xref: /openbmc/linux/net/sctp/socket.c (revision 9d56dd3b083a3bec56e9da35ce07baca81030b03)
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 struct percpu_counter sctp_sockets_allocated;
118 
119 static void sctp_enter_memory_pressure(struct sock *sk)
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 = sk_wmem_alloc_get(asoc->base.sk);
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 	/* V4 mapped address are really of AF_INET family */
312 	if (addr->sa.sa_family == AF_INET6 &&
313 	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 		if (!opt->pf->af_supported(AF_INET, opt))
315 			return NULL;
316 	} else {
317 		/* Does this PF support this AF? */
318 		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
319 			return NULL;
320 	}
321 
322 	/* If we get this far, af is valid. */
323 	af = sctp_get_af_specific(addr->sa.sa_family);
324 
325 	if (len < af->sockaddr_len)
326 		return NULL;
327 
328 	return af;
329 }
330 
331 /* Bind a local address either to an endpoint or to an association.  */
332 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
333 {
334 	struct sctp_sock *sp = sctp_sk(sk);
335 	struct sctp_endpoint *ep = sp->ep;
336 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
337 	struct sctp_af *af;
338 	unsigned short snum;
339 	int ret = 0;
340 
341 	/* Common sockaddr verification. */
342 	af = sctp_sockaddr_af(sp, addr, len);
343 	if (!af) {
344 		SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
345 				  sk, addr, len);
346 		return -EINVAL;
347 	}
348 
349 	snum = ntohs(addr->v4.sin_port);
350 
351 	SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
352 				 ", port: %d, new port: %d, len: %d)\n",
353 				 sk,
354 				 addr,
355 				 bp->port, snum,
356 				 len);
357 
358 	/* PF specific bind() address verification. */
359 	if (!sp->pf->bind_verify(sp, addr))
360 		return -EADDRNOTAVAIL;
361 
362 	/* We must either be unbound, or bind to the same port.
363 	 * It's OK to allow 0 ports if we are already bound.
364 	 * We'll just inhert an already bound port in this case
365 	 */
366 	if (bp->port) {
367 		if (!snum)
368 			snum = bp->port;
369 		else if (snum != bp->port) {
370 			SCTP_DEBUG_PRINTK("sctp_do_bind:"
371 				  " New port %d does not match existing port "
372 				  "%d.\n", snum, bp->port);
373 			return -EINVAL;
374 		}
375 	}
376 
377 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
378 		return -EACCES;
379 
380 	/* See if the address matches any of the addresses we may have
381 	 * already bound before checking against other endpoints.
382 	 */
383 	if (sctp_bind_addr_match(bp, addr, sp))
384 		return -EINVAL;
385 
386 	/* Make sure we are allowed to bind here.
387 	 * The function sctp_get_port_local() does duplicate address
388 	 * detection.
389 	 */
390 	addr->v4.sin_port = htons(snum);
391 	if ((ret = sctp_get_port_local(sk, addr))) {
392 		return -EADDRINUSE;
393 	}
394 
395 	/* Refresh ephemeral port.  */
396 	if (!bp->port)
397 		bp->port = inet_sk(sk)->inet_num;
398 
399 	/* Add the address to the bind address list.
400 	 * Use GFP_ATOMIC since BHs will be disabled.
401 	 */
402 	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
403 
404 	/* Copy back into socket for getsockname() use. */
405 	if (!ret) {
406 		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
407 		af->to_sk_saddr(addr, sk);
408 	}
409 
410 	return ret;
411 }
412 
413  /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
414  *
415  * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
416  * at any one time.  If a sender, after sending an ASCONF chunk, decides
417  * it needs to transfer another ASCONF Chunk, it MUST wait until the
418  * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
419  * subsequent ASCONF. Note this restriction binds each side, so at any
420  * time two ASCONF may be in-transit on any given association (one sent
421  * from each endpoint).
422  */
423 static int sctp_send_asconf(struct sctp_association *asoc,
424 			    struct sctp_chunk *chunk)
425 {
426 	int		retval = 0;
427 
428 	/* If there is an outstanding ASCONF chunk, queue it for later
429 	 * transmission.
430 	 */
431 	if (asoc->addip_last_asconf) {
432 		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
433 		goto out;
434 	}
435 
436 	/* Hold the chunk until an ASCONF_ACK is received. */
437 	sctp_chunk_hold(chunk);
438 	retval = sctp_primitive_ASCONF(asoc, chunk);
439 	if (retval)
440 		sctp_chunk_free(chunk);
441 	else
442 		asoc->addip_last_asconf = chunk;
443 
444 out:
445 	return retval;
446 }
447 
448 /* Add a list of addresses as bind addresses to local endpoint or
449  * association.
450  *
451  * Basically run through each address specified in the addrs/addrcnt
452  * array/length pair, determine if it is IPv6 or IPv4 and call
453  * sctp_do_bind() on it.
454  *
455  * If any of them fails, then the operation will be reversed and the
456  * ones that were added will be removed.
457  *
458  * Only sctp_setsockopt_bindx() is supposed to call this function.
459  */
460 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
461 {
462 	int cnt;
463 	int retval = 0;
464 	void *addr_buf;
465 	struct sockaddr *sa_addr;
466 	struct sctp_af *af;
467 
468 	SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
469 			  sk, addrs, addrcnt);
470 
471 	addr_buf = addrs;
472 	for (cnt = 0; cnt < addrcnt; cnt++) {
473 		/* The list may contain either IPv4 or IPv6 address;
474 		 * determine the address length for walking thru the list.
475 		 */
476 		sa_addr = (struct sockaddr *)addr_buf;
477 		af = sctp_get_af_specific(sa_addr->sa_family);
478 		if (!af) {
479 			retval = -EINVAL;
480 			goto err_bindx_add;
481 		}
482 
483 		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
484 				      af->sockaddr_len);
485 
486 		addr_buf += af->sockaddr_len;
487 
488 err_bindx_add:
489 		if (retval < 0) {
490 			/* Failed. Cleanup the ones that have been added */
491 			if (cnt > 0)
492 				sctp_bindx_rem(sk, addrs, cnt);
493 			return retval;
494 		}
495 	}
496 
497 	return retval;
498 }
499 
500 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
501  * associations that are part of the endpoint indicating that a list of local
502  * addresses are added to the endpoint.
503  *
504  * If any of the addresses is already in the bind address list of the
505  * association, we do not send the chunk for that association.  But it will not
506  * affect other associations.
507  *
508  * Only sctp_setsockopt_bindx() is supposed to call this function.
509  */
510 static int sctp_send_asconf_add_ip(struct sock		*sk,
511 				   struct sockaddr	*addrs,
512 				   int 			addrcnt)
513 {
514 	struct sctp_sock		*sp;
515 	struct sctp_endpoint		*ep;
516 	struct sctp_association		*asoc;
517 	struct sctp_bind_addr		*bp;
518 	struct sctp_chunk		*chunk;
519 	struct sctp_sockaddr_entry	*laddr;
520 	union sctp_addr			*addr;
521 	union sctp_addr			saveaddr;
522 	void				*addr_buf;
523 	struct sctp_af			*af;
524 	struct list_head		*p;
525 	int 				i;
526 	int 				retval = 0;
527 
528 	if (!sctp_addip_enable)
529 		return retval;
530 
531 	sp = sctp_sk(sk);
532 	ep = sp->ep;
533 
534 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
535 			  __func__, sk, addrs, addrcnt);
536 
537 	list_for_each_entry(asoc, &ep->asocs, asocs) {
538 
539 		if (!asoc->peer.asconf_capable)
540 			continue;
541 
542 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
543 			continue;
544 
545 		if (!sctp_state(asoc, ESTABLISHED))
546 			continue;
547 
548 		/* Check if any address in the packed array of addresses is
549 		 * in the bind address list of the association. If so,
550 		 * do not send the asconf chunk to its peer, but continue with
551 		 * other associations.
552 		 */
553 		addr_buf = addrs;
554 		for (i = 0; i < addrcnt; i++) {
555 			addr = (union sctp_addr *)addr_buf;
556 			af = sctp_get_af_specific(addr->v4.sin_family);
557 			if (!af) {
558 				retval = -EINVAL;
559 				goto out;
560 			}
561 
562 			if (sctp_assoc_lookup_laddr(asoc, addr))
563 				break;
564 
565 			addr_buf += af->sockaddr_len;
566 		}
567 		if (i < addrcnt)
568 			continue;
569 
570 		/* Use the first valid address in bind addr list of
571 		 * association as Address Parameter of ASCONF CHUNK.
572 		 */
573 		bp = &asoc->base.bind_addr;
574 		p = bp->address_list.next;
575 		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
576 		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
577 						   addrcnt, SCTP_PARAM_ADD_IP);
578 		if (!chunk) {
579 			retval = -ENOMEM;
580 			goto out;
581 		}
582 
583 		retval = sctp_send_asconf(asoc, chunk);
584 		if (retval)
585 			goto out;
586 
587 		/* Add the new addresses to the bind address list with
588 		 * use_as_src set to 0.
589 		 */
590 		addr_buf = addrs;
591 		for (i = 0; i < addrcnt; i++) {
592 			addr = (union sctp_addr *)addr_buf;
593 			af = sctp_get_af_specific(addr->v4.sin_family);
594 			memcpy(&saveaddr, addr, af->sockaddr_len);
595 			retval = sctp_add_bind_addr(bp, &saveaddr,
596 						    SCTP_ADDR_NEW, GFP_ATOMIC);
597 			addr_buf += af->sockaddr_len;
598 		}
599 	}
600 
601 out:
602 	return retval;
603 }
604 
605 /* Remove a list of addresses from bind addresses list.  Do not remove the
606  * last address.
607  *
608  * Basically run through each address specified in the addrs/addrcnt
609  * array/length pair, determine if it is IPv6 or IPv4 and call
610  * sctp_del_bind() on it.
611  *
612  * If any of them fails, then the operation will be reversed and the
613  * ones that were removed will be added back.
614  *
615  * At least one address has to be left; if only one address is
616  * available, the operation will return -EBUSY.
617  *
618  * Only sctp_setsockopt_bindx() is supposed to call this function.
619  */
620 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
621 {
622 	struct sctp_sock *sp = sctp_sk(sk);
623 	struct sctp_endpoint *ep = sp->ep;
624 	int cnt;
625 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
626 	int retval = 0;
627 	void *addr_buf;
628 	union sctp_addr *sa_addr;
629 	struct sctp_af *af;
630 
631 	SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
632 			  sk, addrs, addrcnt);
633 
634 	addr_buf = addrs;
635 	for (cnt = 0; cnt < addrcnt; cnt++) {
636 		/* If the bind address list is empty or if there is only one
637 		 * bind address, there is nothing more to be removed (we need
638 		 * at least one address here).
639 		 */
640 		if (list_empty(&bp->address_list) ||
641 		    (sctp_list_single_entry(&bp->address_list))) {
642 			retval = -EBUSY;
643 			goto err_bindx_rem;
644 		}
645 
646 		sa_addr = (union sctp_addr *)addr_buf;
647 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
648 		if (!af) {
649 			retval = -EINVAL;
650 			goto err_bindx_rem;
651 		}
652 
653 		if (!af->addr_valid(sa_addr, sp, NULL)) {
654 			retval = -EADDRNOTAVAIL;
655 			goto err_bindx_rem;
656 		}
657 
658 		if (sa_addr->v4.sin_port != htons(bp->port)) {
659 			retval = -EINVAL;
660 			goto err_bindx_rem;
661 		}
662 
663 		/* FIXME - There is probably a need to check if sk->sk_saddr and
664 		 * sk->sk_rcv_addr are currently set to one of the addresses to
665 		 * be removed. This is something which needs to be looked into
666 		 * when we are fixing the outstanding issues with multi-homing
667 		 * socket routing and failover schemes. Refer to comments in
668 		 * sctp_do_bind(). -daisy
669 		 */
670 		retval = sctp_del_bind_addr(bp, sa_addr);
671 
672 		addr_buf += af->sockaddr_len;
673 err_bindx_rem:
674 		if (retval < 0) {
675 			/* Failed. Add the ones that has been removed back */
676 			if (cnt > 0)
677 				sctp_bindx_add(sk, addrs, cnt);
678 			return retval;
679 		}
680 	}
681 
682 	return retval;
683 }
684 
685 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
686  * the associations that are part of the endpoint indicating that a list of
687  * local addresses are removed from the endpoint.
688  *
689  * If any of the addresses is already in the bind address list of the
690  * association, we do not send the chunk for that association.  But it will not
691  * affect other associations.
692  *
693  * Only sctp_setsockopt_bindx() is supposed to call this function.
694  */
695 static int sctp_send_asconf_del_ip(struct sock		*sk,
696 				   struct sockaddr	*addrs,
697 				   int			addrcnt)
698 {
699 	struct sctp_sock	*sp;
700 	struct sctp_endpoint	*ep;
701 	struct sctp_association	*asoc;
702 	struct sctp_transport	*transport;
703 	struct sctp_bind_addr	*bp;
704 	struct sctp_chunk	*chunk;
705 	union sctp_addr		*laddr;
706 	void			*addr_buf;
707 	struct sctp_af		*af;
708 	struct sctp_sockaddr_entry *saddr;
709 	int 			i;
710 	int 			retval = 0;
711 
712 	if (!sctp_addip_enable)
713 		return retval;
714 
715 	sp = sctp_sk(sk);
716 	ep = sp->ep;
717 
718 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
719 			  __func__, sk, addrs, addrcnt);
720 
721 	list_for_each_entry(asoc, &ep->asocs, asocs) {
722 
723 		if (!asoc->peer.asconf_capable)
724 			continue;
725 
726 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
727 			continue;
728 
729 		if (!sctp_state(asoc, ESTABLISHED))
730 			continue;
731 
732 		/* Check if any address in the packed array of addresses is
733 		 * not present in the bind address list of the association.
734 		 * If so, do not send the asconf chunk to its peer, but
735 		 * continue with other associations.
736 		 */
737 		addr_buf = addrs;
738 		for (i = 0; i < addrcnt; i++) {
739 			laddr = (union sctp_addr *)addr_buf;
740 			af = sctp_get_af_specific(laddr->v4.sin_family);
741 			if (!af) {
742 				retval = -EINVAL;
743 				goto out;
744 			}
745 
746 			if (!sctp_assoc_lookup_laddr(asoc, laddr))
747 				break;
748 
749 			addr_buf += af->sockaddr_len;
750 		}
751 		if (i < addrcnt)
752 			continue;
753 
754 		/* Find one address in the association's bind address list
755 		 * that is not in the packed array of addresses. This is to
756 		 * make sure that we do not delete all the addresses in the
757 		 * association.
758 		 */
759 		bp = &asoc->base.bind_addr;
760 		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
761 					       addrcnt, sp);
762 		if (!laddr)
763 			continue;
764 
765 		/* We do not need RCU protection throughout this loop
766 		 * because this is done under a socket lock from the
767 		 * setsockopt call.
768 		 */
769 		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
770 						   SCTP_PARAM_DEL_IP);
771 		if (!chunk) {
772 			retval = -ENOMEM;
773 			goto out;
774 		}
775 
776 		/* Reset use_as_src flag for the addresses in the bind address
777 		 * list that are to be deleted.
778 		 */
779 		addr_buf = addrs;
780 		for (i = 0; i < addrcnt; i++) {
781 			laddr = (union sctp_addr *)addr_buf;
782 			af = sctp_get_af_specific(laddr->v4.sin_family);
783 			list_for_each_entry(saddr, &bp->address_list, list) {
784 				if (sctp_cmp_addr_exact(&saddr->a, laddr))
785 					saddr->state = SCTP_ADDR_DEL;
786 			}
787 			addr_buf += af->sockaddr_len;
788 		}
789 
790 		/* Update the route and saddr entries for all the transports
791 		 * as some of the addresses in the bind address list are
792 		 * about to be deleted and cannot be used as source addresses.
793 		 */
794 		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
795 					transports) {
796 			dst_release(transport->dst);
797 			sctp_transport_route(transport, NULL,
798 					     sctp_sk(asoc->base.sk));
799 		}
800 
801 		retval = sctp_send_asconf(asoc, chunk);
802 	}
803 out:
804 	return retval;
805 }
806 
807 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
808  *
809  * API 8.1
810  * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
811  *                int flags);
812  *
813  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
814  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
815  * or IPv6 addresses.
816  *
817  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
818  * Section 3.1.2 for this usage.
819  *
820  * addrs is a pointer to an array of one or more socket addresses. Each
821  * address is contained in its appropriate structure (i.e. struct
822  * sockaddr_in or struct sockaddr_in6) the family of the address type
823  * must be used to distinguish the address length (note that this
824  * representation is termed a "packed array" of addresses). The caller
825  * specifies the number of addresses in the array with addrcnt.
826  *
827  * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
828  * -1, and sets errno to the appropriate error code.
829  *
830  * For SCTP, the port given in each socket address must be the same, or
831  * sctp_bindx() will fail, setting errno to EINVAL.
832  *
833  * The flags parameter is formed from the bitwise OR of zero or more of
834  * the following currently defined flags:
835  *
836  * SCTP_BINDX_ADD_ADDR
837  *
838  * SCTP_BINDX_REM_ADDR
839  *
840  * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
841  * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
842  * addresses from the association. The two flags are mutually exclusive;
843  * if both are given, sctp_bindx() will fail with EINVAL. A caller may
844  * not remove all addresses from an association; sctp_bindx() will
845  * reject such an attempt with EINVAL.
846  *
847  * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
848  * additional addresses with an endpoint after calling bind().  Or use
849  * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
850  * socket is associated with so that no new association accepted will be
851  * associated with those addresses. If the endpoint supports dynamic
852  * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
853  * endpoint to send the appropriate message to the peer to change the
854  * peers address lists.
855  *
856  * Adding and removing addresses from a connected association is
857  * optional functionality. Implementations that do not support this
858  * functionality should return EOPNOTSUPP.
859  *
860  * Basically do nothing but copying the addresses from user to kernel
861  * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
862  * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
863  * from userspace.
864  *
865  * We don't use copy_from_user() for optimization: we first do the
866  * sanity checks (buffer size -fast- and access check-healthy
867  * pointer); if all of those succeed, then we can alloc the memory
868  * (expensive operation) needed to copy the data to kernel. Then we do
869  * the copying without checking the user space area
870  * (__copy_from_user()).
871  *
872  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
873  * it.
874  *
875  * sk        The sk of the socket
876  * addrs     The pointer to the addresses in user land
877  * addrssize Size of the addrs buffer
878  * op        Operation to perform (add or remove, see the flags of
879  *           sctp_bindx)
880  *
881  * Returns 0 if ok, <0 errno code on error.
882  */
883 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
884 				      struct sockaddr __user *addrs,
885 				      int addrs_size, int op)
886 {
887 	struct sockaddr *kaddrs;
888 	int err;
889 	int addrcnt = 0;
890 	int walk_size = 0;
891 	struct sockaddr *sa_addr;
892 	void *addr_buf;
893 	struct sctp_af *af;
894 
895 	SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
896 			  " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
897 
898 	if (unlikely(addrs_size <= 0))
899 		return -EINVAL;
900 
901 	/* Check the user passed a healthy pointer.  */
902 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
903 		return -EFAULT;
904 
905 	/* Alloc space for the address array in kernel memory.  */
906 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
907 	if (unlikely(!kaddrs))
908 		return -ENOMEM;
909 
910 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
911 		kfree(kaddrs);
912 		return -EFAULT;
913 	}
914 
915 	/* Walk through the addrs buffer and count the number of addresses. */
916 	addr_buf = kaddrs;
917 	while (walk_size < addrs_size) {
918 		sa_addr = (struct sockaddr *)addr_buf;
919 		af = sctp_get_af_specific(sa_addr->sa_family);
920 
921 		/* If the address family is not supported or if this address
922 		 * causes the address buffer to overflow return EINVAL.
923 		 */
924 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
925 			kfree(kaddrs);
926 			return -EINVAL;
927 		}
928 		addrcnt++;
929 		addr_buf += af->sockaddr_len;
930 		walk_size += af->sockaddr_len;
931 	}
932 
933 	/* Do the work. */
934 	switch (op) {
935 	case SCTP_BINDX_ADD_ADDR:
936 		err = sctp_bindx_add(sk, kaddrs, addrcnt);
937 		if (err)
938 			goto out;
939 		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
940 		break;
941 
942 	case SCTP_BINDX_REM_ADDR:
943 		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
944 		if (err)
945 			goto out;
946 		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
947 		break;
948 
949 	default:
950 		err = -EINVAL;
951 		break;
952 	}
953 
954 out:
955 	kfree(kaddrs);
956 
957 	return err;
958 }
959 
960 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
961  *
962  * Common routine for handling connect() and sctp_connectx().
963  * Connect will come in with just a single address.
964  */
965 static int __sctp_connect(struct sock* sk,
966 			  struct sockaddr *kaddrs,
967 			  int addrs_size,
968 			  sctp_assoc_t *assoc_id)
969 {
970 	struct sctp_sock *sp;
971 	struct sctp_endpoint *ep;
972 	struct sctp_association *asoc = NULL;
973 	struct sctp_association *asoc2;
974 	struct sctp_transport *transport;
975 	union sctp_addr to;
976 	struct sctp_af *af;
977 	sctp_scope_t scope;
978 	long timeo;
979 	int err = 0;
980 	int addrcnt = 0;
981 	int walk_size = 0;
982 	union sctp_addr *sa_addr = NULL;
983 	void *addr_buf;
984 	unsigned short port;
985 	unsigned int f_flags = 0;
986 
987 	sp = sctp_sk(sk);
988 	ep = sp->ep;
989 
990 	/* connect() cannot be done on a socket that is already in ESTABLISHED
991 	 * state - UDP-style peeled off socket or a TCP-style socket that
992 	 * is already connected.
993 	 * It cannot be done even on a TCP-style listening socket.
994 	 */
995 	if (sctp_sstate(sk, ESTABLISHED) ||
996 	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
997 		err = -EISCONN;
998 		goto out_free;
999 	}
1000 
1001 	/* Walk through the addrs buffer and count the number of addresses. */
1002 	addr_buf = kaddrs;
1003 	while (walk_size < addrs_size) {
1004 		sa_addr = (union sctp_addr *)addr_buf;
1005 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1006 		port = ntohs(sa_addr->v4.sin_port);
1007 
1008 		/* If the address family is not supported or if this address
1009 		 * causes the address buffer to overflow return EINVAL.
1010 		 */
1011 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012 			err = -EINVAL;
1013 			goto out_free;
1014 		}
1015 
1016 		/* Save current address so we can work with it */
1017 		memcpy(&to, sa_addr, af->sockaddr_len);
1018 
1019 		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1020 		if (err)
1021 			goto out_free;
1022 
1023 		/* Make sure the destination port is correctly set
1024 		 * in all addresses.
1025 		 */
1026 		if (asoc && asoc->peer.port && asoc->peer.port != port)
1027 			goto out_free;
1028 
1029 
1030 		/* Check if there already is a matching association on the
1031 		 * endpoint (other than the one created here).
1032 		 */
1033 		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1034 		if (asoc2 && asoc2 != asoc) {
1035 			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1036 				err = -EISCONN;
1037 			else
1038 				err = -EALREADY;
1039 			goto out_free;
1040 		}
1041 
1042 		/* If we could not find a matching association on the endpoint,
1043 		 * make sure that there is no peeled-off association matching
1044 		 * the peer address even on another socket.
1045 		 */
1046 		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1047 			err = -EADDRNOTAVAIL;
1048 			goto out_free;
1049 		}
1050 
1051 		if (!asoc) {
1052 			/* If a bind() or sctp_bindx() is not called prior to
1053 			 * an sctp_connectx() call, the system picks an
1054 			 * ephemeral port and will choose an address set
1055 			 * equivalent to binding with a wildcard address.
1056 			 */
1057 			if (!ep->base.bind_addr.port) {
1058 				if (sctp_autobind(sk)) {
1059 					err = -EAGAIN;
1060 					goto out_free;
1061 				}
1062 			} else {
1063 				/*
1064 				 * If an unprivileged user inherits a 1-many
1065 				 * style socket with open associations on a
1066 				 * privileged port, it MAY be permitted to
1067 				 * accept new associations, but it SHOULD NOT
1068 				 * be permitted to open new associations.
1069 				 */
1070 				if (ep->base.bind_addr.port < PROT_SOCK &&
1071 				    !capable(CAP_NET_BIND_SERVICE)) {
1072 					err = -EACCES;
1073 					goto out_free;
1074 				}
1075 			}
1076 
1077 			scope = sctp_scope(&to);
1078 			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1079 			if (!asoc) {
1080 				err = -ENOMEM;
1081 				goto out_free;
1082 			}
1083 
1084 			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1085 							      GFP_KERNEL);
1086 			if (err < 0) {
1087 				goto out_free;
1088 			}
1089 
1090 		}
1091 
1092 		/* Prime the peer's transport structures.  */
1093 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1094 						SCTP_UNKNOWN);
1095 		if (!transport) {
1096 			err = -ENOMEM;
1097 			goto out_free;
1098 		}
1099 
1100 		addrcnt++;
1101 		addr_buf += af->sockaddr_len;
1102 		walk_size += af->sockaddr_len;
1103 	}
1104 
1105 	/* In case the user of sctp_connectx() wants an association
1106 	 * id back, assign one now.
1107 	 */
1108 	if (assoc_id) {
1109 		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1110 		if (err < 0)
1111 			goto out_free;
1112 	}
1113 
1114 	err = sctp_primitive_ASSOCIATE(asoc, NULL);
1115 	if (err < 0) {
1116 		goto out_free;
1117 	}
1118 
1119 	/* Initialize sk's dport and daddr for getpeername() */
1120 	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1121 	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1122 	af->to_sk_daddr(sa_addr, sk);
1123 	sk->sk_err = 0;
1124 
1125 	/* in-kernel sockets don't generally have a file allocated to them
1126 	 * if all they do is call sock_create_kern().
