xref: /linux/net/sctp/socket.c (revision d39d0ed196aa1685bb24771e92f78633c66ac9cb)
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 #include <linux/slab.h>
71 
72 #include <net/ip.h>
73 #include <net/icmp.h>
74 #include <net/route.h>
75 #include <net/ipv6.h>
76 #include <net/inet_common.h>
77 
78 #include <linux/socket.h> /* for sa_family_t */
79 #include <net/sock.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
82 
83 /* WARNING:  Please do not remove the SCTP_STATIC attribute to
84  * any of the functions below as they are used to export functions
85  * used by a project regression testsuite.
86  */
87 
88 /* Forward declarations for internal helper functions. */
89 static int sctp_writeable(struct sock *sk);
90 static void sctp_wfree(struct sk_buff *skb);
91 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 				size_t msg_len);
93 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
94 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
95 static int sctp_wait_for_accept(struct sock *sk, long timeo);
96 static void sctp_wait_for_close(struct sock *sk, long timeo);
97 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
98 					union sctp_addr *addr, int len);
99 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
100 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf(struct sctp_association *asoc,
104 			    struct sctp_chunk *chunk);
105 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
106 static int sctp_autobind(struct sock *sk);
107 static void sctp_sock_migrate(struct sock *, struct sock *,
108 			      struct sctp_association *, sctp_socket_type_t);
109 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110 
111 extern struct kmem_cache *sctp_bucket_cachep;
112 extern int sysctl_sctp_mem[3];
113 extern int sysctl_sctp_rmem[3];
114 extern int sysctl_sctp_wmem[3];
115 
116 static int sctp_memory_pressure;
117 static atomic_t sctp_memory_allocated;
118 struct percpu_counter sctp_sockets_allocated;
119 
120 static void sctp_enter_memory_pressure(struct sock *sk)
121 {
122 	sctp_memory_pressure = 1;
123 }
124 
125 
126 /* Get the sndbuf space available at the time on the association.  */
127 static inline int sctp_wspace(struct sctp_association *asoc)
128 {
129 	int amt;
130 
131 	if (asoc->ep->sndbuf_policy)
132 		amt = asoc->sndbuf_used;
133 	else
134 		amt = sk_wmem_alloc_get(asoc->base.sk);
135 
136 	if (amt >= asoc->base.sk->sk_sndbuf) {
137 		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
138 			amt = 0;
139 		else {
140 			amt = sk_stream_wspace(asoc->base.sk);
141 			if (amt < 0)
142 				amt = 0;
143 		}
144 	} else {
145 		amt = asoc->base.sk->sk_sndbuf - amt;
146 	}
147 	return amt;
148 }
149 
150 /* Increment the used sndbuf space count of the corresponding association by
151  * the size of the outgoing data chunk.
152  * Also, set the skb destructor for sndbuf accounting later.
153  *
154  * Since it is always 1-1 between chunk and skb, and also a new skb is always
155  * allocated for chunk bundling in sctp_packet_transmit(), we can use the
156  * destructor in the data chunk skb for the purpose of the sndbuf space
157  * tracking.
158  */
159 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 {
161 	struct sctp_association *asoc = chunk->asoc;
162 	struct sock *sk = asoc->base.sk;
163 
164 	/* The sndbuf space is tracked per association.  */
165 	sctp_association_hold(asoc);
166 
167 	skb_set_owner_w(chunk->skb, sk);
168 
169 	chunk->skb->destructor = sctp_wfree;
170 	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
171 	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 
173 	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
174 				sizeof(struct sk_buff) +
175 				sizeof(struct sctp_chunk);
176 
177 	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
178 	sk->sk_wmem_queued += chunk->skb->truesize;
179 	sk_mem_charge(sk, chunk->skb->truesize);
180 }
181 
182 /* Verify that this is a valid address. */
183 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
184 				   int len)
185 {
186 	struct sctp_af *af;
187 
188 	/* Verify basic sockaddr. */
189 	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
190 	if (!af)
191 		return -EINVAL;
192 
193 	/* Is this a valid SCTP address?  */
194 	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
195 		return -EINVAL;
196 
197 	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
198 		return -EINVAL;
199 
200 	return 0;
201 }
202 
203 /* Look up the association by its id.  If this is not a UDP-style
204  * socket, the ID field is always ignored.
205  */
206 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 {
208 	struct sctp_association *asoc = NULL;
209 
210 	/* If this is not a UDP-style socket, assoc id should be ignored. */
211 	if (!sctp_style(sk, UDP)) {
212 		/* Return NULL if the socket state is not ESTABLISHED. It
213 		 * could be a TCP-style listening socket or a socket which
214 		 * hasn't yet called connect() to establish an association.
215 		 */
216 		if (!sctp_sstate(sk, ESTABLISHED))
217 			return NULL;
218 
219 		/* Get the first and the only association from the list. */
220 		if (!list_empty(&sctp_sk(sk)->ep->asocs))
221 			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
222 					  struct sctp_association, asocs);
223 		return asoc;
224 	}
225 
226 	/* Otherwise this is a UDP-style socket. */
227 	if (!id || (id == (sctp_assoc_t)-1))
228 		return NULL;
229 
230 	spin_lock_bh(&sctp_assocs_id_lock);
231 	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
232 	spin_unlock_bh(&sctp_assocs_id_lock);
233 
234 	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
235 		return NULL;
236 
237 	return asoc;
238 }
239 
240 /* Look up the transport from an address and an assoc id. If both address and
241  * id are specified, the associations matching the address and the id should be
242  * the same.
243  */
244 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
245 					      struct sockaddr_storage *addr,
246 					      sctp_assoc_t id)
247 {
248 	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
249 	struct sctp_transport *transport;
250 	union sctp_addr *laddr = (union sctp_addr *)addr;
251 
252 	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
253 					       laddr,
254 					       &transport);
255 
256 	if (!addr_asoc)
257 		return NULL;
258 
259 	id_asoc = sctp_id2assoc(sk, id);
260 	if (id_asoc && (id_asoc != addr_asoc))
261 		return NULL;
262 
263 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
264 						(union sctp_addr *)addr);
265 
266 	return transport;
267 }
268 
269 /* API 3.1.2 bind() - UDP Style Syntax
270  * The syntax of bind() is,
271  *
272  *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
273  *
274  *   sd      - the socket descriptor returned by socket().
275  *   addr    - the address structure (struct sockaddr_in or struct
276  *             sockaddr_in6 [RFC 2553]),
277  *   addr_len - the size of the address structure.
278  */
279 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
280 {
281 	int retval = 0;
282 
283 	sctp_lock_sock(sk);
284 
285 	SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
286 			  sk, addr, addr_len);
287 
288 	/* Disallow binding twice. */
289 	if (!sctp_sk(sk)->ep->base.bind_addr.port)
290 		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
291 				      addr_len);
292 	else
293 		retval = -EINVAL;
294 
295 	sctp_release_sock(sk);
296 
297 	return retval;
298 }
299 
300 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 
302 /* Verify this is a valid sockaddr. */
303 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
304 					union sctp_addr *addr, int len)
305 {
306 	struct sctp_af *af;
307 
308 	/* Check minimum size.  */
309 	if (len < sizeof (struct sockaddr))
310 		return NULL;
311 
312 	/* V4 mapped address are really of AF_INET family */
313 	if (addr->sa.sa_family == AF_INET6 &&
314 	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
315 		if (!opt->pf->af_supported(AF_INET, opt))
316 			return NULL;
317 	} else {
318 		/* Does this PF support this AF? */
319 		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
320 			return NULL;
321 	}
322 
323 	/* If we get this far, af is valid. */
324 	af = sctp_get_af_specific(addr->sa.sa_family);
325 
326 	if (len < af->sockaddr_len)
327 		return NULL;
328 
329 	return af;
330 }
331 
332 /* Bind a local address either to an endpoint or to an association.  */
333 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 {
335 	struct sctp_sock *sp = sctp_sk(sk);
336 	struct sctp_endpoint *ep = sp->ep;
337 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
338 	struct sctp_af *af;
339 	unsigned short snum;
340 	int ret = 0;
341 
342 	/* Common sockaddr verification. */
343 	af = sctp_sockaddr_af(sp, addr, len);
344 	if (!af) {
345 		SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
346 				  sk, addr, len);
347 		return -EINVAL;
348 	}
349 
350 	snum = ntohs(addr->v4.sin_port);
351 
352 	SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
353 				 ", port: %d, new port: %d, len: %d)\n",
354 				 sk,
355 				 addr,
356 				 bp->port, snum,
357 				 len);
358 
359 	/* PF specific bind() address verification. */
360 	if (!sp->pf->bind_verify(sp, addr))
361 		return -EADDRNOTAVAIL;
362 
363 	/* We must either be unbound, or bind to the same port.
