlen);
3125  *
3126  *   sd      - the socket descript.
3127  *   level   - set to IPPROTO_SCTP for all SCTP options.
3128  *   optname - the option name.
3129  *   optval  - the buffer to store the value of the option.
3130  *   optlen  - the size of the buffer.
3131  */
3132 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3133 				char __user *optval, int optlen)
3134 {
3135 	int retval = 0;
3136 
3137 	SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3138 			  sk, optname);
3139 
3140 	/* I can hardly begin to describe how wrong this is.  This is
3141 	 * so broken as to be worse than useless.  The API draft
3142 	 * REALLY is NOT helpful here...  I am not convinced that the
3143 	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3144 	 * are at all well-founded.
3145 	 */
3146 	if (level != SOL_SCTP) {
3147 		struct sctp_af *af = sctp_sk(sk)->pf->af;
3148 		retval = af->setsockopt(sk, level, optname, optval, optlen);
3149 		goto out_nounlock;
3150 	}
3151 
3152 	sctp_lock_sock(sk);
3153 
3154 	switch (optname) {
3155 	case SCTP_SOCKOPT_BINDX_ADD:
3156 		/* 'optlen' is the size of the addresses buffer. */
3157 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3158 					       optlen, SCTP_BINDX_ADD_ADDR);
3159 		break;
3160 
3161 	case SCTP_SOCKOPT_BINDX_REM:
3162 		/* 'optlen' is the size of the addresses buffer. */
3163 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3164 					       optlen, SCTP_BINDX_REM_ADDR);
3165 		break;
3166 
3167 	case SCTP_SOCKOPT_CONNECTX:
3168 		/* 'optlen' is the size of the addresses buffer. */
3169 		retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3170 					       optlen);
3171 		break;
3172 
3173 	case SCTP_DISABLE_FRAGMENTS:
3174 		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3175 		break;
3176 
3177 	case SCTP_EVENTS:
3178 		retval = sctp_setsockopt_events(sk, optval, optlen);
3179 		break;
3180 
3181 	case SCTP_AUTOCLOSE:
3182 		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3183 		break;
3184 
3185 	case SCTP_PEER_ADDR_PARAMS:
3186 		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3187 		break;
3188 
3189 	case SCTP_DELAYED_ACK_TIME:
3190 		retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3191 		break;
3192 	case SCTP_PARTIAL_DELIVERY_POINT:
3193 		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3194 		break;
3195 
3196 	case SCTP_INITMSG:
3197 		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3198 		break;
3199 	case SCTP_DEFAULT_SEND_PARAM:
3200 		retval = sctp_setsockopt_default_send_param(sk, optval,
3201 							    optlen);
3202 		break;
3203 	case SCTP_PRIMARY_ADDR:
3204 		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3205 		break;
3206 	case SCTP_SET_PEER_PRIMARY_ADDR:
3207 		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3208 		break;
3209 	case SCTP_NODELAY:
3210 		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3211 		break;
3212 	case SCTP_RTOINFO:
3213 		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3214 		break;
3215 	case SCTP_ASSOCINFO:
3216 		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3217 		break;
3218 	case SCTP_I_WANT_MAPPED_V4_ADDR:
3219 		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3220 		break;
3221 	case SCTP_MAXSEG:
3222 		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3223 		break;
3224 	case SCTP_ADAPTATION_LAYER:
3225 		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3226 		break;
3227 	case SCTP_CONTEXT:
3228 		retval = sctp_setsockopt_context(sk, optval, optlen);
3229 		break;
3230 	case SCTP_FRAGMENT_INTERLEAVE:
3231 		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3232 		break;
3233 	case SCTP_MAX_BURST:
3234 		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3235 		break;
3236 	case SCTP_AUTH_CHUNK:
3237 		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3238 		break;
3239 	case SCTP_HMAC_IDENT:
3240 		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3241 		break;
3242 	case SCTP_AUTH_KEY:
3243 		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3244 		break;
3245 	case SCTP_AUTH_ACTIVE_KEY:
3246 		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3247 		break;
3248 	case SCTP_AUTH_DELETE_KEY:
3249 		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3250 		break;
3251 	default:
3252 		retval = -ENOPROTOOPT;
3253 		break;
3254 	}
3255 
3256 	sctp_release_sock(sk);
3257 
3258 out_nounlock:
3259 	return retval;
3260 }
3261 
3262 /* API 3.1.6 connect() - UDP Style Syntax
3263  *
3264  * An application may use the connect() call in the UDP model to initiate an
3265  * association without sending data.
3266  *
3267  * The syntax is:
3268  *
3269  * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3270  *
3271  * sd: the socket descriptor to have a new association added to.
3272  *
3273  * nam: the address structure (either struct sockaddr_in or struct
3274  *    sockaddr_in6 defined in RFC2553 [7]).
3275  *
3276  * len: the size of the address.
3277  */
3278 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3279 			     int addr_len)
3280 {
3281 	int err = 0;
3282 	struct sctp_af *af;
3283 
3284 	sctp_lock_sock(sk);
3285 
3286 	SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3287 			  __func__, sk, addr, addr_len);
3288 
3289 	/* Validate addr_len before calling common connect/connectx routine. */
3290 	af = sctp_get_af_specific(addr->sa_family);
3291 	if (!af || addr_len < af->sockaddr_len) {
3292 		err = -EINVAL;
3293 	} else {
3294 		/* Pass correct addr len to common routine (so it knows there
3295 		 * is only one address being passed.
3296 		 */
3297 		err = __sctp_connect(sk, addr, af->sockaddr_len);
3298 	}
3299 
3300 	sctp_release_sock(sk);
3301 	return err;
3302 }
3303 
3304 /* FIXME: Write comments. */
3305 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3306 {
3307 	return -EOPNOTSUPP; /* STUB */
3308 }
3309 
3310 /* 4.1.4 accept() - TCP Style Syntax
3311  *
3312  * Applications use accept() call to remove an established SCTP
3313  * association from the accept queue of the endpoint.  A new socket
3314  * descriptor will be returned from accept() to represent the newly
3315  * formed association.
3316  */
3317 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3318 {
3319 	struct sctp_sock *sp;
3320 	struct sctp_endpoint *ep;
3321 	struct sock *newsk = NULL;
3322 	struct sctp_association *asoc;
3323 	long timeo;
3324 	int error = 0;
3325 
3326 	sctp_lock_sock(sk);
3327 
3328 	sp = sctp_sk(sk);
3329 	ep = sp->ep;
3330 
3331 	if (!sctp_style(sk, TCP)) {
3332 		error = -EOPNOTSUPP;
3333 		goto out;
3334 	}
3335 
3336 	if (!sctp_sstate(sk, LISTENING)) {
3337 		error = -EINVAL;
3338 		goto out;
3339 	}
3340 
3341 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3342 
3343 	error = sctp_wait_for_accept(sk, timeo);
3344 	if (error)
3345 		goto out;
3346 
3347 	/* We treat the list of associations on the endpoint as the accept
3348 	 * queue and pick the first association on the list.
3349 	 */
3350 	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3351 
3352 	newsk = sp->pf->create_accept_sk(sk, asoc);
3353 	if (!newsk) {
3354 		error = -ENOMEM;
3355 		goto out;
3356 	}
3357 
3358 	/* Populate the fields of the newsk from the oldsk and migrate the
3359 	 * asoc to the newsk.
3360 	 */
3361 	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3362 
3363 out:
3364 	sctp_release_sock(sk);
3365 	*err = error;
3366 	return newsk;
3367 }
3368 
3369 /* The SCTP ioctl handler. */
3370 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3371 {
3372 	return -ENOIOCTLCMD;
3373 }
3374 
3375 /* This is the function which gets called during socket creation to
3376  * initialized the SCTP-specific portion of the sock.
3377  * The sock structure should already be zero-filled memory.
3378  */
3379 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3380 {
3381 	struct sctp_endpoint *ep;
3382 	struct sctp_sock *sp;
3383 
3384 	SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3385 
3386 	sp = sctp_sk(sk);
3387 
3388 	/* Initialize the SCTP per socket area.  */
3389 	switch (sk->sk_type) {
3390 	case SOCK_SEQPACKET:
3391 		sp->type = SCTP_SOCKET_UDP;
3392 		break;
3393 	case SOCK_STREAM:
3394 		sp->type = SCTP_SOCKET_TCP;
3395 		break;
3396 	default:
3397 		return -ESOCKTNOSUPPORT;
3398 	}
3399 
3400 	/* Initialize default send parameters. These parameters can be
3401 	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3402 	 */
3403 	sp->default_stream = 0;
3404 	sp->default_ppid = 0;
3405 	sp->default_flags = 0;
3406 	sp->default_context = 0;
3407 	sp->default_timetolive = 0;
3408 
3409 	sp->default_rcv_context = 0;
3410 	sp->max_burst = sctp_max_burst;
3411 
3412 	/* Initialize default setup parameters. These parameters
3413 	 * can be modified with the SCTP_INITMSG socket option or
3414 	 * overridden by the SCTP_INIT CMSG.
3415 	 */
3416 	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3417 	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3418 	sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3419 	sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3420 
3421 	/* Initialize default RTO related parameters.  These parameters can
3422 	 * be modified for with the SCTP_RTOINFO socket option.
3423 	 */
3424 	sp->rtoinfo.srto_initial = sctp_rto_initial;
3425 	sp->rtoinfo.srto_max     = sctp_rto_max;
3426 	sp->rtoinfo.srto_min     = sctp_rto_min;
3427 
3428 	/* Initialize default association related parameters. These parameters
3429 	 * can be modified with the SCTP_ASSOCINFO socket option.
3430 	 */
3431 	sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3432 	sp->assocparams.sasoc_number_peer_destinations = 0;
3433 	sp->assocparams.sasoc_peer_rwnd = 0;
3434 	sp->assocparams.sasoc_local_rwnd = 0;
3435 	sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3436 
3437 	/* Initialize default event subscriptions. By default, all the
3438 	 * options are off.
3439 	 */
3440 	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3441 
3442 	/* Default Peer Address Parameters.  These defaults can
3443 	 * be modified via SCTP_PEER_ADDR_PARAMS
3444 	 */
3445 	sp->hbinterval  = sctp_hb_interval;
3446 	sp->pathmaxrxt  = sctp_max_retrans_path;
3447 	sp->pathmtu     = 0; // allow default discovery
3448 	sp->sackdelay   = sctp_sack_timeout;
3449 	sp->param_flags = SPP_HB_ENABLE |
3450 			  SPP_PMTUD_ENABLE |
3451 			  SPP_SACKDELAY_ENABLE;
3452 
3453 	/* If enabled no SCTP message fragmentation will be performed.
3454 	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3455 	 */
3456 	sp->disable_fragments = 0;
3457 
3458 	/* Enable Nagle algorithm by default.  */
3459 	sp->nodelay           = 0;
3460 
3461 	/* Enable by default. */
3462 	sp->v4mapped          = 1;
3463 
3464 	/* Auto-close idle associations after the configured
3465 	 * number of seconds.  A value of 0 disables this
3466 	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3467 	 * for UDP-style sockets only.
3468 	 */
3469 	sp->autoclose         = 0;
3470 
3471 	/* User specified fragmentation limit. */
3472 	sp->user_frag         = 0;
3473 
3474 	sp->adaptation_ind = 0;
3475 
3476 	sp->pf = sctp_get_pf_specific(sk->sk_family);
3477 
3478 	/* Control variables for partial data delivery. */
3479 	atomic_set(&sp->pd_mode, 0);
3480 	skb_queue_head_init(&sp->pd_lobby);
3481 	sp->frag_interleave = 0;
3482 
3483 	/* Create a per socket endpoint structure.  Even if we
3484 	 * change the data structure relationships, this may still
3485 	 * be useful for storing pre-connect address information.
3486 	 */
3487 	ep = sctp_endpoint_new(sk, GFP_KERNEL);
3488 	if (!ep)
3489 		return -ENOMEM;
3490 
3491 	sp->ep = ep;
3492 	sp->hmac = NULL;
3493 
3494 	SCTP_DBG_OBJCNT_INC(sock);
3495 	atomic_inc(&sctp_sockets_allocated);
3496 	return 0;
3497 }
3498 
3499 /* Cleanup any SCTP per socket resources.  */
3500 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3501 {
3502 	struct sctp_endpoint *ep;
3503 
3504 	SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3505 
3506 	/* Release our hold on the endpoint. */
3507 	ep = sctp_sk(sk)->ep;
3508 	sctp_endpoint_free(ep);
3509 	atomic_dec(&sctp_sockets_allocated);
3510 	return 0;
3511 }
3512 
3513 /* API 4.1.7 shutdown() - TCP Style Syntax
3514  *     int shutdown(int socket, int how);
3515  *
3516  *     sd      - the socket descriptor of the association to be closed.
3517  *     how     - Specifies the type of shutdown.  The  values  are
3518  *               as follows:
3519  *               SHUT_RD
3520  *                     Disables further receive operations. No SCTP
3521  *                     protocol action is taken.
3522  *               SHUT_WR
3523  *                     Disables further send operations, and initiates
3524  *                     the SCTP shutdown sequence.
3525  *               SHUT_RDWR
3526  *                     Disables further send  and  receive  operations
3527  *                     and initiates the SCTP shutdown sequence.
3528  */
3529 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3530 {
3531 	struct sctp_endpoint *ep;
3532 	struct sctp_association *asoc;
3533 
3534 	if (!sctp_style(sk, TCP))
3535 		return;
3536 
3537 	if (how & SEND_SHUTDOWN) {
3538 		ep = sctp_sk(sk)->ep;
3539 		if (!list_empty(&ep->asocs)) {
3540 			asoc = list_entry(ep->asocs.next,
3541 					  struct sctp_association, asocs);
3542 			sctp_primitive_SHUTDOWN(asoc, NULL);
3543 		}
3544 	}
3545 }
3546 
3547 /* 7.2.1 Association Status (SCTP_STATUS)
3548 
3549  * Applications can retrieve current status information about an
3550  * association, including association state, peer receiver window size,
3551  * number of unacked data chunks, and number of data chunks pending
3552  * receipt.  This information is read-only.