1127 	 */
1128 	if (sk->sk_socket->file)
1129 		f_flags = sk->sk_socket->file->f_flags;
1130 
1131 	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1132 
1133 	err = sctp_wait_for_connect(asoc, &timeo);
1134 	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1135 		*assoc_id = asoc->assoc_id;
1136 
1137 	/* Don't free association on exit. */
1138 	asoc = NULL;
1139 
1140 out_free:
1141 
1142 	SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1143 			  " kaddrs: %p err: %d\n",
1144 			  asoc, kaddrs, err);
1145 	if (asoc)
1146 		sctp_association_free(asoc);
1147 	return err;
1148 }
1149 
1150 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1151  *
1152  * API 8.9
1153  * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1154  * 			sctp_assoc_t *asoc);
1155  *
1156  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1157  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1158  * or IPv6 addresses.
1159  *
1160  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1161  * Section 3.1.2 for this usage.
1162  *
1163  * addrs is a pointer to an array of one or more socket addresses. Each
1164  * address is contained in its appropriate structure (i.e. struct
1165  * sockaddr_in or struct sockaddr_in6) the family of the address type
1166  * must be used to distengish the address length (note that this
1167  * representation is termed a "packed array" of addresses). The caller
1168  * specifies the number of addresses in the array with addrcnt.
1169  *
1170  * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1171  * the association id of the new association.  On failure, sctp_connectx()
1172  * returns -1, and sets errno to the appropriate error code.  The assoc_id
1173  * is not touched by the kernel.
1174  *
1175  * For SCTP, the port given in each socket address must be the same, or
1176  * sctp_connectx() will fail, setting errno to EINVAL.
1177  *
1178  * An application can use sctp_connectx to initiate an association with
1179  * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1180  * allows a caller to specify multiple addresses at which a peer can be
1181  * reached.  The way the SCTP stack uses the list of addresses to set up
1182  * the association is implementation dependant.  This function only
1183  * specifies that the stack will try to make use of all the addresses in
1184  * the list when needed.
1185  *
1186  * Note that the list of addresses passed in is only used for setting up
1187  * the association.  It does not necessarily equal the set of addresses
1188  * the peer uses for the resulting association.  If the caller wants to
1189  * find out the set of peer addresses, it must use sctp_getpaddrs() to
1190  * retrieve them after the association has been set up.
1191  *
1192  * Basically do nothing but copying the addresses from user to kernel
1193  * land and invoking either sctp_connectx(). This is used for tunneling
1194  * the sctp_connectx() request through sctp_setsockopt() from userspace.
1195  *
1196  * We don't use copy_from_user() for optimization: we first do the
1197  * sanity checks (buffer size -fast- and access check-healthy
1198  * pointer); if all of those succeed, then we can alloc the memory
1199  * (expensive operation) needed to copy the data to kernel. Then we do
1200  * the copying without checking the user space area
1201  * (__copy_from_user()).
1202  *
1203  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1204  * it.
1205  *
1206  * sk        The sk of the socket
1207  * addrs     The pointer to the addresses in user land
1208  * addrssize Size of the addrs buffer
1209  *
1210  * Returns >=0 if ok, <0 errno code on error.
1211  */
1212 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1213 				      struct sockaddr __user *addrs,
1214 				      int addrs_size,
1215 				      sctp_assoc_t *assoc_id)
1216 {
1217 	int err = 0;
1218 	struct sockaddr *kaddrs;
1219 
1220 	SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1221 			  __func__, sk, addrs, addrs_size);
1222 
1223 	if (unlikely(addrs_size <= 0))
1224 		return -EINVAL;
1225 
1226 	/* Check the user passed a healthy pointer.  */
1227 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1228 		return -EFAULT;
1229 
1230 	/* Alloc space for the address array in kernel memory.  */
1231 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1232 	if (unlikely(!kaddrs))
1233 		return -ENOMEM;
1234 
1235 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1236 		err = -EFAULT;
1237 	} else {
1238 		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1239 	}
1240 
1241 	kfree(kaddrs);
1242 
1243 	return err;
1244 }
1245 
1246 /*
1247  * This is an older interface.  It's kept for backward compatibility
1248  * to the option that doesn't provide association id.
1249  */
1250 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1251 				      struct sockaddr __user *addrs,
1252 				      int addrs_size)
1253 {
1254 	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1255 }
1256 
1257 /*
1258  * New interface for the API.  The since the API is done with a socket
1259  * option, to make it simple we feed back the association id is as a return
1260  * indication to the call.  Error is always negative and association id is
1261  * always positive.
1262  */
1263 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1264 				      struct sockaddr __user *addrs,
1265 				      int addrs_size)
1266 {
1267 	sctp_assoc_t assoc_id = 0;
1268 	int err = 0;
1269 
1270 	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1271 
1272 	if (err)
1273 		return err;
1274 	else
1275 		return assoc_id;
1276 }
1277 
1278 /*
1279  * New (hopefully final) interface for the API.
1280  * We use the sctp_getaddrs_old structure so that use-space library
1281  * can avoid any unnecessary allocations.   The only defferent part
1282  * is that we store the actual length of the address buffer into the
1283  * addrs_num structure member.  That way we can re-use the existing
1284  * code.
1285  */
1286 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1287 					char __user *optval,
1288 					int __user *optlen)
1289 {
1290 	struct sctp_getaddrs_old param;
1291 	sctp_assoc_t assoc_id = 0;
1292 	int err = 0;
1293 
1294 	if (len < sizeof(param))
1295 		return -EINVAL;
1296 
1297 	if (copy_from_user(&param, optval, sizeof(param)))
1298 		return -EFAULT;
1299 
1300 	err = __sctp_setsockopt_connectx(sk,
1301 			(struct sockaddr __user *)param.addrs,
1302 			param.addr_num, &assoc_id);
1303 
1304 	if (err == 0 || err == -EINPROGRESS) {
1305 		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1306 			return -EFAULT;
1307 		if (put_user(sizeof(assoc_id), optlen))
1308 			return -EFAULT;
1309 	}
1310 
1311 	return err;
1312 }
1313 
1314 /* API 3.1.4 close() - UDP Style Syntax
1315  * Applications use close() to perform graceful shutdown (as described in
1316  * Section 10.1 of [SCTP]) on ALL the associations currently represented
1317  * by a UDP-style socket.
1318  *
1319  * The syntax is
1320  *
1321  *   ret = close(int sd);
1322  *
1323  *   sd      - the socket descriptor of the associations to be closed.
1324  *
1325  * To gracefully shutdown a specific association represented by the
1326  * UDP-style socket, an application should use the sendmsg() call,
1327  * passing no user data, but including the appropriate flag in the
1328  * ancillary data (see Section xxxx).
1329  *
1330  * If sd in the close() call is a branched-off socket representing only
1331  * one association, the shutdown is performed on that association only.
1332  *
1333  * 4.1.6 close() - TCP Style Syntax
1334  *
1335  * Applications use close() to gracefully close down an association.
1336  *
1337  * The syntax is:
1338  *
1339  *    int close(int sd);
1340  *
1341  *      sd      - the socket descriptor of the association to be closed.
1342  *
1343  * After an application calls close() on a socket descriptor, no further
1344  * socket operations will succeed on that descriptor.
1345  *
1346  * API 7.1.4 SO_LINGER
1347  *
1348  * An application using the TCP-style socket can use this option to
1349  * perform the SCTP ABORT primitive.  The linger option structure is:
1350  *
1351  *  struct  linger {
1352  *     int     l_onoff;                // option on/off
1353  *     int     l_linger;               // linger time
1354  * };
1355  *
1356  * To enable the option, set l_onoff to 1.  If the l_linger value is set
1357  * to 0, calling close() is the same as the ABORT primitive.  If the
1358  * value is set to a negative value, the setsockopt() call will return
1359  * an error.  If the value is set to a positive value linger_time, the
1360  * close() can be blocked for at most linger_time ms.  If the graceful
1361  * shutdown phase does not finish during this period, close() will
1362  * return but the graceful shutdown phase continues in the system.
1363  */
1364 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1365 {
1366 	struct sctp_endpoint *ep;
1367 	struct sctp_association *asoc;
1368 	struct list_head *pos, *temp;
1369 
1370 	SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1371 
1372 	sctp_lock_sock(sk);
1373 	sk->sk_shutdown = SHUTDOWN_MASK;
1374 	sk->sk_state = SCTP_SS_CLOSING;
1375 
1376 	ep = sctp_sk(sk)->ep;
1377 
1378 	/* Walk all associations on an endpoint.  */
1379 	list_for_each_safe(pos, temp, &ep->asocs) {
1380 		asoc = list_entry(pos, struct sctp_association, asocs);
1381 
1382 		if (sctp_style(sk, TCP)) {
1383 			/* A closed association can still be in the list if
1384 			 * it belongs to a TCP-style listening socket that is
1385 			 * not yet accepted. If so, free it. If not, send an
1386 			 * ABORT or SHUTDOWN based on the linger options.
1387 			 */
1388 			if (sctp_state(asoc, CLOSED)) {
1389 				sctp_unhash_established(asoc);
1390 				sctp_association_free(asoc);
1391 				continue;
1392 			}
1393 		}
1394 
1395 		if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1396 			struct sctp_chunk *chunk;
1397 
1398 			chunk = sctp_make_abort_user(asoc, NULL, 0);
1399 			if (chunk)
1400 				sctp_primitive_ABORT(asoc, chunk);
1401 		} else
1402 			sctp_primitive_SHUTDOWN(asoc, NULL);
1403 	}
1404 
1405 	/* Clean up any skbs sitting on the receive queue.  */
1406 	sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1407 	sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1408 
1409 	/* On a TCP-style socket, block for at most linger_time if set. */
1410 	if (sctp_style(sk, TCP) && timeout)
1411 		sctp_wait_for_close(sk, timeout);
1412 
1413 	/* This will run the backlog queue.  */
1414 	sctp_release_sock(sk);
1415 
1416 	/* Supposedly, no process has access to the socket, but
1417 	 * the net layers still may.
1418 	 */
1419 	sctp_local_bh_disable();
1420 	sctp_bh_lock_sock(sk);
1421 
1422 	/* Hold the sock, since sk_common_release() will put sock_put()
1423 	 * and we have just a little more cleanup.
1424 	 */
1425 	sock_hold(sk);
1426 	sk_common_release(sk);
1427 
1428 	sctp_bh_unlock_sock(sk);
1429 	sctp_local_bh_enable();
1430 
1431 	sock_put(sk);
1432 
1433 	SCTP_DBG_OBJCNT_DEC(sock);
1434 }
1435 
1436 /* Handle EPIPE error. */
1437 static int sctp_error(struct sock *sk, int flags, int err)
1438 {
1439 	if (err == -EPIPE)
1440 		err = sock_error(sk) ? : -EPIPE;
1441 	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1442 		send_sig(SIGPIPE, current, 0);
1443 	return err;
1444 }
1445 
1446 /* API 3.1.3 sendmsg() - UDP Style Syntax
1447  *
1448  * An application uses sendmsg() and recvmsg() calls to transmit data to
1449  * and receive data from its peer.
1450  *
1451  *  ssize_t sendmsg(int socket, const struct msghdr *message,
1452  *                  int flags);
1453  *
1454  *  socket  - the socket descriptor of the endpoint.
1455  *  message - pointer to the msghdr structure which contains a single
1456  *            user message and possibly some ancillary data.
1457  *
1458  *            See Section 5 for complete description of the data
1459  *            structures.
1460  *
1461  *  flags   - flags sent or received with the user message, see Section
1462  *            5 for complete description of the flags.
1463  *
1464  * Note:  This function could use a rewrite especially when explicit
1465  * connect support comes in.
1466  */
1467 /* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1468 
1469 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1470 
1471 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1472 			     struct msghdr *msg, size_t msg_len)
1473 {
1474 	struct sctp_sock *sp;
1475 	struct sctp_endpoint *ep;
1476 	struct sctp_association *new_asoc=NULL, *asoc=NULL;
1477 	struct sctp_transport *transport, *chunk_tp;
1478 	struct sctp_chunk *chunk;
1479 	union sctp_addr to;
1480 	struct sockaddr *msg_name = NULL;
1481 	struct sctp_sndrcvinfo default_sinfo = { 0 };
1482 	struct sctp_sndrcvinfo *sinfo;
1483 	struct sctp_initmsg *sinit;
1484 	sctp_assoc_t associd = 0;
1485 	sctp_cmsgs_t cmsgs = { NULL };
1486 	int err;
1487 	sctp_scope_t scope;
1488 	long timeo;
1489 	__u16 sinfo_flags = 0;
1490 	struct sctp_datamsg *datamsg;
1491 	int msg_flags = msg->msg_flags;
1492 
1493 	SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1494 			  sk, msg, msg_len);
1495 
1496 	err = 0;
1497 	sp = sctp_sk(sk);
1498 	ep = sp->ep;
1499 
1500 	SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1501 
1502 	/* We cannot send a message over a TCP-style listening socket. */
1503 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1504 		err = -EPIPE;
1505 		goto out_nounlock;
1506 	}
1507 
1508 	/* Parse out the SCTP CMSGs.  */
1509 	err = sctp_msghdr_parse(msg, &cmsgs);
1510 
1511 	if (err) {
1512 		SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1513 		goto out_nounlock;
1514 	}
1515 
1516 	/* Fetch the destination address for this packet.  This
1517 	 * address only selects the association--it is not necessarily
1518 	 * the address we will send to.
1519 	 * For a peeled-off socket, msg_name is ignored.
1520 	 */
1521 	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1522 		int msg_namelen = msg->msg_namelen;
1523 
1524 		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1525 				       msg_namelen);
1526 		if (err)
1527 			return err;
1528 
1529 		if (msg_namelen > sizeof(to))
1530 			msg_namelen = sizeof(to);
1531 		memcpy(&to, msg->msg_name, msg_namelen);
1532 		msg_name = msg->msg_name;
1533 	}
1534 
1535 	sinfo = cmsgs.info;
1536 	sinit = cmsgs.init;
1537 
1538 	/* Did the user specify SNDRCVINFO?  */
1539 	if (sinfo) {
1540 		sinfo_flags = sinfo->sinfo_flags;
1541 		associd = sinfo->sinfo_assoc_id;
1542 	}
1543 
1544 	SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1545 			  msg_len, sinfo_flags);
1546 
1547 	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1548 	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1549 		err = -EINVAL;
1550 		goto out_nounlock;
1551 	}
1552 
1553 	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1554 	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1555 	 * If SCTP_ABORT is set, the message length could be non zero with
1556 	 * the msg_iov set to the user abort reason.
1557 	 */
1558 	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1559 	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1560 		err = -EINVAL;
1561 		goto out_nounlock;
1562 	}
1563 
1564 	/* If SCTP_ADDR_OVER is set, there must be an address
1565 	 * specified in msg_name.
1566 	 */
1567 	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1568 		err = -EINVAL;
1569 		goto out_nounlock;
1570 	}
1571 
1572 	transport = NULL;
1573 
1574 	SCTP_DEBUG_PRINTK("About to look up association.\n");
1575 
1576 	sctp_lock_sock(sk);
1577 
1578 	/* If a msg_name has been specified, assume this is to be used.  */
1579 	if (msg_name) {
1580 		/* Look for a matching association on the endpoint. */
1581 		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1582 		if (!asoc) {
1583 			/* If we could not find a matching association on the
1584 			 * endpoint, make sure that it is not a TCP-style
1585 			 * socket that already has an association or there is
1586 			 * no peeled-off association on another socket.
1587 			 */
1588 			if ((sctp_style(sk, TCP) &&
1589 			     sctp_sstate(sk, ESTABLISHED)) ||
1590 			    sctp_endpoint_is_peeled_off(ep, &to)) {
1591 				err = -EADDRNOTAVAIL;
1592 				goto out_unlock;
1593 			}
1594 		}
1595 	} else {
1596 		asoc = sctp_id2assoc(sk, associd);
1597 		if (!asoc) {
1598 			err = -EPIPE;
1599 			goto out_unlock;
1600 		}
1601 	}
1602 
1603 	if (asoc) {
1604 		SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1605 
1606 		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1607 		 * socket that has an association in CLOSED state. This can
1608 		 * happen when an accepted socket has an association that is
1609 		 * already CLOSED.
1610 		 */
1611 		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1612 			err = -EPIPE;
1613 			goto out_unlock;
1614 		}
1615 
1616 		if (sinfo_flags & SCTP_EOF) {
1617 			SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1618 					  asoc);
1619 			sctp_primitive_SHUTDOWN(asoc, NULL);
1620 			err = 0;
1621 			goto out_unlock;
1622 		}
1623 		if (sinfo_flags & SCTP_ABORT) {
1624 
1625 			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1626 			if (!chunk) {
1627 				err = -ENOMEM;
1628 				goto out_unlock;
1629 			}
1630 
1631 			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1632 			sctp_primitive_ABORT(asoc, chunk);
1633 			err = 0;
1634 			goto out_unlock;
1635 		}
1636 	}
1637 
1638 	/* Do we need to create the association?  */
1639 	if (!asoc) {
1640 		SCTP_DEBUG_PRINTK("There is no association yet.\n");
1641 
1642 		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1643 			err = -EINVAL;
1644 			goto out_unlock;
1645 		}
1646 
1647 		/* Check for invalid stream against the stream counts,
1648 		 * either the default or the user specified stream counts.
1649 		 */
1650 		if (sinfo) {
1651 			if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1652 				/* Check against the defaults. */
1653 				if (sinfo->sinfo_stream >=
1654 				    sp->initmsg.sinit_num_ostreams) {
1655 					err = -EINVAL;
1656 					goto out_unlock;
1657 				}
1658 			} else {
1659 				/* Check against the requested.  */
1660 				if (sinfo->sinfo_stream >=
1661 				    sinit->sinit_num_ostreams) {
1662 					err = -EINVAL;
1663 					goto out_unlock;
1664 				}
1665 			}
1666 		}
1667 
1668 		/*
1669 		 * API 3.1.2 bind() - UDP Style Syntax
1670 		 * If a bind() or sctp_bindx() is not called prior to a
1671 		 * sendmsg() call that initiates a new association, the
1672 		 * system picks an ephemeral port and will choose an address
1673 		 * set equivalent to binding with a wildcard address.
1674 		 */
1675 		if (!ep->base.bind_addr.port) {
1676 			if (sctp_autobind(sk)) {
1677 				err = -EAGAIN;
1678 				goto out_unlock;
1679 			}
1680 		} else {
1681 			/*
1682 			 * If an unprivileged user inherits a one-to-many
1683 			 * style socket with open associations on a privileged
1684 			 * port, it MAY be permitted to accept new associations,
1685 			 * but it SHOULD NOT be permitted to open new
1686 			 * associations.
1687 			 */
1688 			if (ep->base.bind_addr.port < PROT_SOCK &&
1689 			    !capable(CAP_NET_BIND_SERVICE)) {
1690 				err = -EACCES;
1691 				goto out_unlock;
1692 			}
1693 		}
1694 
1695 		scope = sctp_scope(&to);
1696 		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1697 		if (!new_asoc) {
1698 			err = -ENOMEM;
1699 			goto out_unlock;
1700 		}
1701 		asoc = new_asoc;
1702 		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1703 		if (err < 0) {
1704 			err = -ENOMEM;
1705 			goto out_free;
1706 		}
1707 
1708 		/* If the SCTP_INIT ancillary data is specified, set all
1709 		 * the association init values accordingly.
1710 		 */
1711 		if (sinit) {
1712 			if (sinit->sinit_num_ostreams) {
1713 				asoc->c.sinit_num_ostreams =
1714 					sinit->sinit_num_ostreams;
1715 			}
1716 			if (sinit->sinit_max_instreams) {
1717 				asoc->c.sinit_max_instreams =
1718 					sinit->sinit_max_instreams;
1719 			}
1720 			if (sinit->sinit_max_attempts) {
1721 				asoc->max_init_attempts
1722 					= sinit->sinit_max_attempts;
1723 			}
1724 			if (sinit->sinit_max_init_timeo) {
1725 				asoc->max_init_timeo =
1726 				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1727 			}
1728 		}
1729 
1730 		/* Prime the peer's transport structures.  */
1731 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1732 		if (!transport) {
1733 			err = -ENOMEM;
1734 			goto out_free;
1735 		}
1736 	}
1737 
1738 	/* ASSERT: we have a valid association at this point.  */
1739 	SCTP_DEBUG_PRINTK("We have a valid association.\n");
1740 
1741 	if (!sinfo) {
1742 		/* If the user didn't specify SNDRCVINFO, make up one with
1743 		 * some defaults.
1744 		 */
1745 		default_sinfo.sinfo_stream = asoc->default_stream;
1746 		default_sinfo.sinfo_flags = asoc->default_flags;
1747 		default_sinfo.sinfo_ppid = asoc->default_ppid;
1748 		default_sinfo.sinfo_context = asoc->default_context;
1749 		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1750 		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1751 		sinfo = &default_sinfo;
1752 	}
1753 
1754 	/* API 7.1.7, the sndbuf size per association bounds the
1755 	 * maximum size of data that can be sent in a single send call.
1756 	 */
1757 	if (msg_len > sk->sk_sndbuf) {
1758 		err = -EMSGSIZE;
1759 		goto out_free;
1760 	}
1761 
1762 	if (asoc->pmtu_pending)
1763 		sctp_assoc_pending_pmtu(asoc);
1764 
1765 	/* If fragmentation is disabled and the message length exceeds the
1766 	 * association fragmentation point, return EMSGSIZE.  The I-D
1767 	 * does not specify what this error is, but this looks like
1768 	 * a great fit.
1769 	 */
1770 	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1771 		err = -EMSGSIZE;
1772 		goto out_free;
1773 	}
1774 
1775 	if (sinfo) {
1776 		/* Check for invalid stream. */
1777 		if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1778 			err = -EINVAL;
1779 			goto out_free;
1780 		}
1781 	}
1782 
1783 	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1784 	if (!sctp_wspace(asoc)) {
1785 		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1786 		if (err)
1787 			goto out_free;
1788 	}
1789 
1790 	/* If an address is passed with the sendto/sendmsg call, it is used
1791 	 * to override the primary destination address in the TCP model, or
1792 	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1793 	 */
1794 	if ((sctp_style(sk, TCP) && msg_name) ||
1795 	    (sinfo_flags & SCTP_ADDR_OVER)) {
1796 		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1797 		if (!chunk_tp) {
1798 			err = -EINVAL;
1799 			goto out_free;
1800 		}
1801 	} else
1802 		chunk_tp = NULL;
1803 
1804 	/* Auto-connect, if we aren't connected already. */
1805 	if (sctp_state(asoc, CLOSED)) {
1806 		err = sctp_primitive_ASSOCIATE(asoc, NULL);
1807 		if (err < 0)
1808 			goto out_free;
1809 		SCTP_DEBUG_PRINTK("We associated primitively.\n");
1810 	}
1811 
1812 	/* Break the message into multiple chunks of maximum size. */
1813 	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1814 	if (!datamsg) {
1815 		err = -ENOMEM;
1816 		goto out_free;
1817 	}
1818 
1819 	/* Now send the (possibly) fragmented message. */
1820 	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1821 		sctp_chunk_hold(chunk);
1822 
1823 		/* Do accounting for the write space.  */
1824 		sctp_set_owner_w(chunk);
1825 
1826 		chunk->transport = chunk_tp;
1827 	}
1828 
1829 	/* Send it to the lower layers.  Note:  all chunks
1830 	 * must either fail or succeed.   The lower layer
1831 	 * works that way today.  Keep it that way or this
1832 	 * breaks.
1833 	 */
1834 	err = sctp_primitive_SEND(asoc, datamsg);
1835 	/* Did the lower layer accept the chunk? */
1836 	if (err)
1837 		sctp_datamsg_free(datamsg);
1838 	else
1839 		sctp_datamsg_put(datamsg);
1840 
1841 	SCTP_DEBUG_PRINTK("We sent primitively.\n");
1842 
1843 	if (err)
1844 		goto out_free;
1845 	else
1846 		err = msg_len;
1847 
1848 	/* If we are already past ASSOCIATE, the lower
1849 	 * layers are responsible for association cleanup.
1850 	 */
1851 	goto out_unlock;
1852 
1853 out_free:
1854 	if (new_asoc)
1855 		sctp_association_free(asoc);
1856 out_unlock:
1857 	sctp_release_sock(sk);
1858 
1859 out_nounlock:
1860 	return sctp_error(sk, msg_flags, err);
1861 
1862 #if 0
1863 do_sock_err:
1864 	if (msg_len)
1865 		err = msg_len;
1866 	else
1867 		err = sock_error(sk);
1868 	goto out;
1869 
1870 do_interrupted:
1871 	if (msg_len)
1872 		err = msg_len;
1873 	goto out;
1874 #endif /* 0 */
1875 }
1876 
1877 /* This is an extended version of skb_pull() that removes the data from the
1878  * start of a skb even when data is spread across the list of skb's in the
1879  * frag_list. len specifies the total amount of data that needs to be removed.
1880  * when 'len' bytes could be removed from the skb, it returns 0.
1881  * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1882  * could not be removed.