364 	 * It's OK to allow 0 ports if we are already bound.
365 	 * We'll just inhert an already bound port in this case
366 	 */
367 	if (bp->port) {
368 		if (!snum)
369 			snum = bp->port;
370 		else if (snum != bp->port) {
371 			SCTP_DEBUG_PRINTK("sctp_do_bind:"
372 				  " New port %d does not match existing port "
373 				  "%d.\n", snum, bp->port);
374 			return -EINVAL;
375 		}
376 	}
377 
378 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
379 		return -EACCES;
380 
381 	/* See if the address matches any of the addresses we may have
382 	 * already bound before checking against other endpoints.
383 	 */
384 	if (sctp_bind_addr_match(bp, addr, sp))
385 		return -EINVAL;
386 
387 	/* Make sure we are allowed to bind here.
388 	 * The function sctp_get_port_local() does duplicate address
389 	 * detection.
390 	 */
391 	addr->v4.sin_port = htons(snum);
392 	if ((ret = sctp_get_port_local(sk, addr))) {
393 		return -EADDRINUSE;
394 	}
395 
396 	/* Refresh ephemeral port.  */
397 	if (!bp->port)
398 		bp->port = inet_sk(sk)->inet_num;
399 
400 	/* Add the address to the bind address list.
401 	 * Use GFP_ATOMIC since BHs will be disabled.
402 	 */
403 	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
404 
405 	/* Copy back into socket for getsockname() use. */
406 	if (!ret) {
407 		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
408 		af->to_sk_saddr(addr, sk);
409 	}
410 
411 	return ret;
412 }
413 
414  /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
415  *
416  * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
417  * at any one time.  If a sender, after sending an ASCONF chunk, decides
418  * it needs to transfer another ASCONF Chunk, it MUST wait until the
419  * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
420  * subsequent ASCONF. Note this restriction binds each side, so at any
421  * time two ASCONF may be in-transit on any given association (one sent
422  * from each endpoint).
423  */
424 static int sctp_send_asconf(struct sctp_association *asoc,
425 			    struct sctp_chunk *chunk)
426 {
427 	int		retval = 0;
428 
429 	/* If there is an outstanding ASCONF chunk, queue it for later
430 	 * transmission.
431 	 */
432 	if (asoc->addip_last_asconf) {
433 		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
434 		goto out;
435 	}
436 
437 	/* Hold the chunk until an ASCONF_ACK is received. */
438 	sctp_chunk_hold(chunk);
439 	retval = sctp_primitive_ASCONF(asoc, chunk);
440 	if (retval)
441 		sctp_chunk_free(chunk);
442 	else
443 		asoc->addip_last_asconf = chunk;
444 
445 out:
446 	return retval;
447 }
448 
449 /* Add a list of addresses as bind addresses to local endpoint or
450  * association.
451  *
452  * Basically run through each address specified in the addrs/addrcnt
453  * array/length pair, determine if it is IPv6 or IPv4 and call
454  * sctp_do_bind() on it.
455  *
456  * If any of them fails, then the operation will be reversed and the
457  * ones that were added will be removed.
458  *
459  * Only sctp_setsockopt_bindx() is supposed to call this function.
460  */
461 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
462 {
463 	int cnt;
464 	int retval = 0;
465 	void *addr_buf;
466 	struct sockaddr *sa_addr;
467 	struct sctp_af *af;
468 
469 	SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
470 			  sk, addrs, addrcnt);
471 
472 	addr_buf = addrs;
473 	for (cnt = 0; cnt < addrcnt; cnt++) {
474 		/* The list may contain either IPv4 or IPv6 address;
475 		 * determine the address length for walking thru the list.
476 		 */
477 		sa_addr = (struct sockaddr *)addr_buf;
478 		af = sctp_get_af_specific(sa_addr->sa_family);
479 		if (!af) {
480 			retval = -EINVAL;
481 			goto err_bindx_add;
482 		}
483 
484 		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
485 				      af->sockaddr_len);
486 
487 		addr_buf += af->sockaddr_len;
488 
489 err_bindx_add:
490 		if (retval < 0) {
491 			/* Failed. Cleanup the ones that have been added */
492 			if (cnt > 0)
493 				sctp_bindx_rem(sk, addrs, cnt);
494 			return retval;
495 		}
496 	}
497 
498 	return retval;
499 }
500 
501 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
502  * associations that are part of the endpoint indicating that a list of local
503  * addresses are added to the endpoint.
504  *
505  * If any of the addresses is already in the bind address list of the
506  * association, we do not send the chunk for that association.  But it will not
507  * affect other associations.
508  *
509  * Only sctp_setsockopt_bindx() is supposed to call this function.
510  */
511 static int sctp_send_asconf_add_ip(struct sock		*sk,
512 				   struct sockaddr	*addrs,
513 				   int 			addrcnt)
514 {
515 	struct sctp_sock		*sp;
516 	struct sctp_endpoint		*ep;
517 	struct sctp_association		*asoc;
518 	struct sctp_bind_addr		*bp;
519 	struct sctp_chunk		*chunk;
520 	struct sctp_sockaddr_entry	*laddr;
521 	union sctp_addr			*addr;
522 	union sctp_addr			saveaddr;
523 	void				*addr_buf;
524 	struct sctp_af			*af;
525 	struct list_head		*p;
526 	int 				i;
527 	int 				retval = 0;
528 
529 	if (!sctp_addip_enable)
530 		return retval;
531 
532 	sp = sctp_sk(sk);
533 	ep = sp->ep;
534 
535 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
536 			  __func__, sk, addrs, addrcnt);
537 
538 	list_for_each_entry(asoc, &ep->asocs, asocs) {
539 
540 		if (!asoc->peer.asconf_capable)
541 			continue;
542 
543 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
544 			continue;
545 
546 		if (!sctp_state(asoc, ESTABLISHED))
547 			continue;
548 
549 		/* Check if any address in the packed array of addresses is
550 		 * in the bind address list of the association. If so,
551 		 * do not send the asconf chunk to its peer, but continue with
552 		 * other associations.
553 		 */
554 		addr_buf = addrs;
555 		for (i = 0; i < addrcnt; i++) {
556 			addr = (union sctp_addr *)addr_buf;
557 			af = sctp_get_af_specific(addr->v4.sin_family);
558 			if (!af) {
559 				retval = -EINVAL;
560 				goto out;
561 			}
562 
563 			if (sctp_assoc_lookup_laddr(asoc, addr))
564 				break;
565 
566 			addr_buf += af->sockaddr_len;
567 		}
568 		if (i < addrcnt)
569 			continue;
570 
571 		/* Use the first valid address in bind addr list of
572 		 * association as Address Parameter of ASCONF CHUNK.
573 		 */
574 		bp = &asoc->base.bind_addr;
575 		p = bp->address_list.next;
576 		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
577 		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
578 						   addrcnt, SCTP_PARAM_ADD_IP);
579 		if (!chunk) {
580 			retval = -ENOMEM;
581 			goto out;
582 		}
583 
584 		retval = sctp_send_asconf(asoc, chunk);
585 		if (retval)
586 			goto out;
587 
588 		/* Add the new addresses to the bind address list with
589 		 * use_as_src set to 0.
590 		 */
591 		addr_buf = addrs;
592 		for (i = 0; i < addrcnt; i++) {
593 			addr = (union sctp_addr *)addr_buf;
594 			af = sctp_get_af_specific(addr->v4.sin_family);
595 			memcpy(&saveaddr, addr, af->sockaddr_len);
596 			retval = sctp_add_bind_addr(bp, &saveaddr,
597 						    SCTP_ADDR_NEW, GFP_ATOMIC);
598 			addr_buf += af->sockaddr_len;
599 		}
600 	}
601 
602 out:
603 	return retval;
604 }
605 
606 /* Remove a list of addresses from bind addresses list.  Do not remove the
607  * last address.