3553  */
3554 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3555 				       char __user *optval,
3556 				       int __user *optlen)
3557 {
3558 	struct sctp_status status;
3559 	struct sctp_association *asoc = NULL;
3560 	struct sctp_transport *transport;
3561 	sctp_assoc_t associd;
3562 	int retval = 0;
3563 
3564 	if (len < sizeof(status)) {
3565 		retval = -EINVAL;
3566 		goto out;
3567 	}
3568 
3569 	len = sizeof(status);
3570 	if (copy_from_user(&status, optval, len)) {
3571 		retval = -EFAULT;
3572 		goto out;
3573 	}
3574 
3575 	associd = status.sstat_assoc_id;
3576 	asoc = sctp_id2assoc(sk, associd);
3577 	if (!asoc) {
3578 		retval = -EINVAL;
3579 		goto out;
3580 	}
3581 
3582 	transport = asoc->peer.primary_path;
3583 
3584 	status.sstat_assoc_id = sctp_assoc2id(asoc);
3585 	status.sstat_state = asoc->state;
3586 	status.sstat_rwnd =  asoc->peer.rwnd;
3587 	status.sstat_unackdata = asoc->unack_data;
3588 
3589 	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3590 	status.sstat_instrms = asoc->c.sinit_max_instreams;
3591 	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3592 	status.sstat_fragmentation_point = asoc->frag_point;
3593 	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3594 	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3595 			transport->af_specific->sockaddr_len);
3596 	/* Map ipv4 address into v4-mapped-on-v6 address.  */
3597 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3598 		(union sctp_addr *)&status.sstat_primary.spinfo_address);
3599 	status.sstat_primary.spinfo_state = transport->state;
3600 	status.sstat_primary.spinfo_cwnd = transport->cwnd;
3601 	status.sstat_primary.spinfo_srtt = transport->srtt;
3602 	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3603 	status.sstat_primary.spinfo_mtu = transport->pathmtu;
3604 
3605 	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3606 		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3607 
3608 	if (put_user(len, optlen)) {
3609 		retval = -EFAULT;
3610 		goto out;
3611 	}
3612 
3613 	SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3614 			  len, status.sstat_state, status.sstat_rwnd,
3615 			  status.sstat_assoc_id);
3616 
3617 	if (copy_to_user(optval, &status, len)) {
3618 		retval = -EFAULT;
3619 		goto out;
3620 	}
3621 
3622 out:
3623 	return (retval);
3624 }
3625 
3626 
3627 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3628  *
3629  * Applications can retrieve information about a specific peer address
3630  * of an association, including its reachability state, congestion
3631  * window, and retransmission timer values.  This information is
3632  * read-only.
3633  */
3634 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3635 					  char __user *optval,
3636 					  int __user *optlen)
3637 {
3638 	struct sctp_paddrinfo pinfo;
3639 	struct sctp_transport *transport;
3640 	int retval = 0;
3641 
3642 	if (len < sizeof(pinfo)) {
3643 		retval = -EINVAL;
3644 		goto out;
3645 	}
3646 
3647 	len = sizeof(pinfo);
3648 	if (copy_from_user(&pinfo, optval, len)) {
3649 		retval = -EFAULT;
3650 		goto out;
3651 	}
3652 
3653 	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3654 					   pinfo.spinfo_assoc_id);
3655 	if (!transport)
3656 		return -EINVAL;
3657 
3658 	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3659 	pinfo.spinfo_state = transport->state;
3660 	pinfo.spinfo_cwnd = transport->cwnd;
3661 	pinfo.spinfo_srtt = transport->srtt;
3662 	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3663 	pinfo.spinfo_mtu = transport->pathmtu;
3664 
3665 	if (pinfo.spinfo_state == SCTP_UNKNOWN)
3666 		pinfo.spinfo_state = SCTP_ACTIVE;
3667 
3668 	if (put_user(len, optlen)) {
3669 		retval = -EFAULT;
3670 		goto out;
3671 	}
3672 
3673 	if (copy_to_user(optval, &pinfo, len)) {
3674 		retval = -EFAULT;
3675 		goto out;
3676 	}
3677 
3678 out:
3679 	return (retval);
3680 }
3681 
3682 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3683  *
3684  * This option is a on/off flag.  If enabled no SCTP message
3685  * fragmentation will be performed.  Instead if a message being sent
3686  * exceeds the current PMTU size, the message will NOT be sent and
3687  * instead a error will be indicated to the user.
3688  */
3689 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3690 					char __user *optval, int __user *optlen)
3691 {
3692 	int val;
3693 
3694 	if (len < sizeof(int))
3695 		return -EINVAL;
3696 
3697 	len = sizeof(int);
3698 	val = (sctp_sk(sk)->disable_fragments == 1);
3699 	if (put_user(len, optlen))
3700 		return -EFAULT;
3701 	if (copy_to_user(optval, &val, len))
3702 		return -EFAULT;
3703 	return 0;
3704 }
3705 
3706 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3707  *
3708  * This socket option is used to specify various notifications and
3709  * ancillary data the user wishes to receive.
3710  */
3711 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3712 				  int __user *optlen)
3713 {
3714 	if (len < sizeof(struct sctp_event_subscribe))
3715 		return -EINVAL;
3716 	len = sizeof(struct sctp_event_subscribe);
3717 	if (put_user(len, optlen))
3718 		return -EFAULT;
3719 	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3720 		return -EFAULT;
3721 	return 0;
3722 }
3723 
3724 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3725  *
3726  * This socket option is applicable to the UDP-style socket only.  When
3727  * set it will cause associations that are idle for more than the
3728  * specified number of seconds to automatically close.  An association
3729  * being idle is defined an association that has NOT sent or received
3730  * user data.  The special value of '0' indicates that no automatic
3731  * close of any associations should be performed.  The option expects an
3732  * integer defining the number of seconds of idle time before an
3733  * association is closed.
3734  */
3735 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3736 {
3737 	/* Applicable to UDP-style socket only */
3738 	if (sctp_style(sk, TCP))
3739 		return -EOPNOTSUPP;
3740 	if (len < sizeof(int))
3741 		return -EINVAL;
3742 	len = sizeof(int);
3743 	if (put_user(len, optlen))
3744 		return -EFAULT;
3745 	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3746 		return -EFAULT;
3747 	return 0;
3748 }
3749 
3750 /* Helper routine to branch off an association to a new socket.  */
3751 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3752 				struct socket **sockp)
3753 {
3754 	struct sock *sk = asoc->base.sk;
3755 	struct socket *sock;
3756 	struct inet_sock *inetsk;
3757 	struct sctp_af *af;
3758 	int err = 0;
3759 
3760 	/* An association cannot be branched off from an already peeled-off
3761 	 * socket, nor is this supported for tcp style sockets.
3762 	 */
3763 	if (!sctp_style(sk, UDP))
3764 		return -EINVAL;
3765 
3766 	/* Create a new socket.  */
3767 	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3768 	if (err < 0)
3769 		return err;
3770 
3771 	/* Populate the fields of the newsk from the oldsk and migrate the
3772 	 * asoc to the newsk.
3773 	 */
3774 	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3775 
3776 	/* Make peeled-off sockets more like 1-1 accepted sockets.
3777 	 * Set the daddr and initialize id to something more random
3778 	 */
3779 	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3780 	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3781 	inetsk = inet_sk(sock->sk);
3782 	inetsk->id = asoc->next_tsn ^ jiffies;
3783 
3784 	*sockp = sock;
3785 
3786 	return err;
3787 }
3788 
3789 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3790 {
3791 	sctp_peeloff_arg_t peeloff;
3792 	struct socket *newsock;
3793 	int retval = 0;
3794 	struct sctp_association *asoc;
3795 
3796 	if (len < sizeof(sctp_peeloff_arg_t))
3797 		return -EINVAL;
3798 	len = sizeof(sctp_peeloff_arg_t);
3799 	if (copy_from_user(&peeloff, optval, len))
3800 		return -EFAULT;
3801 
3802 	asoc = sctp_id2assoc(sk, peeloff.associd);
3803 	if (!asoc) {
3804 		retval = -EINVAL;
3805 		goto out;
3806 	}
3807 
3808 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3809 
3810 	retval = sctp_do_peeloff(asoc, &newsock);
3811 	if (retval < 0)
3812 		goto out;
3813 
3814 	/* Map the socket to an unused fd that can be returned to the user.  */
3815 	retval = sock_map_fd(newsock);
3816 	if (retval < 0) {
3817 		sock_release(newsock);
3818 		goto out;
3819 	}
3820 
3821 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3822 			  __func__, sk, asoc, newsock->sk, retval);
3823 
3824 	/* Return the fd mapped to the new socket.  */
3825 	peeloff.sd = retval;
3826 	if (put_user(len, optlen))
3827 		return -EFAULT;
3828 	if (copy_to_user(optval, &peeloff, len))
3829 		retval = -EFAULT;
3830 
3831 out:
3832 	return retval;
3833 }
3834 
3835 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3836  *
3837  * Applications can enable or disable heartbeats for any peer address of
3838  * an association, modify an address's heartbeat interval, force a
3839  * heartbeat to be sent immediately, and adjust the address's maximum
3840  * number of retransmissions sent before an address is considered
3841  * unreachable.  The following structure is used to access and modify an
3842  * address's parameters:
3843  *
3844  *  struct sctp_paddrparams {
3845  *     sctp_assoc_t            spp_assoc_id;
3846  *     struct sockaddr_storage spp_address;
3847  *     uint32_t                spp_hbinterval;
3848  *     uint16_t                spp_pathmaxrxt;
3849  *     uint32_t                spp_pathmtu;
3850  *     uint32_t                spp_sackdelay;
3851  *     uint32_t                spp_flags;
3852  * };
3853  *
3854  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
3855  *                     application, and identifies the association for
3856  *                     this query.
3857  *   spp_address     - This specifies which address is of interest.
3858  *   spp_hbinterval  - This contains the value of the heartbeat interval,
3859  *                     in milliseconds.  If a  value of zero
3860  *                     is present in this field then no changes are to
3861  *                     be made to this parameter.
3862  *   spp_pathmaxrxt  - This contains the maximum number of
3863  *                     retransmissions before this address shall be
3864  *                     considered unreachable. If a  value of zero
3865  *                     is present in this field then no changes are to
3866  *                     be made to this parameter.
3867  *   spp_pathmtu     - When Path MTU discovery is disabled the value
3868  *                     specified here will be the "fixed" path mtu.
3869  *                     Note that if the spp_address field is empty
3870  *                     then all associations on this address will
3871  *                     have this fixed path mtu set upon them.
3872  *
3873  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
3874  *                     the number of milliseconds that sacks will be delayed
3875  *                     for. This value will apply to all addresses of an
3876  *                     association if the spp_address field is empty. Note
3877  *                     also, that if delayed sack is enabled and this
3878  *                     value is set to 0, no change is made to the last
3879  *                     recorded delayed sack timer value.
3880  *
3881  *   spp_flags       - These flags are used to control various features
3882  *                     on an association. The flag field may contain
3883  *                     zero or more of the following options.
3884  *
3885  *                     SPP_HB_ENABLE  - Enable heartbeats on the
3886  *                     specified address. Note that if the address
3887  *                     field is empty all addresses for the association
3888  *                     have heartbeats enabled upon them.
3889  *
3890  *                     SPP_HB_DISABLE - Disable heartbeats on the
3891  *                     speicifed address. Note that if the address
3892  *                     field is empty all addresses for the association
3893  *                     will have their heartbeats disabled. Note also
3894  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
3895  *                     mutually exclusive, only one of these two should
3896  *                     be specified. Enabling both fields will have
3897  *                     undetermined results.
3898  *
3899  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
3900  *                     to be made immediately.
3901  *
3902  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
3903  *                     discovery upon the specified address. Note that
3904  *                     if the address feild is empty then all addresses
3905  *                     on the association are effected.
3906  *
3907  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
3908  *                     discovery upon the specified address. Note that
3909  *                     if the address feild is empty then all addresses
3910  *                     on the association are effected. Not also that
3911  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3912  *                     exclusive. Enabling both will have undetermined
3913  *                     results.
3914  *
3915  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
3916  *                     on delayed sack. The time specified in spp_sackdelay
3917  *                     is used to specify the sack delay for this address. Note
3918  *                     that if spp_address is empty then all addresses will
3919  *                     enable delayed sack and take on the sack delay
3920  *                     value specified in spp_sackdelay.
3921  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
3922  *                     off delayed sack. If the spp_address field is blank then
3923  *                     delayed sack is disabled for the entire association. Note
3924  *                     also that this field is mutually exclusive to
3925  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
3926  *                     results.
3927  */
3928 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3929 					    char __user *optval, int __user *optlen)
3930 {
3931 	struct sctp_paddrparams  params;
3932 	struct sctp_transport   *trans = NULL;
3933 	struct sctp_association *asoc = NULL;
3934 	struct sctp_sock        *sp = sctp_sk(sk);
3935 
3936 	if (len < sizeof(struct sctp_paddrparams))
3937 		return -EINVAL;
3938 	len = sizeof(struct sctp_paddrparams);
3939 	if (copy_from_user(&params, optval, len))
3940 		return -EFAULT;
3941 
3942 	/* If an address other than INADDR_ANY is specified, and
3943 	 * no transport is found, then the request is invalid.
3944 	 */
3945 	if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3946 		trans = sctp_addr_id2transport(sk, &params.spp_address,
3947 					       params.spp_assoc_id);
3948 		if (!trans) {
3949 			SCTP_DEBUG_PRINTK("Failed no transport\n");
3950 			return -EINVAL;
3951 		}
3952 	}
3953 
3954 	/* Get association, if assoc_id != 0 and the socket is a one
3955 	 * to many style socket, and an association was not found, then
3956 	 * the id was invalid.