1883  */
1884 static int sctp_skb_pull(struct sk_buff *skb, int len)
1885 {
1886 	struct sk_buff *list;
1887 	int skb_len = skb_headlen(skb);
1888 	int rlen;
1889 
1890 	if (len <= skb_len) {
1891 		__skb_pull(skb, len);
1892 		return 0;
1893 	}
1894 	len -= skb_len;
1895 	__skb_pull(skb, skb_len);
1896 
1897 	skb_walk_frags(skb, list) {
1898 		rlen = sctp_skb_pull(list, len);
1899 		skb->len -= (len-rlen);
1900 		skb->data_len -= (len-rlen);
1901 
1902 		if (!rlen)
1903 			return 0;
1904 
1905 		len = rlen;
1906 	}
1907 
1908 	return len;
1909 }
1910 
1911 /* API 3.1.3  recvmsg() - UDP Style Syntax
1912  *
1913  *  ssize_t recvmsg(int socket, struct msghdr *message,
1914  *                    int flags);
1915  *
1916  *  socket  - the socket descriptor of the endpoint.
1917  *  message - pointer to the msghdr structure which contains a single
1918  *            user message and possibly some ancillary data.
1919  *
1920  *            See Section 5 for complete description of the data
1921  *            structures.
1922  *
1923  *  flags   - flags sent or received with the user message, see Section
1924  *            5 for complete description of the flags.
1925  */
1926 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1927 
1928 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1929 			     struct msghdr *msg, size_t len, int noblock,
1930 			     int flags, int *addr_len)
1931 {
1932 	struct sctp_ulpevent *event = NULL;
1933 	struct sctp_sock *sp = sctp_sk(sk);
1934 	struct sk_buff *skb;
1935 	int copied;
1936 	int err = 0;
1937 	int skb_len;
1938 
1939 	SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1940 			  "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1941 			  "len", len, "knoblauch", noblock,
1942 			  "flags", flags, "addr_len", addr_len);
1943 
1944 	sctp_lock_sock(sk);
1945 
1946 	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1947 		err = -ENOTCONN;
1948 		goto out;
1949 	}
1950 
1951 	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1952 	if (!skb)
1953 		goto out;
1954 
1955 	/* Get the total length of the skb including any skb's in the
1956 	 * frag_list.
1957 	 */
1958 	skb_len = skb->len;
1959 
1960 	copied = skb_len;
1961 	if (copied > len)
1962 		copied = len;
1963 
1964 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1965 
1966 	event = sctp_skb2event(skb);
1967 
1968 	if (err)
1969 		goto out_free;
1970 
1971 	sock_recv_ts_and_drops(msg, sk, skb);
1972 	if (sctp_ulpevent_is_notification(event)) {
1973 		msg->msg_flags |= MSG_NOTIFICATION;
1974 		sp->pf->event_msgname(event, msg->msg_name, addr_len);
1975 	} else {
1976 		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1977 	}
1978 
1979 	/* Check if we allow SCTP_SNDRCVINFO. */
1980 	if (sp->subscribe.sctp_data_io_event)
1981 		sctp_ulpevent_read_sndrcvinfo(event, msg);
1982 #if 0
1983 	/* FIXME: we should be calling IP/IPv6 layers.  */
1984 	if (sk->sk_protinfo.af_inet.cmsg_flags)
1985 		ip_cmsg_recv(msg, skb);
1986 #endif
1987 
1988 	err = copied;
1989 
1990 	/* If skb's length exceeds the user's buffer, update the skb and
1991 	 * push it back to the receive_queue so that the next call to
1992 	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1993 	 */
1994 	if (skb_len > copied) {
1995 		msg->msg_flags &= ~MSG_EOR;
1996 		if (flags & MSG_PEEK)
1997 			goto out_free;
1998 		sctp_skb_pull(skb, copied);
1999 		skb_queue_head(&sk->sk_receive_queue, skb);
2000 
2001 		/* When only partial message is copied to the user, increase
2002 		 * rwnd by that amount. If all the data in the skb is read,
2003 		 * rwnd is updated when the event is freed.
2004 		 */
2005 		if (!sctp_ulpevent_is_notification(event))
2006 			sctp_assoc_rwnd_increase(event->asoc, copied);
2007 		goto out;
2008 	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2009 		   (event->msg_flags & MSG_EOR))
2010 		msg->msg_flags |= MSG_EOR;
2011 	else
2012 		msg->msg_flags &= ~MSG_EOR;
2013 
2014 out_free:
2015 	if (flags & MSG_PEEK) {
2016 		/* Release the skb reference acquired after peeking the skb in
2017 		 * sctp_skb_recv_datagram().
2018 		 */
2019 		kfree_skb(skb);
2020 	} else {
2021 		/* Free the event which includes releasing the reference to
2022 		 * the owner of the skb, freeing the skb and updating the
2023 		 * rwnd.
2024 		 */
2025 		sctp_ulpevent_free(event);
2026 	}
2027 out:
2028 	sctp_release_sock(sk);
2029 	return err;
2030 }
2031 
2032 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2033  *
2034  * This option is a on/off flag.  If enabled no SCTP message
2035  * fragmentation will be performed.  Instead if a message being sent
2036  * exceeds the current PMTU size, the message will NOT be sent and
2037  * instead a error will be indicated to the user.
2038  */
2039 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2040 					     char __user *optval,
2041 					     unsigned int optlen)
2042 {
2043 	int val;
2044 
2045 	if (optlen < sizeof(int))
2046 		return -EINVAL;
2047 
2048 	if (get_user(val, (int __user *)optval))
2049 		return -EFAULT;
2050 
2051 	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2052 
2053 	return 0;
2054 }
2055 
2056 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2057 				  unsigned int optlen)
2058 {
2059 	if (optlen > sizeof(struct sctp_event_subscribe))
2060 		return -EINVAL;
2061 	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2062 		return -EFAULT;
2063 	return 0;
2064 }
2065 
2066 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2067  *
2068  * This socket option is applicable to the UDP-style socket only.  When
2069  * set it will cause associations that are idle for more than the
2070  * specified number of seconds to automatically close.  An association
2071  * being idle is defined an association that has NOT sent or received
2072  * user data.  The special value of '0' indicates that no automatic
2073  * close of any associations should be performed.  The option expects an
2074  * integer defining the number of seconds of idle time before an
2075  * association is closed.
2076  */
2077 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2078 				     unsigned int optlen)
2079 {
2080 	struct sctp_sock *sp = sctp_sk(sk);
2081 
2082 	/* Applicable to UDP-style socket only */
2083 	if (sctp_style(sk, TCP))
2084 		return -EOPNOTSUPP;
2085 	if (optlen != sizeof(int))
2086 		return -EINVAL;
2087 	if (copy_from_user(&sp->autoclose, optval, optlen))
2088 		return -EFAULT;
2089 	/* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2090 	sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2091 
2092 	return 0;
2093 }
2094 
2095 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2096  *
2097  * Applications can enable or disable heartbeats for any peer address of
2098  * an association, modify an address's heartbeat interval, force a
2099  * heartbeat to be sent immediately, and adjust the address's maximum
2100  * number of retransmissions sent before an address is considered
2101  * unreachable.  The following structure is used to access and modify an
2102  * address's parameters:
2103  *
2104  *  struct sctp_paddrparams {
2105  *     sctp_assoc_t            spp_assoc_id;
2106  *     struct sockaddr_storage spp_address;
2107  *     uint32_t                spp_hbinterval;
2108  *     uint16_t                spp_pathmaxrxt;
2109  *     uint32_t                spp_pathmtu;
2110  *     uint32_t                spp_sackdelay;
2111  *     uint32_t                spp_flags;
2112  * };
2113  *
2114  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2115  *                     application, and identifies the association for
2116  *                     this query.
2117  *   spp_address     - This specifies which address is of interest.
2118  *   spp_hbinterval  - This contains the value of the heartbeat interval,
2119  *                     in milliseconds.  If a  value of zero
2120  *                     is present in this field then no changes are to
2121  *                     be made to this parameter.
2122  *   spp_pathmaxrxt  - This contains the maximum number of
2123  *                     retransmissions before this address shall be
2124  *                     considered unreachable. If a  value of zero
2125  *                     is present in this field then no changes are to
2126  *                     be made to this parameter.
2127  *   spp_pathmtu     - When Path MTU discovery is disabled the value
2128  *                     specified here will be the "fixed" path mtu.
2129  *                     Note that if the spp_address field is empty
2130  *                     then all associations on this address will
2131  *                     have this fixed path mtu set upon them.
2132  *
2133  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2134  *                     the number of milliseconds that sacks will be delayed
2135  *                     for. This value will apply to all addresses of an
2136  *                     association if the spp_address field is empty. Note
2137  *                     also, that if delayed sack is enabled and this
2138  *                     value is set to 0, no change is made to the last
2139  *                     recorded delayed sack timer value.
2140  *
2141  *   spp_flags       - These flags are used to control various features
2142  *                     on an association. The flag field may contain
2143  *                     zero or more of the following options.
2144  *
2145  *                     SPP_HB_ENABLE  - Enable heartbeats on the
2146  *                     specified address. Note that if the address
2147  *                     field is empty all addresses for the association
2148  *                     have heartbeats enabled upon them.
2149  *
2150  *                     SPP_HB_DISABLE - Disable heartbeats on the
2151  *                     speicifed address. Note that if the address
2152  *                     field is empty all addresses for the association
2153  *                     will have their heartbeats disabled. Note also
2154  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2155  *                     mutually exclusive, only one of these two should
2156  *                     be specified. Enabling both fields will have
2157  *                     undetermined results.
2158  *
2159  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2160  *                     to be made immediately.
2161  *
2162  *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2163  *                     heartbeat delayis to be set to the value of 0
2164  *                     milliseconds.
2165  *
2166  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2167  *                     discovery upon the specified address. Note that
2168  *                     if the address feild is empty then all addresses
2169  *                     on the association are effected.
2170  *
2171  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2172  *                     discovery upon the specified address. Note that
2173  *                     if the address feild is empty then all addresses
2174  *                     on the association are effected. Not also that
2175  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2176  *                     exclusive. Enabling both will have undetermined
2177  *                     results.
2178  *
2179  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2180  *                     on delayed sack. The time specified in spp_sackdelay
2181  *                     is used to specify the sack delay for this address. Note
2182  *                     that if spp_address is empty then all addresses will
2183  *                     enable delayed sack and take on the sack delay
2184  *                     value specified in spp_sackdelay.
2185  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2186  *                     off delayed sack. If the spp_address field is blank then
2187  *                     delayed sack is disabled for the entire association. Note
2188  *                     also that this field is mutually exclusive to
2189  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2190  *                     results.
2191  */
2192 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2193 				       struct sctp_transport   *trans,
2194 				       struct sctp_association *asoc,
2195 				       struct sctp_sock        *sp,
2196 				       int                      hb_change,
2197 				       int                      pmtud_change,
2198 				       int                      sackdelay_change)
2199 {
2200 	int error;
2201 
2202 	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2203 		error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2204 		if (error)
2205 			return error;
2206 	}
2207 
2208 	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2209 	 * this field is ignored.  Note also that a value of zero indicates
2210 	 * the current setting should be left unchanged.
2211 	 */
2212 	if (params->spp_flags & SPP_HB_ENABLE) {
2213 
2214 		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2215 		 * set.  This lets us use 0 value when this flag
2216 		 * is set.
2217 		 */
2218 		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2219 			params->spp_hbinterval = 0;
2220 
2221 		if (params->spp_hbinterval ||
2222 		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2223 			if (trans) {
2224 				trans->hbinterval =
2225 				    msecs_to_jiffies(params->spp_hbinterval);
2226 			} else if (asoc) {
2227 				asoc->hbinterval =
2228 				    msecs_to_jiffies(params->spp_hbinterval);
2229 			} else {
2230 				sp->hbinterval = params->spp_hbinterval;
2231 			}
2232 		}
2233 	}
2234 
2235 	if (hb_change) {
2236 		if (trans) {
2237 			trans->param_flags =
2238 				(trans->param_flags & ~SPP_HB) | hb_change;
2239 		} else if (asoc) {
2240 			asoc->param_flags =
2241 				(asoc->param_flags & ~SPP_HB) | hb_change;
2242 		} else {
2243 			sp->param_flags =
2244 				(sp->param_flags & ~SPP_HB) | hb_change;
2245 		}
2246 	}
2247 
2248 	/* When Path MTU discovery is disabled the value specified here will
2249 	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2250 	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2251 	 * effect).
2252 	 */
2253 	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2254 		if (trans) {
2255 			trans->pathmtu = params->spp_pathmtu;
2256 			sctp_assoc_sync_pmtu(asoc);
2257 		} else if (asoc) {
2258 			asoc->pathmtu = params->spp_pathmtu;
2259 			sctp_frag_point(asoc, params->spp_pathmtu);
2260 		} else {
2261 			sp->pathmtu = params->spp_pathmtu;
2262 		}
2263 	}
2264 
2265 	if (pmtud_change) {
2266 		if (trans) {
2267 			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2268 				(params->spp_flags & SPP_PMTUD_ENABLE);
2269 			trans->param_flags =
2270 				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2271 			if (update) {
2272 				sctp_transport_pmtu(trans);
2273 				sctp_assoc_sync_pmtu(asoc);
2274 			}
2275 		} else if (asoc) {
2276 			asoc->param_flags =
2277 				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2278 		} else {
2279 			sp->param_flags =
2280 				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2281 		}
2282 	}
2283 
2284 	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2285 	 * value of this field is ignored.  Note also that a value of zero
2286 	 * indicates the current setting should be left unchanged.
2287 	 */
2288 	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2289 		if (trans) {
2290 			trans->sackdelay =
2291 				msecs_to_jiffies(params->spp_sackdelay);
2292 		} else if (asoc) {
2293 			asoc->sackdelay =
2294 				msecs_to_jiffies(params->spp_sackdelay);
2295 		} else {
2296 			sp->sackdelay = params->spp_sackdelay;
2297 		}
2298 	}
2299 
2300 	if (sackdelay_change) {
2301 		if (trans) {
2302 			trans->param_flags =
2303 				(trans->param_flags & ~SPP_SACKDELAY) |
2304 				sackdelay_change;
2305 		} else if (asoc) {
2306 			asoc->param_flags =
2307 				(asoc->param_flags & ~SPP_SACKDELAY) |
2308 				sackdelay_change;
2309 		} else {
2310 			sp->param_flags =
2311 				(sp->param_flags & ~SPP_SACKDELAY) |
2312 				sackdelay_change;
2313 		}
2314 	}
2315 
2316 	/* Note that a value of zero indicates the current setting should be
2317 	   left unchanged.
2318 	 */
2319 	if (params->spp_pathmaxrxt) {
2320 		if (trans) {
2321 			trans->pathmaxrxt = params->spp_pathmaxrxt;
2322 		} else if (asoc) {
2323 			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2324 		} else {
2325 			sp->pathmaxrxt = params->spp_pathmaxrxt;
2326 		}
2327 	}
2328 
2329 	return 0;
2330 }
2331 
2332 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2333 					    char __user *optval,
2334 					    unsigned int optlen)
2335 {
2336 	struct sctp_paddrparams  params;
2337 	struct sctp_transport   *trans = NULL;
2338 	struct sctp_association *asoc = NULL;
2339 	struct sctp_sock        *sp = sctp_sk(sk);
2340 	int error;
2341 	int hb_change, pmtud_change, sackdelay_change;
2342 
2343 	if (optlen != sizeof(struct sctp_paddrparams))
2344 		return - EINVAL;
2345 
2346 	if (copy_from_user(&params, optval, optlen))
2347 		return -EFAULT;
2348 
2349 	/* Validate flags and value parameters. */
2350 	hb_change        = params.spp_flags & SPP_HB;
2351 	pmtud_change     = params.spp_flags & SPP_PMTUD;
2352 	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2353 
2354 	if (hb_change        == SPP_HB ||
2355 	    pmtud_change     == SPP_PMTUD ||
2356 	    sackdelay_change == SPP_SACKDELAY ||
2357 	    params.spp_sackdelay > 500 ||
2358 	    (params.spp_pathmtu &&
2359 	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2360 		return -EINVAL;
2361 
2362 	/* If an address other than INADDR_ANY is specified, and
2363 	 * no transport is found, then the request is invalid.
2364 	 */
2365 	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2366 		trans = sctp_addr_id2transport(sk, &params.spp_address,
2367 					       params.spp_assoc_id);
2368 		if (!trans)
2369 			return -EINVAL;
2370 	}
2371 
2372 	/* Get association, if assoc_id != 0 and the socket is a one
2373 	 * to many style socket, and an association was not found, then
2374 	 * the id was invalid.
2375 	 */
2376 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2377 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2378 		return -EINVAL;
2379 
2380 	/* Heartbeat demand can only be sent on a transport or
2381 	 * association, but not a socket.
2382 	 */
2383 	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2384 		return -EINVAL;
2385 
2386 	/* Process parameters. */
2387 	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2388 					    hb_change, pmtud_change,
2389 					    sackdelay_change);
2390 
2391 	if (error)
2392 		return error;
2393 
2394 	/* If changes are for association, also apply parameters to each
2395 	 * transport.
2396 	 */
2397 	if (!trans && asoc) {
2398 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2399 				transports) {
2400 			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2401 						    hb_change, pmtud_change,
2402 						    sackdelay_change);
2403 		}
2404 	}
2405 
2406 	return 0;
2407 }
2408 
2409 /*
2410  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2411  *
2412  * This option will effect the way delayed acks are performed.  This
2413  * option allows you to get or set the delayed ack time, in
2414  * milliseconds.  It also allows changing the delayed ack frequency.
2415  * Changing the frequency to 1 disables the delayed sack algorithm.  If
2416  * the assoc_id is 0, then this sets or gets the endpoints default
2417  * values.  If the assoc_id field is non-zero, then the set or get
2418  * effects the specified association for the one to many model (the
2419  * assoc_id field is ignored by the one to one model).  Note that if
2420  * sack_delay or sack_freq are 0 when setting this option, then the
2421  * current values will remain unchanged.
2422  *
2423  * struct sctp_sack_info {
2424  *     sctp_assoc_t            sack_assoc_id;
2425  *     uint32_t                sack_delay;
2426  *     uint32_t                sack_freq;
2427  * };
2428  *
2429  * sack_assoc_id -  This parameter, indicates which association the user
2430  *    is performing an action upon.  Note that if this field's value is
2431  *    zero then the endpoints default value is changed (effecting future
2432  *    associations only).
2433  *
2434  * sack_delay -  This parameter contains the number of milliseconds that
2435  *    the user is requesting the delayed ACK timer be set to.  Note that
2436  *    this value is defined in the standard to be between 200 and 500
2437  *    milliseconds.
2438  *
2439  * sack_freq -  This parameter contains the number of packets that must
2440  *    be received before a sack is sent without waiting for the delay
2441  *    timer to expire.  The default value for this is 2, setting this
2442  *    value to 1 will disable the delayed sack algorithm.
2443  */
2444 
2445 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2446 				       char __user *optval, unsigned int optlen)
2447 {
2448 	struct sctp_sack_info    params;
2449 	struct sctp_transport   *trans = NULL;
2450 	struct sctp_association *asoc = NULL;
2451 	struct sctp_sock        *sp = sctp_sk(sk);
2452 
2453 	if (optlen == sizeof(struct sctp_sack_info)) {
2454 		if (copy_from_user(&params, optval, optlen))
2455 			return -EFAULT;
2456 
2457 		if (params.sack_delay == 0 && params.sack_freq == 0)
2458 			return 0;
2459 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2460 		printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2461 		       "in delayed_ack socket option deprecated\n");
2462 		printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2463 		if (copy_from_user(&params, optval, optlen))
2464 			return -EFAULT;
2465 
2466 		if (params.sack_delay == 0)
2467 			params.sack_freq = 1;
2468 		else
2469 			params.sack_freq = 0;
2470 	} else
2471 		return - EINVAL;
2472 
2473 	/* Validate value parameter. */
2474 	if (params.sack_delay > 500)
2475 		return -EINVAL;
2476 
2477 	/* Get association, if sack_assoc_id != 0 and the socket is a one
2478 	 * to many style socket, and an association was not found, then
2479 	 * the id was invalid.
2480 	 */
2481 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2482 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2483 		return -EINVAL;
2484 
2485 	if (params.sack_delay) {
2486 		if (asoc) {
2487 			asoc->sackdelay =
2488 				msecs_to_jiffies(params.sack_delay);
2489 			asoc->param_flags =
2490 				(asoc->param_flags & ~SPP_SACKDELAY) |
2491 				SPP_SACKDELAY_ENABLE;
2492 		} else {
2493 			sp->sackdelay = params.sack_delay;
2494 			sp->param_flags =
2495 				(sp->param_flags & ~SPP_SACKDELAY) |
2496 				SPP_SACKDELAY_ENABLE;
2497 		}
2498 	}
2499 
2500 	if (params.sack_freq == 1) {
2501 		if (asoc) {
2502 			asoc->param_flags =
2503 				(asoc->param_flags & ~SPP_SACKDELAY) |
2504 				SPP_SACKDELAY_DISABLE;
2505 		} else {
2506 			sp->param_flags =
2507 				(sp->param_flags & ~SPP_SACKDELAY) |
2508 				SPP_SACKDELAY_DISABLE;
2509 		}
2510 	} else if (params.sack_freq > 1) {
2511 		if (asoc) {
2512 			asoc->sackfreq = params.sack_freq;
2513 			asoc->param_flags =
2514 				(asoc->param_flags & ~SPP_SACKDELAY) |
2515 				SPP_SACKDELAY_ENABLE;
2516 		} else {
2517 			sp->sackfreq = params.sack_freq;
2518 			sp->param_flags =
2519 				(sp->param_flags & ~SPP_SACKDELAY) |
2520 				SPP_SACKDELAY_ENABLE;
2521 		}
2522 	}
2523 
2524 	/* If change is for association, also apply to each transport. */
2525 	if (asoc) {
2526 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2527 				transports) {
2528 			if (params.sack_delay) {
2529 				trans->sackdelay =
2530 					msecs_to_jiffies(params.sack_delay);
2531 				trans->param_flags =
2532 					(trans->param_flags & ~SPP_SACKDELAY) |
2533 					SPP_SACKDELAY_ENABLE;
2534 			}
2535 			if (params.sack_freq == 1) {
2536 				trans->param_flags =
2537 					(trans->param_flags & ~SPP_SACKDELAY) |
2538 					SPP_SACKDELAY_DISABLE;
2539 			} else if (params.sack_freq > 1) {
2540 				trans->sackfreq = params.sack_freq;
2541 				trans->param_flags =
2542 					(trans->param_flags & ~SPP_SACKDELAY) |
2543 					SPP_SACKDELAY_ENABLE;
2544 			}
2545 		}
2546 	}
2547 
2548 	return 0;
2549 }
2550 
2551 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2552  *
2553  * Applications can specify protocol parameters for the default association
2554  * initialization.  The option name argument to setsockopt() and getsockopt()
2555  * is SCTP_INITMSG.
2556  *
2557  * Setting initialization parameters is effective only on an unconnected
2558  * socket (for UDP-style sockets only future associations are effected
2559  * by the change).  With TCP-style sockets, this option is inherited by
2560  * sockets derived from a listener socket.
2561  */
2562 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2563 {
2564 	struct sctp_initmsg sinit;
2565 	struct sctp_sock *sp = sctp_sk(sk);
2566 
2567 	if (optlen != sizeof(struct sctp_initmsg))
2568 		return -EINVAL;
2569 	if (copy_from_user(&sinit, optval, optlen))
2570 		return -EFAULT;
2571 
2572 	if (sinit.sinit_num_ostreams)
2573 		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2574 	if (sinit.sinit_max_instreams)
2575 		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2576 	if (sinit.sinit_max_attempts)
2577 		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2578 	if (sinit.sinit_max_init_timeo)
2579 		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2580 
2581 	return 0;
2582 }
2583 
2584 /*
2585  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2586  *
2587  *   Applications that wish to use the sendto() system call may wish to
2588  *   specify a default set of parameters that would normally be supplied
2589  *   through the inclusion of ancillary data.  This socket option allows
2590  *   such an application to set the default sctp_sndrcvinfo structure.
2591  *   The application that wishes to use this socket option simply passes
2592  *   in to this call the sctp_sndrcvinfo structure defined in Section
2593  *   5.2.2) The input parameters accepted by this call include
2594  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2595  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2596  *   to this call if the caller is using the UDP model.
2597  */
2598 static int sctp_setsockopt_default_send_param(struct sock *sk,
2599 					      char __user *optval,
2600 					      unsigned int optlen)
2601 {
2602 	struct sctp_sndrcvinfo info;
2603 	struct sctp_association *asoc;
2604 	struct sctp_sock *sp = sctp_sk(sk);
2605 
2606 	if (optlen != sizeof(struct sctp_sndrcvinfo))
2607 		return -EINVAL;
2608 	if (copy_from_user(&info, optval, optlen))
2609 		return -EFAULT;
2610 
2611 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2612 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2613 		return -EINVAL;
2614 
2615 	if (asoc) {
2616 		asoc->default_stream = info.sinfo_stream;
2617 		asoc->default_flags = info.sinfo_flags;
2618 		asoc->default_ppid = info.sinfo_ppid;
2619 		asoc->default_context = info.sinfo_context;
2620 		asoc->default_timetolive = info.sinfo_timetolive;
2621 	} else {
2622 		sp->default_stream = info.sinfo_stream;
2623 		sp->default_flags = info.sinfo_flags;
2624 		sp->default_ppid = info.sinfo_ppid;
2625 		sp->default_context = info.sinfo_context;
2626 		sp->default_timetolive = info.sinfo_timetolive;
2627 	}
2628 
2629 	return 0;
2630 }
2631 
2632 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2633  *
2634  * Requests that the local SCTP stack use the enclosed peer address as
2635  * the association primary.  The enclosed address must be one of the
2636  * association peer's addresses.