608  *
609  * Basically run through each address specified in the addrs/addrcnt
610  * array/length pair, determine if it is IPv6 or IPv4 and call
611  * sctp_del_bind() on it.
612  *
613  * If any of them fails, then the operation will be reversed and the
614  * ones that were removed will be added back.
615  *
616  * At least one address has to be left; if only one address is
617  * available, the operation will return -EBUSY.
618  *
619  * Only sctp_setsockopt_bindx() is supposed to call this function.
620  */
621 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
622 {
623 	struct sctp_sock *sp = sctp_sk(sk);
624 	struct sctp_endpoint *ep = sp->ep;
625 	int cnt;
626 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
627 	int retval = 0;
628 	void *addr_buf;
629 	union sctp_addr *sa_addr;
630 	struct sctp_af *af;
631 
632 	SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
633 			  sk, addrs, addrcnt);
634 
635 	addr_buf = addrs;
636 	for (cnt = 0; cnt < addrcnt; cnt++) {
637 		/* If the bind address list is empty or if there is only one
638 		 * bind address, there is nothing more to be removed (we need
639 		 * at least one address here).
640 		 */
641 		if (list_empty(&bp->address_list) ||
642 		    (sctp_list_single_entry(&bp->address_list))) {
643 			retval = -EBUSY;
644 			goto err_bindx_rem;
645 		}
646 
647 		sa_addr = (union sctp_addr *)addr_buf;
648 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
649 		if (!af) {
650 			retval = -EINVAL;
651 			goto err_bindx_rem;
652 		}
653 
654 		if (!af->addr_valid(sa_addr, sp, NULL)) {
655 			retval = -EADDRNOTAVAIL;
656 			goto err_bindx_rem;
657 		}
658 
659 		if (sa_addr->v4.sin_port != htons(bp->port)) {
660 			retval = -EINVAL;
661 			goto err_bindx_rem;
662 		}
663 
664 		/* FIXME - There is probably a need to check if sk->sk_saddr and
665 		 * sk->sk_rcv_addr are currently set to one of the addresses to
666 		 * be removed. This is something which needs to be looked into
667 		 * when we are fixing the outstanding issues with multi-homing
668 		 * socket routing and failover schemes. Refer to comments in
669 		 * sctp_do_bind(). -daisy
670 		 */
671 		retval = sctp_del_bind_addr(bp, sa_addr);
672 
673 		addr_buf += af->sockaddr_len;
674 err_bindx_rem:
675 		if (retval < 0) {
676 			/* Failed. Add the ones that has been removed back */
677 			if (cnt > 0)
678 				sctp_bindx_add(sk, addrs, cnt);
679 			return retval;
680 		}
681 	}
682 
683 	return retval;
684 }
685 
686 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
687  * the associations that are part of the endpoint indicating that a list of
688  * local addresses are removed from the endpoint.
689  *
690  * If any of the addresses is already in the bind address list of the
691  * association, we do not send the chunk for that association.  But it will not
692  * affect other associations.
693  *
694  * Only sctp_setsockopt_bindx() is supposed to call this function.
695  */
696 static int sctp_send_asconf_del_ip(struct sock		*sk,
697 				   struct sockaddr	*addrs,
698 				   int			addrcnt)
699 {
700 	struct sctp_sock	*sp;
701 	struct sctp_endpoint	*ep;
702 	struct sctp_association	*asoc;
703 	struct sctp_transport	*transport;
704 	struct sctp_bind_addr	*bp;
705 	struct sctp_chunk	*chunk;
706 	union sctp_addr		*laddr;
707 	void			*addr_buf;
708 	struct sctp_af		*af;
709 	struct sctp_sockaddr_entry *saddr;
710 	int 			i;
711 	int 			retval = 0;
712 
713 	if (!sctp_addip_enable)
714 		return retval;
715 
716 	sp = sctp_sk(sk);
717 	ep = sp->ep;
718 
719 	SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
720 			  __func__, sk, addrs, addrcnt);
721 
722 	list_for_each_entry(asoc, &ep->asocs, asocs) {
723 
724 		if (!asoc->peer.asconf_capable)
725 			continue;
726 
727 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
728 			continue;
729 
730 		if (!sctp_state(asoc, ESTABLISHED))
731 			continue;
732 
733 		/* Check if any address in the packed array of addresses is
734 		 * not present in the bind address list of the association.
735 		 * If so, do not send the asconf chunk to its peer, but
736 		 * continue with other associations.
737 		 */
738 		addr_buf = addrs;
739 		for (i = 0; i < addrcnt; i++) {
740 			laddr = (union sctp_addr *)addr_buf;
741 			af = sctp_get_af_specific(laddr->v4.sin_family);
742 			if (!af) {
743 				retval = -EINVAL;
744 				goto out;
745 			}
746 
747 			if (!sctp_assoc_lookup_laddr(asoc, laddr))
748 				break;
749 
750 			addr_buf += af->sockaddr_len;
751 		}
752 		if (i < addrcnt)
753 			continue;
754 
755 		/* Find one address in the association's bind address list
756 		 * that is not in the packed array of addresses. This is to
757 		 * make sure that we do not delete all the addresses in the
758 		 * association.
759 		 */
760 		bp = &asoc->base.bind_addr;
761 		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
762 					       addrcnt, sp);
763 		if (!laddr)
764 			continue;
765 
766 		/* We do not need RCU protection throughout this loop
767 		 * because this is done under a socket lock from the
768 		 * setsockopt call.
769 		 */
770 		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
771 						   SCTP_PARAM_DEL_IP);
772 		if (!chunk) {
773 			retval = -ENOMEM;
774 			goto out;
775 		}
776 
777 		/* Reset use_as_src flag for the addresses in the bind address
778 		 * list that are to be deleted.
779 		 */
780 		addr_buf = addrs;
781 		for (i = 0; i < addrcnt; i++) {
782 			laddr = (union sctp_addr *)addr_buf;
783 			af = sctp_get_af_specific(laddr->v4.sin_family);
784 			list_for_each_entry(saddr, &bp->address_list, list) {
785 				if (sctp_cmp_addr_exact(&saddr->a, laddr))
786 					saddr->state = SCTP_ADDR_DEL;
787 			}
788 			addr_buf += af->sockaddr_len;
789 		}
790 
791 		/* Update the route and saddr entries for all the transports
792 		 * as some of the addresses in the bind address list are
793 		 * about to be deleted and cannot be used as source addresses.
794 		 */
795 		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
796 					transports) {
797 			dst_release(transport->dst);
798 			sctp_transport_route(transport, NULL,
799 					     sctp_sk(asoc->base.sk));
800 		}
801 
802 		retval = sctp_send_asconf(asoc, chunk);
803 	}
804 out:
805 	return retval;
806 }
807 
808 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809  *
810  * API 8.1
811  * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
812  *                int flags);
813  *
814  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
815  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
816  * or IPv6 addresses.
817  *
818  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
819  * Section 3.1.2 for this usage.
820  *
821  * addrs is a pointer to an array of one or more socket addresses. Each
822  * address is contained in its appropriate structure (i.e. struct
823  * sockaddr_in or struct sockaddr_in6) the family of the address type
824  * must be used to distinguish the address length (note that this
825  * representation is termed a "packed array" of addresses). The caller
826  * specifies the number of addresses in the array with addrcnt.
827  *
828  * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
829  * -1, and sets errno to the appropriate error code.
830  *
831  * For SCTP, the port given in each socket address must be the same, or
832  * sctp_bindx() will fail, setting errno to EINVAL.
833  *
834  * The flags parameter is formed from the bitwise OR of zero or more of
835  * the following currently defined flags:
836  *
837  * SCTP_BINDX_ADD_ADDR
838  *
839  * SCTP_BINDX_REM_ADDR
840  *
841  * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
842  * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
843  * addresses from the association. The two flags are mutually exclusive;
844  * if both are given, sctp_bindx() will fail with EINVAL. A caller may
845  * not remove all addresses from an association; sctp_bindx() will
846  * reject such an attempt with EINVAL.
847  *
848  * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
849  * additional addresses with an endpoint after calling bind().  Or use
850  * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
851  * socket is associated with so that no new association accepted will be
852  * associated with those addresses. If the endpoint supports dynamic
853  * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
854  * endpoint to send the appropriate message to the peer to change the
855  * peers address lists.