3957 	 */
3958 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3959 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3960 		SCTP_DEBUG_PRINTK("Failed no association\n");
3961 		return -EINVAL;
3962 	}
3963 
3964 	if (trans) {
3965 		/* Fetch transport values. */
3966 		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3967 		params.spp_pathmtu    = trans->pathmtu;
3968 		params.spp_pathmaxrxt = trans->pathmaxrxt;
3969 		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
3970 
3971 		/*draft-11 doesn't say what to return in spp_flags*/
3972 		params.spp_flags      = trans->param_flags;
3973 	} else if (asoc) {
3974 		/* Fetch association values. */
3975 		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3976 		params.spp_pathmtu    = asoc->pathmtu;
3977 		params.spp_pathmaxrxt = asoc->pathmaxrxt;
3978 		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
3979 
3980 		/*draft-11 doesn't say what to return in spp_flags*/
3981 		params.spp_flags      = asoc->param_flags;
3982 	} else {
3983 		/* Fetch socket values. */
3984 		params.spp_hbinterval = sp->hbinterval;
3985 		params.spp_pathmtu    = sp->pathmtu;
3986 		params.spp_sackdelay  = sp->sackdelay;
3987 		params.spp_pathmaxrxt = sp->pathmaxrxt;
3988 
3989 		/*draft-11 doesn't say what to return in spp_flags*/
3990 		params.spp_flags      = sp->param_flags;
3991 	}
3992 
3993 	if (copy_to_user(optval, &params, len))
3994 		return -EFAULT;
3995 
3996 	if (put_user(len, optlen))
3997 		return -EFAULT;
3998 
3999 	return 0;
4000 }
4001 
4002 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
4003  *
4004  *   This options will get or set the delayed ack timer.  The time is set
4005  *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
4006  *   endpoints default delayed ack timer value.  If the assoc_id field is
4007  *   non-zero, then the set or get effects the specified association.
4008  *
4009  *   struct sctp_assoc_value {
4010  *       sctp_assoc_t            assoc_id;
4011  *       uint32_t                assoc_value;
4012  *   };
4013  *
4014  *     assoc_id    - This parameter, indicates which association the
4015  *                   user is preforming an action upon. Note that if
4016  *                   this field's value is zero then the endpoints
4017  *                   default value is changed (effecting future
4018  *                   associations only).
4019  *
4020  *     assoc_value - This parameter contains the number of milliseconds
4021  *                   that the user is requesting the delayed ACK timer
4022  *                   be set to. Note that this value is defined in
4023  *                   the standard to be between 200 and 500 milliseconds.
4024  *
4025  *                   Note: a value of zero will leave the value alone,
4026  *                   but disable SACK delay. A non-zero value will also
4027  *                   enable SACK delay.
4028  */
4029 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
4030 					    char __user *optval,
4031 					    int __user *optlen)
4032 {
4033 	struct sctp_assoc_value  params;
4034 	struct sctp_association *asoc = NULL;
4035 	struct sctp_sock        *sp = sctp_sk(sk);
4036 
4037 	if (len < sizeof(struct sctp_assoc_value))
4038 		return - EINVAL;
4039 
4040 	len = sizeof(struct sctp_assoc_value);
4041 
4042 	if (copy_from_user(&params, optval, len))
4043 		return -EFAULT;
4044 
4045 	/* Get association, if assoc_id != 0 and the socket is a one
4046 	 * to many style socket, and an association was not found, then
4047 	 * the id was invalid.
4048 	 */
4049 	asoc = sctp_id2assoc(sk, params.assoc_id);
4050 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4051 		return -EINVAL;
4052 
4053 	if (asoc) {
4054 		/* Fetch association values. */
4055 		if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
4056 			params.assoc_value = jiffies_to_msecs(
4057 				asoc->sackdelay);
4058 		else
4059 			params.assoc_value = 0;
4060 	} else {
4061 		/* Fetch socket values. */
4062 		if (sp->param_flags & SPP_SACKDELAY_ENABLE)
4063 			params.assoc_value  = sp->sackdelay;
4064 		else
4065 			params.assoc_value  = 0;
4066 	}
4067 
4068 	if (copy_to_user(optval, &params, len))
4069 		return -EFAULT;
4070 
4071 	if (put_user(len, optlen))
4072 		return -EFAULT;
4073 
4074 	return 0;
4075 }
4076 
4077 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4078  *
4079  * Applications can specify protocol parameters for the default association
4080  * initialization.  The option name argument to setsockopt() and getsockopt()
4081  * is SCTP_INITMSG.
4082  *
4083  * Setting initialization parameters is effective only on an unconnected
4084  * socket (for UDP-style sockets only future associations are effected
4085  * by the change).  With TCP-style sockets, this option is inherited by
4086  * sockets derived from a listener socket.
4087  */
4088 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4089 {
4090 	if (len < sizeof(struct sctp_initmsg))
4091 		return -EINVAL;
4092 	len = sizeof(struct sctp_initmsg);
4093 	if (put_user(len, optlen))
4094 		return -EFAULT;
4095 	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4096 		return -EFAULT;
4097 	return 0;
4098 }
4099 
4100 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4101 					      char __user *optval,
4102 					      int __user *optlen)
4103 {
4104 	sctp_assoc_t id;
4105 	struct sctp_association *asoc;
4106 	struct list_head *pos;
4107 	int cnt = 0;
4108 
4109 	if (len < sizeof(sctp_assoc_t))
4110 		return -EINVAL;
4111 
4112 	if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4113 		return -EFAULT;
4114 
4115 	/* For UDP-style sockets, id specifies the association to query.  */
4116 	asoc = sctp_id2assoc(sk, id);
4117 	if (!asoc)
4118 		return -EINVAL;
4119 
4120 	list_for_each(pos, &asoc->peer.transport_addr_list) {
4121 		cnt ++;
4122 	}
4123 
4124 	return cnt;
4125 }
4126 
4127 /*
4128  * Old API for getting list of peer addresses. Does not work for 32-bit
4129  * programs running on a 64-bit kernel
4130  */
4131 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4132 					  char __user *optval,
4133 					  int __user *optlen)
4134 {
4135 	struct sctp_association *asoc;
4136 	int cnt = 0;
4137 	struct sctp_getaddrs_old getaddrs;
4138 	struct sctp_transport *from;
4139 	void __user *to;
4140 	union sctp_addr temp;
4141 	struct sctp_sock *sp = sctp_sk(sk);
4142 	int addrlen;
4143 
4144 	if (len < sizeof(struct sctp_getaddrs_old))
4145 		return -EINVAL;
4146 
4147 	len = sizeof(struct sctp_getaddrs_old);
4148 
4149 	if (copy_from_user(&getaddrs, optval, len))
4150 		return -EFAULT;
4151 
4152 	if (getaddrs.addr_num <= 0) return -EINVAL;
4153 
4154 	/* For UDP-style sockets, id specifies the association to query.  */
4155 	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4156 	if (!asoc)
4157 		return -EINVAL;
4158 
4159 	to = (void __user *)getaddrs.addrs;
4160 	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4161 				transports) {
4162 		memcpy(&temp, &from->ipaddr, sizeof(temp));
4163 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4164 		addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4165 		if (copy_to_user(to, &temp, addrlen))
4166 			return -EFAULT;
4167 		to += addrlen ;
4168 		cnt ++;
4169 		if (cnt >= getaddrs.addr_num) break;
4170 	}
4171 	getaddrs.addr_num = cnt;
4172 	if (put_user(len, optlen))
4173 		return -EFAULT;
4174 	if (copy_to_user(optval, &getaddrs, len))
4175 		return -EFAULT;
4176 
4177 	return 0;
4178 }
4179 
4180 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4181 				      char __user *optval, int __user *optlen)
4182 {
4183 	struct sctp_association *asoc;
4184 	int cnt = 0;
4185 	struct sctp_getaddrs getaddrs;
4186 	struct sctp_transport *from;
4187 	void __user *to;
4188 	union sctp_addr temp;
4189 	struct sctp_sock *sp = sctp_sk(sk);
4190 	int addrlen;
4191 	size_t space_left;
4192 	int bytes_copied;
4193 
4194 	if (len < sizeof(struct sctp_getaddrs))
4195 		return -EINVAL;
4196 
4197 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4198 		return -EFAULT;
4199 
4200 	/* For UDP-style sockets, id specifies the association to query.  */
4201 	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4202 	if (!asoc)
4203 		return -EINVAL;
4204 
4205 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4206 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4207 
4208 	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4209 				transports) {
4210 		memcpy(&temp, &from->ipaddr, sizeof(temp));
4211 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4212 		addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4213 		if (space_left < addrlen)
4214 			return -ENOMEM;
4215 		if (copy_to_user(to, &temp, addrlen))
4216 			return -EFAULT;
4217 		to += addrlen;
4218 		cnt++;
4219 		space_left -= addrlen;
4220 	}
4221 
4222 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4223 		return -EFAULT;
4224 	bytes_copied = ((char __user *)to) - optval;
4225 	if (put_user(bytes_copied, optlen))
4226 		return -EFAULT;
4227 
4228 	return 0;
4229 }
4230 
4231 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4232 					       char __user *optval,
4233 					       int __user *optlen)
4234 {
4235 	sctp_assoc_t id;
4236 	struct sctp_bind_addr *bp;
4237 	struct sctp_association *asoc;
4238 	struct sctp_sockaddr_entry *addr;
4239 	int cnt = 0;
4240 
4241 	if (len < sizeof(sctp_assoc_t))
4242 		return -EINVAL;
4243 
4244 	if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4245 		return -EFAULT;
4246 
4247 	/*
4248 	 *  For UDP-style sockets, id specifies the association to query.
4249 	 *  If the id field is set to the value '0' then the locally bound
4250 	 *  addresses are returned without regard to any particular
4251 	 *  association.
4252 	 */
4253 	if (0 == id) {
4254 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4255 	} else {
4256 		asoc = sctp_id2assoc(sk, id);
4257 		if (!asoc)
4258 			return -EINVAL;
4259 		bp = &asoc->base.bind_addr;
4260 	}
4261 
4262 	/* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4263 	 * addresses from the global local address list.
4264 	 */
4265 	if (sctp_list_single_entry(&bp->address_list)) {
4266 		addr = list_entry(bp->address_list.next,
4267 				  struct sctp_sockaddr_entry, list);
4268 		if (sctp_is_any(&addr->a)) {
4269 			rcu_read_lock();
4270 			list_for_each_entry_rcu(addr,
4271 						&sctp_local_addr_list, list) {
4272 				if (!addr->valid)
4273 					continue;
4274 
4275 				if ((PF_INET == sk->sk_family) &&
4276 				    (AF_INET6 == addr->a.sa.sa_family))
4277 					continue;
4278 
4279 				cnt++;
4280 			}
4281 			rcu_read_unlock();
4282 		} else {
4283 			cnt = 1;
4284 		}
4285 		goto done;
4286 	}
4287 
4288 	/* Protection on the bound address list is not needed,
4289 	 * since in the socket option context we hold the socket lock,
4290 	 * so there is no way that the bound address list can change.
4291 	 */
4292 	list_for_each_entry(addr, &bp->address_list, list) {
4293 		cnt ++;
4294 	}
4295 done:
4296 	return cnt;
4297 }
4298 
4299 /* Helper function that copies local addresses to user and returns the number
4300  * of addresses copied.
4301  */
4302 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4303 					int max_addrs, void *to,
4304 					int *bytes_copied)
4305 {
4306 	struct sctp_sockaddr_entry *addr;
4307 	union sctp_addr temp;
4308 	int cnt = 0;
4309 	int addrlen;
4310 
4311 	rcu_read_lock();
4312 	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4313 		if (!addr->valid)
4314 			continue;
4315 
4316 		if ((PF_INET == sk->sk_family) &&
4317 		    (AF_INET6 == addr->a.sa.sa_family))
4318 			continue;
4319 		memcpy(&temp, &addr->a, sizeof(temp));
4320 		if (!temp.v4.sin_port)
4321 			temp.v4.sin_port = htons(port);
4322 
4323 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4324 								&temp);
4325 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4326 		memcpy(to, &temp, addrlen);
4327 
4328 		to += addrlen;
4329 		*bytes_copied += addrlen;
4330 		cnt ++;
4331 		if (cnt >= max_addrs) break;
4332 	}
4333 	rcu_read_unlock();
4334 
4335 	return cnt;
4336 }
4337 
4338 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4339 			    size_t space_left, int *bytes_copied)
4340 {
4341 	struct sctp_sockaddr_entry *addr;
4342 	union sctp_addr temp;
4343 	int cnt = 0;
4344 	int addrlen;
4345 
4346 	rcu_read_lock();
4347 	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4348 		if (!addr->valid)
4349 			continue;
4350 
4351 		if ((PF_INET == sk->sk_family) &&
4352 		    (AF_INET6 == addr->a.sa.sa_family))
4353 			continue;
4354 		memcpy(&temp, &addr->a, sizeof(temp));
4355 		if (!temp.v4.sin_port)
4356 			temp.v4.sin_port = htons(port);
4357 
4358 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4359 								&temp);
4360 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4361 		if (space_left < addrlen) {
4362 			cnt =  -ENOMEM;
4363 			break;
4364 		}
4365 		memcpy(to, &temp, addrlen);
4366 
4367 		to += addrlen;
4368 		cnt ++;
4369 		space_left -= addrlen;
4370 		*bytes_copied += addrlen;
4371 	}
4372 	rcu_read_unlock();
4373 
4374 	return cnt;
4375 }
4376 
4377 /* Old API for getting list of local addresses. Does not work for 32-bit
4378  * programs running on a 64-bit kernel
4379  */
4380 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4381 					   char __user *optval, int __user *optlen)
4382 {
4383 	struct sctp_bind_addr *bp;
4384 	struct sctp_association *asoc;
4385 	int cnt = 0;
4386 	struct sctp_getaddrs_old getaddrs;
4387 	struct sctp_sockaddr_entry *addr;
4388 	void __user *to;
4389 	union sctp_addr temp;
4390 	struct sctp_sock *sp = sctp_sk(sk);
4391 	int addrlen;
4392 	int err = 0;
4393 	void *addrs;
4394 	void *buf;
4395 	int bytes_copied = 0;
4396 
4397 	if (len < sizeof(struct sctp_getaddrs_old))
4398 		return -EINVAL;
4399 
4400 	len = sizeof(struct sctp_getaddrs_old);
4401 	if (copy_from_user(&getaddrs, optval, len))
4402 		return -EFAULT;
4403 
4404 	if (getaddrs.addr_num <= 0) return -EINVAL;
4405 	/*
4406 	 *  For UDP-style sockets, id specifies the association to query.