2637  */
2638 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2639 					unsigned int optlen)
2640 {
2641 	struct sctp_prim prim;
2642 	struct sctp_transport *trans;
2643 
2644 	if (optlen != sizeof(struct sctp_prim))
2645 		return -EINVAL;
2646 
2647 	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2648 		return -EFAULT;
2649 
2650 	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2651 	if (!trans)
2652 		return -EINVAL;
2653 
2654 	sctp_assoc_set_primary(trans->asoc, trans);
2655 
2656 	return 0;
2657 }
2658 
2659 /*
2660  * 7.1.5 SCTP_NODELAY
2661  *
2662  * Turn on/off any Nagle-like algorithm.  This means that packets are
2663  * generally sent as soon as possible and no unnecessary delays are
2664  * introduced, at the cost of more packets in the network.  Expects an
2665  *  integer boolean flag.
2666  */
2667 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2668 				   unsigned int optlen)
2669 {
2670 	int val;
2671 
2672 	if (optlen < sizeof(int))
2673 		return -EINVAL;
2674 	if (get_user(val, (int __user *)optval))
2675 		return -EFAULT;
2676 
2677 	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2678 	return 0;
2679 }
2680 
2681 /*
2682  *
2683  * 7.1.1 SCTP_RTOINFO
2684  *
2685  * The protocol parameters used to initialize and bound retransmission
2686  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2687  * and modify these parameters.
2688  * All parameters are time values, in milliseconds.  A value of 0, when
2689  * modifying the parameters, indicates that the current value should not
2690  * be changed.
2691  *
2692  */
2693 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2694 {
2695 	struct sctp_rtoinfo rtoinfo;
2696 	struct sctp_association *asoc;
2697 
2698 	if (optlen != sizeof (struct sctp_rtoinfo))
2699 		return -EINVAL;
2700 
2701 	if (copy_from_user(&rtoinfo, optval, optlen))
2702 		return -EFAULT;
2703 
2704 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2705 
2706 	/* Set the values to the specific association */
2707 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2708 		return -EINVAL;
2709 
2710 	if (asoc) {
2711 		if (rtoinfo.srto_initial != 0)
2712 			asoc->rto_initial =
2713 				msecs_to_jiffies(rtoinfo.srto_initial);
2714 		if (rtoinfo.srto_max != 0)
2715 			asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2716 		if (rtoinfo.srto_min != 0)
2717 			asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2718 	} else {
2719 		/* If there is no association or the association-id = 0
2720 		 * set the values to the endpoint.
2721 		 */
2722 		struct sctp_sock *sp = sctp_sk(sk);
2723 
2724 		if (rtoinfo.srto_initial != 0)
2725 			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2726 		if (rtoinfo.srto_max != 0)
2727 			sp->rtoinfo.srto_max = rtoinfo.srto_max;
2728 		if (rtoinfo.srto_min != 0)
2729 			sp->rtoinfo.srto_min = rtoinfo.srto_min;
2730 	}
2731 
2732 	return 0;
2733 }
2734 
2735 /*
2736  *
2737  * 7.1.2 SCTP_ASSOCINFO
2738  *
2739  * This option is used to tune the maximum retransmission attempts
2740  * of the association.
2741  * Returns an error if the new association retransmission value is
2742  * greater than the sum of the retransmission value  of the peer.
2743  * See [SCTP] for more information.
2744  *
2745  */
2746 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2747 {
2748 
2749 	struct sctp_assocparams assocparams;
2750 	struct sctp_association *asoc;
2751 
2752 	if (optlen != sizeof(struct sctp_assocparams))
2753 		return -EINVAL;
2754 	if (copy_from_user(&assocparams, optval, optlen))
2755 		return -EFAULT;
2756 
2757 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2758 
2759 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2760 		return -EINVAL;
2761 
2762 	/* Set the values to the specific association */
2763 	if (asoc) {
2764 		if (assocparams.sasoc_asocmaxrxt != 0) {
2765 			__u32 path_sum = 0;
2766 			int   paths = 0;
2767 			struct sctp_transport *peer_addr;
2768 
2769 			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2770 					transports) {
2771 				path_sum += peer_addr->pathmaxrxt;
2772 				paths++;
2773 			}
2774 
2775 			/* Only validate asocmaxrxt if we have more than
2776 			 * one path/transport.  We do this because path
2777 			 * retransmissions are only counted when we have more
2778 			 * then one path.
2779 			 */
2780 			if (paths > 1 &&
2781 			    assocparams.sasoc_asocmaxrxt > path_sum)
2782 				return -EINVAL;
2783 
2784 			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2785 		}
2786 
2787 		if (assocparams.sasoc_cookie_life != 0) {
2788 			asoc->cookie_life.tv_sec =
2789 					assocparams.sasoc_cookie_life / 1000;
2790 			asoc->cookie_life.tv_usec =
2791 					(assocparams.sasoc_cookie_life % 1000)
2792 					* 1000;
2793 		}
2794 	} else {
2795 		/* Set the values to the endpoint */
2796 		struct sctp_sock *sp = sctp_sk(sk);
2797 
2798 		if (assocparams.sasoc_asocmaxrxt != 0)
2799 			sp->assocparams.sasoc_asocmaxrxt =
2800 						assocparams.sasoc_asocmaxrxt;
2801 		if (assocparams.sasoc_cookie_life != 0)
2802 			sp->assocparams.sasoc_cookie_life =
2803 						assocparams.sasoc_cookie_life;
2804 	}
2805 	return 0;
2806 }
2807 
2808 /*
2809  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2810  *
2811  * This socket option is a boolean flag which turns on or off mapped V4
2812  * addresses.  If this option is turned on and the socket is type
2813  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2814  * If this option is turned off, then no mapping will be done of V4
2815  * addresses and a user will receive both PF_INET6 and PF_INET type
2816  * addresses on the socket.
2817  */
2818 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2819 {
2820 	int val;
2821 	struct sctp_sock *sp = sctp_sk(sk);
2822 
2823 	if (optlen < sizeof(int))
2824 		return -EINVAL;
2825 	if (get_user(val, (int __user *)optval))
2826 		return -EFAULT;
2827 	if (val)
2828 		sp->v4mapped = 1;
2829 	else
2830 		sp->v4mapped = 0;
2831 
2832 	return 0;
2833 }
2834 
2835 /*
2836  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2837  * This option will get or set the maximum size to put in any outgoing
2838  * SCTP DATA chunk.  If a message is larger than this size it will be
2839  * fragmented by SCTP into the specified size.  Note that the underlying
2840  * SCTP implementation may fragment into smaller sized chunks when the
2841  * PMTU of the underlying association is smaller than the value set by
2842  * the user.  The default value for this option is '0' which indicates
2843  * the user is NOT limiting fragmentation and only the PMTU will effect
2844  * SCTP's choice of DATA chunk size.  Note also that values set larger
2845  * than the maximum size of an IP datagram will effectively let SCTP
2846  * control fragmentation (i.e. the same as setting this option to 0).
2847  *
2848  * The following structure is used to access and modify this parameter:
2849  *
2850  * struct sctp_assoc_value {
2851  *   sctp_assoc_t assoc_id;
2852  *   uint32_t assoc_value;
2853  * };
2854  *
2855  * assoc_id:  This parameter is ignored for one-to-one style sockets.
2856  *    For one-to-many style sockets this parameter indicates which
2857  *    association the user is performing an action upon.  Note that if
2858  *    this field's value is zero then the endpoints default value is
2859  *    changed (effecting future associations only).
2860  * assoc_value:  This parameter specifies the maximum size in bytes.
2861  */
2862 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2863 {
2864 	struct sctp_assoc_value params;
2865 	struct sctp_association *asoc;
2866 	struct sctp_sock *sp = sctp_sk(sk);
2867 	int val;
2868 
2869 	if (optlen == sizeof(int)) {
2870 		printk(KERN_WARNING
2871 		   "SCTP: Use of int in maxseg socket option deprecated\n");
2872 		printk(KERN_WARNING
2873 		   "SCTP: Use struct sctp_assoc_value instead\n");
2874 		if (copy_from_user(&val, optval, optlen))
2875 			return -EFAULT;
2876 		params.assoc_id = 0;
2877 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2878 		if (copy_from_user(&params, optval, optlen))
2879 			return -EFAULT;
2880 		val = params.assoc_value;
2881 	} else
2882 		return -EINVAL;
2883 
2884 	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2885 		return -EINVAL;
2886 
2887 	asoc = sctp_id2assoc(sk, params.assoc_id);
2888 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2889 		return -EINVAL;
2890 
2891 	if (asoc) {
2892 		if (val == 0) {
2893 			val = asoc->pathmtu;
2894 			val -= sp->pf->af->net_header_len;
2895 			val -= sizeof(struct sctphdr) +
2896 					sizeof(struct sctp_data_chunk);
2897 		}
2898 		asoc->user_frag = val;
2899 		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2900 	} else {
2901 		sp->user_frag = val;
2902 	}
2903 
2904 	return 0;
2905 }
2906 
2907 
2908 /*
2909  *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2910  *
2911  *   Requests that the peer mark the enclosed address as the association
2912  *   primary. The enclosed address must be one of the association's
2913  *   locally bound addresses. The following structure is used to make a
2914  *   set primary request:
2915  */
2916 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2917 					     unsigned int optlen)
2918 {
2919 	struct sctp_sock	*sp;
2920 	struct sctp_endpoint	*ep;
2921 	struct sctp_association	*asoc = NULL;
2922 	struct sctp_setpeerprim	prim;
2923 	struct sctp_chunk	*chunk;
2924 	int 			err;
2925 
2926 	sp = sctp_sk(sk);
2927 	ep = sp->ep;
2928 
2929 	if (!sctp_addip_enable)
2930 		return -EPERM;
2931 
2932 	if (optlen != sizeof(struct sctp_setpeerprim))
2933 		return -EINVAL;
2934 
2935 	if (copy_from_user(&prim, optval, optlen))
2936 		return -EFAULT;
2937 
2938 	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2939 	if (!asoc)
2940 		return -EINVAL;
2941 
2942 	if (!asoc->peer.asconf_capable)
2943 		return -EPERM;
2944 
2945 	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2946 		return -EPERM;
2947 
2948 	if (!sctp_state(asoc, ESTABLISHED))
2949 		return -ENOTCONN;
2950 
2951 	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2952 		return -EADDRNOTAVAIL;
2953 
2954 	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
2955 	chunk = sctp_make_asconf_set_prim(asoc,
2956 					  (union sctp_addr *)&prim.sspp_addr);
2957 	if (!chunk)
2958 		return -ENOMEM;
2959 
2960 	err = sctp_send_asconf(asoc, chunk);
2961 
2962 	SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2963 
2964 	return err;
2965 }
2966 
2967 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2968 					    unsigned int optlen)
2969 {
2970 	struct sctp_setadaptation adaptation;
2971 
2972 	if (optlen != sizeof(struct sctp_setadaptation))
2973 		return -EINVAL;
2974 	if (copy_from_user(&adaptation, optval, optlen))
2975 		return -EFAULT;
2976 
2977 	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2978 
2979 	return 0;
2980 }
2981 
2982 /*
2983  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
2984  *
2985  * The context field in the sctp_sndrcvinfo structure is normally only
2986  * used when a failed message is retrieved holding the value that was
2987  * sent down on the actual send call.  This option allows the setting of
2988  * a default context on an association basis that will be received on
2989  * reading messages from the peer.  This is especially helpful in the
2990  * one-2-many model for an application to keep some reference to an
2991  * internal state machine that is processing messages on the
2992  * association.  Note that the setting of this value only effects
2993  * received messages from the peer and does not effect the value that is
2994  * saved with outbound messages.
2995  */
2996 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2997 				   unsigned int optlen)
2998 {
2999 	struct sctp_assoc_value params;
3000 	struct sctp_sock *sp;
3001 	struct sctp_association *asoc;
3002 
3003 	if (optlen != sizeof(struct sctp_assoc_value))
3004 		return -EINVAL;
3005 	if (copy_from_user(&params, optval, optlen))
3006 		return -EFAULT;
3007 
3008 	sp = sctp_sk(sk);
3009 
3010 	if (params.assoc_id != 0) {
3011 		asoc = sctp_id2assoc(sk, params.assoc_id);
3012 		if (!asoc)
3013 			return -EINVAL;
3014 		asoc->default_rcv_context = params.assoc_value;
3015 	} else {
3016 		sp->default_rcv_context = params.assoc_value;
3017 	}
3018 
3019 	return 0;
3020 }
3021 
3022 /*
3023  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3024  *
3025  * This options will at a minimum specify if the implementation is doing
3026  * fragmented interleave.  Fragmented interleave, for a one to many
3027  * socket, is when subsequent calls to receive a message may return
3028  * parts of messages from different associations.  Some implementations
3029  * may allow you to turn this value on or off.  If so, when turned off,
3030  * no fragment interleave will occur (which will cause a head of line
3031  * blocking amongst multiple associations sharing the same one to many
3032  * socket).  When this option is turned on, then each receive call may
3033  * come from a different association (thus the user must receive data
3034  * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3035  * association each receive belongs to.
3036  *
3037  * This option takes a boolean value.  A non-zero value indicates that
3038  * fragmented interleave is on.  A value of zero indicates that
3039  * fragmented interleave is off.
3040  *
3041  * Note that it is important that an implementation that allows this
3042  * option to be turned on, have it off by default.  Otherwise an unaware
3043  * application using the one to many model may become confused and act
3044  * incorrectly.
3045  */
3046 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3047 					       char __user *optval,
3048 					       unsigned int optlen)
3049 {
3050 	int val;
3051 
3052 	if (optlen != sizeof(int))
3053 		return -EINVAL;
3054 	if (get_user(val, (int __user *)optval))
3055 		return -EFAULT;
3056 
3057 	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3058 
3059 	return 0;
3060 }
3061 
3062 /*
3063  * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3064  *       (SCTP_PARTIAL_DELIVERY_POINT)
3065  *
3066  * This option will set or get the SCTP partial delivery point.  This
3067  * point is the size of a message where the partial delivery API will be
3068  * invoked to help free up rwnd space for the peer.  Setting this to a
3069  * lower value will cause partial deliveries to happen more often.  The
3070  * calls argument is an integer that sets or gets the partial delivery
3071  * point.  Note also that the call will fail if the user attempts to set
3072  * this value larger than the socket receive buffer size.
3073  *
3074  * Note that any single message having a length smaller than or equal to
3075  * the SCTP partial delivery point will be delivered in one single read
3076  * call as long as the user provided buffer is large enough to hold the
3077  * message.
3078  */
3079 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3080 						  char __user *optval,
3081 						  unsigned int optlen)
3082 {
3083 	u32 val;
3084 
3085 	if (optlen != sizeof(u32))
3086 		return -EINVAL;
3087 	if (get_user(val, (int __user *)optval))
3088 		return -EFAULT;
3089 
3090 	/* Note: We double the receive buffer from what the user sets
3091 	 * it to be, also initial rwnd is based on rcvbuf/2.
3092 	 */
3093 	if (val > (sk->sk_rcvbuf >> 1))
3094 		return -EINVAL;
3095 
3096 	sctp_sk(sk)->pd_point = val;
3097 
3098 	return 0; /* is this the right error code? */
3099 }
3100 
3101 /*
3102  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3103  *
3104  * This option will allow a user to change the maximum burst of packets
3105  * that can be emitted by this association.  Note that the default value
3106  * is 4, and some implementations may restrict this setting so that it
3107  * can only be lowered.
3108  *
3109  * NOTE: This text doesn't seem right.  Do this on a socket basis with
3110  * future associations inheriting the socket value.
3111  */
3112 static int sctp_setsockopt_maxburst(struct sock *sk,
3113 				    char __user *optval,
3114 				    unsigned int optlen)
3115 {
3116 	struct sctp_assoc_value params;
3117 	struct sctp_sock *sp;
3118 	struct sctp_association *asoc;
3119 	int val;
3120 	int assoc_id = 0;
3121 
3122 	if (optlen == sizeof(int)) {
3123 		printk(KERN_WARNING
3124 		   "SCTP: Use of int in max_burst socket option deprecated\n");
3125 		printk(KERN_WARNING
3126 		   "SCTP: Use struct sctp_assoc_value instead\n");
3127 		if (copy_from_user(&val, optval, optlen))
3128 			return -EFAULT;
3129 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3130 		if (copy_from_user(&params, optval, optlen))
3131 			return -EFAULT;
3132 		val = params.assoc_value;
3133 		assoc_id = params.assoc_id;
3134 	} else
3135 		return -EINVAL;
3136 
3137 	sp = sctp_sk(sk);
3138 
3139 	if (assoc_id != 0) {
3140 		asoc = sctp_id2assoc(sk, assoc_id);
3141 		if (!asoc)
3142 			return -EINVAL;
3143 		asoc->max_burst = val;
3144 	} else
3145 		sp->max_burst = val;
3146 
3147 	return 0;
3148 }
3149 
3150 /*
3151  * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3152  *
3153  * This set option adds a chunk type that the user is requesting to be
3154  * received only in an authenticated way.  Changes to the list of chunks
3155  * will only effect future associations on the socket.
3156  */
3157 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3158 				      char __user *optval,
3159 				      unsigned int optlen)
3160 {
3161 	struct sctp_authchunk val;
3162 
3163 	if (!sctp_auth_enable)
3164 		return -EACCES;
3165 
3166 	if (optlen != sizeof(struct sctp_authchunk))
3167 		return -EINVAL;
3168 	if (copy_from_user(&val, optval, optlen))
3169 		return -EFAULT;
3170 
3171 	switch (val.sauth_chunk) {
3172 		case SCTP_CID_INIT:
3173 		case SCTP_CID_INIT_ACK:
3174 		case SCTP_CID_SHUTDOWN_COMPLETE:
3175 		case SCTP_CID_AUTH:
3176 			return -EINVAL;
3177 	}
3178 
3179 	/* add this chunk id to the endpoint */
3180 	return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3181 }
3182 
3183 /*
3184  * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3185  *
3186  * This option gets or sets the list of HMAC algorithms that the local
3187  * endpoint requires the peer to use.
3188  */
3189 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3190 				      char __user *optval,
3191 				      unsigned int optlen)
3192 {
3193 	struct sctp_hmacalgo *hmacs;
3194 	u32 idents;
3195 	int err;
3196 
3197 	if (!sctp_auth_enable)
3198 		return -EACCES;
3199 
3200 	if (optlen < sizeof(struct sctp_hmacalgo))
3201 		return -EINVAL;
3202 
3203 	hmacs = kmalloc(optlen, GFP_KERNEL);
3204 	if (!hmacs)
3205 		return -ENOMEM;
3206 
3207 	if (copy_from_user(hmacs, optval, optlen)) {
3208 		err = -EFAULT;
3209 		goto out;
3210 	}
3211 
3212 	idents = hmacs->shmac_num_idents;
3213 	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3214 	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3215 		err = -EINVAL;
3216 		goto out;
3217 	}
3218 
3219 	err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3220 out:
3221 	kfree(hmacs);
3222 	return err;
3223 }
3224 
3225 /*
3226  * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3227  *
3228  * This option will set a shared secret key which is used to build an
3229  * association shared key.
3230  */
3231 static int sctp_setsockopt_auth_key(struct sock *sk,
3232 				    char __user *optval,
3233 				    unsigned int optlen)
3234 {
3235 	struct sctp_authkey *authkey;
3236 	struct sctp_association *asoc;
3237 	int ret;
3238 
3239 	if (!sctp_auth_enable)
3240 		return -EACCES;
3241 
3242 	if (optlen <= sizeof(struct sctp_authkey))
3243 		return -EINVAL;
3244 
3245 	authkey = kmalloc(optlen, GFP_KERNEL);
3246 	if (!authkey)
3247 		return -ENOMEM;
3248 
3249 	if (copy_from_user(authkey, optval, optlen)) {
3250 		ret = -EFAULT;
3251 		goto out;
3252 	}
3253 
3254 	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3255 		ret = -EINVAL;
3256 		goto out;
3257 	}
3258 
3259 	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3260 	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3261 		ret = -EINVAL;
3262 		goto out;
3263 	}
3264 
3265 	ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3266 out:
3267 	kfree(authkey);
3268 	return ret;
3269 }
3270 
3271 /*
3272  * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3273  *
3274  * This option will get or set the active shared key to be used to build
3275  * the association shared key.
3276  */
3277 static int sctp_setsockopt_active_key(struct sock *sk,
3278 				      char __user *optval,
3279 				      unsigned int optlen)
3280 {
3281 	struct sctp_authkeyid val;
3282 	struct sctp_association *asoc;
3283 
3284 	if (!sctp_auth_enable)
3285 		return -EACCES;
3286 
3287 	if (optlen != sizeof(struct sctp_authkeyid))
3288 		return -EINVAL;
3289 	if (copy_from_user(&val, optval, optlen))
3290 		return -EFAULT;
3291 
3292 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3293 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3294 		return -EINVAL;
3295 
3296 	return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3297 					val.scact_keynumber);
3298 }
3299 
3300 /*
3301  * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3302  *
3303  * This set option will delete a shared secret key from use.
3304  */
3305 static int sctp_setsockopt_del_key(struct sock *sk,
3306 				   char __user *optval,
3307 				   unsigned int optlen)
3308 {
3309 	struct sctp_authkeyid val;
3310 	struct sctp_association *asoc;
3311 
3312 	if (!sctp_auth_enable)
3313 		return -EACCES;
3314 
3315 	if (optlen != sizeof(struct sctp_authkeyid))
3316 		return -EINVAL;
3317 	if (copy_from_user(&val, optval, optlen))
3318 		return -EFAULT;
3319 
3320 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3321 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3322 		return -EINVAL;
3323 
3324 	return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3325 				    val.scact_keynumber);
3326 
3327 }
3328 
3329 
3330 /* API 6.2 setsockopt(), getsockopt()
3331  *
3332  * Applications use setsockopt() and getsockopt() to set or retrieve
3333  * socket options.  Socket options are used to change the default
3334  * behavior of sockets calls.  They are described in Section 7.
3335  *
3336  * The syntax is:
3337  *
3338  *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3339  *                    int __user *optlen);
3340  *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3341  *                    int optlen);
3342  *
3343  *   sd      - the socket descript.
3344  *   level   - set to IPPROTO_SCTP for all SCTP options.
3345  *   optname - the option name.
3346  *   optval  - the buffer to store the value of the option.
3347  *   optlen  - the size of the buffer.
3348  */
3349 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3350 				char __user *optval, unsigned int optlen)
3351 {
3352 	int retval = 0;
3353 
3354 	SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3355 			  sk, optname);
3356 
3357 	/* I can hardly begin to describe how wrong this is.  This is
3358 	 * so broken as to be worse than useless.  The API draft
3359 	 * REALLY is NOT helpful here...  I am not convinced that the
3360 	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3361 	 * are at all well-founded.
3362 	 */
3363 	if (level != SOL_SCTP) {
3364 		struct sctp_af *af = sctp_sk(sk)->pf->af;
3365 		retval = af->setsockopt(sk, level, optname, optval, optlen);
3366 		goto out_nounlock;
3367 	}
3368 
3369 	sctp_lock_sock(sk);
3370 
3371 	switch (optname) {
3372 	case SCTP_SOCKOPT_BINDX_ADD:
3373 		/* 'optlen' is the size of the addresses buffer. */
3374 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3375 					       optlen, SCTP_BINDX_ADD_ADDR);
3376 		break;
3377 
3378 	case SCTP_SOCKOPT_BINDX_REM:
3379 		/* 'optlen' is the size of the addresses buffer. */
3380 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3381 					       optlen, SCTP_BINDX_REM_ADDR);
3382 		break;
3383 
3384 	case SCTP_SOCKOPT_CONNECTX_OLD:
3385 		/* 'optlen' is the size of the addresses buffer. */
3386 		retval = sctp_setsockopt_connectx_old(sk,
3387 					    (struct sockaddr __user *)optval,
3388 					    optlen);
3389 		break;
3390 
3391 	case SCTP_SOCKOPT_CONNECTX:
3392 		/* 'optlen' is the size of the addresses buffer. */
3393 		retval = sctp_setsockopt_connectx(sk,
3394 					    (struct sockaddr __user *)optval,
3395 					    optlen);
3396 		break;
3397 
3398 	case SCTP_DISABLE_FRAGMENTS:
3399 		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3400 		break;
3401 
3402 	case SCTP_EVENTS:
3403 		retval = sctp_setsockopt_events(sk, optval, optlen);
3404 		break;
3405 
3406 	case SCTP_AUTOCLOSE:
3407 		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3408 		break;
3409 
3410 	case SCTP_PEER_ADDR_PARAMS:
3411 		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3412 		break;
3413 
3414 	case SCTP_DELAYED_ACK:
3415 		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3416 		break;
3417 	case SCTP_PARTIAL_DELIVERY_POINT:
3418 		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3419 		break;
3420 
3421 	case SCTP_INITMSG:
3422 		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3423 		break;
3424 	case SCTP_DEFAULT_SEND_PARAM:
3425 		retval = sctp_setsockopt_default_send_param(sk, optval,
3426 							    optlen);
3427 		break;
3428 	case SCTP_PRIMARY_ADDR:
3429 		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3430 		break;
3431 	case SCTP_SET_PEER_PRIMARY_ADDR:
3432 		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3433 		break;
3434 	case SCTP_NODELAY:
3435 		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3436 		break;
3437 	case SCTP_RTOINFO:
3438 		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3439 		break;
3440 	case SCTP_ASSOCINFO:
3441 		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3442 		break;
3443 	case SCTP_I_WANT_MAPPED_V4_ADDR:
3444 		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3445 		break;
3446 	case SCTP_MAXSEG:
3447 		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3448 		break;
3449 	case SCTP_ADAPTATION_LAYER:
3450 		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3451 		break;
3452 	case SCTP_CONTEXT:
3453 		retval = sctp_setsockopt_context(sk, optval, optlen);
3454 		break;
3455 	case SCTP_FRAGMENT_INTERLEAVE:
3456 		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3457 		break;
3458 	case SCTP_MAX_BURST:
3459 		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3460 		break;
3461 	case SCTP_AUTH_CHUNK:
3462 		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3463 		break;
3464 	case SCTP_HMAC_IDENT:
3465 		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3466 		break;
3467 	case SCTP_AUTH_KEY:
3468 		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3469 		break;
3470 	case SCTP_AUTH_ACTIVE_KEY:
3471 		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3472 		break;
3473 	case SCTP_AUTH_DELETE_KEY:
3474 		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3475 		break;
3476 	default:
3477 		retval = -ENOPROTOOPT;
3478 		break;
3479 	}
3480 
3481 	sctp_release_sock(sk);
3482 
3483 out_nounlock:
3484 	return retval;
3485 }
3486 
3487 /* API 3.1.6 connect() - UDP Style Syntax
3488  *
3489  * An application may use the connect() call in the UDP model to initiate an
3490  * association without sending data.