856  *
857  * Adding and removing addresses from a connected association is
858  * optional functionality. Implementations that do not support this
859  * functionality should return EOPNOTSUPP.
860  *
861  * Basically do nothing but copying the addresses from user to kernel
862  * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
863  * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
864  * from userspace.
865  *
866  * We don't use copy_from_user() for optimization: we first do the
867  * sanity checks (buffer size -fast- and access check-healthy
868  * pointer); if all of those succeed, then we can alloc the memory
869  * (expensive operation) needed to copy the data to kernel. Then we do
870  * the copying without checking the user space area
871  * (__copy_from_user()).
872  *
873  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
874  * it.
875  *
876  * sk        The sk of the socket
877  * addrs     The pointer to the addresses in user land
878  * addrssize Size of the addrs buffer
879  * op        Operation to perform (add or remove, see the flags of
880  *           sctp_bindx)
881  *
882  * Returns 0 if ok, <0 errno code on error.
883  */
884 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
885 				      struct sockaddr __user *addrs,
886 				      int addrs_size, int op)
887 {
888 	struct sockaddr *kaddrs;
889 	int err;
890 	int addrcnt = 0;
891 	int walk_size = 0;
892 	struct sockaddr *sa_addr;
893 	void *addr_buf;
894 	struct sctp_af *af;
895 
896 	SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
897 			  " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
898 
899 	if (unlikely(addrs_size <= 0))
900 		return -EINVAL;
901 
902 	/* Check the user passed a healthy pointer.  */
903 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
904 		return -EFAULT;
905 
906 	/* Alloc space for the address array in kernel memory.  */
907 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
908 	if (unlikely(!kaddrs))
909 		return -ENOMEM;
910 
911 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
912 		kfree(kaddrs);
913 		return -EFAULT;
914 	}
915 
916 	/* Walk through the addrs buffer and count the number of addresses. */
917 	addr_buf = kaddrs;
918 	while (walk_size < addrs_size) {
919 		sa_addr = (struct sockaddr *)addr_buf;
920 		af = sctp_get_af_specific(sa_addr->sa_family);
921 
922 		/* If the address family is not supported or if this address
923 		 * causes the address buffer to overflow return EINVAL.
924 		 */
925 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
926 			kfree(kaddrs);
927 			return -EINVAL;
928 		}
929 		addrcnt++;
930 		addr_buf += af->sockaddr_len;
931 		walk_size += af->sockaddr_len;
932 	}
933 
934 	/* Do the work. */
935 	switch (op) {
936 	case SCTP_BINDX_ADD_ADDR:
937 		err = sctp_bindx_add(sk, kaddrs, addrcnt);
938 		if (err)
939 			goto out;
940 		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
941 		break;
942 
943 	case SCTP_BINDX_REM_ADDR:
944 		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
945 		if (err)
946 			goto out;
947 		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
948 		break;
949 
950 	default:
951 		err = -EINVAL;
952 		break;
953 	}
954 
955 out:
956 	kfree(kaddrs);
957 
958 	return err;
959 }
960 
961 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962  *
963  * Common routine for handling connect() and sctp_connectx().
964  * Connect will come in with just a single address.
965  */
966 static int __sctp_connect(struct sock* sk,
967 			  struct sockaddr *kaddrs,
968 			  int addrs_size,
969 			  sctp_assoc_t *assoc_id)
970 {
971 	struct sctp_sock *sp;
972 	struct sctp_endpoint *ep;
973 	struct sctp_association *asoc = NULL;
974 	struct sctp_association *asoc2;
975 	struct sctp_transport *transport;
976 	union sctp_addr to;
977 	struct sctp_af *af;
978 	sctp_scope_t scope;
979 	long timeo;
980 	int err = 0;
981 	int addrcnt = 0;
982 	int walk_size = 0;
983 	union sctp_addr *sa_addr = NULL;
984 	void *addr_buf;
985 	unsigned short port;
986 	unsigned int f_flags = 0;
987 
988 	sp = sctp_sk(sk);
989 	ep = sp->ep;
990 
991 	/* connect() cannot be done on a socket that is already in ESTABLISHED
992 	 * state - UDP-style peeled off socket or a TCP-style socket that
993 	 * is already connected.
994 	 * It cannot be done even on a TCP-style listening socket.
995 	 */
996 	if (sctp_sstate(sk, ESTABLISHED) ||
997 	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
998 		err = -EISCONN;
999 		goto out_free;
1000 	}
1001 
1002 	/* Walk through the addrs buffer and count the number of addresses. */
1003 	addr_buf = kaddrs;
1004 	while (walk_size < addrs_size) {
1005 		sa_addr = (union sctp_addr *)addr_buf;
1006 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1007 		port = ntohs(sa_addr->v4.sin_port);
1008 
1009 		/* If the address family is not supported or if this address
1010 		 * causes the address buffer to overflow return EINVAL.
1011 		 */
1012 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1013 			err = -EINVAL;
1014 			goto out_free;
1015 		}
1016 
1017 		/* Save current address so we can work with it */
1018 		memcpy(&to, sa_addr, af->sockaddr_len);
1019 
1020 		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1021 		if (err)
1022 			goto out_free;
1023 
1024 		/* Make sure the destination port is correctly set
1025 		 * in all addresses.
1026 		 */
1027 		if (asoc && asoc->peer.port && asoc->peer.port != port)
1028 			goto out_free;
1029 
1030 
1031 		/* Check if there already is a matching association on the
1032 		 * endpoint (other than the one created here).
1033 		 */
1034 		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1035 		if (asoc2 && asoc2 != asoc) {
1036 			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1037 				err = -EISCONN;
1038 			else
1039 				err = -EALREADY;
1040 			goto out_free;
1041 		}
1042 
1043 		/* If we could not find a matching association on the endpoint,
1044 		 * make sure that there is no peeled-off association matching
1045 		 * the peer address even on another socket.
1046 		 */
1047 		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1048 			err = -EADDRNOTAVAIL;
1049 			goto out_free;
1050 		}
1051 
1052 		if (!asoc) {
1053 			/* If a bind() or sctp_bindx() is not called prior to
1054 			 * an sctp_connectx() call, the system picks an
1055 			 * ephemeral port and will choose an address set
1056 			 * equivalent to binding with a wildcard address.
1057 			 */
1058 			if (!ep->base.bind_addr.port) {
1059 				if (sctp_autobind(sk)) {
1060 					err = -EAGAIN;
1061 					goto out_free;
1062 				}
1063 			} else {
1064 				/*
1065 				 * If an unprivileged user inherits a 1-many
1066 				 * style socket with open associations on a
1067 				 * privileged port, it MAY be permitted to
1068 				 * accept new associations, but it SHOULD NOT
1069 				 * be permitted to open new associations.
1070 				 */
1071 				if (ep->base.bind_addr.port < PROT_SOCK &&
1072 				    !capable(CAP_NET_BIND_SERVICE)) {
1073 					err = -EACCES;
1074 					goto out_free;
1075 				}
1076 			}
1077 
1078 			scope = sctp_scope(&to);
1079 			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1080 			if (!asoc) {
1081 				err = -ENOMEM;
1082 				goto out_free;
1083 			}
1084 
1085 			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1086 							      GFP_KERNEL);
1087 			if (err < 0) {
1088 				goto out_free;
1089 			}
1090 
1091 		}
1092 
1093 		/* Prime the peer's transport structures.  */
1094 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1095 						SCTP_UNKNOWN);
1096 		if (!transport) {
1097 			err = -ENOMEM;
1098 			goto out_free;
1099 		}
1100 
1101 		addrcnt++;
1102 		addr_buf += af->sockaddr_len;
1103 		walk_size += af->sockaddr_len;
1104 	}
1105 
1106 	/* In case the user of sctp_connectx() wants an association
1107 	 * id back, assign one now.
1108 	 */
1109 	if (assoc_id) {
1110 		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1111 		if (err < 0)
1112 			goto out_free;
1113 	}
1114 
1115 	err = sctp_primitive_ASSOCIATE(asoc, NULL);
1116 	if (err < 0) {
1117 		goto out_free;
1118 	}
1119 
1120 	/* Initialize sk's dport and daddr for getpeername() */
1121 	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1122 	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1123 	af->to_sk_daddr(sa_addr, sk);
1124 	sk->sk_err = 0;
1125 
1126 	/* in-kernel sockets don't generally have a file allocated to them
1127 	 * if all they do is call sock_create_kern().