4407 	 *  If the id field is set to the value '0' then the locally bound
4408 	 *  addresses are returned without regard to any particular
4409 	 *  association.
4410 	 */
4411 	if (0 == getaddrs.assoc_id) {
4412 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4413 	} else {
4414 		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4415 		if (!asoc)
4416 			return -EINVAL;
4417 		bp = &asoc->base.bind_addr;
4418 	}
4419 
4420 	to = getaddrs.addrs;
4421 
4422 	/* Allocate space for a local instance of packed array to hold all
4423 	 * the data.  We store addresses here first and then put write them
4424 	 * to the user in one shot.
4425 	 */
4426 	addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4427 			GFP_KERNEL);
4428 	if (!addrs)
4429 		return -ENOMEM;
4430 
4431 	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4432 	 * addresses from the global local address list.
4433 	 */
4434 	if (sctp_list_single_entry(&bp->address_list)) {
4435 		addr = list_entry(bp->address_list.next,
4436 				  struct sctp_sockaddr_entry, list);
4437 		if (sctp_is_any(&addr->a)) {
4438 			cnt = sctp_copy_laddrs_old(sk, bp->port,
4439 						   getaddrs.addr_num,
4440 						   addrs, &bytes_copied);
4441 			goto copy_getaddrs;
4442 		}
4443 	}
4444 
4445 	buf = addrs;
4446 	/* Protection on the bound address list is not needed since
4447 	 * in the socket option context we hold a socket lock and
4448 	 * thus the bound address list can't change.
4449 	 */
4450 	list_for_each_entry(addr, &bp->address_list, list) {
4451 		memcpy(&temp, &addr->a, sizeof(temp));
4452 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4453 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4454 		memcpy(buf, &temp, addrlen);
4455 		buf += addrlen;
4456 		bytes_copied += addrlen;
4457 		cnt ++;
4458 		if (cnt >= getaddrs.addr_num) break;
4459 	}
4460 
4461 copy_getaddrs:
4462 	/* copy the entire address list into the user provided space */
4463 	if (copy_to_user(to, addrs, bytes_copied)) {
4464 		err = -EFAULT;
4465 		goto error;
4466 	}
4467 
4468 	/* copy the leading structure back to user */
4469 	getaddrs.addr_num = cnt;
4470 	if (copy_to_user(optval, &getaddrs, len))
4471 		err = -EFAULT;
4472 
4473 error:
4474 	kfree(addrs);
4475 	return err;
4476 }
4477 
4478 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4479 				       char __user *optval, int __user *optlen)
4480 {
4481 	struct sctp_bind_addr *bp;
4482 	struct sctp_association *asoc;
4483 	int cnt = 0;
4484 	struct sctp_getaddrs getaddrs;
4485 	struct sctp_sockaddr_entry *addr;
4486 	void __user *to;
4487 	union sctp_addr temp;
4488 	struct sctp_sock *sp = sctp_sk(sk);
4489 	int addrlen;
4490 	int err = 0;
4491 	size_t space_left;
4492 	int bytes_copied = 0;
4493 	void *addrs;
4494 	void *buf;
4495 
4496 	if (len < sizeof(struct sctp_getaddrs))
4497 		return -EINVAL;
4498 
4499 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4500 		return -EFAULT;
4501 
4502 	/*
4503 	 *  For UDP-style sockets, id specifies the association to query.
4504 	 *  If the id field is set to the value '0' then the locally bound
4505 	 *  addresses are returned without regard to any particular
4506 	 *  association.
4507 	 */
4508 	if (0 == getaddrs.assoc_id) {
4509 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4510 	} else {
4511 		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4512 		if (!asoc)
4513 			return -EINVAL;
4514 		bp = &asoc->base.bind_addr;
4515 	}
4516 
4517 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4518 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4519 
4520 	addrs = kmalloc(space_left, GFP_KERNEL);
4521 	if (!addrs)
4522 		return -ENOMEM;
4523 
4524 	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4525 	 * addresses from the global local address list.
4526 	 */
4527 	if (sctp_list_single_entry(&bp->address_list)) {
4528 		addr = list_entry(bp->address_list.next,
4529 				  struct sctp_sockaddr_entry, list);
4530 		if (sctp_is_any(&addr->a)) {
4531 			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4532 						space_left, &bytes_copied);
4533 			if (cnt < 0) {
4534 				err = cnt;
4535 				goto out;
4536 			}
4537 			goto copy_getaddrs;
4538 		}
4539 	}
4540 
4541 	buf = addrs;
4542 	/* Protection on the bound address list is not needed since
4543 	 * in the socket option context we hold a socket lock and
4544 	 * thus the bound address list can't change.
4545 	 */
4546 	list_for_each_entry(addr, &bp->address_list, list) {
4547 		memcpy(&temp, &addr->a, sizeof(temp));
4548 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4549 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4550 		if (space_left < addrlen) {
4551 			err =  -ENOMEM; /*fixme: right error?*/
4552 			goto out;
4553 		}
4554 		memcpy(buf, &temp, addrlen);
4555 		buf += addrlen;
4556 		bytes_copied += addrlen;
4557 		cnt ++;
4558 		space_left -= addrlen;
4559 	}
4560 
4561 copy_getaddrs:
4562 	if (copy_to_user(to, addrs, bytes_copied)) {
4563 		err = -EFAULT;
4564 		goto out;
4565 	}
4566 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4567 		err = -EFAULT;
4568 		goto out;
4569 	}
4570 	if (put_user(bytes_copied, optlen))
4571 		err = -EFAULT;
4572 out:
4573 	kfree(addrs);
4574 	return err;
4575 }
4576 
4577 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4578  *
4579  * Requests that the local SCTP stack use the enclosed peer address as
4580  * the association primary.  The enclosed address must be one of the
4581  * association peer's addresses.
4582  */
4583 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4584 					char __user *optval, int __user *optlen)
4585 {
4586 	struct sctp_prim prim;
4587 	struct sctp_association *asoc;
4588 	struct sctp_sock *sp = sctp_sk(sk);
4589 
4590 	if (len < sizeof(struct sctp_prim))
4591 		return -EINVAL;
4592 
4593 	len = sizeof(struct sctp_prim);
4594 
4595 	if (copy_from_user(&prim, optval, len))
4596 		return -EFAULT;
4597 
4598 	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4599 	if (!asoc)
4600 		return -EINVAL;
4601 
4602 	if (!asoc->peer.primary_path)
4603 		return -ENOTCONN;
4604 
4605 	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4606 		asoc->peer.primary_path->af_specific->sockaddr_len);
4607 
4608 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4609 			(union sctp_addr *)&prim.ssp_addr);
4610 
4611 	if (put_user(len, optlen))
4612 		return -EFAULT;
4613 	if (copy_to_user(optval, &prim, len))
4614 		return -EFAULT;
4615 
4616 	return 0;
4617 }
4618 
4619 /*
4620  * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4621  *
4622  * Requests that the local endpoint set the specified Adaptation Layer
4623  * Indication parameter for all future INIT and INIT-ACK exchanges.
4624  */
4625 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4626 				  char __user *optval, int __user *optlen)
4627 {
4628 	struct sctp_setadaptation adaptation;
4629 
4630 	if (len < sizeof(struct sctp_setadaptation))
4631 		return -EINVAL;
4632 
4633 	len = sizeof(struct sctp_setadaptation);
4634 
4635 	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4636 
4637 	if (put_user(len, optlen))
4638 		return -EFAULT;
4639 	if (copy_to_user(optval, &adaptation, len))
4640 		return -EFAULT;
4641 
4642 	return 0;
4643 }
4644 
4645 /*
4646  *
4647  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4648  *
4649  *   Applications that wish to use the sendto() system call may wish to
4650  *   specify a default set of parameters that would normally be supplied
4651  *   through the inclusion of ancillary data.  This socket option allows
4652  *   such an application to set the default sctp_sndrcvinfo structure.
4653 
4654 
4655  *   The application that wishes to use this socket option simply passes
4656  *   in to this call the sctp_sndrcvinfo structure defined in Section
4657  *   5.2.2) The input parameters accepted by this call include
4658  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4659  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4660  *   to this call if the caller is using the UDP model.
4661  *
4662  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4663  */
4664 static int sctp_getsockopt_default_send_param(struct sock *sk,
4665 					int len, char __user *optval,
4666 					int __user *optlen)
4667 {
4668 	struct sctp_sndrcvinfo info;
4669 	struct sctp_association *asoc;
4670 	struct sctp_sock *sp = sctp_sk(sk);
4671 
4672 	if (len < sizeof(struct sctp_sndrcvinfo))
4673 		return -EINVAL;
4674 
4675 	len = sizeof(struct sctp_sndrcvinfo);
4676 
4677 	if (copy_from_user(&info, optval, len))
4678 		return -EFAULT;
4679 
4680 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4681 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4682 		return -EINVAL;
4683 
4684 	if (asoc) {
4685 		info.sinfo_stream = asoc->default_stream;
4686 		info.sinfo_flags = asoc->default_flags;
4687 		info.sinfo_ppid = asoc->default_ppid;
4688 		info.sinfo_context = asoc->default_context;
4689 		info.sinfo_timetolive = asoc->default_timetolive;
4690 	} else {
4691 		info.sinfo_stream = sp->default_stream;
4692 		info.sinfo_flags = sp->default_flags;
4693 		info.sinfo_ppid = sp->default_ppid;
4694 		info.sinfo_context = sp->default_context;
4695 		info.sinfo_timetolive = sp->default_timetolive;
4696 	}
4697 
4698 	if (put_user(len, optlen))
4699 		return -EFAULT;
4700 	if (copy_to_user(optval, &info, len))
4701 		return -EFAULT;
4702 
4703 	return 0;
4704 }
4705 
4706 /*
4707  *
4708  * 7.1.5 SCTP_NODELAY
4709  *
4710  * Turn on/off any Nagle-like algorithm.  This means that packets are
4711  * generally sent as soon as possible and no unnecessary delays are
4712  * introduced, at the cost of more packets in the network.  Expects an
4713  * integer boolean flag.
4714  */
4715 
4716 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4717 				   char __user *optval, int __user *optlen)
4718 {
4719 	int val;
4720 
4721 	if (len < sizeof(int))
4722 		return -EINVAL;
4723 
4724 	len = sizeof(int);
4725 	val = (sctp_sk(sk)->nodelay == 1);
4726 	if (put_user(len, optlen))
4727 		return -EFAULT;
4728 	if (copy_to_user(optval, &val, len))
4729 		return -EFAULT;
4730 	return 0;
4731 }
4732 
4733 /*
4734  *
4735  * 7.1.1 SCTP_RTOINFO
4736  *
4737  * The protocol parameters used to initialize and bound retransmission
4738  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4739  * and modify these parameters.
4740  * All parameters are time values, in milliseconds.  A value of 0, when
4741  * modifying the parameters, indicates that the current value should not
4742  * be changed.
4743  *
4744  */
4745 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4746 				char __user *optval,
4747 				int __user *optlen) {
4748 	struct sctp_rtoinfo rtoinfo;
4749 	struct sctp_association *asoc;
4750 
4751 	if (len < sizeof (struct sctp_rtoinfo))
4752 		return -EINVAL;
4753 
4754 	len = sizeof(struct sctp_rtoinfo);
4755 
4756 	if (copy_from_user(&rtoinfo, optval, len))
4757 		return -EFAULT;
4758 
4759 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4760 
4761 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4762 		return -EINVAL;
4763 
4764 	/* Values corresponding to the specific association. */
4765 	if (asoc) {
4766 		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4767 		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4768 		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4769 	} else {
4770 		/* Values corresponding to the endpoint. */
4771 		struct sctp_sock *sp = sctp_sk(sk);
4772 
4773 		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4774 		rtoinfo.srto_max = sp->rtoinfo.srto_max;
4775 		rtoinfo.srto_min = sp->rtoinfo.srto_min;
4776 	}
4777 
4778 	if (put_user(len, optlen))
4779 		return -EFAULT;
4780 
4781 	if (copy_to_user(optval, &rtoinfo, len))
4782 		return -EFAULT;
4783 
4784 	return 0;
4785 }
4786 
4787 /*
4788  *
4789  * 7.1.2 SCTP_ASSOCINFO
4790  *
4791  * This option is used to tune the maximum retransmission attempts
4792  * of the association.
4793  * Returns an error if the new association retransmission value is
4794  * greater than the sum of the retransmission value  of the peer.
4795  * See [SCTP] for more information.
4796  *
4797  */
4798 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4799 				     char __user *optval,
4800 				     int __user *optlen)
4801 {
4802 
4803 	struct sctp_assocparams assocparams;
4804 	struct sctp_association *asoc;
4805 	struct list_head *pos;
4806 	int cnt = 0;
4807 
4808 	if (len < sizeof (struct sctp_assocparams))
4809 		return -EINVAL;
4810 
4811 	len = sizeof(struct sctp_assocparams);
4812 
4813 	if (copy_from_user(&assocparams, optval, len))
4814 		return -EFAULT;
4815 
4816 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4817 
4818 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4819 		return -EINVAL;
4820 
4821 	/* Values correspoinding to the specific association */
4822 	if (asoc) {
4823 		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4824 		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4825 		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4826 		assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4827 						* 1000) +
4828 						(asoc->cookie_life.tv_usec
4829 						/ 1000);
4830 
4831 		list_for_each(pos, &asoc->peer.transport_addr_list) {
4832 			cnt ++;
4833 		}
4834 
4835 		assocparams.sasoc_number_peer_destinations = cnt;
4836 	} else {
4837 		/* Values corresponding to the endpoint */
4838 		struct sctp_sock *sp = sctp_sk(sk);
4839 
4840 		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4841 		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4842 		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4843 		assocparams.sasoc_cookie_life =
4844 					sp->assocparams.sasoc_cookie_life;
4845 		assocparams.sasoc_number_peer_destinations =
4846 					sp->assocparams.