3491  *
3492  * The syntax is:
3493  *
3494  * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3495  *
3496  * sd: the socket descriptor to have a new association added to.
3497  *
3498  * nam: the address structure (either struct sockaddr_in or struct
3499  *    sockaddr_in6 defined in RFC2553 [7]).
3500  *
3501  * len: the size of the address.
3502  */
3503 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3504 			     int addr_len)
3505 {
3506 	int err = 0;
3507 	struct sctp_af *af;
3508 
3509 	sctp_lock_sock(sk);
3510 
3511 	SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3512 			  __func__, sk, addr, addr_len);
3513 
3514 	/* Validate addr_len before calling common connect/connectx routine. */
3515 	af = sctp_get_af_specific(addr->sa_family);
3516 	if (!af || addr_len < af->sockaddr_len) {
3517 		err = -EINVAL;
3518 	} else {
3519 		/* Pass correct addr len to common routine (so it knows there
3520 		 * is only one address being passed.
3521 		 */
3522 		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3523 	}
3524 
3525 	sctp_release_sock(sk);
3526 	return err;
3527 }
3528 
3529 /* FIXME: Write comments. */
3530 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3531 {
3532 	return -EOPNOTSUPP; /* STUB */
3533 }
3534 
3535 /* 4.1.4 accept() - TCP Style Syntax
3536  *
3537  * Applications use accept() call to remove an established SCTP
3538  * association from the accept queue of the endpoint.  A new socket
3539  * descriptor will be returned from accept() to represent the newly
3540  * formed association.
3541  */
3542 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3543 {
3544 	struct sctp_sock *sp;
3545 	struct sctp_endpoint *ep;
3546 	struct sock *newsk = NULL;
3547 	struct sctp_association *asoc;
3548 	long timeo;
3549 	int error = 0;
3550 
3551 	sctp_lock_sock(sk);
3552 
3553 	sp = sctp_sk(sk);
3554 	ep = sp->ep;
3555 
3556 	if (!sctp_style(sk, TCP)) {
3557 		error = -EOPNOTSUPP;
3558 		goto out;
3559 	}
3560 
3561 	if (!sctp_sstate(sk, LISTENING)) {
3562 		error = -EINVAL;
3563 		goto out;
3564 	}
3565 
3566 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3567 
3568 	error = sctp_wait_for_accept(sk, timeo);
3569 	if (error)
3570 		goto out;
3571 
3572 	/* We treat the list of associations on the endpoint as the accept
3573 	 * queue and pick the first association on the list.
3574 	 */
3575 	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3576 
3577 	newsk = sp->pf->create_accept_sk(sk, asoc);
3578 	if (!newsk) {
3579 		error = -ENOMEM;
3580 		goto out;
3581 	}
3582 
3583 	/* Populate the fields of the newsk from the oldsk and migrate the
3584 	 * asoc to the newsk.
3585 	 */
3586 	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3587 
3588 out:
3589 	sctp_release_sock(sk);
3590 	*err = error;
3591 	return newsk;
3592 }
3593 
3594 /* The SCTP ioctl handler. */
3595 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3596 {
3597 	return -ENOIOCTLCMD;
3598 }
3599 
3600 /* This is the function which gets called during socket creation to
3601  * initialized the SCTP-specific portion of the sock.
3602  * The sock structure should already be zero-filled memory.
3603  */
3604 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3605 {
3606 	struct sctp_endpoint *ep;
3607 	struct sctp_sock *sp;
3608 
3609 	SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3610 
3611 	sp = sctp_sk(sk);
3612 
3613 	/* Initialize the SCTP per socket area.  */
3614 	switch (sk->sk_type) {
3615 	case SOCK_SEQPACKET:
3616 		sp->type = SCTP_SOCKET_UDP;
3617 		break;
3618 	case SOCK_STREAM:
3619 		sp->type = SCTP_SOCKET_TCP;
3620 		break;
3621 	default:
3622 		return -ESOCKTNOSUPPORT;
3623 	}
3624 
3625 	/* Initialize default send parameters. These parameters can be
3626 	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3627 	 */
3628 	sp->default_stream = 0;
3629 	sp->default_ppid = 0;
3630 	sp->default_flags = 0;
3631 	sp->default_context = 0;
3632 	sp->default_timetolive = 0;
3633 
3634 	sp->default_rcv_context = 0;
3635 	sp->max_burst = sctp_max_burst;
3636 
3637 	/* Initialize default setup parameters. These parameters
3638 	 * can be modified with the SCTP_INITMSG socket option or
3639 	 * overridden by the SCTP_INIT CMSG.
3640 	 */
3641 	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3642 	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3643 	sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3644 	sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3645 
3646 	/* Initialize default RTO related parameters.  These parameters can
3647 	 * be modified for with the SCTP_RTOINFO socket option.
3648 	 */
3649 	sp->rtoinfo.srto_initial = sctp_rto_initial;
3650 	sp->rtoinfo.srto_max     = sctp_rto_max;
3651 	sp->rtoinfo.srto_min     = sctp_rto_min;
3652 
3653 	/* Initialize default association related parameters. These parameters
3654 	 * can be modified with the SCTP_ASSOCINFO socket option.
3655 	 */
3656 	sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3657 	sp->assocparams.sasoc_number_peer_destinations = 0;
3658 	sp->assocparams.sasoc_peer_rwnd = 0;
3659 	sp->assocparams.sasoc_local_rwnd = 0;
3660 	sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3661 
3662 	/* Initialize default event subscriptions. By default, all the
3663 	 * options are off.
3664 	 */
3665 	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3666 
3667 	/* Default Peer Address Parameters.  These defaults can
3668 	 * be modified via SCTP_PEER_ADDR_PARAMS
3669 	 */
3670 	sp->hbinterval  = sctp_hb_interval;
3671 	sp->pathmaxrxt  = sctp_max_retrans_path;
3672 	sp->pathmtu     = 0; // allow default discovery
3673 	sp->sackdelay   = sctp_sack_timeout;
3674 	sp->sackfreq	= 2;
3675 	sp->param_flags = SPP_HB_ENABLE |
3676 			  SPP_PMTUD_ENABLE |
3677 			  SPP_SACKDELAY_ENABLE;
3678 
3679 	/* If enabled no SCTP message fragmentation will be performed.
3680 	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3681 	 */
3682 	sp->disable_fragments = 0;
3683 
3684 	/* Enable Nagle algorithm by default.  */
3685 	sp->nodelay           = 0;
3686 
3687 	/* Enable by default. */
3688 	sp->v4mapped          = 1;
3689 
3690 	/* Auto-close idle associations after the configured
3691 	 * number of seconds.  A value of 0 disables this
3692 	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3693 	 * for UDP-style sockets only.
3694 	 */
3695 	sp->autoclose         = 0;
3696 
3697 	/* User specified fragmentation limit. */
3698 	sp->user_frag         = 0;
3699 
3700 	sp->adaptation_ind = 0;
3701 
3702 	sp->pf = sctp_get_pf_specific(sk->sk_family);
3703 
3704 	/* Control variables for partial data delivery. */
3705 	atomic_set(&sp->pd_mode, 0);
3706 	skb_queue_head_init(&sp->pd_lobby);
3707 	sp->frag_interleave = 0;
3708 
3709 	/* Create a per socket endpoint structure.  Even if we
3710 	 * change the data structure relationships, this may still
3711 	 * be useful for storing pre-connect address information.
3712 	 */
3713 	ep = sctp_endpoint_new(sk, GFP_KERNEL);
3714 	if (!ep)
3715 		return -ENOMEM;
3716 
3717 	sp->ep = ep;
3718 	sp->hmac = NULL;
3719 
3720 	SCTP_DBG_OBJCNT_INC(sock);
3721 	percpu_counter_inc(&sctp_sockets_allocated);
3722 
3723 	local_bh_disable();
3724 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3725 	local_bh_enable();
3726 
3727 	return 0;
3728 }
3729 
3730 /* Cleanup any SCTP per socket resources.  */
3731 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3732 {
3733 	struct sctp_endpoint *ep;
3734 
3735 	SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3736 
3737 	/* Release our hold on the endpoint. */
3738 	ep = sctp_sk(sk)->ep;
3739 	sctp_endpoint_free(ep);
3740 	percpu_counter_dec(&sctp_sockets_allocated);
3741 	local_bh_disable();
3742 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3743 	local_bh_enable();
3744 }
3745 
3746 /* API 4.1.7 shutdown() - TCP Style Syntax
3747  *     int shutdown(int socket, int how);
3748  *
3749  *     sd      - the socket descriptor of the association to be closed.
3750  *     how     - Specifies the type of shutdown.  The  values  are
3751  *               as follows:
3752  *               SHUT_RD
3753  *                     Disables further receive operations. No SCTP
3754  *                     protocol action is taken.
3755  *               SHUT_WR
3756  *                     Disables further send operations, and initiates
3757  *                     the SCTP shutdown sequence.
3758  *               SHUT_RDWR
3759  *                     Disables further send  and  receive  operations
3760  *                     and initiates the SCTP shutdown sequence.
3761  */
3762 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3763 {
3764 	struct sctp_endpoint *ep;
3765 	struct sctp_association *asoc;
3766 
3767 	if (!sctp_style(sk, TCP))
3768 		return;
3769 
3770 	if (how & SEND_SHUTDOWN) {
3771 		ep = sctp_sk(sk)->ep;
3772 		if (!list_empty(&ep->asocs)) {
3773 			asoc = list_entry(ep->asocs.next,
3774 					  struct sctp_association, asocs);
3775 			sctp_primitive_SHUTDOWN(asoc, NULL);
3776 		}
3777 	}
3778 }
3779 
3780 /* 7.2.1 Association Status (SCTP_STATUS)
3781 
3782  * Applications can retrieve current status information about an
3783  * association, including association state, peer receiver window size,
3784  * number of unacked data chunks, and number of data chunks pending
3785  * receipt.  This information is read-only.
3786  */
3787 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3788 				       char __user *optval,
3789 				       int __user *optlen)
3790 {
3791 	struct sctp_status status;
3792 	struct sctp_association *asoc = NULL;
3793 	struct sctp_transport *transport;
3794 	sctp_assoc_t associd;
3795 	int retval = 0;
3796 
3797 	if (len < sizeof(status)) {
3798 		retval = -EINVAL;
3799 		goto out;
3800 	}
3801 
3802 	len = sizeof(status);
3803 	if (copy_from_user(&status, optval, len)) {
3804 		retval = -EFAULT;
3805 		goto out;
3806 	}
3807 
3808 	associd = status.sstat_assoc_id;
3809 	asoc = sctp_id2assoc(sk, associd);
3810 	if (!asoc) {
3811 		retval = -EINVAL;
3812 		goto out;
3813 	}
3814 
3815 	transport = asoc->peer.primary_path;
3816 
3817 	status.sstat_assoc_id = sctp_assoc2id(asoc);
3818 	status.sstat_state = asoc->state;
3819 	status.sstat_rwnd =  asoc->peer.rwnd;
3820 	status.sstat_unackdata = asoc->unack_data;
3821 
3822 	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3823 	status.sstat_instrms = asoc->c.sinit_max_instreams;
3824 	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3825 	status.sstat_fragmentation_point = asoc->frag_point;
3826 	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3827 	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3828 			transport->af_specific->sockaddr_len);
3829 	/* Map ipv4 address into v4-mapped-on-v6 address.  */
3830 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3831 		(union sctp_addr *)&status.sstat_primary.spinfo_address);
3832 	status.sstat_primary.spinfo_state = transport->state;
3833 	status.sstat_primary.spinfo_cwnd = transport->cwnd;
3834 	status.sstat_primary.spinfo_srtt = transport->srtt;
3835 	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3836 	status.sstat_primary.spinfo_mtu = transport->pathmtu;
3837 
3838 	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3839 		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3840 
3841 	if (put_user(len, optlen)) {
3842 		retval = -EFAULT;
3843 		goto out;
3844 	}
3845 
3846 	SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3847 			  len, status.sstat_state, status.sstat_rwnd,
3848 			  status.sstat_assoc_id);
3849 
3850 	if (copy_to_user(optval, &status, len)) {
3851 		retval = -EFAULT;
3852 		goto out;
3853 	}
3854 
3855 out:
3856 	return (retval);
3857 }
3858 
3859 
3860 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3861  *
3862  * Applications can retrieve information about a specific peer address
3863  * of an association, including its reachability state, congestion
3864  * window, and retransmission timer values.  This information is
3865  * read-only.
3866  */
3867 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3868 					  char __user *optval,
3869 					  int __user *optlen)
3870 {
3871 	struct sctp_paddrinfo pinfo;
3872 	struct sctp_transport *transport;
3873 	int retval = 0;
3874 
3875 	if (len < sizeof(pinfo)) {
3876 		retval = -EINVAL;
3877 		goto out;
3878 	}
3879 
3880 	len = sizeof(pinfo);
3881 	if (copy_from_user(&pinfo, optval, len)) {
3882 		retval = -EFAULT;
3883 		goto out;
3884 	}
3885 
3886 	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3887 					   pinfo.spinfo_assoc_id);
3888 	if (!transport)
3889 		return -EINVAL;
3890 
3891 	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3892 	pinfo.spinfo_state = transport->state;
3893 	pinfo.spinfo_cwnd = transport->cwnd;
3894 	pinfo.spinfo_srtt = transport->srtt;
3895 	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3896 	pinfo.spinfo_mtu = transport->pathmtu;
3897 
3898 	if (pinfo.spinfo_state == SCTP_UNKNOWN)
3899 		pinfo.spinfo_state = SCTP_ACTIVE;
3900 
3901 	if (put_user(len, optlen)) {
3902 		retval = -EFAULT;
3903 		goto out;
3904 	}
3905 
3906 	if (copy_to_user(optval, &pinfo, len)) {
3907 		retval = -EFAULT;
3908 		goto out;
3909 	}
3910 
3911 out:
3912 	return (retval);
3913 }
3914 
3915 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3916  *
3917  * This option is a on/off flag.  If enabled no SCTP message
3918  * fragmentation will be performed.  Instead if a message being sent
3919  * exceeds the current PMTU size, the message will NOT be sent and
3920  * instead a error will be indicated to the user.
3921  */
3922 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3923 					char __user *optval, int __user *optlen)
3924 {
3925 	int val;
3926 
3927 	if (len < sizeof(int))
3928 		return -EINVAL;
3929 
3930 	len = sizeof(int);
3931 	val = (sctp_sk(sk)->disable_fragments == 1);
3932 	if (put_user(len, optlen))
3933 		return -EFAULT;
3934 	if (copy_to_user(optval, &val, len))
3935 		return -EFAULT;
3936 	return 0;
3937 }
3938 
3939 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3940  *
3941  * This socket option is used to specify various notifications and
3942  * ancillary data the user wishes to receive.
3943  */
3944 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3945 				  int __user *optlen)
3946 {
3947 	if (len < sizeof(struct sctp_event_subscribe))
3948 		return -EINVAL;
3949 	len = sizeof(struct sctp_event_subscribe);
3950 	if (put_user(len, optlen))
3951 		return -EFAULT;
3952 	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3953 		return -EFAULT;
3954 	return 0;
3955 }
3956 
3957 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3958  *
3959  * This socket option is applicable to the UDP-style socket only.  When
3960  * set it will cause associations that are idle for more than the
3961  * specified number of seconds to automatically close.  An association
3962  * being idle is defined an association that has NOT sent or received
3963  * user data.  The special value of '0' indicates that no automatic
3964  * close of any associations should be performed.  The option expects an
3965  * integer defining the number of seconds of idle time before an
3966  * association is closed.
3967  */
3968 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3969 {
3970 	/* Applicable to UDP-style socket only */
3971 	if (sctp_style(sk, TCP))
3972 		return -EOPNOTSUPP;
3973 	if (len < sizeof(int))
3974 		return -EINVAL;
3975 	len = sizeof(int);
3976 	if (put_user(len, optlen))
3977 		return -EFAULT;
3978 	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3979 		return -EFAULT;
3980 	return 0;
3981 }
3982 
3983 /* Helper routine to branch off an association to a new socket.  */
3984 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3985 				struct socket **sockp)
3986 {
3987 	struct sock *sk = asoc->base.sk;
3988 	struct socket *sock;
3989 	struct sctp_af *af;
3990 	int err = 0;
3991 
3992 	/* An association cannot be branched off from an already peeled-off
3993 	 * socket, nor is this supported for tcp style sockets.
3994 	 */
3995 	if (!sctp_style(sk, UDP))
3996 		return -EINVAL;
3997 
3998 	/* Create a new socket.  */
3999 	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4000 	if (err < 0)
4001 		return err;
4002 
4003 	sctp_copy_sock(sock->sk, sk, asoc);
4004 
4005 	/* Make peeled-off sockets more like 1-1 accepted sockets.
4006 	 * Set the daddr and initialize id to something more random
4007 	 */
4008 	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4009 	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4010 
4011 	/* Populate the fields of the newsk from the oldsk and migrate the
4012 	 * asoc to the newsk.
4013 	 */
4014 	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4015 
4016 	*sockp = sock;
4017 
4018 	return err;
4019 }
4020 
4021 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4022 {
4023 	sctp_peeloff_arg_t peeloff;
4024 	struct socket *newsock;
4025 	int retval = 0;
4026 	struct sctp_association *asoc;
4027 
4028 	if (len < sizeof(sctp_peeloff_arg_t))
4029 		return -EINVAL;
4030 	len = sizeof(sctp_peeloff_arg_t);
4031 	if (copy_from_user(&peeloff, optval, len))
4032 		return -EFAULT;
4033 
4034 	asoc = sctp_id2assoc(sk, peeloff.associd);
4035 	if (!asoc) {
4036 		retval = -EINVAL;
4037 		goto out;
4038 	}
4039 
4040 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4041 
4042 	retval = sctp_do_peeloff(asoc, &newsock);
4043 	if (retval < 0)
4044 		goto out;
4045 
4046 	/* Map the socket to an unused fd that can be returned to the user.  */
4047 	retval = sock_map_fd(newsock, 0);
4048 	if (retval < 0) {
4049 		sock_release(newsock);
4050 		goto out;
4051 	}
4052 
4053 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4054 			  __func__, sk, asoc, newsock->sk, retval);
4055 
4056 	/* Return the fd mapped to the new socket.  */
4057 	peeloff.sd = retval;
4058 	if (put_user(len, optlen))
4059 		return -EFAULT;
4060 	if (copy_to_user(optval, &peeloff, len))
4061 		retval = -EFAULT;
4062 
4063 out:
4064 	return retval;
4065 }
4066 
4067 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4068  *
4069  * Applications can enable or disable heartbeats for any peer address of
4070  * an association, modify an address's heartbeat interval, force a
4071  * heartbeat to be sent immediately, and adjust the address's maximum
4072  * number of retransmissions sent before an address is considered
4073  * unreachable.  The following structure is used to access and modify an
4074  * address's parameters:
4075  *
4076  *  struct sctp_paddrparams {
4077  *     sctp_assoc_t            spp_assoc_id;
4078  *     struct sockaddr_storage spp_address;
4079  *     uint32_t                spp_hbinterval;
4080  *     uint16_t                spp_pathmaxrxt;
4081  *     uint32_t                spp_pathmtu;
4082  *     uint32_t                spp_sackdelay;
4083  *     uint32_t                spp_flags;
4084  * };
4085  *
4086  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4087  *                     application, and identifies the association for
4088  *                     this query.
4089  *   spp_address     - This specifies which address is of interest.
4090  *   spp_hbinterval  - This contains the value of the heartbeat interval,
4091  *                     in milliseconds.  If a  value of zero
4092  *                     is present in this field then no changes are to
4093  *                     be made to this parameter.
4094  *   spp_pathmaxrxt  - This contains the maximum number of
4095  *                     retransmissions before this address shall be
4096  *                     considered unreachable. If a  value of zero
4097  *                     is present in this field then no changes are to
4098  *                     be made to this parameter.
4099  *   spp_pathmtu     - When Path MTU discovery is disabled the value
4100  *                     specified here will be the "fixed" path mtu.
4101  *                     Note that if the spp_address field is empty
4102  *                     then all associations on this address will
4103  *                     have this fixed path mtu set upon them.
4104  *
4105  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4106  *                     the number of milliseconds that sacks will be delayed
4107  *                     for. This value will apply to all addresses of an
4108  *                     association if the spp_address field is empty. Note
4109  *                     also, that if delayed sack is enabled and this
4110  *                     value is set to 0, no change is made to the last
4111  *                     recorded delayed sack timer value.
4112  *
4113  *   spp_flags       - These flags are used to control various features
4114  *                     on an association. The flag field may contain
4115  *                     zero or more of the following options.
4116  *
4117  *                     SPP_HB_ENABLE  - Enable heartbeats on the
4118  *                     specified address. Note that if the address
4119  *                     field is empty all addresses for the association
4120  *                     have heartbeats enabled upon them.
4121  *
4122  *                     SPP_HB_DISABLE - Disable heartbeats on the
4123  *                     speicifed address. Note that if the address
4124  *                     field is empty all addresses for the association
4125  *                     will have their heartbeats disabled. Note also
4126  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4127  *                     mutually exclusive, only one of these two should
4128  *                     be specified. Enabling both fields will have
4129  *                     undetermined results.
4130  *
4131  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4132  *                     to be made immediately.
4133  *
4134  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4135  *                     discovery upon the specified address. Note that
4136  *                     if the address feild is empty then all addresses
4137  *                     on the association are effected.
4138  *
4139  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4140  *                     discovery upon the specified address. Note that
4141  *                     if the address feild is empty then all addresses
4142  *                     on the association are effected. Not also that
4143  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4144  *                     exclusive. Enabling both will have undetermined
4145  *                     results.
4146  *
4147  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4148  *                     on delayed sack. The time specified in spp_sackdelay
4149  *                     is used to specify the sack delay for this address. Note
4150  *                     that if spp_address is empty then all addresses will
4151  *                     enable delayed sack and take on the sack delay
4152  *                     value specified in spp_sackdelay.
4153  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4154  *                     off delayed sack. If the spp_address field is blank then
4155  *                     delayed sack is disabled for the entire association. Note
4156  *                     also that this field is mutually exclusive to
4157  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4158  *                     results.
4159  */
4160 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4161 					    char __user *optval, int __user *optlen)
4162 {
4163 	struct sctp_paddrparams  params;
4164 	struct sctp_transport   *trans = NULL;
4165 	struct sctp_association *asoc = NULL;
4166 	struct sctp_sock        *sp = sctp_sk(sk);
4167 
4168 	if (len < sizeof(struct sctp_paddrparams))
4169 		return -EINVAL;
4170 	len = sizeof(struct sctp_paddrparams);
4171 	if (copy_from_user(&params, optval, len))
4172 		return -EFAULT;
4173 
4174 	/* If an address other than INADDR_ANY is specified, and
4175 	 * no transport is found, then the request is invalid.
4176 	 */
4177 	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4178 		trans = sctp_addr_id2transport(sk, &params.spp_address,
4179 					       params.spp_assoc_id);
4180 		if (!trans) {
4181 			SCTP_DEBUG_PRINTK("Failed no transport\n");
4182 			return -EINVAL;
4183 		}
4184 	}
4185 
4186 	/* Get association, if assoc_id != 0 and the socket is a one
4187 	 * to many style socket, and an association was not found, then
4188 	 * the id was invalid.
4189 	 */
4190 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4191 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4192 		SCTP_DEBUG_PRINTK("Failed no association\n");
4193 		return -EINVAL;
4194 	}
4195 
4196 	if (trans) {
4197 		/* Fetch transport values. */
4198 		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4199 		params.spp_pathmtu    = trans->pathmtu;
4200 		params.spp_pathmaxrxt = trans->pathmaxrxt;
4201 		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4202 
4203 		/*draft-11 doesn't say what to return in spp_flags*/
4204 		params.spp_flags      = trans->param_flags;
4205 	} else if (asoc) {
4206 		/* Fetch association values. */
4207 		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4208 		params.spp_pathmtu    = asoc->pathmtu;
4209 		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4210 		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4211 
4212 		/*draft-11 doesn't say what to return in spp_flags*/
4213 		params.spp_flags      = asoc->param_flags;
4214 	} else {
4215 		/* Fetch socket values. */
4216 		params.spp_hbinterval = sp->hbinterval;
4217 		params.spp_pathmtu    = sp->pathmtu;
4218 		params.spp_sackdelay  = sp->sackdelay;
4219 		params.spp_pathmaxrxt = sp->pathmaxrxt;
4220 
4221 		/*draft-11 doesn't say what to return in spp_flags*/
4222 		params.spp_flags      = sp->param_flags;
4223 	}
4224 
4225 	if (copy_to_user(optval, &params, len))
4226 		return -EFAULT;
4227 
4228 	if (put_user(len, optlen))
4229 		return -EFAULT;
4230 
4231 	return 0;
4232 }
4233 
4234 /*
4235  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4236  *
4237  * This option will effect the way delayed acks are performed.  This
4238  * option allows you to get or set the delayed ack time, in
4239  * milliseconds.  It also allows changing the delayed ack frequency.