1128 	 */
1129 	if (sk->sk_socket->file)
1130 		f_flags = sk->sk_socket->file->f_flags;
1131 
1132 	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1133 
1134 	err = sctp_wait_for_connect(asoc, &timeo);
1135 	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1136 		*assoc_id = asoc->assoc_id;
1137 
1138 	/* Don't free association on exit. */
1139 	asoc = NULL;
1140 
1141 out_free:
1142 
1143 	SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1144 			  " kaddrs: %p err: %d\n",
1145 			  asoc, kaddrs, err);
1146 	if (asoc)
1147 		sctp_association_free(asoc);
1148 	return err;
1149 }
1150 
1151 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1152  *
1153  * API 8.9
1154  * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1155  * 			sctp_assoc_t *asoc);
1156  *
1157  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1158  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1159  * or IPv6 addresses.
1160  *
1161  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1162  * Section 3.1.2 for this usage.
1163  *
1164  * addrs is a pointer to an array of one or more socket addresses. Each
1165  * address is contained in its appropriate structure (i.e. struct
1166  * sockaddr_in or struct sockaddr_in6) the family of the address type
1167  * must be used to distengish the address length (note that this
1168  * representation is termed a "packed array" of addresses). The caller
1169  * specifies the number of addresses in the array with addrcnt.
1170  *
1171  * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1172  * the association id of the new association.  On failure, sctp_connectx()
1173  * returns -1, and sets errno to the appropriate error code.  The assoc_id
1174  * is not touched by the kernel.
1175  *
1176  * For SCTP, the port given in each socket address must be the same, or
1177  * sctp_connectx() will fail, setting errno to EINVAL.
1178  *
1179  * An application can use sctp_connectx to initiate an association with
1180  * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1181  * allows a caller to specify multiple addresses at which a peer can be
1182  * reached.  The way the SCTP stack uses the list of addresses to set up
1183  * the association is implementation dependant.  This function only
1184  * specifies that the stack will try to make use of all the addresses in
1185  * the list when needed.
1186  *
1187  * Note that the list of addresses passed in is only used for setting up
1188  * the association.  It does not necessarily equal the set of addresses
1189  * the peer uses for the resulting association.  If the caller wants to
1190  * find out the set of peer addresses, it must use sctp_getpaddrs() to
1191  * retrieve them after the association has been set up.
1192  *
1193  * Basically do nothing but copying the addresses from user to kernel
1194  * land and invoking either sctp_connectx(). This is used for tunneling
1195  * the sctp_connectx() request through sctp_setsockopt() from userspace.
1196  *
1197  * We don't use copy_from_user() for optimization: we first do the
1198  * sanity checks (buffer size -fast- and access check-healthy
1199  * pointer); if all of those succeed, then we can alloc the memory
1200  * (expensive operation) needed to copy the data to kernel. Then we do
1201  * the copying without checking the user space area
1202  * (__copy_from_user()).
1203  *
1204  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1205  * it.
1206  *
1207  * sk        The sk of the socket
1208  * addrs     The pointer to the addresses in user land
1209  * addrssize Size of the addrs buffer
1210  *
1211  * Returns >=0 if ok, <0 errno code on error.
1212  */
1213 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1214 				      struct sockaddr __user *addrs,
1215 				      int addrs_size,
1216 				      sctp_assoc_t *assoc_id)
1217 {
1218 	int err = 0;
1219 	struct sockaddr *kaddrs;
1220 
1221 	SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1222 			  __func__, sk, addrs, addrs_size);
1223 
1224 	if (unlikely(addrs_size <= 0))
1225 		return -EINVAL;
1226 
1227 	/* Check the user passed a healthy pointer.  */
1228 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1229 		return -EFAULT;
1230 
1231 	/* Alloc space for the address array in kernel memory.  */
1232 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1233 	if (unlikely(!kaddrs))
1234 		return -ENOMEM;
1235 
1236 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1237 		err = -EFAULT;
1238 	} else {
1239 		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1240 	}
1241 
1242 	kfree(kaddrs);
1243 
1244 	return err;
1245 }
1246 
1247 /*
1248  * This is an older interface.  It's kept for backward compatibility
1249  * to the option that doesn't provide association id.
1250  */
1251 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1252 				      struct sockaddr __user *addrs,
1253 				      int addrs_size)
1254 {
1255 	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1256 }
1257 
1258 /*
1259  * New interface for the API.  The since the API is done with a socket
1260  * option, to make it simple we feed back the association id is as a return
1261  * indication to the call.  Error is always negative and association id is
1262  * always positive.
1263  */
1264 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1265 				      struct sockaddr __user *addrs,
1266 				      int addrs_size)
1267 {
1268 	sctp_assoc_t assoc_id = 0;
1269 	int err = 0;
1270 
1271 	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1272 
1273 	if (err)
1274 		return err;
1275 	else
1276 		return assoc_id;
1277 }
1278 
1279 /*
1280  * New (hopefully final) interface for the API.
1281  * We use the sctp_getaddrs_old structure so that use-space library
1282  * can avoid any unnecessary allocations.   The only defferent part
1283  * is that we store the actual length of the address buffer into the
1284  * addrs_num structure member.  That way we can re-use the existing
1285  * code.
1286  */
1287 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1288 					char __user *optval,
1289 					int __user *optlen)
1290 {
1291 	struct sctp_getaddrs_old param;
1292 	sctp_assoc_t assoc_id = 0;
1293 	int err = 0;
1294 
1295 	if (len < sizeof(param))
1296 		return -EINVAL;
1297 
1298 	if (copy_from_user(&param, optval, sizeof(param)))
1299 		return -EFAULT;
1300 
1301 	err = __sctp_setsockopt_connectx(sk,
1302 			(struct sockaddr __user *)param.addrs,
1303 			param.addr_num, &assoc_id);
1304 
1305 	if (err == 0 || err == -EINPROGRESS) {
1306 		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1307 			return -EFAULT;
1308 		if (put_user(sizeof(assoc_id), optlen))
1309 			return -EFAULT;
1310 	}
1311 
1312 	return err;
1313 }
1314 
1315 /* API 3.1.4 close() - UDP Style Syntax
1316  * Applications use close() to perform graceful shutdown (as described in
1317  * Section 10.1 of [SCTP]) on ALL the associations currently represented
1318  * by a UDP-style socket.
1319  *
1320  * The syntax is
1321  *
1322  *   ret = close(int sd);
1323  *
1324  *   sd      - the socket descriptor of the associations to be closed.
1325  *
1326  * To gracefully shutdown a specific association represented by the
1327  * UDP-style socket, an application should use the sendmsg() call,
1328  * passing no user data, but including the appropriate flag in the
1329  * ancillary data (see Section xxxx).
1330  *
1331  * If sd in the close() call is a branched-off socket representing only
1332  * one association, the shutdown is performed on that association only.
1333  *
1334  * 4.1.6 close() - TCP Style Syntax
1335  *
1336  * Applications use close() to gracefully close down an association.
1337  *
1338  * The syntax is:
1339  *
1340  *    int close(int sd);
1341  *
1342  *      sd      - the socket descriptor of the association to be closed.
1343  *
1344  * After an application calls close() on a socket descriptor, no further
1345  * socket operations will succeed on that descriptor.
1346  *
1347  * API 7.1.4 SO_LINGER
1348  *
1349  * An application using the TCP-style socket can use this option to
1350  * perform the SCTP ABORT primitive.  The linger option structure is:
1351  *
1352  *  struct  linger {
1353  *     int     l_onoff;                // option on/off
1354  *     int     l_linger;               // linger time
1355  * };
1356  *
1357  * To enable the option, set l_onoff to 1.  If the l_linger value is set
1358  * to 0, calling close() is the same as the ABORT primitive.  If the
1359  * value is set to a negative value, the setsockopt() call will return
1360  * an error.  If the value is set to a positive value linger_time, the
1361  * close() can be blocked for at most linger_time ms.  If the graceful
1362  * shutdown phase does not finish during this period, close() will
1363  * return but the graceful shutdown phase continues in the system.