4847 					sasoc_number_peer_destinations;
4848 	}
4849 
4850 	if (put_user(len, optlen))
4851 		return -EFAULT;
4852 
4853 	if (copy_to_user(optval, &assocparams, len))
4854 		return -EFAULT;
4855 
4856 	return 0;
4857 }
4858 
4859 /*
4860  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4861  *
4862  * This socket option is a boolean flag which turns on or off mapped V4
4863  * addresses.  If this option is turned on and the socket is type
4864  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4865  * If this option is turned off, then no mapping will be done of V4
4866  * addresses and a user will receive both PF_INET6 and PF_INET type
4867  * addresses on the socket.
4868  */
4869 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4870 				    char __user *optval, int __user *optlen)
4871 {
4872 	int val;
4873 	struct sctp_sock *sp = sctp_sk(sk);
4874 
4875 	if (len < sizeof(int))
4876 		return -EINVAL;
4877 
4878 	len = sizeof(int);
4879 	val = sp->v4mapped;
4880 	if (put_user(len, optlen))
4881 		return -EFAULT;
4882 	if (copy_to_user(optval, &val, len))
4883 		return -EFAULT;
4884 
4885 	return 0;
4886 }
4887 
4888 /*
4889  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
4890  * (chapter and verse is quoted at sctp_setsockopt_context())
4891  */
4892 static int sctp_getsockopt_context(struct sock *sk, int len,
4893 				   char __user *optval, int __user *optlen)
4894 {
4895 	struct sctp_assoc_value params;
4896 	struct sctp_sock *sp;
4897 	struct sctp_association *asoc;
4898 
4899 	if (len < sizeof(struct sctp_assoc_value))
4900 		return -EINVAL;
4901 
4902 	len = sizeof(struct sctp_assoc_value);
4903 
4904 	if (copy_from_user(&params, optval, len))
4905 		return -EFAULT;
4906 
4907 	sp = sctp_sk(sk);
4908 
4909 	if (params.assoc_id != 0) {
4910 		asoc = sctp_id2assoc(sk, params.assoc_id);
4911 		if (!asoc)
4912 			return -EINVAL;
4913 		params.assoc_value = asoc->default_rcv_context;
4914 	} else {
4915 		params.assoc_value = sp->default_rcv_context;
4916 	}
4917 
4918 	if (put_user(len, optlen))
4919 		return -EFAULT;
4920 	if (copy_to_user(optval, &params, len))
4921 		return -EFAULT;
4922 
4923 	return 0;
4924 }
4925 
4926 /*
4927  * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4928  *
4929  * This socket option specifies the maximum size to put in any outgoing
4930  * SCTP chunk.  If a message is larger than this size it will be
4931  * fragmented by SCTP into the specified size.  Note that the underlying
4932  * SCTP implementation may fragment into smaller sized chunks when the
4933  * PMTU of the underlying association is smaller than the value set by
4934  * the user.
4935  */
4936 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4937 				  char __user *optval, int __user *optlen)
4938 {
4939 	int val;
4940 
4941 	if (len < sizeof(int))
4942 		return -EINVAL;
4943 
4944 	len = sizeof(int);
4945 
4946 	val = sctp_sk(sk)->user_frag;
4947 	if (put_user(len, optlen))
4948 		return -EFAULT;
4949 	if (copy_to_user(optval, &val, len))
4950 		return -EFAULT;
4951 
4952 	return 0;
4953 }
4954 
4955 /*
4956  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4957  * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4958  */
4959 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4960 					       char __user *optval, int __user *optlen)
4961 {
4962 	int val;
4963 
4964 	if (len < sizeof(int))
4965 		return -EINVAL;
4966 
4967 	len = sizeof(int);
4968 
4969 	val = sctp_sk(sk)->frag_interleave;
4970 	if (put_user(len, optlen))
4971 		return -EFAULT;
4972 	if (copy_to_user(optval, &val, len))
4973 		return -EFAULT;
4974 
4975 	return 0;
4976 }
4977 
4978 /*
4979  * 7.1.25.  Set or Get the sctp partial delivery point
4980  * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4981  */
4982 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4983 						  char __user *optval,
4984 						  int __user *optlen)
4985 {
4986 	u32 val;
4987 
4988 	if (len < sizeof(u32))
4989 		return -EINVAL;
4990 
4991 	len = sizeof(u32);
4992 
4993 	val = sctp_sk(sk)->pd_point;
4994 	if (put_user(len, optlen))
4995 		return -EFAULT;
4996 	if (copy_to_user(optval, &val, len))
4997 		return -EFAULT;
4998 
4999 	return -ENOTSUPP;
5000 }
5001 
5002 /*
5003  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5004  * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5005  */
5006 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5007 				    char __user *optval,
5008 				    int __user *optlen)
5009 {
5010 	struct sctp_assoc_value params;
5011 	struct sctp_sock *sp;
5012 	struct sctp_association *asoc;
5013 
5014 	if (len < sizeof(int))
5015 		return -EINVAL;
5016 
5017 	if (len == sizeof(int)) {
5018 		printk(KERN_WARNING
5019 		   "SCTP: Use of int in max_burst socket option deprecated\n");
5020 		printk(KERN_WARNING
5021 		   "SCTP: Use struct sctp_assoc_value instead\n");
5022 		params.assoc_id = 0;
5023 	} else if (len == sizeof (struct sctp_assoc_value)) {
5024 		if (copy_from_user(&params, optval, len))
5025 			return -EFAULT;
5026 	} else
5027 		return -EINVAL;
5028 
5029 	sp = sctp_sk(sk);
5030 
5031 	if (params.assoc_id != 0) {
5032 		asoc = sctp_id2assoc(sk, params.assoc_id);
5033 		if (!asoc)
5034 			return -EINVAL;
5035 		params.assoc_value = asoc->max_burst;
5036 	} else
5037 		params.assoc_value = sp->max_burst;
5038 
5039 	if (len == sizeof(int)) {
5040 		if (copy_to_user(optval, &params.assoc_value, len))
5041 			return -EFAULT;
5042 	} else {
5043 		if (copy_to_user(optval, &params, len))
5044 			return -EFAULT;
5045 	}
5046 
5047 	return 0;
5048 
5049 }
5050 
5051 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5052 				    char __user *optval, int __user *optlen)
5053 {
5054 	struct sctp_hmac_algo_param *hmacs;
5055 	__u16 param_len;
5056 
5057 	hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5058 	param_len = ntohs(hmacs->param_hdr.length);
5059 
5060 	if (len < param_len)
5061 		return -EINVAL;
5062 	if (put_user(len, optlen))
5063 		return -EFAULT;
5064 	if (copy_to_user(optval, hmacs->hmac_ids, len))
5065 		return -EFAULT;
5066 
5067 	return 0;
5068 }
5069 
5070 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5071 				    char __user *optval, int __user *optlen)
5072 {
5073 	struct sctp_authkeyid val;
5074 	struct sctp_association *asoc;
5075 
5076 	if (len < sizeof(struct sctp_authkeyid))
5077 		return -EINVAL;
5078 	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5079 		return -EFAULT;
5080 
5081 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5082 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5083 		return -EINVAL;
5084 
5085 	if (asoc)
5086 		val.scact_keynumber = asoc->active_key_id;
5087 	else
5088 		val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5089 
5090 	return 0;
5091 }
5092 
5093 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5094 				    char __user *optval, int __user *optlen)
5095 {
5096 	struct sctp_authchunks __user *p = (void __user *)optval;
5097 	struct sctp_authchunks val;
5098 	struct sctp_association *asoc;
5099 	struct sctp_chunks_param *ch;
5100 	u32    num_chunks;
5101 	char __user *to;
5102 
5103 	if (len <= sizeof(struct sctp_authchunks))
5104 		return -EINVAL;
5105 
5106 	if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5107 		return -EFAULT;
5108 
5109 	to = p->gauth_chunks;
5110 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5111 	if (!asoc)
5112 		return -EINVAL;
5113 
5114 	ch = asoc->peer.peer_chunks;
5115 
5116 	/* See if the user provided enough room for all the data */
5117 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5118 	if (len < num_chunks)
5119 		return -EINVAL;
5120 
5121 	len = num_chunks;
5122 	if (put_user(len, optlen))
5123 		return -EFAULT;
5124 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5125 		return -EFAULT;
5126 	if (copy_to_user(to, ch->chunks, len))
5127 		return -EFAULT;
5128 
5129 	return 0;
5130 }
5131 
5132 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5133 				    char __user *optval, int __user *optlen)
5134 {
5135 	struct sctp_authchunks __user *p = (void __user *)optval;
5136 	struct sctp_authchunks val;
5137 	struct sctp_association *asoc;
5138 	struct sctp_chunks_param *ch;
5139 	u32    num_chunks;
5140 	char __user *to;
5141 
5142 	if (len <= sizeof(struct sctp_authchunks))
5143 		return -EINVAL;
5144 
5145 	if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5146 		return -EFAULT;
5147 
5148 	to = p->gauth_chunks;
5149 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5150 	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5151 		return -EINVAL;
5152 
5153 	if (asoc)
5154 		ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5155 	else
5156 		ch = sctp_sk(sk)->ep->auth_chunk_list;
5157 
5158 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5159 	if (len < num_chunks)
5160 		return -EINVAL;
5161 
5162 	len = num_chunks;
5163 	if (put_user(len, optlen))
5164 		return -EFAULT;
5165 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5166 		return -EFAULT;
5167 	if (copy_to_user(to, ch->chunks, len))
5168 		return -EFAULT;
5169 
5170 	return 0;
5171 }
5172 
5173 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5174 				char __user *optval, int __user *optlen)
5175 {
5176 	int retval = 0;
5177 	int len;
5178 
5179 	SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5180 			  sk, optname);
5181 
5182 	/* I can hardly begin to describe how wrong this is.  This is
5183 	 * so broken as to be worse than useless.  The API draft
5184 	 * REALLY is NOT helpful here...  I am not convinced that the
5185 	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5186 	 * are at all well-founded.
5187 	 */
5188 	if (level != SOL_SCTP) {
5189 		struct sctp_af *af = sctp_sk(sk)->pf->af;
5190 
5191 		retval = af->getsockopt(sk, level, optname, optval, optlen);
5192 		return retval;
5193 	}
5194 
5195 	if (get_user(len, optlen))
5196 		return -EFAULT;
5197 
5198 	sctp_lock_sock(sk);
5199 
5200 	switch (optname) {
5201 	case SCTP_STATUS:
5202 		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5203 		break;
5204 	case SCTP_DISABLE_FRAGMENTS:
5205 		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5206 							   optlen);
5207 		break;
5208 	case SCTP_EVENTS:
5209 		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5210 		break;
5211 	case SCTP_AUTOCLOSE:
5212 		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5213 		break;
5214 	case SCTP_SOCKOPT_PEELOFF:
5215 		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5216 		break;
5217 	case SCTP_PEER_ADDR_PARAMS:
5218 		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5219 							  optlen);
5220 		break;
5221 	case SCTP_DELAYED_ACK_TIME:
5222 		retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
5223 							  optlen);
5224 		break;
5225 	case SCTP_INITMSG:
5226 		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5227 		break;
5228 	case SCTP_GET_PEER_ADDRS_NUM_OLD:
5229 		retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5230 							    optlen);
5231 		break;
5232 	case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5233 		retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5234 							     optlen);
5235 		break;
5236 	case SCTP_GET_PEER_ADDRS_OLD:
5237 		retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5238 							optlen);
5239 		break;
5240 	case SCTP_GET_LOCAL_ADDRS_OLD:
5241 		retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5242 							 optlen);
5243 		break;
5244 	case SCTP_GET_PEER_ADDRS:
5245 		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5246 						    optlen);
5247 		break;
5248 	case SCTP_GET_LOCAL_ADDRS:
5249 		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5250 						     optlen);
5251 		break;
5252 	case SCTP_DEFAULT_SEND_PARAM:
5253 		retval = sctp_getsockopt_default_send_param(sk, len,
5254 							    optval, optlen);
5255 		break;
5256 	case SCTP_PRIMARY_ADDR:
5257 		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5258 		break;
5259 	case SCTP_NODELAY:
5260 		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5261 		break;
5262 	case SCTP_RTOINFO:
5263 		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5264 		break;
5265 	case SCTP_ASSOCINFO:
5266 		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5267 		break;
5268 	case SCTP_I_WANT_MAPPED_V4_ADDR:
5269 		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5270 		break;
5271 	case SCTP_MAXSEG:
5272 		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5273 		break;
5274 	case SCTP_GET_PEER_ADDR_INFO:
5275 		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5276 							optlen);
5277 		break;
5278 	case SCTP_ADAPTATION_LAYER:
5279 		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5280 							optlen);
5281 		break;
5282 	case SCTP_CONTEXT:
5283 		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5284 		break;
5285 	case SCTP_FRAGMENT_INTERLEAVE:
5286 		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5287 							     optlen);
5288 		break;
5289 	case SCTP_PARTIAL_DELIVERY_POINT:
5290 		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5291 								optlen);
5292 		break;
5293 	case SCTP_MAX_BURST:
5294 		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5295 		break;
5296 	case SCTP_AUTH_KEY:
5297 	case SCTP_AUTH_CHUNK:
5298 	case SCTP_AUTH_DELETE_KEY:
5299 		retval = -EOPNOTSUPP;
5300 		break;
5301 	case SCTP_HMAC_IDENT:
5302 		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5303 		break;
5304 	case SCTP_AUTH_ACTIVE_KEY:
5305 		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5306 		break;
5307 	case SCTP_PEER_AUTH_CHUNKS:
5308 		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5309 							optlen);
5310 		break;
5311 	case SCTP_LOCAL_AUTH_CHUNKS:
5312 		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5313 							optlen);
5314 		break;
5315 	default:
5316 		retval = -ENOPROTOOPT;
5317 		break;
5318 	}
5319 
5320 	sctp_release_sock(sk);
5321 	return retval;
5322 }
5323 
5324 static void sctp_hash(struct sock *sk)
5325 {
5326 	/* STUB */
5327 }
5328 
5329 static void sctp_unhash(struct sock *sk)
5330 {
5331 	/* STUB */
5332 }
5333 
5334 /* Check if port is acceptable.  Possibly find first available port.