4240  * Changing the frequency to 1 disables the delayed sack algorithm.  If
4241  * the assoc_id is 0, then this sets or gets the endpoints default
4242  * values.  If the assoc_id field is non-zero, then the set or get
4243  * effects the specified association for the one to many model (the
4244  * assoc_id field is ignored by the one to one model).  Note that if
4245  * sack_delay or sack_freq are 0 when setting this option, then the
4246  * current values will remain unchanged.
4247  *
4248  * struct sctp_sack_info {
4249  *     sctp_assoc_t            sack_assoc_id;
4250  *     uint32_t                sack_delay;
4251  *     uint32_t                sack_freq;
4252  * };
4253  *
4254  * sack_assoc_id -  This parameter, indicates which association the user
4255  *    is performing an action upon.  Note that if this field's value is
4256  *    zero then the endpoints default value is changed (effecting future
4257  *    associations only).
4258  *
4259  * sack_delay -  This parameter contains the number of milliseconds that
4260  *    the user is requesting the delayed ACK timer be set to.  Note that
4261  *    this value is defined in the standard to be between 200 and 500
4262  *    milliseconds.
4263  *
4264  * sack_freq -  This parameter contains the number of packets that must
4265  *    be received before a sack is sent without waiting for the delay
4266  *    timer to expire.  The default value for this is 2, setting this
4267  *    value to 1 will disable the delayed sack algorithm.
4268  */
4269 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4270 					    char __user *optval,
4271 					    int __user *optlen)
4272 {
4273 	struct sctp_sack_info    params;
4274 	struct sctp_association *asoc = NULL;
4275 	struct sctp_sock        *sp = sctp_sk(sk);
4276 
4277 	if (len >= sizeof(struct sctp_sack_info)) {
4278 		len = sizeof(struct sctp_sack_info);
4279 
4280 		if (copy_from_user(&params, optval, len))
4281 			return -EFAULT;
4282 	} else if (len == sizeof(struct sctp_assoc_value)) {
4283 		printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4284 		       "in delayed_ack socket option deprecated\n");
4285 		printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4286 		if (copy_from_user(&params, optval, len))
4287 			return -EFAULT;
4288 	} else
4289 		return - EINVAL;
4290 
4291 	/* Get association, if sack_assoc_id != 0 and the socket is a one
4292 	 * to many style socket, and an association was not found, then
4293 	 * the id was invalid.
4294 	 */
4295 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4296 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4297 		return -EINVAL;
4298 
4299 	if (asoc) {
4300 		/* Fetch association values. */
4301 		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4302 			params.sack_delay = jiffies_to_msecs(
4303 				asoc->sackdelay);
4304 			params.sack_freq = asoc->sackfreq;
4305 
4306 		} else {
4307 			params.sack_delay = 0;
4308 			params.sack_freq = 1;
4309 		}
4310 	} else {
4311 		/* Fetch socket values. */
4312 		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4313 			params.sack_delay  = sp->sackdelay;
4314 			params.sack_freq = sp->sackfreq;
4315 		} else {
4316 			params.sack_delay  = 0;
4317 			params.sack_freq = 1;
4318 		}
4319 	}
4320 
4321 	if (copy_to_user(optval, &params, len))
4322 		return -EFAULT;
4323 
4324 	if (put_user(len, optlen))
4325 		return -EFAULT;
4326 
4327 	return 0;
4328 }
4329 
4330 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4331  *
4332  * Applications can specify protocol parameters for the default association
4333  * initialization.  The option name argument to setsockopt() and getsockopt()
4334  * is SCTP_INITMSG.
4335  *
4336  * Setting initialization parameters is effective only on an unconnected
4337  * socket (for UDP-style sockets only future associations are effected
4338  * by the change).  With TCP-style sockets, this option is inherited by
4339  * sockets derived from a listener socket.
4340  */
4341 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4342 {
4343 	if (len < sizeof(struct sctp_initmsg))
4344 		return -EINVAL;
4345 	len = sizeof(struct sctp_initmsg);
4346 	if (put_user(len, optlen))
4347 		return -EFAULT;
4348 	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4349 		return -EFAULT;
4350 	return 0;
4351 }
4352 
4353 
4354 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4355 				      char __user *optval, int __user *optlen)
4356 {
4357 	struct sctp_association *asoc;
4358 	int cnt = 0;
4359 	struct sctp_getaddrs getaddrs;
4360 	struct sctp_transport *from;
4361 	void __user *to;
4362 	union sctp_addr temp;
4363 	struct sctp_sock *sp = sctp_sk(sk);
4364 	int addrlen;
4365 	size_t space_left;
4366 	int bytes_copied;
4367 
4368 	if (len < sizeof(struct sctp_getaddrs))
4369 		return -EINVAL;
4370 
4371 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4372 		return -EFAULT;
4373 
4374 	/* For UDP-style sockets, id specifies the association to query.  */
4375 	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4376 	if (!asoc)
4377 		return -EINVAL;
4378 
4379 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4380 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4381 
4382 	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4383 				transports) {
4384 		memcpy(&temp, &from->ipaddr, sizeof(temp));
4385 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4386 		addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4387 		if (space_left < addrlen)
4388 			return -ENOMEM;
4389 		if (copy_to_user(to, &temp, addrlen))
4390 			return -EFAULT;
4391 		to += addrlen;
4392 		cnt++;
4393 		space_left -= addrlen;
4394 	}
4395 
4396 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4397 		return -EFAULT;
4398 	bytes_copied = ((char __user *)to) - optval;
4399 	if (put_user(bytes_copied, optlen))
4400 		return -EFAULT;
4401 
4402 	return 0;
4403 }
4404 
4405 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4406 			    size_t space_left, int *bytes_copied)
4407 {
4408 	struct sctp_sockaddr_entry *addr;
4409 	union sctp_addr temp;
4410 	int cnt = 0;
4411 	int addrlen;
4412 
4413 	rcu_read_lock();
4414 	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4415 		if (!addr->valid)
4416 			continue;
4417 
4418 		if ((PF_INET == sk->sk_family) &&
4419 		    (AF_INET6 == addr->a.sa.sa_family))
4420 			continue;
4421 		if ((PF_INET6 == sk->sk_family) &&
4422 		    inet_v6_ipv6only(sk) &&
4423 		    (AF_INET == addr->a.sa.sa_family))
4424 			continue;
4425 		memcpy(&temp, &addr->a, sizeof(temp));
4426 		if (!temp.v4.sin_port)
4427 			temp.v4.sin_port = htons(port);
4428 
4429 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4430 								&temp);
4431 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4432 		if (space_left < addrlen) {
4433 			cnt =  -ENOMEM;
4434 			break;
4435 		}
4436 		memcpy(to, &temp, addrlen);
4437 
4438 		to += addrlen;
4439 		cnt ++;
4440 		space_left -= addrlen;
4441 		*bytes_copied += addrlen;
4442 	}
4443 	rcu_read_unlock();
4444 
4445 	return cnt;
4446 }
4447 
4448 
4449 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4450 				       char __user *optval, int __user *optlen)
4451 {
4452 	struct sctp_bind_addr *bp;
4453 	struct sctp_association *asoc;
4454 	int cnt = 0;
4455 	struct sctp_getaddrs getaddrs;
4456 	struct sctp_sockaddr_entry *addr;
4457 	void __user *to;
4458 	union sctp_addr temp;
4459 	struct sctp_sock *sp = sctp_sk(sk);
4460 	int addrlen;
4461 	int err = 0;
4462 	size_t space_left;
4463 	int bytes_copied = 0;
4464 	void *addrs;
4465 	void *buf;
4466 
4467 	if (len < sizeof(struct sctp_getaddrs))
4468 		return -EINVAL;
4469 
4470 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4471 		return -EFAULT;
4472 
4473 	/*
4474 	 *  For UDP-style sockets, id specifies the association to query.
4475 	 *  If the id field is set to the value '0' then the locally bound
4476 	 *  addresses are returned without regard to any particular
4477 	 *  association.
4478 	 */
4479 	if (0 == getaddrs.assoc_id) {
4480 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4481 	} else {
4482 		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4483 		if (!asoc)
4484 			return -EINVAL;
4485 		bp = &asoc->base.bind_addr;
4486 	}
4487 
4488 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4489 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4490 
4491 	addrs = kmalloc(space_left, GFP_KERNEL);
4492 	if (!addrs)
4493 		return -ENOMEM;
4494 
4495 	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4496 	 * addresses from the global local address list.
4497 	 */
4498 	if (sctp_list_single_entry(&bp->address_list)) {
4499 		addr = list_entry(bp->address_list.next,
4500 				  struct sctp_sockaddr_entry, list);
4501 		if (sctp_is_any(sk, &addr->a)) {
4502 			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4503 						space_left, &bytes_copied);
4504 			if (cnt < 0) {
4505 				err = cnt;
4506 				goto out;
4507 			}
4508 			goto copy_getaddrs;
4509 		}
4510 	}
4511 
4512 	buf = addrs;
4513 	/* Protection on the bound address list is not needed since
4514 	 * in the socket option context we hold a socket lock and
4515 	 * thus the bound address list can't change.
4516 	 */
4517 	list_for_each_entry(addr, &bp->address_list, list) {
4518 		memcpy(&temp, &addr->a, sizeof(temp));
4519 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4520 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4521 		if (space_left < addrlen) {
4522 			err =  -ENOMEM; /*fixme: right error?*/
4523 			goto out;
4524 		}
4525 		memcpy(buf, &temp, addrlen);
4526 		buf += addrlen;
4527 		bytes_copied += addrlen;
4528 		cnt ++;
4529 		space_left -= addrlen;
4530 	}
4531 
4532 copy_getaddrs:
4533 	if (copy_to_user(to, addrs, bytes_copied)) {
4534 		err = -EFAULT;
4535 		goto out;
4536 	}
4537 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4538 		err = -EFAULT;
4539 		goto out;
4540 	}
4541 	if (put_user(bytes_copied, optlen))
4542 		err = -EFAULT;
4543 out:
4544 	kfree(addrs);
4545 	return err;
4546 }
4547 
4548 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4549  *
4550  * Requests that the local SCTP stack use the enclosed peer address as
4551  * the association primary.  The enclosed address must be one of the
4552  * association peer's addresses.
4553  */
4554 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4555 					char __user *optval, int __user *optlen)
4556 {
4557 	struct sctp_prim prim;
4558 	struct sctp_association *asoc;
4559 	struct sctp_sock *sp = sctp_sk(sk);
4560 
4561 	if (len < sizeof(struct sctp_prim))
4562 		return -EINVAL;
4563 
4564 	len = sizeof(struct sctp_prim);
4565 
4566 	if (copy_from_user(&prim, optval, len))
4567 		return -EFAULT;
4568 
4569 	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4570 	if (!asoc)
4571 		return -EINVAL;
4572 
4573 	if (!asoc->peer.primary_path)
4574 		return -ENOTCONN;
4575 
4576 	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4577 		asoc->peer.primary_path->af_specific->sockaddr_len);
4578 
4579 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4580 			(union sctp_addr *)&prim.ssp_addr);
4581 
4582 	if (put_user(len, optlen))
4583 		return -EFAULT;
4584 	if (copy_to_user(optval, &prim, len))
4585 		return -EFAULT;
4586 
4587 	return 0;
4588 }
4589 
4590 /*
4591  * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4592  *
4593  * Requests that the local endpoint set the specified Adaptation Layer
4594  * Indication parameter for all future INIT and INIT-ACK exchanges.
4595  */
4596 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4597 				  char __user *optval, int __user *optlen)
4598 {
4599 	struct sctp_setadaptation adaptation;
4600 
4601 	if (len < sizeof(struct sctp_setadaptation))
4602 		return -EINVAL;
4603 
4604 	len = sizeof(struct sctp_setadaptation);
4605 
4606 	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4607 
4608 	if (put_user(len, optlen))
4609 		return -EFAULT;
4610 	if (copy_to_user(optval, &adaptation, len))
4611 		return -EFAULT;
4612 
4613 	return 0;
4614 }
4615 
4616 /*
4617  *
4618  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4619  *
4620  *   Applications that wish to use the sendto() system call may wish to
4621  *   specify a default set of parameters that would normally be supplied
4622  *   through the inclusion of ancillary data.  This socket option allows
4623  *   such an application to set the default sctp_sndrcvinfo structure.
4624 
4625 
4626  *   The application that wishes to use this socket option simply passes
4627  *   in to this call the sctp_sndrcvinfo structure defined in Section
4628  *   5.2.2) The input parameters accepted by this call include
4629  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4630  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4631  *   to this call if the caller is using the UDP model.
4632  *
4633  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4634  */
4635 static int sctp_getsockopt_default_send_param(struct sock *sk,
4636 					int len, char __user *optval,
4637 					int __user *optlen)
4638 {
4639 	struct sctp_sndrcvinfo info;
4640 	struct sctp_association *asoc;
4641 	struct sctp_sock *sp = sctp_sk(sk);
4642 
4643 	if (len < sizeof(struct sctp_sndrcvinfo))
4644 		return -EINVAL;
4645 
4646 	len = sizeof(struct sctp_sndrcvinfo);
4647 
4648 	if (copy_from_user(&info, optval, len))
4649 		return -EFAULT;
4650 
4651 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4652 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4653 		return -EINVAL;
4654 
4655 	if (asoc) {
4656 		info.sinfo_stream = asoc->default_stream;
4657 		info.sinfo_flags = asoc->default_flags;
4658 		info.sinfo_ppid = asoc->default_ppid;
4659 		info.sinfo_context = asoc->default_context;
4660 		info.sinfo_timetolive = asoc->default_timetolive;
4661 	} else {
4662 		info.sinfo_stream = sp->default_stream;
4663 		info.sinfo_flags = sp->default_flags;
4664 		info.sinfo_ppid = sp->default_ppid;
4665 		info.sinfo_context = sp->default_context;
4666 		info.sinfo_timetolive = sp->default_timetolive;
4667 	}
4668 
4669 	if (put_user(len, optlen))
4670 		return -EFAULT;
4671 	if (copy_to_user(optval, &info, len))
4672 		return -EFAULT;
4673 
4674 	return 0;
4675 }
4676 
4677 /*
4678  *
4679  * 7.1.5 SCTP_NODELAY
4680  *
4681  * Turn on/off any Nagle-like algorithm.  This means that packets are
4682  * generally sent as soon as possible and no unnecessary delays are
4683  * introduced, at the cost of more packets in the network.  Expects an
4684  * integer boolean flag.
4685  */
4686 
4687 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4688 				   char __user *optval, int __user *optlen)
4689 {
4690 	int val;
4691 
4692 	if (len < sizeof(int))
4693 		return -EINVAL;
4694 
4695 	len = sizeof(int);
4696 	val = (sctp_sk(sk)->nodelay == 1);
4697 	if (put_user(len, optlen))
4698 		return -EFAULT;
4699 	if (copy_to_user(optval, &val, len))
4700 		return -EFAULT;
4701 	return 0;
4702 }
4703 
4704 /*
4705  *
4706  * 7.1.1 SCTP_RTOINFO
4707  *
4708  * The protocol parameters used to initialize and bound retransmission
4709  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4710  * and modify these parameters.
4711  * All parameters are time values, in milliseconds.  A value of 0, when
4712  * modifying the parameters, indicates that the current value should not
4713  * be changed.
4714  *
4715  */
4716 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4717 				char __user *optval,
4718 				int __user *optlen) {
4719 	struct sctp_rtoinfo rtoinfo;
4720 	struct sctp_association *asoc;
4721 
4722 	if (len < sizeof (struct sctp_rtoinfo))
4723 		return -EINVAL;
4724 
4725 	len = sizeof(struct sctp_rtoinfo);
4726 
4727 	if (copy_from_user(&rtoinfo, optval, len))
4728 		return -EFAULT;
4729 
4730 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4731 
4732 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4733 		return -EINVAL;
4734 
4735 	/* Values corresponding to the specific association. */
4736 	if (asoc) {
4737 		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4738 		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4739 		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4740 	} else {
4741 		/* Values corresponding to the endpoint. */
4742 		struct sctp_sock *sp = sctp_sk(sk);
4743 
4744 		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4745 		rtoinfo.srto_max = sp->rtoinfo.srto_max;
4746 		rtoinfo.srto_min = sp->rtoinfo.srto_min;
4747 	}
4748 
4749 	if (put_user(len, optlen))
4750 		return -EFAULT;
4751 
4752 	if (copy_to_user(optval, &rtoinfo, len))
4753 		return -EFAULT;
4754 
4755 	return 0;
4756 }
4757 
4758 /*
4759  *
4760  * 7.1.2 SCTP_ASSOCINFO
4761  *
4762  * This option is used to tune the maximum retransmission attempts
4763  * of the association.
4764  * Returns an error if the new association retransmission value is
4765  * greater than the sum of the retransmission value  of the peer.
4766  * See [SCTP] for more information.
4767  *
4768  */
4769 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4770 				     char __user *optval,
4771 				     int __user *optlen)
4772 {
4773 
4774 	struct sctp_assocparams assocparams;
4775 	struct sctp_association *asoc;
4776 	struct list_head *pos;
4777 	int cnt = 0;
4778 
4779 	if (len < sizeof (struct sctp_assocparams))
4780 		return -EINVAL;
4781 
4782 	len = sizeof(struct sctp_assocparams);
4783 
4784 	if (copy_from_user(&assocparams, optval, len))
4785 		return -EFAULT;
4786 
4787 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4788 
4789 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4790 		return -EINVAL;
4791 
4792 	/* Values correspoinding to the specific association */
4793 	if (asoc) {
4794 		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4795 		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4796 		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4797 		assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4798 						* 1000) +
4799 						(asoc->cookie_life.tv_usec
4800 						/ 1000);
4801 
4802 		list_for_each(pos, &asoc->peer.transport_addr_list) {
4803 			cnt ++;
4804 		}
4805 
4806 		assocparams.sasoc_number_peer_destinations = cnt;
4807 	} else {
4808 		/* Values corresponding to the endpoint */
4809 		struct sctp_sock *sp = sctp_sk(sk);
4810 
4811 		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4812 		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4813 		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4814 		assocparams.sasoc_cookie_life =
4815 					sp->assocparams.sasoc_cookie_life;
4816 		assocparams.sasoc_number_peer_destinations =
4817 					sp->assocparams.
4818 					sasoc_number_peer_destinations;
4819 	}
4820 
4821 	if (put_user(len, optlen))
4822 		return -EFAULT;
4823 
4824 	if (copy_to_user(optval, &assocparams, len))
4825 		return -EFAULT;
4826 
4827 	return 0;
4828 }
4829 
4830 /*
4831  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4832  *
4833  * This socket option is a boolean flag which turns on or off mapped V4
4834  * addresses.  If this option is turned on and the socket is type
4835  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4836  * If this option is turned off, then no mapping will be done of V4
4837  * addresses and a user will receive both PF_INET6 and PF_INET type
4838  * addresses on the socket.
4839  */
4840 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4841 				    char __user *optval, int __user *optlen)
4842 {
4843 	int val;
4844 	struct sctp_sock *sp = sctp_sk(sk);
4845 
4846 	if (len < sizeof(int))
4847 		return -EINVAL;
4848 
4849 	len = sizeof(int);
4850 	val = sp->v4mapped;
4851 	if (put_user(len, optlen))
4852 		return -EFAULT;
4853 	if (copy_to_user(optval, &val, len))
4854 		return -EFAULT;
4855 
4856 	return 0;
4857 }
4858 
4859 /*
4860  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
4861  * (chapter and verse is quoted at sctp_setsockopt_context())
4862  */
4863 static int sctp_getsockopt_context(struct sock *sk, int len,
4864 				   char __user *optval, int __user *optlen)
4865 {
4866 	struct sctp_assoc_value params;
4867 	struct sctp_sock *sp;
4868 	struct sctp_association *asoc;
4869 
4870 	if (len < sizeof(struct sctp_assoc_value))
4871 		return -EINVAL;
4872 
4873 	len = sizeof(struct sctp_assoc_value);
4874 
4875 	if (copy_from_user(&params, optval, len))
4876 		return -EFAULT;
4877 
4878 	sp = sctp_sk(sk);
4879 
4880 	if (params.assoc_id != 0) {
4881 		asoc = sctp_id2assoc(sk, params.assoc_id);
4882 		if (!asoc)
4883 			return -EINVAL;
4884 		params.assoc_value = asoc->default_rcv_context;
4885 	} else {
4886 		params.assoc_value = sp->default_rcv_context;
4887 	}
4888 
4889 	if (put_user(len, optlen))
4890 		return -EFAULT;
4891 	if (copy_to_user(optval, &params, len))
4892 		return -EFAULT;
4893 
4894 	return 0;
4895 }
4896 
4897 /*
4898  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4899  * This option will get or set the maximum size to put in any outgoing
4900  * SCTP DATA chunk.  If a message is larger than this size it will be
4901  * fragmented by SCTP into the specified size.  Note that the underlying
4902  * SCTP implementation may fragment into smaller sized chunks when the
4903  * PMTU of the underlying association is smaller than the value set by
4904  * the user.  The default value for this option is '0' which indicates
4905  * the user is NOT limiting fragmentation and only the PMTU will effect
4906  * SCTP's choice of DATA chunk size.  Note also that values set larger
4907  * than the maximum size of an IP datagram will effectively let SCTP
4908  * control fragmentation (i.e. the same as setting this option to 0).
4909  *
4910  * The following structure is used to access and modify this parameter:
4911  *
4912  * struct sctp_assoc_value {
4913  *   sctp_assoc_t assoc_id;
4914  *   uint32_t assoc_value;
4915  * };
4916  *
4917  * assoc_id:  This parameter is ignored for one-to-one style sockets.
4918  *    For one-to-many style sockets this parameter indicates which
4919  *    association the user is performing an action upon.  Note that if
4920  *    this field's value is zero then the endpoints default value is
4921  *    changed (effecting future associations only).
4922  * assoc_value:  This parameter specifies the maximum size in bytes.