1364  */
1365 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1366 {
1367 	struct sctp_endpoint *ep;
1368 	struct sctp_association *asoc;
1369 	struct list_head *pos, *temp;
1370 
1371 	SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1372 
1373 	sctp_lock_sock(sk);
1374 	sk->sk_shutdown = SHUTDOWN_MASK;
1375 	sk->sk_state = SCTP_SS_CLOSING;
1376 
1377 	ep = sctp_sk(sk)->ep;
1378 
1379 	/* Walk all associations on an endpoint.  */
1380 	list_for_each_safe(pos, temp, &ep->asocs) {
1381 		asoc = list_entry(pos, struct sctp_association, asocs);
1382 
1383 		if (sctp_style(sk, TCP)) {
1384 			/* A closed association can still be in the list if
1385 			 * it belongs to a TCP-style listening socket that is
1386 			 * not yet accepted. If so, free it. If not, send an
1387 			 * ABORT or SHUTDOWN based on the linger options.
1388 			 */
1389 			if (sctp_state(asoc, CLOSED)) {
1390 				sctp_unhash_established(asoc);
1391 				sctp_association_free(asoc);
1392 				continue;
1393 			}
1394 		}
1395 
1396 		if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1397 			struct sctp_chunk *chunk;
1398 
1399 			chunk = sctp_make_abort_user(asoc, NULL, 0);
1400 			if (chunk)
1401 				sctp_primitive_ABORT(asoc, chunk);
1402 		} else
1403 			sctp_primitive_SHUTDOWN(asoc, NULL);
1404 	}
1405 
1406 	/* Clean up any skbs sitting on the receive queue.  */
1407 	sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1408 	sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1409 
1410 	/* On a TCP-style socket, block for at most linger_time if set. */
1411 	if (sctp_style(sk, TCP) && timeout)
1412 		sctp_wait_for_close(sk, timeout);
1413 
1414 	/* This will run the backlog queue.  */
1415 	sctp_release_sock(sk);
1416 
1417 	/* Supposedly, no process has access to the socket, but
1418 	 * the net layers still may.
1419 	 */
1420 	sctp_local_bh_disable();
1421 	sctp_bh_lock_sock(sk);
1422 
1423 	/* Hold the sock, since sk_common_release() will put sock_put()
1424 	 * and we have just a little more cleanup.
1425 	 */
1426 	sock_hold(sk);
1427 	sk_common_release(sk);
1428 
1429 	sctp_bh_unlock_sock(sk);
1430 	sctp_local_bh_enable();
1431 
1432 	sock_put(sk);
1433 
1434 	SCTP_DBG_OBJCNT_DEC(sock);
1435 }
1436 
1437 /* Handle EPIPE error. */
1438 static int sctp_error(struct sock *sk, int flags, int err)
1439 {
1440 	if (err == -EPIPE)
1441 		err = sock_error(sk) ? : -EPIPE;
1442 	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1443 		send_sig(SIGPIPE, current, 0);
1444 	return err;
1445 }
1446 
1447 /* API 3.1.3 sendmsg() - UDP Style Syntax
1448  *
1449  * An application uses sendmsg() and recvmsg() calls to transmit data to
1450  * and receive data from its peer.
1451  *
1452  *  ssize_t sendmsg(int socket, const struct msghdr *message,
1453  *                  int flags);
1454  *
1455  *  socket  - the socket descriptor of the endpoint.
1456  *  message - pointer to the msghdr structure which contains a single
1457  *            user message and possibly some ancillary data.
1458  *
1459  *            See Section 5 for complete description of the data
1460  *            structures.
1461  *
1462  *  flags   - flags sent or received with the user message, see Section
1463  *            5 for complete description of the flags.
1464  *
1465  * Note:  This function could use a rewrite especially when explicit
1466  * connect support comes in.
1467  */
1468 /* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1469 
1470 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1471 
1472 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1473 			     struct msghdr *msg, size_t msg_len)
1474 {
1475 	struct sctp_sock *sp;
1476 	struct sctp_endpoint *ep;
1477 	struct sctp_association *new_asoc=NULL, *asoc=NULL;
1478 	struct sctp_transport *transport, *chunk_tp;
1479 	struct sctp_chunk *chunk;
1480 	union sctp_addr to;
1481 	struct sockaddr *msg_name = NULL;
1482 	struct sctp_sndrcvinfo default_sinfo = { 0 };
1483 	struct sctp_sndrcvinfo *sinfo;
1484 	struct sctp_initmsg *sinit;
1485 	sctp_assoc_t associd = 0;
1486 	sctp_cmsgs_t cmsgs = { NULL };
1487 	int err;
1488 	sctp_scope_t scope;
1489 	long timeo;
1490 	__u16 sinfo_flags = 0;
1491 	struct sctp_datamsg *datamsg;
1492 	int msg_flags = msg->msg_flags;
1493 
1494 	SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1495 			  sk, msg, msg_len);
1496 
1497 	err = 0;
1498 	sp = sctp_sk(sk);
1499 	ep = sp->ep;
1500 
1501 	SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1502 
1503 	/* We cannot send a message over a TCP-style listening socket. */
1504 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1505 		err = -EPIPE;
1506 		goto out_nounlock;
1507 	}
1508 
1509 	/* Parse out the SCTP CMSGs.  */
1510 	err = sctp_msghdr_parse(msg, &cmsgs);
1511 
1512 	if (err) {
1513 		SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1514 		goto out_nounlock;
1515 	}
1516 
1517 	/* Fetch the destination address for this packet.  This
1518 	 * address only selects the association--it is not necessarily
1519 	 * the address we will send to.
1520 	 * For a peeled-off socket, msg_name is ignored.
1521 	 */
1522 	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1523 		int msg_namelen = msg->msg_namelen;
1524 
1525 		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1526 				       msg_namelen);
1527 		if (err)
1528 			return err;
1529 
1530 		if (msg_namelen > sizeof(to))
1531 			msg_namelen = sizeof(to);
1532 		memcpy(&to, msg->msg_name, msg_namelen);
1533 		msg_name = msg->msg_name;
1534 	}
1535 
1536 	sinfo = cmsgs.info;
1537 	sinit = cmsgs.init;
1538 
1539 	/* Did the user specify SNDRCVINFO?  */
1540 	if (sinfo) {
1541 		sinfo_flags = sinfo->sinfo_flags;
1542 		associd = sinfo->sinfo_assoc_id;
1543 	}
1544 
1545 	SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1546 			  msg_len, sinfo_flags);
1547 
1548 	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1549 	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1550 		err = -EINVAL;
1551 		goto out_nounlock;
1552 	}
1553 
1554 	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1555 	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1556 	 * If SCTP_ABORT is set, the message length could be non zero with
1557 	 * the msg_iov set to the user abort reason.
1558 	 */
1559 	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1560 	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1561 		err = -EINVAL;
1562 		goto out_nounlock;
1563 	}
1564 
1565 	/* If SCTP_ADDR_OVER is set, there must be an address
1566 	 * specified in msg_name.
1567 	 */
1568 	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1569 		err = -EINVAL;
1570 		goto out_nounlock;
1571 	}
1572 
1573 	transport = NULL;
1574 
1575 	SCTP_DEBUG_PRINTK("About to look up association.\n");
1576 
1577 	sctp_lock_sock(sk);
1578 
1579 	/* If a msg_name has been specified, assume this is to be used.  */
1580 	if (msg_name) {
1581 		/* Look for a matching association on the endpoint. */
1582 		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1583 		if (!asoc) {
1584 			/* If we could not find a matching association on the
1585 			 * endpoint, make sure that it is not a TCP-style
1586 			 * socket that already has an association or there is
1587 			 * no peeled-off association on another socket.
1588 			 */
1589 			if ((sctp_style(sk, TCP) &&
1590 			     sctp_sstate(sk, ESTABLISHED)) ||
1591 			    sctp_endpoint_is_peeled_off(ep, &to)) {
1592 				err = -EADDRNOTAVAIL;
1593 				goto out_unlock;
1594 			}
1595 		}
1596 	} else {
1597 		asoc = sctp_id2assoc(sk, associd);
1598 		if (!asoc) {
1599 			err = -EPIPE;
1600 			goto out_unlock;
1601 		}
1602 	}
1603 
1604 	if (asoc) {
1605 		SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1606 
1607 		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1608 		 * socket that has an association in CLOSED state. This can
1609 		 * happen when an accepted socket has an association that is
1610 		 * already CLOSED.