5335  *
5336  * The port hash table (contained in the 'global' SCTP protocol storage
5337  * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5338  * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5339  * list (the list number is the port number hashed out, so as you
5340  * would expect from a hash function, all the ports in a given list have
5341  * such a number that hashes out to the same list number; you were
5342  * expecting that, right?); so each list has a set of ports, with a
5343  * link to the socket (struct sock) that uses it, the port number and
5344  * a fastreuse flag (FIXME: NPI ipg).
5345  */
5346 static struct sctp_bind_bucket *sctp_bucket_create(
5347 	struct sctp_bind_hashbucket *head, unsigned short snum);
5348 
5349 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5350 {
5351 	struct sctp_bind_hashbucket *head; /* hash list */
5352 	struct sctp_bind_bucket *pp; /* hash list port iterator */
5353 	struct hlist_node *node;
5354 	unsigned short snum;
5355 	int ret;
5356 
5357 	snum = ntohs(addr->v4.sin_port);
5358 
5359 	SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5360 	sctp_local_bh_disable();
5361 
5362 	if (snum == 0) {
5363 		/* Search for an available port. */
5364 		int low, high, remaining, index;
5365 		unsigned int rover;
5366 
5367 		inet_get_local_port_range(&low, &high);
5368 		remaining = (high - low) + 1;
5369 		rover = net_random() % remaining + low;
5370 
5371 		do {
5372 			rover++;
5373 			if ((rover < low) || (rover > high))
5374 				rover = low;
5375 			index = sctp_phashfn(rover);
5376 			head = &sctp_port_hashtable[index];
5377 			sctp_spin_lock(&head->lock);
5378 			sctp_for_each_hentry(pp, node, &head->chain)
5379 				if (pp->port == rover)
5380 					goto next;
5381 			break;
5382 		next:
5383 			sctp_spin_unlock(&head->lock);
5384 		} while (--remaining > 0);
5385 
5386 		/* Exhausted local port range during search? */
5387 		ret = 1;
5388 		if (remaining <= 0)
5389 			goto fail;
5390 
5391 		/* OK, here is the one we will use.  HEAD (the port
5392 		 * hash table list entry) is non-NULL and we hold it's
5393 		 * mutex.
5394 		 */
5395 		snum = rover;
5396 	} else {
5397 		/* We are given an specific port number; we verify
5398 		 * that it is not being used. If it is used, we will
5399 		 * exahust the search in the hash list corresponding
5400 		 * to the port number (snum) - we detect that with the
5401 		 * port iterator, pp being NULL.
5402 		 */
5403 		head = &sctp_port_hashtable[sctp_phashfn(snum)];
5404 		sctp_spin_lock(&head->lock);
5405 		sctp_for_each_hentry(pp, node, &head->chain) {
5406 			if (pp->port == snum)
5407 				goto pp_found;
5408 		}
5409 	}
5410 	pp = NULL;
5411 	goto pp_not_found;
5412 pp_found:
5413 	if (!hlist_empty(&pp->owner)) {
5414 		/* We had a port hash table hit - there is an
5415 		 * available port (pp != NULL) and it is being
5416 		 * used by other socket (pp->owner not empty); that other
5417 		 * socket is going to be sk2.
5418 		 */
5419 		int reuse = sk->sk_reuse;
5420 		struct sock *sk2;
5421 		struct hlist_node *node;
5422 
5423 		SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5424 		if (pp->fastreuse && sk->sk_reuse &&
5425 			sk->sk_state != SCTP_SS_LISTENING)
5426 			goto success;
5427 
5428 		/* Run through the list of sockets bound to the port
5429 		 * (pp->port) [via the pointers bind_next and
5430 		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5431 		 * we get the endpoint they describe and run through
5432 		 * the endpoint's list of IP (v4 or v6) addresses,
5433 		 * comparing each of the addresses with the address of
5434 		 * the socket sk. If we find a match, then that means
5435 		 * that this port/socket (sk) combination are already
5436 		 * in an endpoint.
5437 		 */
5438 		sk_for_each_bound(sk2, node, &pp->owner) {
5439 			struct sctp_endpoint *ep2;
5440 			ep2 = sctp_sk(sk2)->ep;
5441 
5442 			if (reuse && sk2->sk_reuse &&
5443 			    sk2->sk_state != SCTP_SS_LISTENING)
5444 				continue;
5445 
5446 			if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5447 						 sctp_sk(sk))) {
5448 				ret = (long)sk2;
5449 				goto fail_unlock;
5450 			}
5451 		}
5452 		SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5453 	}
5454 pp_not_found:
5455 	/* If there was a hash table miss, create a new port.  */
5456 	ret = 1;
5457 	if (!pp && !(pp = sctp_bucket_create(head, snum)))
5458 		goto fail_unlock;
5459 
5460 	/* In either case (hit or miss), make sure fastreuse is 1 only
5461 	 * if sk->sk_reuse is too (that is, if the caller requested
5462 	 * SO_REUSEADDR on this socket -sk-).
5463 	 */
5464 	if (hlist_empty(&pp->owner)) {
5465 		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5466 			pp->fastreuse = 1;
5467 		else
5468 			pp->fastreuse = 0;
5469 	} else if (pp->fastreuse &&
5470 		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5471 		pp->fastreuse = 0;
5472 
5473 	/* We are set, so fill up all the data in the hash table
5474 	 * entry, tie the socket list information with the rest of the
5475 	 * sockets FIXME: Blurry, NPI (ipg).
5476 	 */
5477 success:
5478 	if (!sctp_sk(sk)->bind_hash) {
5479 		inet_sk(sk)->num = snum;
5480 		sk_add_bind_node(sk, &pp->owner);
5481 		sctp_sk(sk)->bind_hash = pp;
5482 	}
5483 	ret = 0;
5484 
5485 fail_unlock:
5486 	sctp_spin_unlock(&head->lock);
5487 
5488 fail:
5489 	sctp_local_bh_enable();
5490 	return ret;
5491 }
5492 
5493 /* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5494  * port is requested.
5495  */
5496 static int sctp_get_port(struct sock *sk, unsigned short snum)
5497 {
5498 	long ret;
5499 	union sctp_addr addr;
5500 	struct sctp_af *af = sctp_sk(sk)->pf->af;
5501 
5502 	/* Set up a dummy address struct from the sk. */
5503 	af->from_sk(&addr, sk);
5504 	addr.v4.sin_port = htons(snum);
5505 
5506 	/* Note: sk->sk_num gets filled in if ephemeral port request. */
5507 	ret = sctp_get_port_local(sk, &addr);
5508 
5509 	return (ret ? 1 : 0);
5510 }
5511 
5512 /*
5513  * 3.1.3 listen() - UDP Style Syntax
5514  *
5515  *   By default, new associations are not accepted for UDP style sockets.
5516  *   An application uses listen() to mark a socket as being able to
5517  *   accept new associations.
5518  */
5519 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5520 {
5521 	struct sctp_sock *sp = sctp_sk(sk);
5522 	struct sctp_endpoint *ep = sp->ep;
5523 
5524 	/* Only UDP style sockets that are not peeled off are allowed to
5525 	 * listen().
5526 	 */
5527 	if (!sctp_style(sk, UDP))
5528 		return -EINVAL;
5529 
5530 	/* If backlog is zero, disable listening. */
5531 	if (!backlog) {
5532 		if (sctp_sstate(sk, CLOSED))
5533 			return 0;
5534 
5535 		sctp_unhash_endpoint(ep);
5536 		sk->sk_state = SCTP_SS_CLOSED;
5537 		return 0;
5538 	}
5539 
5540 	/* Return if we are already listening. */
5541 	if (sctp_sstate(sk, LISTENING))
5542 		return 0;
5543 
5544 	/*
5545 	 * If a bind() or sctp_bindx() is not called prior to a listen()
5546 	 * call that allows new associations to be accepted, the system
5547 	 * picks an ephemeral port and will choose an address set equivalent
5548 	 * to binding with a wildcard address.
5549 	 *
5550 	 * This is not currently spelled out in the SCTP sockets
5551 	 * extensions draft, but follows the practice as seen in TCP
5552 	 * sockets.
5553 	 *
5554 	 * Additionally, turn off fastreuse flag since we are not listening
5555 	 */
5556 	sk->sk_state = SCTP_SS_LISTENING;
5557 	if (!ep->base.bind_addr.port) {
5558 		if (sctp_autobind(sk))
5559 			return -EAGAIN;
5560 	} else
5561 		sctp_sk(sk)->bind_hash->fastreuse = 0;
5562 
5563 	sctp_hash_endpoint(ep);
5564 	return 0;
5565 }
5566 
5567 /*
5568  * 4.1.3 listen() - TCP Style Syntax
5569  *
5570  *   Applications uses listen() to ready the SCTP endpoint for accepting
5571  *   inbound associations.
5572  */
5573 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5574 {
5575 	struct sctp_sock *sp = sctp_sk(sk);
5576 	struct sctp_endpoint *ep = sp->ep;
5577 
5578 	/* If backlog is zero, disable listening. */
5579 	if (!backlog) {
5580 		if (sctp_sstate(sk, CLOSED))
5581 			return 0;
5582 
5583 		sctp_unhash_endpoint(ep);
5584 		sk->sk_state = SCTP_SS_CLOSED;
5585 		return 0;
5586 	}
5587 
5588 	if (sctp_sstate(sk, LISTENING))
5589 		return 0;
5590 
5591 	/*
5592 	 * If a bind() or sctp_bindx() is not called prior to a listen()
5593 	 * call that allows new associations to be accepted, the system
5594 	 * picks an ephemeral port and will choose an address set equivalent
5595 	 * to binding with a wildcard address.
5596 	 *
5597 	 * This is not currently spelled out in the SCTP sockets
5598 	 * extensions draft, but follows the practice as seen in TCP
5599 	 * sockets.
5600 	 */
5601 	sk->sk_state = SCTP_SS_LISTENING;
5602 	if (!ep->base.bind_addr.port) {
5603 		if (sctp_autobind(sk))
5604 			return -EAGAIN;
5605 	} else
5606 		sctp_sk(sk)->bind_hash->fastreuse = 0;
5607 
5608 	sk->sk_max_ack_backlog = backlog;
5609 	sctp_hash_endpoint(ep);
5610 	return 0;
5611 }
5612 
5613 /*
5614  *  Move a socket to LISTENING state.
5615  */
5616 int sctp_inet_listen(struct socket *sock, int backlog)
5617 {
5618 	struct sock *sk = sock->sk;
5619 	struct crypto_hash *tfm = NULL;
5620 	int err = -EINVAL;
5621 
5622 	if (unlikely(backlog < 0))
5623 		goto out;
5624 
5625 	sctp_lock_sock(sk);
5626 
5627 	if (sock->state != SS_UNCONNECTED)
5628 		goto out;
5629 
5630 	/* Allocate HMAC for generating cookie. */
5631 	if (sctp_hmac_alg) {
5632 		tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5633 		if (IS_ERR(tfm)) {
5634 			if (net_ratelimit()) {
5635 				printk(KERN_INFO
5636 				       "SCTP: failed to load transform for %s: %ld\n",
5637 					sctp_hmac_alg, PTR_ERR(tfm));
5638 			}
5639 			err = -ENOSYS;
5640 			goto out;
5641 		}
5642 	}
5643 
5644 	switch (sock->type) {
5645 	case SOCK_SEQPACKET:
5646 		err = sctp_seqpacket_listen(sk, backlog);
5647 		break;
5648 	case SOCK_STREAM:
5649 		err = sctp_stream_listen(sk, backlog);
5650 		break;
5651 	default:
5652 		break;
5653 	}
5654 
5655 	if (err)
5656 		goto cleanup;
5657 
5658 	/* Store away the transform reference. */
5659 	sctp_sk(sk)->hmac = tfm;
5660 out:
5661 	sctp_release_sock(sk);
5662 	return err;
5663 cleanup:
5664 	crypto_free_hash(tfm);
5665 	goto out;
5666 }
5667 
5668 /*
5669  * This function is done by modeling the current datagram_poll() and the
5670  * tcp_poll().  Note that, based on these implementations, we don't
5671  * lock the socket in this function, even though it seems that,
5672  * ideally, locking or some other mechanisms can be used to ensure
5673  * the integrity of the counters (sndbuf and wmem_alloc) used
5674  * in this place.  We assume that we don't need locks either until proven
5675  * otherwise.
5676  *
5677  * Another thing to note is that we include the Async I/O support
5678  * here, again, by modeling the current TCP/UDP code.  We don't have
5679  * a good way to test with it yet.
5680  */
5681 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5682 {
5683 	struct sock *sk = sock->sk;
5684 	struct sctp_sock *sp = sctp_sk(sk);
5685 	unsigned int mask;
5686 
5687 	poll_wait(file, sk->sk_sleep, wait);
5688 
5689 	/* A TCP-style listening socket becomes readable when the accept queue
5690 	 * is not empty.
5691 	 */
5692 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5693 		return (!list_empty(&sp->ep->asocs)) ?
5694 			(POLLIN | POLLRDNORM) : 0;
5695 
5696 	mask = 0;
5697 
5698 	/* Is there any exceptional events?  */
5699 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5700 		mask |= POLLERR;
5701 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5702 		mask |= POLLRDHUP;
5703 	if (sk->sk_shutdown == SHUTDOWN_MASK)
5704 		mask |= POLLHUP;
5705 
5706 	/* Is it readable?  Reconsider this code with TCP-style support.  */
5707 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
5708 	    (sk->sk_shutdown & RCV_SHUTDOWN))
5709 		mask |= POLLIN | POLLRDNORM;
5710 
5711 	/* The association is either gone or not ready.  */
5712 	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5713 		return mask;
5714 
5715 	/* Is it writable?  */
5716 	if (sctp_writeable(sk)) {
5717 		mask |= POLLOUT | POLLWRNORM;
5718 	} else {
5719 		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5720 		/*
5721 		 * Since the socket is not locked, the buffer
5722 		 * might be made available after the writeable check and
5723 		 * before the bit is set.  This could cause a lost I/O
5724 		 * signal.  tcp_poll() has a race breaker for this race
5725 		 * condition.  Based on their implementation, we put
5726 		 * in the following code to cover it as well.