4923  */
4924 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4925 				  char __user *optval, int __user *optlen)
4926 {
4927 	struct sctp_assoc_value params;
4928 	struct sctp_association *asoc;
4929 
4930 	if (len == sizeof(int)) {
4931 		printk(KERN_WARNING
4932 		   "SCTP: Use of int in maxseg socket option deprecated\n");
4933 		printk(KERN_WARNING
4934 		   "SCTP: Use struct sctp_assoc_value instead\n");
4935 		params.assoc_id = 0;
4936 	} else if (len >= sizeof(struct sctp_assoc_value)) {
4937 		len = sizeof(struct sctp_assoc_value);
4938 		if (copy_from_user(&params, optval, sizeof(params)))
4939 			return -EFAULT;
4940 	} else
4941 		return -EINVAL;
4942 
4943 	asoc = sctp_id2assoc(sk, params.assoc_id);
4944 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4945 		return -EINVAL;
4946 
4947 	if (asoc)
4948 		params.assoc_value = asoc->frag_point;
4949 	else
4950 		params.assoc_value = sctp_sk(sk)->user_frag;
4951 
4952 	if (put_user(len, optlen))
4953 		return -EFAULT;
4954 	if (len == sizeof(int)) {
4955 		if (copy_to_user(optval, &params.assoc_value, len))
4956 			return -EFAULT;
4957 	} else {
4958 		if (copy_to_user(optval, &params, len))
4959 			return -EFAULT;
4960 	}
4961 
4962 	return 0;
4963 }
4964 
4965 /*
4966  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4967  * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4968  */
4969 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4970 					       char __user *optval, int __user *optlen)
4971 {
4972 	int val;
4973 
4974 	if (len < sizeof(int))
4975 		return -EINVAL;
4976 
4977 	len = sizeof(int);
4978 
4979 	val = sctp_sk(sk)->frag_interleave;
4980 	if (put_user(len, optlen))
4981 		return -EFAULT;
4982 	if (copy_to_user(optval, &val, len))
4983 		return -EFAULT;
4984 
4985 	return 0;
4986 }
4987 
4988 /*
4989  * 7.1.25.  Set or Get the sctp partial delivery point
4990  * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4991  */
4992 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4993 						  char __user *optval,
4994 						  int __user *optlen)
4995 {
4996 	u32 val;
4997 
4998 	if (len < sizeof(u32))
4999 		return -EINVAL;
5000 
5001 	len = sizeof(u32);
5002 
5003 	val = sctp_sk(sk)->pd_point;
5004 	if (put_user(len, optlen))
5005 		return -EFAULT;
5006 	if (copy_to_user(optval, &val, len))
5007 		return -EFAULT;
5008 
5009 	return -ENOTSUPP;
5010 }
5011 
5012 /*
5013  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5014  * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5015  */
5016 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5017 				    char __user *optval,
5018 				    int __user *optlen)
5019 {
5020 	struct sctp_assoc_value params;
5021 	struct sctp_sock *sp;
5022 	struct sctp_association *asoc;
5023 
5024 	if (len == sizeof(int)) {
5025 		printk(KERN_WARNING
5026 		   "SCTP: Use of int in max_burst socket option deprecated\n");
5027 		printk(KERN_WARNING
5028 		   "SCTP: Use struct sctp_assoc_value instead\n");
5029 		params.assoc_id = 0;
5030 	} else if (len >= sizeof(struct sctp_assoc_value)) {
5031 		len = sizeof(struct sctp_assoc_value);
5032 		if (copy_from_user(&params, optval, len))
5033 			return -EFAULT;
5034 	} else
5035 		return -EINVAL;
5036 
5037 	sp = sctp_sk(sk);
5038 
5039 	if (params.assoc_id != 0) {
5040 		asoc = sctp_id2assoc(sk, params.assoc_id);
5041 		if (!asoc)
5042 			return -EINVAL;
5043 		params.assoc_value = asoc->max_burst;
5044 	} else
5045 		params.assoc_value = sp->max_burst;
5046 
5047 	if (len == sizeof(int)) {
5048 		if (copy_to_user(optval, &params.assoc_value, len))
5049 			return -EFAULT;
5050 	} else {
5051 		if (copy_to_user(optval, &params, len))
5052 			return -EFAULT;
5053 	}
5054 
5055 	return 0;
5056 
5057 }
5058 
5059 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5060 				    char __user *optval, int __user *optlen)
5061 {
5062 	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5063 	struct sctp_hmac_algo_param *hmacs;
5064 	__u16 data_len = 0;
5065 	u32 num_idents;
5066 
5067 	if (!sctp_auth_enable)
5068 		return -EACCES;
5069 
5070 	hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5071 	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5072 
5073 	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5074 		return -EINVAL;
5075 
5076 	len = sizeof(struct sctp_hmacalgo) + data_len;
5077 	num_idents = data_len / sizeof(u16);
5078 
5079 	if (put_user(len, optlen))
5080 		return -EFAULT;
5081 	if (put_user(num_idents, &p->shmac_num_idents))
5082 		return -EFAULT;
5083 	if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5084 		return -EFAULT;
5085 	return 0;
5086 }
5087 
5088 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5089 				    char __user *optval, int __user *optlen)
5090 {
5091 	struct sctp_authkeyid val;
5092 	struct sctp_association *asoc;
5093 
5094 	if (!sctp_auth_enable)
5095 		return -EACCES;
5096 
5097 	if (len < sizeof(struct sctp_authkeyid))
5098 		return -EINVAL;
5099 	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5100 		return -EFAULT;
5101 
5102 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5103 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5104 		return -EINVAL;
5105 
5106 	if (asoc)
5107 		val.scact_keynumber = asoc->active_key_id;
5108 	else
5109 		val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5110 
5111 	len = sizeof(struct sctp_authkeyid);
5112 	if (put_user(len, optlen))
5113 		return -EFAULT;
5114 	if (copy_to_user(optval, &val, len))
5115 		return -EFAULT;
5116 
5117 	return 0;
5118 }
5119 
5120 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5121 				    char __user *optval, int __user *optlen)
5122 {
5123 	struct sctp_authchunks __user *p = (void __user *)optval;
5124 	struct sctp_authchunks val;
5125 	struct sctp_association *asoc;
5126 	struct sctp_chunks_param *ch;
5127 	u32    num_chunks = 0;
5128 	char __user *to;
5129 
5130 	if (!sctp_auth_enable)
5131 		return -EACCES;
5132 
5133 	if (len < sizeof(struct sctp_authchunks))
5134 		return -EINVAL;
5135 
5136 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5137 		return -EFAULT;
5138 
5139 	to = p->gauth_chunks;
5140 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5141 	if (!asoc)
5142 		return -EINVAL;
5143 
5144 	ch = asoc->peer.peer_chunks;
5145 	if (!ch)
5146 		goto num;
5147 
5148 	/* See if the user provided enough room for all the data */
5149 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5150 	if (len < num_chunks)
5151 		return -EINVAL;
5152 
5153 	if (copy_to_user(to, ch->chunks, num_chunks))
5154 		return -EFAULT;
5155 num:
5156 	len = sizeof(struct sctp_authchunks) + num_chunks;
5157 	if (put_user(len, optlen)) return -EFAULT;
5158 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5159 		return -EFAULT;
5160 	return 0;
5161 }
5162 
5163 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5164 				    char __user *optval, int __user *optlen)
5165 {
5166 	struct sctp_authchunks __user *p = (void __user *)optval;
5167 	struct sctp_authchunks val;
5168 	struct sctp_association *asoc;
5169 	struct sctp_chunks_param *ch;
5170 	u32    num_chunks = 0;
5171 	char __user *to;
5172 
5173 	if (!sctp_auth_enable)
5174 		return -EACCES;
5175 
5176 	if (len < sizeof(struct sctp_authchunks))
5177 		return -EINVAL;
5178 
5179 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5180 		return -EFAULT;
5181 
5182 	to = p->gauth_chunks;
5183 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5184 	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5185 		return -EINVAL;
5186 
5187 	if (asoc)
5188 		ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5189 	else
5190 		ch = sctp_sk(sk)->ep->auth_chunk_list;
5191 
5192 	if (!ch)
5193 		goto num;
5194 
5195 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5196 	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5197 		return -EINVAL;
5198 
5199 	if (copy_to_user(to, ch->chunks, num_chunks))
5200 		return -EFAULT;
5201 num:
5202 	len = sizeof(struct sctp_authchunks) + num_chunks;
5203 	if (put_user(len, optlen))
5204 		return -EFAULT;
5205 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5206 		return -EFAULT;
5207 
5208 	return 0;
5209 }
5210 
5211 /*
5212  * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5213  * This option gets the current number of associations that are attached
5214  * to a one-to-many style socket.  The option value is an uint32_t.
5215  */
5216 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5217 				    char __user *optval, int __user *optlen)
5218 {
5219 	struct sctp_sock *sp = sctp_sk(sk);
5220 	struct sctp_association *asoc;
5221 	u32 val = 0;
5222 
5223 	if (sctp_style(sk, TCP))
5224 		return -EOPNOTSUPP;
5225 
5226 	if (len < sizeof(u32))
5227 		return -EINVAL;
5228 
5229 	len = sizeof(u32);
5230 
5231 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5232 		val++;
5233 	}
5234 
5235 	if (put_user(len, optlen))
5236 		return -EFAULT;
5237 	if (copy_to_user(optval, &val, len))
5238 		return -EFAULT;
5239 
5240 	return 0;
5241 }
5242 
5243 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5244 				char __user *optval, int __user *optlen)
5245 {
5246 	int retval = 0;
5247 	int len;
5248 
5249 	SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5250 			  sk, optname);
5251 
5252 	/* I can hardly begin to describe how wrong this is.  This is
5253 	 * so broken as to be worse than useless.  The API draft
5254 	 * REALLY is NOT helpful here...  I am not convinced that the
5255 	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5256 	 * are at all well-founded.
5257 	 */
5258 	if (level != SOL_SCTP) {
5259 		struct sctp_af *af = sctp_sk(sk)->pf->af;
5260 
5261 		retval = af->getsockopt(sk, level, optname, optval, optlen);
5262 		return retval;
5263 	}
5264 
5265 	if (get_user(len, optlen))
5266 		return -EFAULT;
5267 
5268 	sctp_lock_sock(sk);
5269 
5270 	switch (optname) {
5271 	case SCTP_STATUS:
5272 		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5273 		break;
5274 	case SCTP_DISABLE_FRAGMENTS:
5275 		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5276 							   optlen);
5277 		break;
5278 	case SCTP_EVENTS:
5279 		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5280 		break;
5281 	case SCTP_AUTOCLOSE:
5282 		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5283 		break;
5284 	case SCTP_SOCKOPT_PEELOFF:
5285 		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5286 		break;
5287 	case SCTP_PEER_ADDR_PARAMS:
5288 		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5289 							  optlen);
5290 		break;
5291 	case SCTP_DELAYED_ACK:
5292 		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5293 							  optlen);
5294 		break;
5295 	case SCTP_INITMSG:
5296 		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5297 		break;
5298 	case SCTP_GET_PEER_ADDRS:
5299 		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5300 						    optlen);
5301 		break;
5302 	case SCTP_GET_LOCAL_ADDRS:
5303 		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5304 						     optlen);
5305 		break;
5306 	case SCTP_SOCKOPT_CONNECTX3:
5307 		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5308 		break;
5309 	case SCTP_DEFAULT_SEND_PARAM:
5310 		retval = sctp_getsockopt_default_send_param(sk, len,
5311 							    optval, optlen);
5312 		break;
5313 	case SCTP_PRIMARY_ADDR:
5314 		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5315 		break;
5316 	case SCTP_NODELAY:
5317 		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5318 		break;
5319 	case SCTP_RTOINFO:
5320 		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5321 		break;
5322 	case SCTP_ASSOCINFO:
5323 		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5324 		break;
5325 	case SCTP_I_WANT_MAPPED_V4_ADDR:
5326 		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5327 		break;
5328 	case SCTP_MAXSEG:
5329 		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5330 		break;
5331 	case SCTP_GET_PEER_ADDR_INFO:
5332 		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5333 							optlen);
5334 		break;
5335 	case SCTP_ADAPTATION_LAYER:
5336 		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5337 							optlen);
5338 		break;
5339 	case SCTP_CONTEXT:
5340 		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5341 		break;
5342 	case SCTP_FRAGMENT_INTERLEAVE:
5343 		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5344 							     optlen);
5345 		break;
5346 	case SCTP_PARTIAL_DELIVERY_POINT:
5347 		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5348 								optlen);
5349 		break;
5350 	case SCTP_MAX_BURST:
5351 		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5352 		break;
5353 	case SCTP_AUTH_KEY:
5354 	case SCTP_AUTH_CHUNK:
5355 	case SCTP_AUTH_DELETE_KEY:
5356 		retval = -EOPNOTSUPP;
5357 		break;
5358 	case SCTP_HMAC_IDENT:
5359 		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5360 		break;
5361 	case SCTP_AUTH_ACTIVE_KEY:
5362 		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5363 		break;
5364 	case SCTP_PEER_AUTH_CHUNKS:
5365 		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5366 							optlen);
5367 		break;
5368 	case SCTP_LOCAL_AUTH_CHUNKS:
5369 		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5370 							optlen);
5371 		break;
5372 	case SCTP_GET_ASSOC_NUMBER:
5373 		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5374 		break;
5375 	default:
5376 		retval = -ENOPROTOOPT;
5377 		break;
5378 	}
5379 
5380 	sctp_release_sock(sk);
5381 	return retval;
5382 }
5383 
5384 static void sctp_hash(struct sock *sk)
5385 {
5386 	/* STUB */
5387 }
5388 
5389 static void sctp_unhash(struct sock *sk)
5390 {
5391 	/* STUB */
5392 }
5393 
5394 /* Check if port is acceptable.  Possibly find first available port.
5395  *
5396  * The port hash table (contained in the 'global' SCTP protocol storage
5397  * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5398  * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5399  * list (the list number is the port number hashed out, so as you
5400  * would expect from a hash function, all the ports in a given list have
5401  * such a number that hashes out to the same list number; you were
5402  * expecting that, right?); so each list has a set of ports, with a
5403  * link to the socket (struct sock) that uses it, the port number and
5404  * a fastreuse flag (FIXME: NPI ipg).
5405  */
5406 static struct sctp_bind_bucket *sctp_bucket_create(
5407 	struct sctp_bind_hashbucket *head, unsigned short snum);
5408 
5409 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5410 {
5411 	struct sctp_bind_hashbucket *head; /* hash list */
5412 	struct sctp_bind_bucket *pp; /* hash list port iterator */
5413 	struct hlist_node *node;
5414 	unsigned short snum;
5415 	int ret;
5416 
5417 	snum = ntohs(addr->v4.sin_port);
5418 
5419 	SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5420 	sctp_local_bh_disable();
5421 
5422 	if (snum == 0) {
5423 		/* Search for an available port. */
5424 		int low, high, remaining, index;
5425 		unsigned int rover;
5426 
5427 		inet_get_local_port_range(&low, &high);
5428 		remaining = (high - low) + 1;
5429 		rover = net_random() % remaining + low;
5430 
5431 		do {
5432 			rover++;
5433 			if ((rover < low) || (rover > high))
5434 				rover = low;
5435 			index = sctp_phashfn(rover);
5436 			head = &sctp_port_hashtable[index];
5437 			sctp_spin_lock(&head->lock);
5438 			sctp_for_each_hentry(pp, node, &head->chain)
5439 				if (pp->port == rover)
5440 					goto next;
5441 			break;
5442 		next:
5443 			sctp_spin_unlock(&head->lock);
5444 		} while (--remaining > 0);
5445 
5446 		/* Exhausted local port range during search? */
5447 		ret = 1;
5448 		if (remaining <= 0)
5449 			goto fail;
5450 
5451 		/* OK, here is the one we will use.  HEAD (the port
5452 		 * hash table list entry) is non-NULL and we hold it's
5453 		 * mutex.
5454 		 */
5455 		snum = rover;
5456 	} else {
5457 		/* We are given an specific port number; we verify
5458 		 * that it is not being used. If it is used, we will
5459 		 * exahust the search in the hash list corresponding
5460 		 * to the port number (snum) - we detect that with the
5461 		 * port iterator, pp being NULL.
5462 		 */
5463 		head = &sctp_port_hashtable[sctp_phashfn(snum)];
5464 		sctp_spin_lock(&head->lock);
5465 		sctp_for_each_hentry(pp, node, &head->chain) {
5466 			if (pp->port == snum)
5467 				goto pp_found;
5468 		}
5469 	}
5470 	pp = NULL;
5471 	goto pp_not_found;
5472 pp_found:
5473 	if (!hlist_empty(&pp->owner)) {
5474 		/* We had a port hash table hit - there is an
5475 		 * available port (pp != NULL) and it is being
5476 		 * used by other socket (pp->owner not empty); that other
5477 		 * socket is going to be sk2.
5478 		 */
5479 		int reuse = sk->sk_reuse;
5480 		struct sock *sk2;
5481 		struct hlist_node *node;
5482 
5483 		SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5484 		if (pp->fastreuse && sk->sk_reuse &&
5485 			sk->sk_state != SCTP_SS_LISTENING)
5486 			goto success;
5487 
5488 		/* Run through the list of sockets bound to the port
5489 		 * (pp->port) [via the pointers bind_next and
5490 		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5491 		 * we get the endpoint they describe and run through
5492 		 * the endpoint's list of IP (v4 or v6) addresses,
5493 		 * comparing each of the addresses with the address of
5494 		 * the socket sk. If we find a match, then that means
5495 		 * that this port/socket (sk) combination are already
5496 		 * in an endpoint.
5497 		 */
5498 		sk_for_each_bound(sk2, node, &pp->owner) {
5499 			struct sctp_endpoint *ep2;
5500 			ep2 = sctp_sk(sk2)->ep;
5501 
5502 			if (sk == sk2 ||
5503 			    (reuse && sk2->sk_reuse &&
5504 			     sk2->sk_state != SCTP_SS_LISTENING))
5505 				continue;
5506 
5507 			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5508 						 sctp_sk(sk2), sctp_sk(sk))) {
5509 				ret = (long)sk2;
5510 				goto fail_unlock;
5511 			}
5512 		}
5513 		SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5514 	}
5515 pp_not_found:
5516 	/* If there was a hash table miss, create a new port.  */
5517 	ret = 1;
5518 	if (!pp && !(pp = sctp_bucket_create(head, snum)))
5519 		goto fail_unlock;
5520 
5521 	/* In either case (hit or miss), make sure fastreuse is 1 only
5522 	 * if sk->sk_reuse is too (that is, if the caller requested
5523 	 * SO_REUSEADDR on this socket -sk-).
5524 	 */
5525 	if (hlist_empty(&pp->owner)) {
5526 		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5527 			pp->fastreuse = 1;
5528 		else
5529 			pp->fastreuse = 0;
5530 	} else if (pp->fastreuse &&
5531 		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5532 		pp->fastreuse = 0;
5533 
5534 	/* We are set, so fill up all the data in the hash table
5535 	 * entry, tie the socket list information with the rest of the
5536 	 * sockets FIXME: Blurry, NPI (ipg).
5537 	 */
5538 success:
5539 	if (!sctp_sk(sk)->bind_hash) {
5540 		inet_sk(sk)->inet_num = snum;
5541 		sk_add_bind_node(sk, &pp->owner);
5542 		sctp_sk(sk)->bind_hash = pp;
5543 	}
5544 	ret = 0;
5545 
5546 fail_unlock:
5547 	sctp_spin_unlock(&head->lock);
5548 
5549 fail:
5550 	sctp_local_bh_enable();
5551 	return ret;
5552 }
5553 
5554 /* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5555  * port is requested.
5556  */
5557 static int sctp_get_port(struct sock *sk, unsigned short snum)
5558 {
5559 	long ret;
5560 	union sctp_addr addr;
5561 	struct sctp_af *af = sctp_sk(sk)->pf->af;
5562 
5563 	/* Set up a dummy address struct from the sk. */
5564 	af->from_sk(&addr, sk);
5565 	addr.v4.sin_port = htons(snum);
5566 
5567 	/* Note: sk->sk_num gets filled in if ephemeral port request. */
5568 	ret = sctp_get_port_local(sk, &addr);
5569 
5570 	return (ret ? 1 : 0);
5571 }
5572 
5573 /*
5574  *  Move a socket to LISTENING state.
5575  */
5576 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5577 {
5578 	struct sctp_sock *sp = sctp_sk(sk);
5579 	struct sctp_endpoint *ep = sp->ep;
5580 	struct crypto_hash *tfm = NULL;
5581 
5582 	/* Allocate HMAC for generating cookie. */
5583 	if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5584 		tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5585 		if (IS_ERR(tfm)) {
5586 			if (net_ratelimit()) {
5587 				printk(KERN_INFO
5588 				       "SCTP: failed to load transform for %s: %ld\n",
5589 					sctp_hmac_alg, PTR_ERR(tfm));
5590 			}
5591 			return -ENOSYS;
5592 		}
5593 		sctp_sk(sk)->hmac = tfm;
5594 	}
5595 
5596 	/*
5597 	 * If a bind() or sctp_bindx() is not called prior to a listen()
5598 	 * call that allows new associations to be accepted, the system
5599 	 * picks an ephemeral port and will choose an address set equivalent
5600 	 * to binding with a wildcard address.
5601 	 *
5602 	 * This is not currently spelled out in the SCTP sockets
5603 	 * extensions draft, but follows the practice as seen in TCP
5604 	 * sockets.
5605 	 *
5606 	 */
5607 	sk->sk_state = SCTP_SS_LISTENING;
5608 	if (!ep->base.bind_addr.port) {
5609 		if (sctp_autobind(sk))
5610 			return -EAGAIN;
5611 	} else {
5612 		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5613 			sk->sk_state = SCTP_SS_CLOSED;
5614 			return -EADDRINUSE;
5615 		}
5616 	}
5617 
5618 	sk->sk_max_ack_backlog = backlog;
5619 	sctp_hash_endpoint(ep);
5620 	return 0;
5621 }
5622 
5623 /*
5624  * 4.1.3 / 5.1.3 listen()
5625  *
5626  *   By default, new associations are not accepted for UDP style sockets.
5627  *   An application uses listen() to mark a socket as being able to
5628  *   accept new associations.
5629  *
5630  *   On TCP style sockets, applications use listen() to ready the SCTP
5631  *   endpoint for accepting inbound associations.
5632  *
5633  *   On both types of endpoints a backlog of '0' disables listening.
5634  *
5635  *  Move a socket to LISTENING state.
5636  */
5637 int sctp_inet_listen(struct socket *sock, int backlog)
5638 {
5639 	struct sock *sk = sock->sk;
5640 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5641 	int err = -EINVAL;
5642 
5643 	if (unlikely(backlog < 0))
5644 		return err;
5645 
5646 	sctp_lock_sock(sk);
5647 
5648 	/* Peeled-off sockets are not allowed to listen().  */
5649 	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5650 		goto out;
5651 
5652 	if (sock->state != SS_UNCONNECTED)
5653 		goto out;
5654 
5655 	/* If backlog is zero, disable listening. */
5656 	if (!backlog) {
5657 		if (sctp_sstate(sk, CLOSED))
5658 			goto out;
5659 
5660 		err = 0;
5661 		sctp_unhash_endpoint(ep);
5662 		sk->sk_state = SCTP_SS_CLOSED;
5663 		if (sk->sk_reuse)
5664 			sctp_sk(sk)->bind_hash->fastreuse = 1;
5665 		goto out;
5666 	}
5667 
5668 	/* If we are already listening, just update the backlog */
5669 	if (sctp_sstate(sk, LISTENING))
5670 		sk->sk_max_ack_backlog = backlog;
5671 	else {
5672 		err = sctp_listen_start(sk, backlog);
5673 		if (err)
5674 			goto out;
5675 	}
5676 
5677 	err = 0;
5678 out:
5679 	sctp_release_sock(sk);
5680 	return err;
5681 }
5682 
5683 /*
5684  * This function is done by modeling the current datagram_poll() and the
5685  * tcp_poll().  Note that, based on these implementations, we don't
5686  * lock the socket in this function, even though it seems that,
5687  * ideally, locking or some other mechanisms can be used to ensure
5688  * the integrity of the counters (sndbuf and wmem_alloc) used
5689  * in this place.  We assume that we don't need locks either until proven
5690  * otherwise.
5691  *
5692  * Another thing to note is that we include the Async I/O support
5693  * here, again, by modeling the current TCP/UDP code.  We don't have
5694  * a good way to test with it yet.
5695  */
5696 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5697 {
5698 	struct sock *sk = sock->sk;
5699 	struct sctp_sock *sp = sctp_sk(sk);
5700 	unsigned int mask;
5701 
5702 	poll_wait(file, sk->sk_sleep, wait);
5703 
5704 	/* A TCP-style listening socket becomes readable when the accept queue
5705 	 * is not empty.
5706 	 */
5707 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5708 		return (!list_empty(&sp->ep->asocs)) ?
5709 			(POLLIN | POLLRDNORM) : 0;
5710 
5711 	mask = 0;
5712 
5713 	/* Is there any exceptional events?  */
5714 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5715 		mask |= POLLERR;
5716 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5717 		mask |= POLLRDHUP;
5718 	if (sk->sk_shutdown == SHUTDOWN_MASK)
5719 		mask |= POLLHUP;
5720 
5721 	/* Is it readable?  Reconsider this code with TCP-style support.  */
5722 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
5723 	    (sk->sk_shutdown & RCV_SHUTDOWN))
5724 		mask |= POLLIN | POLLRDNORM;
5725 
5726 	/* The association is either gone or not ready.  */
5727 	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5728 		return mask;
5729 
5730 	/* Is it writable?  */
5731 	if (sctp_writeable(sk)) {
5732 		mask |= POLLOUT | POLLWRNORM;
5733 	} else {
5734 		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5735 		/*
5736 		 * Since the socket is not locked, the buffer
5737 		 * might be made available after the writeable check and
5738 		 * before the bit is set.  This could cause a lost I/O
5739 		 * signal.  tcp_poll() has a race breaker for this race
5740 		 * condition.  Based on their implementation, we put
5741 		 * in the following code to cover it as well.
5742 		 */
5743 		if (sctp_writeable(sk))
5744 			mask |= POLLOUT | POLLWRNORM;
5745 	}
5746 	return mask;
5747 }
5748 
5749 /********************************************************************
5750  * 2nd Level Abstractions
5751  ********************************************************************/
5752 
5753 static struct sctp_bind_bucket *sctp_bucket_create(
5754 	struct sctp_bind_hashbucket *head, unsigned short snum)
5755 {
5756 	struct sctp_bind_bucket *pp;
5757 
5758 	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5759 	if (pp) {
5760 		SCTP_DBG_OBJCNT_INC(bind_bucket);
5761 		pp->port = snum;
5762 		pp->fastreuse = 0;
5763 		INIT_HLIST_HEAD(&pp->owner);
5764 		hlist_add_head(&pp->node, &head->chain);
5765 	}
5766 	return pp;
5767 }
5768 
5769 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5770 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5771 {
5772 	if (pp && hlist_empty(&pp->owner)) {
5773 		__hlist_del(&pp->node);
5774 		kmem_cache_free(sctp_bucket_cachep, pp);
5775 		SCTP_DBG_OBJCNT_DEC(bind_bucket);
5776 	}
5777 }
5778 
5779 /* Release this socket's reference to a local port.  */
5780 static inline void __sctp_put_port(struct sock *sk)
5781 {
5782 	struct sctp_bind_hashbucket *head =
5783 		&sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5784 	struct sctp_bind_bucket *pp;
5785 
5786 	sctp_spin_lock(&head->lock);
5787 	pp = sctp_sk(sk)->bind_hash;
5788 	__sk_del_bind_node(sk);
5789 	sctp_sk(sk)->bind_hash = NULL;
5790 	inet_sk(sk)->inet_num = 0;
5791 	sctp_bucket_destroy(pp);
5792 	sctp_spin_unlock(&head->lock);
5793 }
5794 
5795 void sctp_put_port(struct sock *sk)
5796 {
5797 	sctp_local_bh_disable();
5798 	__sctp_put_port(sk);
5799 	sctp_local_bh_enable();
5800 }
5801 
5802 /*
5803  * The system picks an ephemeral port and choose an address set equivalent
5804  * to binding with a wildcard address.
5805  * One of those addresses will be the primary address for the association.
5806  * This automatically enables the multihoming capability of SCTP.
5807  */
5808 static int sctp_autobind(struct sock *sk)
5809 {
5810 	union sctp_addr autoaddr;
5811 	struct sctp_af *af;
5812 	__be16 port;
5813 
5814 	/* Initialize a local sockaddr structure to INADDR_ANY. */
5815 	af = sctp_sk(sk)->pf->af;
5816 
5817 	port = htons(inet_sk(sk)->inet_num);
5818 	af->inaddr_any(&autoaddr, port);
5819 
5820 	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5821 }
5822 
5823 /* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
5824  *
5825  * From RFC 2292
5826  * 4.2 The cmsghdr Structure *
5827  *
5828  * When ancillary data is sent or received, any number of ancillary data
5829  * objects can be specified by the msg_control and msg_controllen members of
5830  * the msghdr structure, because each object is preceded by
5831  * a cmsghdr structure defining the object's length (the cmsg_len member).