1611 		 */
1612 		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1613 			err = -EPIPE;
1614 			goto out_unlock;
1615 		}
1616 
1617 		if (sinfo_flags & SCTP_EOF) {
1618 			SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1619 					  asoc);
1620 			sctp_primitive_SHUTDOWN(asoc, NULL);
1621 			err = 0;
1622 			goto out_unlock;
1623 		}
1624 		if (sinfo_flags & SCTP_ABORT) {
1625 
1626 			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1627 			if (!chunk) {
1628 				err = -ENOMEM;
1629 				goto out_unlock;
1630 			}
1631 
1632 			SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1633 			sctp_primitive_ABORT(asoc, chunk);
1634 			err = 0;
1635 			goto out_unlock;
1636 		}
1637 	}
1638 
1639 	/* Do we need to create the association?  */
1640 	if (!asoc) {
1641 		SCTP_DEBUG_PRINTK("There is no association yet.\n");
1642 
1643 		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1644 			err = -EINVAL;
1645 			goto out_unlock;
1646 		}
1647 
1648 		/* Check for invalid stream against the stream counts,
1649 		 * either the default or the user specified stream counts.
1650 		 */
1651 		if (sinfo) {
1652 			if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1653 				/* Check against the defaults. */
1654 				if (sinfo->sinfo_stream >=
1655 				    sp->initmsg.sinit_num_ostreams) {
1656 					err = -EINVAL;
1657 					goto out_unlock;
1658 				}
1659 			} else {
1660 				/* Check against the requested.  */
1661 				if (sinfo->sinfo_stream >=
1662 				    sinit->sinit_num_ostreams) {
1663 					err = -EINVAL;
1664 					goto out_unlock;
1665 				}
1666 			}
1667 		}
1668 
1669 		/*
1670 		 * API 3.1.2 bind() - UDP Style Syntax
1671 		 * If a bind() or sctp_bindx() is not called prior to a
1672 		 * sendmsg() call that initiates a new association, the
1673 		 * system picks an ephemeral port and will choose an address
1674 		 * set equivalent to binding with a wildcard address.
1675 		 */
1676 		if (!ep->base.bind_addr.port) {
1677 			if (sctp_autobind(sk)) {
1678 				err = -EAGAIN;
1679 				goto out_unlock;
1680 			}
1681 		} else {
1682 			/*
1683 			 * If an unprivileged user inherits a one-to-many
1684 			 * style socket with open associations on a privileged
1685 			 * port, it MAY be permitted to accept new associations,
1686 			 * but it SHOULD NOT be permitted to open new
1687 			 * associations.
1688 			 */
1689 			if (ep->base.bind_addr.port < PROT_SOCK &&
1690 			    !capable(CAP_NET_BIND_SERVICE)) {
1691 				err = -EACCES;
1692 				goto out_unlock;
1693 			}
1694 		}
1695 
1696 		scope = sctp_scope(&to);
1697 		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1698 		if (!new_asoc) {
1699 			err = -ENOMEM;
1700 			goto out_unlock;
1701 		}
1702 		asoc = new_asoc;
1703 		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1704 		if (err < 0) {
1705 			err = -ENOMEM;
1706 			goto out_free;
1707 		}
1708 
1709 		/* If the SCTP_INIT ancillary data is specified, set all
1710 		 * the association init values accordingly.
1711 		 */
1712 		if (sinit) {
1713 			if (sinit->sinit_num_ostreams) {
1714 				asoc->c.sinit_num_ostreams =
1715 					sinit->sinit_num_ostreams;
1716 			}
1717 			if (sinit->sinit_max_instreams) {
1718 				asoc->c.sinit_max_instreams =
1719 					sinit->sinit_max_instreams;
1720 			}
1721 			if (sinit->sinit_max_attempts) {
1722 				asoc->max_init_attempts
1723 					= sinit->sinit_max_attempts;
1724 			}
1725 			if (sinit->sinit_max_init_timeo) {
1726 				asoc->max_init_timeo =
1727 				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1728 			}
1729 		}
1730 
1731 		/* Prime the peer's transport structures.  */
1732 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1733 		if (!transport) {
1734 			err = -ENOMEM;
1735 			goto out_free;
1736 		}
1737 	}
1738 
1739 	/* ASSERT: we have a valid association at this point.  */
1740 	SCTP_DEBUG_PRINTK("We have a valid association.\n");
1741 
1742 	if (!sinfo) {
1743 		/* If the user didn't specify SNDRCVINFO, make up one with
1744 		 * some defaults.
1745 		 */
1746 		default_sinfo.sinfo_stream = asoc->default_stream;
1747 		default_sinfo.sinfo_flags = asoc->default_flags;
1748 		default_sinfo.sinfo_ppid = asoc->default_ppid;
1749 		default_sinfo.sinfo_context = asoc->default_context;
1750 		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1751 		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1752 		sinfo = &default_sinfo;
1753 	}
1754 
1755 	/* API 7.1.7, the sndbuf size per association bounds the
1756 	 * maximum size of data that can be sent in a single send call.
1757 	 */
1758 	if (msg_len > sk->sk_sndbuf) {
1759 		err = -EMSGSIZE;
1760 		goto out_free;
1761 	}
1762 
1763 	if (asoc->pmtu_pending)
1764 		sctp_assoc_pending_pmtu(asoc);
1765 
1766 	/* If fragmentation is disabled and the message length exceeds the
1767 	 * association fragmentation point, return EMSGSIZE.  The I-D
1768 	 * does not specify what this error is, but this looks like
1769 	 * a great fit.
1770 	 */
1771 	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1772 		err = -EMSGSIZE;
1773 		goto out_free;
1774 	}
1775 
1776 	if (sinfo) {
1777 		/* Check for invalid stream. */
1778 		if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1779 			err = -EINVAL;
1780 			goto out_free;
1781 		}
1782 	}
1783 
1784 	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1785 	if (!sctp_wspace(asoc)) {
1786 		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1787 		if (err)
1788 			goto out_free;
1789 	}
1790 
1791 	/* If an address is passed with the sendto/sendmsg call, it is used
1792 	 * to override the primary destination address in the TCP model, or
1793 	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1794 	 */
1795 	if ((sctp_style(sk, TCP) && msg_name) ||
1796 	    (sinfo_flags & SCTP_ADDR_OVER)) {
1797 		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1798 		if (!chunk_tp) {
1799 			err = -EINVAL;
1800 			goto out_free;
1801 		}
1802 	} else
1803 		chunk_tp = NULL;
1804 
1805 	/* Auto-connect, if we aren't connected already. */
1806 	if (sctp_state(asoc, CLOSED)) {
1807 		err = sctp_primitive_ASSOCIATE(asoc, NULL);
1808 		if (err < 0)
1809 			goto out_free;
1810 		SCTP_DEBUG_PRINTK("We associated primitively.\n");
1811 	}
1812 
1813 	/* Break the message into multiple chunks of maximum size. */
1814 	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1815 	if (!datamsg) {
1816 		err = -ENOMEM;
1817 		goto out_free;
1818 	}
1819 
1820 	/* Now send the (possibly) fragmented message. */
1821 	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1822 		sctp_chunk_hold(chunk);
1823 
1824 		/* Do accounting for the write space.  */
1825 		sctp_set_owner_w(chunk);
1826 
1827 		chunk->transport = chunk_tp;
1828 	}
1829 
1830 	/* Send it to the lower layers.  Note:  all chunks
1831 	 * must either fail or succeed.   The lower layer
1832 	 * works that way today.  Keep it that way or this
1833 	 * breaks.
1834 	 */
1835 	err = sctp_primitive_SEND(asoc, datamsg);
1836 	/* Did the lower layer accept the chunk? */
1837 	if (err)
1838 		sctp_datamsg_free(datamsg);
1839 	else
1840 		sctp_datamsg_put(datamsg);
1841 
1842 	SCTP_DEBUG_PRINTK("We sent primitively.\n");
1843 
1844 	if (err)
1845 		goto out_free;
1846 	else
1847 		err = msg_len;
1848 
1849 	/* If we are already past ASSOCIATE, the lower
1850 	 * layers are responsible for association cleanup.