5727 		 */
5728 		if (sctp_writeable(sk))
5729 			mask |= POLLOUT | POLLWRNORM;
5730 	}
5731 	return mask;
5732 }
5733 
5734 /********************************************************************
5735  * 2nd Level Abstractions
5736  ********************************************************************/
5737 
5738 static struct sctp_bind_bucket *sctp_bucket_create(
5739 	struct sctp_bind_hashbucket *head, unsigned short snum)
5740 {
5741 	struct sctp_bind_bucket *pp;
5742 
5743 	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5744 	if (pp) {
5745 		SCTP_DBG_OBJCNT_INC(bind_bucket);
5746 		pp->port = snum;
5747 		pp->fastreuse = 0;
5748 		INIT_HLIST_HEAD(&pp->owner);
5749 		hlist_add_head(&pp->node, &head->chain);
5750 	}
5751 	return pp;
5752 }
5753 
5754 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5755 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5756 {
5757 	if (pp && hlist_empty(&pp->owner)) {
5758 		__hlist_del(&pp->node);
5759 		kmem_cache_free(sctp_bucket_cachep, pp);
5760 		SCTP_DBG_OBJCNT_DEC(bind_bucket);
5761 	}
5762 }
5763 
5764 /* Release this socket's reference to a local port.  */
5765 static inline void __sctp_put_port(struct sock *sk)
5766 {
5767 	struct sctp_bind_hashbucket *head =
5768 		&sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5769 	struct sctp_bind_bucket *pp;
5770 
5771 	sctp_spin_lock(&head->lock);
5772 	pp = sctp_sk(sk)->bind_hash;
5773 	__sk_del_bind_node(sk);
5774 	sctp_sk(sk)->bind_hash = NULL;
5775 	inet_sk(sk)->num = 0;
5776 	sctp_bucket_destroy(pp);
5777 	sctp_spin_unlock(&head->lock);
5778 }
5779 
5780 void sctp_put_port(struct sock *sk)
5781 {
5782 	sctp_local_bh_disable();
5783 	__sctp_put_port(sk);
5784 	sctp_local_bh_enable();
5785 }
5786 
5787 /*
5788  * The system picks an ephemeral port and choose an address set equivalent
5789  * to binding with a wildcard address.
5790  * One of those addresses will be the primary address for the association.
5791  * This automatically enables the multihoming capability of SCTP.
5792  */
5793 static int sctp_autobind(struct sock *sk)
5794 {
5795 	union sctp_addr autoaddr;
5796 	struct sctp_af *af;
5797 	__be16 port;
5798 
5799 	/* Initialize a local sockaddr structure to INADDR_ANY. */
5800 	af = sctp_sk(sk)->pf->af;
5801 
5802 	port = htons(inet_sk(sk)->num);
5803 	af->inaddr_any(&autoaddr, port);
5804 
5805 	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5806 }
5807 
5808 /* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
5809  *
5810  * From RFC 2292
5811  * 4.2 The cmsghdr Structure *
5812  *
5813  * When ancillary data is sent or received, any number of ancillary data
5814  * objects can be specified by the msg_control and msg_controllen members of
5815  * the msghdr structure, because each object is preceded by
5816  * a cmsghdr structure defining the object's length (the cmsg_len member).
5817  * Historically Berkeley-derived implementations have passed only one object
5818  * at a time, but this API allows multiple objects to be
5819  * passed in a single call to sendmsg() or recvmsg(). The following example
5820  * shows two ancillary data objects in a control buffer.
5821  *
5822  *   |<--------------------------- msg_controllen -------------------------->|
5823  *   |                                                                       |
5824  *
5825  *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
5826  *
5827  *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5828  *   |                                   |                                   |
5829  *
5830  *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
5831  *
5832  *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
5833  *   |                                |  |                                |  |
5834  *
5835  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5836  *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
5837  *
5838  *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
5839  *
5840  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5841  *    ^
5842  *    |
5843  *
5844  * msg_control
5845  * points here
5846  */
5847 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5848 				  sctp_cmsgs_t *cmsgs)
5849 {
5850 	struct cmsghdr *cmsg;
5851 	struct msghdr *my_msg = (struct msghdr *)msg;
5852 
5853 	for (cmsg = CMSG_FIRSTHDR(msg);
5854 	     cmsg != NULL;
5855 	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5856 		if (!CMSG_OK(my_msg, cmsg))
5857 			return -EINVAL;
5858 
5859 		/* Should we parse this header or ignore?  */
5860 		if (cmsg->cmsg_level != IPPROTO_SCTP)
5861 			continue;
5862 
5863 		/* Strictly check lengths following example in SCM code.  */
5864 		switch (cmsg->cmsg_type) {
5865 		case SCTP_INIT:
5866 			/* SCTP Socket API Extension
5867 			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5868 			 *
5869 			 * This cmsghdr structure provides information for
5870 			 * initializing new SCTP associations with sendmsg().
5871 			 * The SCTP_INITMSG socket option uses this same data
5872 			 * structure.  This structure is not used for
5873 			 * recvmsg().
5874 			 *
5875 			 * cmsg_level    cmsg_type      cmsg_data[]
5876 			 * ------------  ------------   ----------------------
5877 			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
5878 			 */
5879 			if (cmsg->cmsg_len !=
5880 			    CMSG_LEN(sizeof(struct sctp_initmsg)))
5881 				return -EINVAL;
5882 			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5883 			break;
5884 
5885 		case SCTP_SNDRCV:
5886 			/* SCTP Socket API Extension
5887 			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5888 			 *
5889 			 * This cmsghdr structure specifies SCTP options for
5890 			 * sendmsg() and describes SCTP header information
5891 			 * about a received message through recvmsg().
5892 			 *
5893 			 * cmsg_level    cmsg_type      cmsg_data[]
5894 			 * ------------  ------------   ----------------------
5895 			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
5896 			 */
5897 			if (cmsg->cmsg_len !=
5898 			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5899 				return -EINVAL;
5900 
5901 			cmsgs->info =
5902 				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5903 
5904 			/* Minimally, validate the sinfo_flags. */
5905 			if (cmsgs->info->sinfo_flags &
5906 			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5907 			      SCTP_ABORT | SCTP_EOF))
5908 				return -EINVAL;
5909 			break;
5910 
5911 		default:
5912 			return -EINVAL;
5913 		}
5914 	}
5915 	return 0;
5916 }
5917 
5918 /*
5919  * Wait for a packet..
5920  * Note: This function is the same function as in core/datagram.c
5921  * with a few modifications to make lksctp work.
5922  */
5923 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5924 {
5925 	int error;
5926 	DEFINE_WAIT(wait);
5927 
5928 	prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5929 
5930 	/* Socket errors? */
5931 	error = sock_error(sk);
5932 	if (error)
5933 		goto out;
5934 
5935 	if (!skb_queue_empty(&sk->sk_receive_queue))
5936 		goto ready;
5937 
5938 	/* Socket shut down?  */
5939 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5940 		goto out;
5941 
5942 	/* Sequenced packets can come disconnected.  If so we report the
5943 	 * problem.
5944 	 */
5945 	error = -ENOTCONN;
5946 
5947 	/* Is there a good reason to think that we may receive some data?  */
5948 	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5949 		goto out;
5950 
5951 	/* Handle signals.  */
5952 	if (signal_pending(current))
5953 		goto interrupted;
5954 
5955 	/* Let another process have a go.  Since we are going to sleep
5956 	 * anyway.  Note: This may cause odd behaviors if the message
5957 	 * does not fit in the user's buffer, but this seems to be the
5958 	 * only way to honor MSG_DONTWAIT realistically.
5959 	 */
5960 	sctp_release_sock(sk);
5961 	*timeo_p = schedule_timeout(*timeo_p);
5962 	sctp_lock_sock(sk);
5963 
5964 ready:
5965 	finish_wait(sk->sk_sleep, &wait);
5966 	return 0;
5967 
5968 interrupted:
5969 	error = sock_intr_errno(*timeo_p);
5970 
5971 out:
5972 	finish_wait(sk->sk_sleep, &wait);
5973 	*err = error;
5974 	return error;
5975 }
5976 
5977 /* Receive a datagram.
5978  * Note: This is pretty much the same routine as in core/datagram.c
5979  * with a few changes to make lksctp work.
5980  */
5981 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5982 					      int noblock, int *err)
5983 {
5984 	int error;
5985 	struct sk_buff *skb;
5986 	long timeo;
5987 
5988 	timeo = sock_rcvtimeo(sk, noblock);
5989 
5990 	SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5991 			  timeo, MAX_SCHEDULE_TIMEOUT);
5992 
5993 	do {
5994 		/* Again only user level code calls this function,
5995 		 * so nothing interrupt level
5996 		 * will suddenly eat the receive_queue.
5997 		 *
5998 		 *  Look at current nfs client by the way...
5999 		 *  However, this function was corrent in any case. 8)
6000 		 */
6001 		if (flags & MSG_PEEK) {
6002 			spin_lock_bh(&sk->sk_receive_queue.lock);
6003 			skb = skb_peek(&sk->sk_receive_queue);
6004 			if (skb)
6005 				atomic_inc(&skb->users);
6006 			spin_unlock_bh(&sk->sk_receive_queue.lock);
6007 		} else {
6008 			skb = skb_dequeue(&sk->sk_receive_queue);
6009 		}
6010 
6011 		if (skb)
6012 			return skb;
6013 
6014 		/* Caller is allowed not to check sk->sk_err before calling. */
6015 		error = sock_error(sk);
6016 		if (error)
6017 			goto no_packet;
6018 
6019 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6020 			break;
6021 
6022 		/* User doesn't want to wait.  */
6023 		error = -EAGAIN;
6024 		if (!timeo)
6025 			goto no_packet;
6026 	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6027 
6028 	return NULL;
6029 
6030 no_packet:
6031 	*err = error;
6032 	return NULL;
6033 }
6034 
6035 /* If sndbuf has changed, wake up per association sndbuf waiters.  */
6036 static void __sctp_write_space(struct sctp_association *asoc)
6037 {
6038 	struct sock *sk = asoc->base.sk;
6039 	struct socket *sock = sk->sk_socket;
6040 
6041 	if ((sctp_wspace(asoc) > 0) && sock) {
6042 		if (waitqueue_active(&asoc->wait))
6043 			wake_up_interruptible(&asoc->wait);
6044 
6045 		if (sctp_writeable(sk)) {
6046 			if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6047 				wake_up_interruptible(sk->sk_sleep);
6048 
6049 			/* Note that we try to include the Async I/O support
6050 			 * here by modeling from the current TCP/UDP code.
6051 			 * We have not tested with it yet.
6052 			 */
6053 			if (sock->fasync_list &&
6054 			    !(sk->sk_shutdown & SEND_SHUTDOWN))
6055 				sock_wake_async(sock,
6056 						SOCK_WAKE_SPACE, POLL_OUT);
6057 		}
6058 	}
6059 }
6060 
6061 /* Do accounting for the sndbuf space.
6062  * Decrement the used sndbuf space of the corresponding association by the
6063  * data size which was just transmitted(freed).
6064  */
6065 static void sctp_wfree(struct sk_buff *skb)
6066 {
6067 	struct sctp_association *asoc;
6068 	struct sctp_chunk *chunk;
6069 	struct sock *sk;
6070 
6071 	/* Get the saved chunk pointer.  */
6072 	chunk = *((struct sctp_chunk **)(skb->cb));
6073 	asoc = chunk->asoc;
6074 	sk = asoc->base.sk;
6075 	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6076 				sizeof(struct sk_buff) +
6077 				sizeof(struct sctp_chunk);
6078 
6079 	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6080 
6081 	/*
6082 	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6083 	 */
6084 	sk->sk_wmem_queued   -= skb->truesize;
6085 	sk_mem_uncharge(sk, skb->truesize);
6086 
6087 	sock_wfree(skb);
6088 	__sctp_write_space(asoc);
6089 
6090 	sctp_association_put(asoc);
6091 }
6092 
6093 /* Do accounting for the receive space on the socket.
6094  * Accounting for the association is done in ulpevent.c
6095  * We set this as a destructor for the cloned data skbs so that
6096  * accounting is done at the correct time.
6097  */
6098 void sctp_sock_rfree(struct sk_buff *skb)
6099 {
6100 	struct sock *sk = skb->sk;
6101 	struct sctp_ulpevent *event = sctp_skb2event(skb);
6102 
6103 	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6104 
6105 	/*
6106 	 * Mimic the behavior of sock_rfree
6107 	 */
6108 	sk_mem_uncharge(sk, event->rmem_len);
6109 }
6110 
6111 
6112 /* Helper function to wait for space in the sndbuf.  */
6113 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6114 				size_t msg_len)
6115 {
6116 	struct sock *sk = asoc->base.sk;
6117 	int err = 0;
6118 	long current_timeo = *timeo_p;
6119 	DEFINE_WAIT(wait);
6120 
6121 	SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6122 			  asoc, (long)(*timeo_p), msg_len);
6123 
6124 	/* Increment the association's refcnt.  */
6125 	sctp_association_hold(asoc);
6126 
6127 	/* Wait on the association specific sndbuf space. */
6128 	for (;;) {
6129 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6130 					  TASK_INTERRUPTIBLE);
6131 		if (!*timeo_p)
6132 			goto do_nonblock;
6133 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6134 		    asoc->base.dead)
6135 			goto do_error;
6136 		if (signal_pending(current))
6137 			goto do_interrupted;
6138 		if (msg_len <= sctp_wspace(asoc))
6139 			break;
6140 
6141 		/* Let another process have a go.  Since we are going
6142 		 * to sleep anyway.