5832  * Historically Berkeley-derived implementations have passed only one object
5833  * at a time, but this API allows multiple objects to be
5834  * passed in a single call to sendmsg() or recvmsg(). The following example
5835  * shows two ancillary data objects in a control buffer.
5836  *
5837  *   |<--------------------------- msg_controllen -------------------------->|
5838  *   |                                                                       |
5839  *
5840  *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
5841  *
5842  *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5843  *   |                                   |                                   |
5844  *
5845  *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
5846  *
5847  *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
5848  *   |                                |  |                                |  |
5849  *
5850  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5851  *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
5852  *
5853  *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
5854  *
5855  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5856  *    ^
5857  *    |
5858  *
5859  * msg_control
5860  * points here
5861  */
5862 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5863 				  sctp_cmsgs_t *cmsgs)
5864 {
5865 	struct cmsghdr *cmsg;
5866 	struct msghdr *my_msg = (struct msghdr *)msg;
5867 
5868 	for (cmsg = CMSG_FIRSTHDR(msg);
5869 	     cmsg != NULL;
5870 	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5871 		if (!CMSG_OK(my_msg, cmsg))
5872 			return -EINVAL;
5873 
5874 		/* Should we parse this header or ignore?  */
5875 		if (cmsg->cmsg_level != IPPROTO_SCTP)
5876 			continue;
5877 
5878 		/* Strictly check lengths following example in SCM code.  */
5879 		switch (cmsg->cmsg_type) {
5880 		case SCTP_INIT:
5881 			/* SCTP Socket API Extension
5882 			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5883 			 *
5884 			 * This cmsghdr structure provides information for
5885 			 * initializing new SCTP associations with sendmsg().
5886 			 * The SCTP_INITMSG socket option uses this same data
5887 			 * structure.  This structure is not used for
5888 			 * recvmsg().
5889 			 *
5890 			 * cmsg_level    cmsg_type      cmsg_data[]
5891 			 * ------------  ------------   ----------------------
5892 			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
5893 			 */
5894 			if (cmsg->cmsg_len !=
5895 			    CMSG_LEN(sizeof(struct sctp_initmsg)))
5896 				return -EINVAL;
5897 			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5898 			break;
5899 
5900 		case SCTP_SNDRCV:
5901 			/* SCTP Socket API Extension
5902 			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5903 			 *
5904 			 * This cmsghdr structure specifies SCTP options for
5905 			 * sendmsg() and describes SCTP header information
5906 			 * about a received message through recvmsg().
5907 			 *
5908 			 * cmsg_level    cmsg_type      cmsg_data[]
5909 			 * ------------  ------------   ----------------------
5910 			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
5911 			 */
5912 			if (cmsg->cmsg_len !=
5913 			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5914 				return -EINVAL;
5915 
5916 			cmsgs->info =
5917 				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5918 
5919 			/* Minimally, validate the sinfo_flags. */
5920 			if (cmsgs->info->sinfo_flags &
5921 			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5922 			      SCTP_ABORT | SCTP_EOF))
5923 				return -EINVAL;
5924 			break;
5925 
5926 		default:
5927 			return -EINVAL;
5928 		}
5929 	}
5930 	return 0;
5931 }
5932 
5933 /*
5934  * Wait for a packet..
5935  * Note: This function is the same function as in core/datagram.c
5936  * with a few modifications to make lksctp work.
5937  */
5938 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5939 {
5940 	int error;
5941 	DEFINE_WAIT(wait);
5942 
5943 	prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5944 
5945 	/* Socket errors? */
5946 	error = sock_error(sk);
5947 	if (error)
5948 		goto out;
5949 
5950 	if (!skb_queue_empty(&sk->sk_receive_queue))
5951 		goto ready;
5952 
5953 	/* Socket shut down?  */
5954 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5955 		goto out;
5956 
5957 	/* Sequenced packets can come disconnected.  If so we report the
5958 	 * problem.
5959 	 */
5960 	error = -ENOTCONN;
5961 
5962 	/* Is there a good reason to think that we may receive some data?  */
5963 	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5964 		goto out;
5965 
5966 	/* Handle signals.  */
5967 	if (signal_pending(current))
5968 		goto interrupted;
5969 
5970 	/* Let another process have a go.  Since we are going to sleep
5971 	 * anyway.  Note: This may cause odd behaviors if the message
5972 	 * does not fit in the user's buffer, but this seems to be the
5973 	 * only way to honor MSG_DONTWAIT realistically.
5974 	 */
5975 	sctp_release_sock(sk);
5976 	*timeo_p = schedule_timeout(*timeo_p);
5977 	sctp_lock_sock(sk);
5978 
5979 ready:
5980 	finish_wait(sk->sk_sleep, &wait);
5981 	return 0;
5982 
5983 interrupted:
5984 	error = sock_intr_errno(*timeo_p);
5985 
5986 out:
5987 	finish_wait(sk->sk_sleep, &wait);
5988 	*err = error;
5989 	return error;
5990 }
5991 
5992 /* Receive a datagram.
5993  * Note: This is pretty much the same routine as in core/datagram.c
5994  * with a few changes to make lksctp work.
5995  */
5996 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5997 					      int noblock, int *err)
5998 {
5999 	int error;
6000 	struct sk_buff *skb;
6001 	long timeo;
6002 
6003 	timeo = sock_rcvtimeo(sk, noblock);
6004 
6005 	SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6006 			  timeo, MAX_SCHEDULE_TIMEOUT);
6007 
6008 	do {
6009 		/* Again only user level code calls this function,
6010 		 * so nothing interrupt level
6011 		 * will suddenly eat the receive_queue.
6012 		 *
6013 		 *  Look at current nfs client by the way...
6014 		 *  However, this function was corrent in any case. 8)
6015 		 */
6016 		if (flags & MSG_PEEK) {
6017 			spin_lock_bh(&sk->sk_receive_queue.lock);
6018 			skb = skb_peek(&sk->sk_receive_queue);
6019 			if (skb)
6020 				atomic_inc(&skb->users);
6021 			spin_unlock_bh(&sk->sk_receive_queue.lock);
6022 		} else {
6023 			skb = skb_dequeue(&sk->sk_receive_queue);
6024 		}
6025 
6026 		if (skb)
6027 			return skb;
6028 
6029 		/* Caller is allowed not to check sk->sk_err before calling. */
6030 		error = sock_error(sk);
6031 		if (error)
6032 			goto no_packet;
6033 
6034 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6035 			break;
6036 
6037 		/* User doesn't want to wait.  */
6038 		error = -EAGAIN;
6039 		if (!timeo)
6040 			goto no_packet;
6041 	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6042 
6043 	return NULL;
6044 
6045 no_packet:
6046 	*err = error;
6047 	return NULL;
6048 }
6049 
6050 /* If sndbuf has changed, wake up per association sndbuf waiters.  */
6051 static void __sctp_write_space(struct sctp_association *asoc)
6052 {
6053 	struct sock *sk = asoc->base.sk;
6054 	struct socket *sock = sk->sk_socket;
6055 
6056 	if ((sctp_wspace(asoc) > 0) && sock) {
6057 		if (waitqueue_active(&asoc->wait))
6058 			wake_up_interruptible(&asoc->wait);
6059 
6060 		if (sctp_writeable(sk)) {
6061 			if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6062 				wake_up_interruptible(sk->sk_sleep);
6063 
6064 			/* Note that we try to include the Async I/O support
6065 			 * here by modeling from the current TCP/UDP code.
6066 			 * We have not tested with it yet.
6067 			 */
6068 			if (sock->fasync_list &&
6069 			    !(sk->sk_shutdown & SEND_SHUTDOWN))
6070 				sock_wake_async(sock,
6071 						SOCK_WAKE_SPACE, POLL_OUT);
6072 		}
6073 	}
6074 }
6075 
6076 /* Do accounting for the sndbuf space.
6077  * Decrement the used sndbuf space of the corresponding association by the
6078  * data size which was just transmitted(freed).
6079  */
6080 static void sctp_wfree(struct sk_buff *skb)
6081 {
6082 	struct sctp_association *asoc;
6083 	struct sctp_chunk *chunk;
6084 	struct sock *sk;
6085 
6086 	/* Get the saved chunk pointer.  */
6087 	chunk = *((struct sctp_chunk **)(skb->cb));
6088 	asoc = chunk->asoc;
6089 	sk = asoc->base.sk;
6090 	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6091 				sizeof(struct sk_buff) +
6092 				sizeof(struct sctp_chunk);
6093 
6094 	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6095 
6096 	/*
6097 	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6098 	 */
6099 	sk->sk_wmem_queued   -= skb->truesize;
6100 	sk_mem_uncharge(sk, skb->truesize);
6101 
6102 	sock_wfree(skb);
6103 	__sctp_write_space(asoc);
6104 
6105 	sctp_association_put(asoc);
6106 }
6107 
6108 /* Do accounting for the receive space on the socket.
6109  * Accounting for the association is done in ulpevent.c
6110  * We set this as a destructor for the cloned data skbs so that
6111  * accounting is done at the correct time.
6112  */
6113 void sctp_sock_rfree(struct sk_buff *skb)
6114 {
6115 	struct sock *sk = skb->sk;
6116 	struct sctp_ulpevent *event = sctp_skb2event(skb);
6117 
6118 	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6119 
6120 	/*
6121 	 * Mimic the behavior of sock_rfree
6122 	 */
6123 	sk_mem_uncharge(sk, event->rmem_len);
6124 }
6125 
6126 
6127 /* Helper function to wait for space in the sndbuf.  */
6128 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6129 				size_t msg_len)
6130 {
6131 	struct sock *sk = asoc->base.sk;
6132 	int err = 0;
6133 	long current_timeo = *timeo_p;
6134 	DEFINE_WAIT(wait);
6135 
6136 	SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6137 			  asoc, (long)(*timeo_p), msg_len);
6138 
6139 	/* Increment the association's refcnt.  */
6140 	sctp_association_hold(asoc);
6141 
6142 	/* Wait on the association specific sndbuf space. */
6143 	for (;;) {
6144 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6145 					  TASK_INTERRUPTIBLE);
6146 		if (!*timeo_p)
6147 			goto do_nonblock;
6148 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6149 		    asoc->base.dead)
6150 			goto do_error;
6151 		if (signal_pending(current))
6152 			goto do_interrupted;
6153 		if (msg_len <= sctp_wspace(asoc))
6154 			break;
6155 
6156 		/* Let another process have a go.  Since we are going
6157 		 * to sleep anyway.
6158 		 */
6159 		sctp_release_sock(sk);
6160 		current_timeo = schedule_timeout(current_timeo);
6161 		BUG_ON(sk != asoc->base.sk);
6162 		sctp_lock_sock(sk);
6163 
6164 		*timeo_p = current_timeo;
6165 	}
6166 
6167 out:
6168 	finish_wait(&asoc->wait, &wait);
6169 
6170 	/* Release the association's refcnt.  */
6171 	sctp_association_put(asoc);
6172 
6173 	return err;
6174 
6175 do_error:
6176 	err = -EPIPE;
6177 	goto out;
6178 
6179 do_interrupted:
6180 	err = sock_intr_errno(*timeo_p);
6181 	goto out;
6182 
6183 do_nonblock:
6184 	err = -EAGAIN;
6185 	goto out;
6186 }
6187 
6188 /* If socket sndbuf has changed, wake up all per association waiters.  */
6189 void sctp_write_space(struct sock *sk)
6190 {
6191 	struct sctp_association *asoc;
6192 
6193 	/* Wake up the tasks in each wait queue.  */
6194 	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6195 		__sctp_write_space(asoc);
6196 	}
6197 }
6198 
6199 /* Is there any sndbuf space available on the socket?
6200  *
6201  * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6202  * associations on the same socket.  For a UDP-style socket with
6203  * multiple associations, it is possible for it to be "unwriteable"
6204  * prematurely.  I assume that this is acceptable because
6205  * a premature "unwriteable" is better than an accidental "writeable" which
6206  * would cause an unwanted block under certain circumstances.  For the 1-1
6207  * UDP-style sockets or TCP-style sockets, this code should work.
6208  *  - Daisy
6209  */
6210 static int sctp_writeable(struct sock *sk)
6211 {
6212 	int amt = 0;
6213 
6214 	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6215 	if (amt < 0)
6216 		amt = 0;
6217 	return amt;
6218 }
6219 
6220 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6221  * returns immediately with EINPROGRESS.
6222  */
6223 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6224 {
6225 	struct sock *sk = asoc->base.sk;
6226 	int err = 0;
6227 	long current_timeo = *timeo_p;
6228 	DEFINE_WAIT(wait);
6229 
6230 	SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6231 			  (long)(*timeo_p));
6232 
6233 	/* Increment the association's refcnt.  */
6234 	sctp_association_hold(asoc);
6235 
6236 	for (;;) {
6237 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6238 					  TASK_INTERRUPTIBLE);
6239 		if (!*timeo_p)
6240 			goto do_nonblock;
6241 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6242 			break;
6243 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6244 		    asoc->base.dead)
6245 			goto do_error;
6246 		if (signal_pending(current))
6247 			goto do_interrupted;
6248 
6249 		if (sctp_state(asoc, ESTABLISHED))
6250 			break;
6251 
6252 		/* Let another process have a go.  Since we are going
6253 		 * to sleep anyway.
6254 		 */
6255 		sctp_release_sock(sk);
6256 		current_timeo = schedule_timeout(current_timeo);
6257 		sctp_lock_sock(sk);
6258 
6259 		*timeo_p = current_timeo;
6260 	}
6261 
6262 out:
6263 	finish_wait(&asoc->wait, &wait);
6264 
6265 	/* Release the association's refcnt.  */
6266 	sctp_association_put(asoc);
6267 
6268 	return err;
6269 
6270 do_error:
6271 	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6272 		err = -ETIMEDOUT;
6273 	else
6274 		err = -ECONNREFUSED;
6275 	goto out;
6276 
6277 do_interrupted:
6278 	err = sock_intr_errno(*timeo_p);
6279 	goto out;
6280 
6281 do_nonblock:
6282 	err = -EINPROGRESS;
6283 	goto out;
6284 }
6285 
6286 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6287 {
6288 	struct sctp_endpoint *ep;
6289 	int err = 0;
6290 	DEFINE_WAIT(wait);
6291 
6292 	ep = sctp_sk(sk)->ep;
6293 
6294 
6295 	for (;;) {
6296 		prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6297 					  TASK_INTERRUPTIBLE);
6298 
6299 		if (list_empty(&ep->asocs)) {
6300 			sctp_release_sock(sk);
6301 			timeo = schedule_timeout(timeo);
6302 			sctp_lock_sock(sk);
6303 		}
6304 
6305 		err = -EINVAL;
6306 		if (!sctp_sstate(sk, LISTENING))
6307 			break;
6308 
6309 		err = 0;
6310 		if (!list_empty(&ep->asocs))
6311 			break;
6312 
6313 		err = sock_intr_errno(timeo);
6314 		if (signal_pending(current))
6315 			break;
6316 
6317 		err = -EAGAIN;
6318 		if (!timeo)
6319 			break;
6320 	}
6321 
6322 	finish_wait(sk->sk_sleep, &wait);
6323 
6324 	return err;
6325 }
6326 
6327 static void sctp_wait_for_close(struct sock *sk, long timeout)
6328 {
6329 	DEFINE_WAIT(wait);
6330 
6331 	do {
6332 		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6333 		if (list_empty(&sctp_sk(sk)->ep->asocs))
6334 			break;
6335 		sctp_release_sock(sk);
6336 		timeout = schedule_timeout(timeout);
6337 		sctp_lock_sock(sk);
6338 	} while (!signal_pending(current) && timeout);
6339 
6340 	finish_wait(sk->sk_sleep, &wait);
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 	skb_walk_frags(skb, frag)
6352 		sctp_skb_set_owner_r_frag(frag, sk);
6353 
6354 done:
6355 	sctp_skb_set_owner_r(skb, sk);
6356 }
6357 
6358 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6359 		    struct sctp_association *asoc)
6360 {
6361 	struct inet_sock *inet = inet_sk(sk);
6362 	struct inet_sock *newinet = inet_sk(newsk);
6363 
6364 	newsk->sk_type = sk->sk_type;
6365 	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6366 	newsk->sk_flags = sk->sk_flags;
6367 	newsk->sk_no_check = sk->sk_no_check;
6368 	newsk->sk_reuse = sk->sk_reuse;
6369 
6370 	newsk->sk_shutdown = sk->sk_shutdown;
6371 	newsk->sk_destruct = inet_sock_destruct;
6372 	newsk->sk_family = sk->sk_family;
6373 	newsk->sk_protocol = IPPROTO_SCTP;
6374 	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6375 	newsk->sk_sndbuf = sk->sk_sndbuf;
6376 	newsk->sk_rcvbuf = sk->sk_rcvbuf;
6377 	newsk->sk_lingertime = sk->sk_lingertime;
6378 	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6379 	newsk->sk_sndtimeo = sk->sk_sndtimeo;
6380 
6381 	newinet = inet_sk(newsk);
6382 
6383 	/* Initialize sk's sport, dport, rcv_saddr and daddr for
6384 	 * getsockname() and getpeername()
6385 	 */
6386 	newinet->inet_sport = inet->inet_sport;
6387 	newinet->inet_saddr = inet->inet_saddr;
6388 	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6389 	newinet->inet_dport = htons(asoc->peer.port);
6390 	newinet->pmtudisc = inet->pmtudisc;
6391 	newinet->inet_id = asoc->next_tsn ^ jiffies;
6392 
6393 	newinet->uc_ttl = inet->uc_ttl;
6394 	newinet->mc_loop = 1;
6395 	newinet->mc_ttl = 1;
6396 	newinet->mc_index = 0;
6397 	newinet->mc_list = NULL;
6398 }
6399 
6400 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6401  * and its messages to the newsk.
6402  */
6403 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6404 			      struct sctp_association *assoc,
6405 			      sctp_socket_type_t type)
6406 {
6407 	struct sctp_sock *oldsp = sctp_sk(oldsk);
6408 	struct sctp_sock *newsp = sctp_sk(newsk);
6409 	struct sctp_bind_bucket *pp; /* hash list port iterator */
6410 	struct sctp_endpoint *newep = newsp->ep;
6411 	struct sk_buff *skb, *tmp;
6412 	struct sctp_ulpevent *event;
6413 	struct sctp_bind_hashbucket *head;
6414 
6415 	/* Migrate socket buffer sizes and all the socket level options to the
6416 	 * new socket.
6417 	 */
6418 	newsk->sk_sndbuf = oldsk->sk_sndbuf;
6419 	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6420 	/* Brute force copy old sctp opt. */
6421 	inet_sk_copy_descendant(newsk, oldsk);
6422 
6423 	/* Restore the ep value that was overwritten with the above structure
6424 	 * copy.
6425 	 */
6426 	newsp->ep = newep;
6427 	newsp->hmac = NULL;
6428 
6429 	/* Hook this new socket in to the bind_hash list. */
6430 	head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6431 	sctp_local_bh_disable();
6432 	sctp_spin_lock(&head->lock);
6433 	pp = sctp_sk(oldsk)->bind_hash;
6434 	sk_add_bind_node(newsk, &pp->owner);
6435 	sctp_sk(newsk)->bind_hash = pp;
6436 	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6437 	sctp_spin_unlock(&head->lock);
6438 	sctp_local_bh_enable();
6439 
6440 	/* Copy the bind_addr list from the original endpoint to the new
6441 	 * endpoint so that we can handle restarts properly
6442 	 */
6443 	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6444 				&oldsp->ep->base.bind_addr, GFP_KERNEL);
6445 
6446 	/* Move any messages in the old socket's receive queue that are for the
6447 	 * peeled off association to the new socket's receive queue.
6448 	 */
6449 	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6450 		event = sctp_skb2event(skb);
6451 		if (event->asoc == assoc) {
6452 			__skb_unlink(skb, &oldsk->sk_receive_queue);
6453 			__skb_queue_tail(&newsk->sk_receive_queue, skb);
6454 			sctp_skb_set_owner_r_frag(skb, newsk);
6455 		}
6456 	}
6457 
6458 	/* Clean up any messages pending delivery due to partial
6459 	 * delivery.   Three cases:
6460 	 * 1) No partial deliver;  no work.
6461 	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6462 	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6463 	 */
6464 	skb_queue_head_init(&newsp->pd_lobby);
6465 	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6466 
6467 	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6468 		struct sk_buff_head *queue;
6469 
6470 		/* Decide which queue to move pd_lobby skbs to. */
6471 		if (assoc->ulpq.pd_mode) {
6472 			queue = &newsp->pd_lobby;
6473 		} else
6474 			queue = &newsk->sk_receive_queue;
6475 
6476 		/* Walk through the pd_lobby, looking for skbs that
6477 		 * need moved to the new socket.
6478 		 */
6479 		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6480 			event = sctp_skb2event(skb);
6481 			if (event->asoc == assoc) {
6482 				__skb_unlink(skb, &oldsp->pd_lobby);
6483 				__skb_queue_tail(queue, skb);
6484 				sctp_skb_set_owner_r_frag(skb, newsk);
6485 			}
6486 		}
6487 
6488 		/* Clear up any skbs waiting for the partial
6489 		 * delivery to finish.
6490 		 */
6491 		if (assoc->ulpq.pd_mode)
6492 			sctp_clear_pd(oldsk, NULL);
6493 
6494 	}
6495 
6496 	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6497 		sctp_skb_set_owner_r_frag(skb, newsk);
6498 
6499 	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6500 		sctp_skb_set_owner_r_frag(skb, newsk);
6501 
6502 	/* Set the type of socket to indicate that it is peeled off from the
6503 	 * original UDP-style socket or created with the accept() call on a
6504 	 * TCP-style socket..
6505 	 */
6506 	newsp->type = type;
6507 
6508 	/* Mark the new socket "in-use" by the user so that any packets
6509 	 * that may arrive on the association after we've moved it are
6510 	 * queued to the backlog.  This prevents a potential race between
6511 	 * backlog processing on the old socket and new-packet processing
6512 	 * on the new socket.
6513 	 *
6514 	 * The caller has just allocated newsk so we can guarantee that other
6515 	 * paths won't try to lock it and then oldsk.
6516 	 */
6517 	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6518 	sctp_assoc_migrate(assoc, newsk);
6519 
6520 	/* If the association on the newsk is already closed before accept()
6521 	 * is called, set RCV_SHUTDOWN flag.
6522 	 */
6523 	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6524 		newsk->sk_shutdown |= RCV_SHUTDOWN;
6525 
6526 	newsk->sk_state = SCTP_SS_ESTABLISHED;
6527 	sctp_release_sock(newsk);
6528 }
6529 
6530 
6531 /* This proto struct describes the ULP interface for SCTP.  */
6532 struct proto sctp_prot = {
6533 	.name        =	"SCTP",
6534 	.owner       =	THIS_MODULE,
6535 	.close       =	sctp_close,
6536 	.connect     =	sctp_connect,
6537 	.disconnect  =	sctp_disconnect,
6538 	.accept      =	sctp_accept,
6539 	.ioctl       =	sctp_ioctl,
6540 	.init        =	sctp_init_sock,
6541 	.destroy     =	sctp_destroy_sock,
6542 	.shutdown    =	sctp_shutdown,
6543 	.setsockopt  =	sctp_setsockopt,
6544 	.getsockopt  =	sctp_getsockopt,
6545 	.sendmsg     =	sctp_sendmsg,
6546 	.recvmsg     =	sctp_recvmsg,
6547 	.bind        =	sctp_bind,
6548 	.backlog_rcv =	sctp_backlog_rcv,
6549 	.hash        =	sctp_hash,
6550 	.unhash      =	sctp_unhash,
6551 	.get_port    =	sctp_get_port,
6552 	.obj_size    =  sizeof(struct sctp_sock),
6553 	.sysctl_mem  =  sysctl_sctp_mem,
6554 	.sysctl_rmem =  sysctl_sctp_rmem,
6555 	.sysctl_wmem =  sysctl_sctp_wmem,
6556 	.memory_pressure = &sctp_memory_pressure,
6557 	.enter_memory_pressure = sctp_enter_memory_pressure,
6558 	.memory_allocated = &sctp_memory_allocated,
6559 	.sockets_allocated = &sctp_sockets_allocated,
6560 };
6561 
6562 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6563 
6564 struct proto sctpv6_prot = {
6565 	.name		= "SCTPv6",
6566 	.owner		= THIS_MODULE,
6567 	.close		= sctp_close,
6568 	.connect	= sctp_connect,
6569 	.disconnect	= sctp_disconnect,
6570 	.accept		= sctp_accept,
6571 	.ioctl		= sctp_ioctl,
6572 	.init		= sctp_init_sock,
6573 	.destroy	= sctp_destroy_sock,
6574 	.shutdown	= sctp_shutdown,
6575 	.setsockopt	= sctp_setsockopt,
6576 	.getsockopt	= sctp_getsockopt,
6577 	.sendmsg	= sctp_sendmsg,
6578 	.recvmsg	= sctp_recvmsg,
6579 	.bind		= sctp_bind,
6580 	.backlog_rcv	= sctp_backlog_rcv,
6581 	.hash		= sctp_hash,
6582 	.unhash		= sctp_unhash,
6583 	.get_port	= sctp_get_port,
6584 	.obj_size	= sizeof(struct sctp6_sock),
6585 	.sysctl_mem	= sysctl_sctp_mem,
6586 	.sysctl_rmem	= sysctl_sctp_rmem,
6587 	.sysctl_wmem	= sysctl_sctp_wmem,
6588 	.memory_pressure = &sctp_memory_pressure,
6589 	.enter_memory_pressure = sctp_enter_memory_pressure,
6590 	.memory_allocated = &sctp_memory_allocated,
6591 	.sockets_allocated = &sctp_sockets_allocated,
6592 };
6593 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
6594