6143 		 */
6144 		sctp_release_sock(sk);
6145 		current_timeo = schedule_timeout(current_timeo);
6146 		BUG_ON(sk != asoc->base.sk);
6147 		sctp_lock_sock(sk);
6148 
6149 		*timeo_p = current_timeo;
6150 	}
6151 
6152 out:
6153 	finish_wait(&asoc->wait, &wait);
6154 
6155 	/* Release the association's refcnt.  */
6156 	sctp_association_put(asoc);
6157 
6158 	return err;
6159 
6160 do_error:
6161 	err = -EPIPE;
6162 	goto out;
6163 
6164 do_interrupted:
6165 	err = sock_intr_errno(*timeo_p);
6166 	goto out;
6167 
6168 do_nonblock:
6169 	err = -EAGAIN;
6170 	goto out;
6171 }
6172 
6173 /* If socket sndbuf has changed, wake up all per association waiters.  */
6174 void sctp_write_space(struct sock *sk)
6175 {
6176 	struct sctp_association *asoc;
6177 
6178 	/* Wake up the tasks in each wait queue.  */
6179 	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6180 		__sctp_write_space(asoc);
6181 	}
6182 }
6183 
6184 /* Is there any sndbuf space available on the socket?
6185  *
6186  * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6187  * associations on the same socket.  For a UDP-style socket with
6188  * multiple associations, it is possible for it to be "unwriteable"
6189  * prematurely.  I assume that this is acceptable because
6190  * a premature "unwriteable" is better than an accidental "writeable" which
6191  * would cause an unwanted block under certain circumstances.  For the 1-1
6192  * UDP-style sockets or TCP-style sockets, this code should work.
6193  *  - Daisy
6194  */
6195 static int sctp_writeable(struct sock *sk)
6196 {
6197 	int amt = 0;
6198 
6199 	amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6200 	if (amt < 0)
6201 		amt = 0;
6202 	return amt;
6203 }
6204 
6205 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6206  * returns immediately with EINPROGRESS.
6207  */
6208 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6209 {
6210 	struct sock *sk = asoc->base.sk;
6211 	int err = 0;
6212 	long current_timeo = *timeo_p;
6213 	DEFINE_WAIT(wait);
6214 
6215 	SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6216 			  (long)(*timeo_p));
6217 
6218 	/* Increment the association's refcnt.  */
6219 	sctp_association_hold(asoc);
6220 
6221 	for (;;) {
6222 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6223 					  TASK_INTERRUPTIBLE);
6224 		if (!*timeo_p)
6225 			goto do_nonblock;
6226 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6227 			break;
6228 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6229 		    asoc->base.dead)
6230 			goto do_error;
6231 		if (signal_pending(current))
6232 			goto do_interrupted;
6233 
6234 		if (sctp_state(asoc, ESTABLISHED))
6235 			break;
6236 
6237 		/* Let another process have a go.  Since we are going
6238 		 * to sleep anyway.
6239 		 */
6240 		sctp_release_sock(sk);
6241 		current_timeo = schedule_timeout(current_timeo);
6242 		sctp_lock_sock(sk);
6243 
6244 		*timeo_p = current_timeo;
6245 	}
6246 
6247 out:
6248 	finish_wait(&asoc->wait, &wait);
6249 
6250 	/* Release the association's refcnt.  */
6251 	sctp_association_put(asoc);
6252 
6253 	return err;
6254 
6255 do_error:
6256 	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6257 		err = -ETIMEDOUT;
6258 	else
6259 		err = -ECONNREFUSED;
6260 	goto out;
6261 
6262 do_interrupted:
6263 	err = sock_intr_errno(*timeo_p);
6264 	goto out;
6265 
6266 do_nonblock:
6267 	err = -EINPROGRESS;
6268 	goto out;
6269 }
6270 
6271 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6272 {
6273 	struct sctp_endpoint *ep;
6274 	int err = 0;
6275 	DEFINE_WAIT(wait);
6276 
6277 	ep = sctp_sk(sk)->ep;
6278 
6279 
6280 	for (;;) {
6281 		prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6282 					  TASK_INTERRUPTIBLE);
6283 
6284 		if (list_empty(&ep->asocs)) {
6285 			sctp_release_sock(sk);
6286 			timeo = schedule_timeout(timeo);
6287 			sctp_lock_sock(sk);
6288 		}
6289 
6290 		err = -EINVAL;
6291 		if (!sctp_sstate(sk, LISTENING))
6292 			break;
6293 
6294 		err = 0;
6295 		if (!list_empty(&ep->asocs))
6296 			break;
6297 
6298 		err = sock_intr_errno(timeo);
6299 		if (signal_pending(current))
6300 			break;
6301 
6302 		err = -EAGAIN;
6303 		if (!timeo)
6304 			break;
6305 	}
6306 
6307 	finish_wait(sk->sk_sleep, &wait);
6308 
6309 	return err;
6310 }
6311 
6312 static void sctp_wait_for_close(struct sock *sk, long timeout)
6313 {
6314 	DEFINE_WAIT(wait);
6315 
6316 	do {
6317 		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6318 		if (list_empty(&sctp_sk(sk)->ep->asocs))
6319 			break;
6320 		sctp_release_sock(sk);
6321 		timeout = schedule_timeout(timeout);
6322 		sctp_lock_sock(sk);
6323 	} while (!signal_pending(current) && timeout);
6324 
6325 	finish_wait(sk->sk_sleep, &wait);
6326 }
6327 
6328 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6329 {
6330 	struct sk_buff *frag;
6331 
6332 	if (!skb->data_len)
6333 		goto done;
6334 
6335 	/* Don't forget the fragments. */
6336 	for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6337 		sctp_sock_rfree_frag(frag);
6338 
6339 done:
6340 	sctp_sock_rfree(skb);
6341 }
6342 
6343 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6344 {
6345 	struct sk_buff *frag;
6346 
6347 	if (!skb->data_len)
6348 		goto done;
6349 
6350 	/* Don't forget the fragments. */
6351 	for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6352 		sctp_skb_set_owner_r_frag(frag, sk);
6353 
6354 done:
6355 	sctp_skb_set_owner_r(skb, sk);
6356 }
6357 
6358 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6359  * and its messages to the newsk.
6360  */
6361 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6362 			      struct sctp_association *assoc,
6363 			      sctp_socket_type_t type)
6364 {
6365 	struct sctp_sock *oldsp = sctp_sk(oldsk);
6366 	struct sctp_sock *newsp = sctp_sk(newsk);
6367 	struct sctp_bind_bucket *pp; /* hash list port iterator */
6368 	struct sctp_endpoint *newep = newsp->ep;
6369 	struct sk_buff *skb, *tmp;
6370 	struct sctp_ulpevent *event;
6371 	struct sctp_bind_hashbucket *head;
6372 
6373 	/* Migrate socket buffer sizes and all the socket level options to the
6374 	 * new socket.
6375 	 */
6376 	newsk->sk_sndbuf = oldsk->sk_sndbuf;
6377 	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6378 	/* Brute force copy old sctp opt. */
6379 	inet_sk_copy_descendant(newsk, oldsk);
6380 
6381 	/* Restore the ep value that was overwritten with the above structure
6382 	 * copy.
6383 	 */
6384 	newsp->ep = newep;
6385 	newsp->hmac = NULL;
6386 
6387 	/* Hook this new socket in to the bind_hash list. */
6388 	head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6389 	sctp_local_bh_disable();
6390 	sctp_spin_lock(&head->lock);
6391 	pp = sctp_sk(oldsk)->bind_hash;
6392 	sk_add_bind_node(newsk, &pp->owner);
6393 	sctp_sk(newsk)->bind_hash = pp;
6394 	inet_sk(newsk)->num = inet_sk(oldsk)->num;
6395 	sctp_spin_unlock(&head->lock);
6396 	sctp_local_bh_enable();
6397 
6398 	/* Copy the bind_addr list from the original endpoint to the new
6399 	 * endpoint so that we can handle restarts properly
6400 	 */
6401 	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6402 				&oldsp->ep->base.bind_addr, GFP_KERNEL);
6403 
6404 	/* Move any messages in the old socket's receive queue that are for the
6405 	 * peeled off association to the new socket's receive queue.
6406 	 */
6407 	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6408 		event = sctp_skb2event(skb);
6409 		if (event->asoc == assoc) {
6410 			sctp_sock_rfree_frag(skb);
6411 			__skb_unlink(skb, &oldsk->sk_receive_queue);
6412 			__skb_queue_tail(&newsk->sk_receive_queue, skb);
6413 			sctp_skb_set_owner_r_frag(skb, newsk);
6414 		}
6415 	}
6416 
6417 	/* Clean up any messages pending delivery due to partial
6418 	 * delivery.   Three cases:
6419 	 * 1) No partial deliver;  no work.
6420 	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6421 	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6422 	 */
6423 	skb_queue_head_init(&newsp->pd_lobby);
6424 	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6425 
6426 	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6427 		struct sk_buff_head *queue;
6428 
6429 		/* Decide which queue to move pd_lobby skbs to. */
6430 		if (assoc->ulpq.pd_mode) {
6431 			queue = &newsp->pd_lobby;
6432 		} else
6433 			queue = &newsk->sk_receive_queue;
6434 
6435 		/* Walk through the pd_lobby, looking for skbs that
6436 		 * need moved to the new socket.
6437 		 */
6438 		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6439 			event = sctp_skb2event(skb);
6440 			if (event->asoc == assoc) {
6441 				sctp_sock_rfree_frag(skb);
6442 				__skb_unlink(skb, &oldsp->pd_lobby);
6443 				__skb_queue_tail(queue, skb);
6444 				sctp_skb_set_owner_r_frag(skb, newsk);
6445 			}
6446 		}
6447 
6448 		/* Clear up any skbs waiting for the partial
6449 		 * delivery to finish.
6450 		 */
6451 		if (assoc->ulpq.pd_mode)
6452 			sctp_clear_pd(oldsk, NULL);
6453 
6454 	}
6455 
6456 	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6457 		sctp_sock_rfree_frag(skb);
6458 		sctp_skb_set_owner_r_frag(skb, newsk);
6459 	}
6460 
6461 	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6462 		sctp_sock_rfree_frag(skb);
6463 		sctp_skb_set_owner_r_frag(skb, newsk);
6464 	}
6465 
6466 	/* Set the type of socket to indicate that it is peeled off from the
6467 	 * original UDP-style socket or created with the accept() call on a
6468 	 * TCP-style socket..
6469 	 */
6470 	newsp->type = type;
6471 
6472 	/* Mark the new socket "in-use" by the user so that any packets
6473 	 * that may arrive on the association after we've moved it are
6474 	 * queued to the backlog.  This prevents a potential race between
6475 	 * backlog processing on the old socket and new-packet processing
6476 	 * on the new socket.
6477 	 *
6478 	 * The caller has just allocated newsk so we can guarantee that other
6479 	 * paths won't try to lock it and then oldsk.
6480 	 */
6481 	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6482 	sctp_assoc_migrate(assoc, newsk);
6483 
6484 	/* If the association on the newsk is already closed before accept()
6485 	 * is called, set RCV_SHUTDOWN flag.
6486 	 */
6487 	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6488 		newsk->sk_shutdown |= RCV_SHUTDOWN;
6489 
6490 	newsk->sk_state = SCTP_SS_ESTABLISHED;
6491 	sctp_release_sock(newsk);
6492 }
6493 
6494 
6495 /* This proto struct describes the ULP interface for SCTP.  */
6496 struct proto sctp_prot = {
6497 	.name        =	"SCTP",
6498 	.owner       =	THIS_MODULE,
6499 	.close       =	sctp_close,
6500 	.connect     =	sctp_connect,
6501 	.disconnect  =	sctp_disconnect,
6502 	.accept      =	sctp_accept,
6503 	.ioctl       =	sctp_ioctl,
6504 	.init        =	sctp_init_sock,
6505 	.destroy     =	sctp_destroy_sock,
6506 	.shutdown    =	sctp_shutdown,
6507 	.setsockopt  =	sctp_setsockopt,
6508 	.getsockopt  =	sctp_getsockopt,
6509 	.sendmsg     =	sctp_sendmsg,
6510 	.recvmsg     =	sctp_recvmsg,
6511 	.bind        =	sctp_bind,
6512 	.backlog_rcv =	sctp_backlog_rcv,
6513 	.hash        =	sctp_hash,
6514 	.unhash      =	sctp_unhash,
6515 	.get_port    =	sctp_get_port,
6516 	.obj_size    =  sizeof(struct sctp_sock),
6517 	.sysctl_mem  =  sysctl_sctp_mem,
6518 	.sysctl_rmem =  sysctl_sctp_rmem,
6519 	.sysctl_wmem =  sysctl_sctp_wmem,
6520 	.memory_pressure = &sctp_memory_pressure,
6521 	.enter_memory_pressure = sctp_enter_memory_pressure,
6522 	.memory_allocated = &sctp_memory_allocated,
6523 	.sockets_allocated = &sctp_sockets_allocated,
6524 };
6525 
6526 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6527 
6528 struct proto sctpv6_prot = {
6529 	.name		= "SCTPv6",
6530 	.owner		= THIS_MODULE,
6531 	.close		= sctp_close,
6532 	.connect	= sctp_connect,
6533 	.disconnect	= sctp_disconnect,
6534 	.accept		= sctp_accept,
6535 	.ioctl		= sctp_ioctl,
6536 	.init		= sctp_init_sock,
6537 	.destroy	= sctp_destroy_sock,
6538 	.shutdown	= sctp_shutdown,
6539 	.setsockopt	= sctp_setsockopt,
6540 	.getsockopt	= sctp_getsockopt,
6541 	.sendmsg	= sctp_sendmsg,
6542 	.recvmsg	= sctp_recvmsg,
6543 	.bind		= sctp_bind,
6544 	.backlog_rcv	= sctp_backlog_rcv,
6545 	.hash		= sctp_hash,
6546 	.unhash		= sctp_unhash,
6547 	.get_port	= sctp_get_port,
6548 	.obj_size	= sizeof(struct sctp6_sock),
6549 	.sysctl_mem	= sysctl_sctp_mem,
6550 	.sysctl_rmem	= sysctl_sctp_rmem,
6551 	.sysctl_wmem	= sysctl_sctp_wmem,
6552 	.memory_pressure = &sctp_memory_pressure,
6553 	.enter_memory_pressure = sctp_enter_memory_pressure,
6554 	.memory_allocated = &sctp_memory_allocated,
6555 	.sockets_allocated = &sctp_sockets_allocated,
6556 };
6557 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
6558