1851 	 */
1852 	goto out_unlock;
1853 
1854 out_free:
1855 	if (new_asoc)
1856 		sctp_association_free(asoc);
1857 out_unlock:
1858 	sctp_release_sock(sk);
1859 
1860 out_nounlock:
1861 	return sctp_error(sk, msg_flags, err);
1862 
1863 #if 0
1864 do_sock_err:
1865 	if (msg_len)
1866 		err = msg_len;
1867 	else
1868 		err = sock_error(sk);
1869 	goto out;
1870 
1871 do_interrupted:
1872 	if (msg_len)
1873 		err = msg_len;
1874 	goto out;
1875 #endif /* 0 */
1876 }
1877 
1878 /* This is an extended version of skb_pull() that removes the data from the
1879  * start of a skb even when data is spread across the list of skb's in the
1880  * frag_list. len specifies the total amount of data that needs to be removed.
1881  * when 'len' bytes could be removed from the skb, it returns 0.
1882  * If 'len' exceeds the total skb length,  it returns the no. of bytes that
1883  * could not be removed.
1884  */
1885 static int sctp_skb_pull(struct sk_buff *skb, int len)
1886 {
1887 	struct sk_buff *list;
1888 	int skb_len = skb_headlen(skb);
1889 	int rlen;
1890 
1891 	if (len <= skb_len) {
1892 		__skb_pull(skb, len);
1893 		return 0;
1894 	}
1895 	len -= skb_len;
1896 	__skb_pull(skb, skb_len);
1897 
1898 	skb_walk_frags(skb, list) {
1899 		rlen = sctp_skb_pull(list, len);
1900 		skb->len -= (len-rlen);
1901 		skb->data_len -= (len-rlen);
1902 
1903 		if (!rlen)
1904 			return 0;
1905 
1906 		len = rlen;
1907 	}
1908 
1909 	return len;
1910 }
1911 
1912 /* API 3.1.3  recvmsg() - UDP Style Syntax
1913  *
1914  *  ssize_t recvmsg(int socket, struct msghdr *message,
1915  *                    int flags);
1916  *
1917  *  socket  - the socket descriptor of the endpoint.
1918  *  message - pointer to the msghdr structure which contains a single
1919  *            user message and possibly some ancillary data.
1920  *
1921  *            See Section 5 for complete description of the data
1922  *            structures.
1923  *
1924  *  flags   - flags sent or received with the user message, see Section
1925  *            5 for complete description of the flags.
1926  */
1927 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1928 
1929 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1930 			     struct msghdr *msg, size_t len, int noblock,
1931 			     int flags, int *addr_len)
1932 {
1933 	struct sctp_ulpevent *event = NULL;
1934 	struct sctp_sock *sp = sctp_sk(sk);
1935 	struct sk_buff *skb;
1936 	int copied;
1937 	int err = 0;
1938 	int skb_len;
1939 
1940 	SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1941 			  "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1942 			  "len", len, "knoblauch", noblock,
1943 			  "flags", flags, "addr_len", addr_len);
1944 
1945 	sctp_lock_sock(sk);
1946 
1947 	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1948 		err = -ENOTCONN;
1949 		goto out;
1950 	}
1951 
1952 	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1953 	if (!skb)
1954 		goto out;
1955 
1956 	/* Get the total length of the skb including any skb's in the
1957 	 * frag_list.
1958 	 */
1959 	skb_len = skb->len;
1960 
1961 	copied = skb_len;
1962 	if (copied > len)
1963 		copied = len;
1964 
1965 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1966 
1967 	event = sctp_skb2event(skb);
1968 
1969 	if (err)
1970 		goto out_free;
1971 
1972 	sock_recv_ts_and_drops(msg, sk, skb);
1973 	if (sctp_ulpevent_is_notification(event)) {
1974 		msg->msg_flags |= MSG_NOTIFICATION;
1975 		sp->pf->event_msgname(event, msg->msg_name, addr_len);
1976 	} else {
1977 		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1978 	}
1979 
1980 	/* Check if we allow SCTP_SNDRCVINFO. */
1981 	if (sp->subscribe.sctp_data_io_event)
1982 		sctp_ulpevent_read_sndrcvinfo(event, msg);
1983 #if 0
1984 	/* FIXME: we should be calling IP/IPv6 layers.  */
1985 	if (sk->sk_protinfo.af_inet.cmsg_flags)
1986 		ip_cmsg_recv(msg, skb);
1987 #endif
1988 
1989 	err = copied;
1990 
1991 	/* If skb's length exceeds the user's buffer, update the skb and
1992 	 * push it back to the receive_queue so that the next call to
1993 	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1994 	 */
1995 	if (skb_len > copied) {
1996 		msg->msg_flags &= ~MSG_EOR;
1997 		if (flags & MSG_PEEK)
1998 			goto out_free;
1999 		sctp_skb_pull(skb, copied);
2000 		skb_queue_head(&sk->sk_receive_queue, skb);
2001 
2002 		/* When only partial message is copied to the user, increase
2003 		 * rwnd by that amount. If all the data in the skb is read,
2004 		 * rwnd is updated when the event is freed.
2005 		 */
2006 		if (!sctp_ulpevent_is_notification(event))
2007 			sctp_assoc_rwnd_increase(event->asoc, copied);
2008 		goto out;
2009 	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2010 		   (event->msg_flags & MSG_EOR))
2011 		msg->msg_flags |= MSG_EOR;
2012 	else
2013 		msg->msg_flags &= ~MSG_EOR;
2014 
2015 out_free:
2016 	if (flags & MSG_PEEK) {
2017 		/* Release the skb reference acquired after peeking the skb in
2018 		 * sctp_skb_recv_datagram().
2019 		 */
2020 		kfree_skb(skb);
2021 	} else {
2022 		/* Free the event which includes releasing the reference to
2023 		 * the owner of the skb, freeing the skb and updating the
2024 		 * rwnd.
2025 		 */
2026 		sctp_ulpevent_free(event);
2027 	}
2028 out:
2029 	sctp_release_sock(sk);
2030 	return err;
2031 }
2032 
2033 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2034  *
2035  * This option is a on/off flag.  If enabled no SCTP message
2036  * fragmentation will be performed.  Instead if a message being sent
2037  * exceeds the current PMTU size, the message will NOT be sent and
2038  * instead a error will be indicated to the user.
2039  */
2040 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2041 					     char __user *optval,
2042 					     unsigned int optlen)
2043 {
2044 	int val;
2045 
2046 	if (optlen < sizeof(int))
2047 		return -EINVAL;
2048 
2049 	if (get_user(val, (int __user *)optval))
2050 		return -EFAULT;
2051 
2052 	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2053 
2054 	return 0;
2055 }
2056 
2057 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2058 				  unsigned int optlen)
2059 {
2060 	if (optlen > sizeof(struct sctp_event_subscribe))
2061 		return -EINVAL;
2062 	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2063 		return -EFAULT;
2064 	return 0;
2065 }
2066 
2067 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2068  *
2069  * This socket option is applicable to the UDP-style socket only.  When
2070  * set it will cause associations that are idle for more than the
2071  * specified number of seconds to automatically close.  An association
2072  * being idle is defined an association that has NOT sent or received
2073  * user data.  The special value of '0' indicates that no automatic
2074  * close of any associations should be performed.  The option expects an
2075  * integer defining the number of seconds of idle time before an
2076  * association is closed.
2077  */
2078 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2079 				     unsigned int optlen)
2080 {
2081 	struct sctp_sock *sp = sctp_sk(sk);
2082 
2083 	/* Applicable to UDP-style socket only */
2084 	if (sctp_style(sk, TCP))
2085 		return -EOPNOTSUPP;
2086 	if (optlen != sizeof(int))
2087 		return -EINVAL;
2088 	if (copy_from_user(&sp->autoclose, optval, optlen))
2089 		return -EFAULT;
2090 	/* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2091 	sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2092 
2093 	return 0;
2094 }
2095 
2096 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2097  *
2098  * Applications can enable or disable heartbeats for any peer address of
2099  * an association, modify an address's heartbeat interval, force a
2100  * heartbeat to be sent immediately, and adjust the address's maximum
2101  * number of retransmissions sent before an address is considered
2102  * unreachable.  The following structure is used to access and modify an
2103  * address's parameters:
2104  *
2105  *  struct sctp_paddrparams {
2106  *     sctp_assoc_t            spp_assoc_id;
2107  *     struct sockaddr_storage spp_address;
2108  *     uint32_t                spp_hbinterval;
2109  *     uint16_t                spp_pathmaxrxt;
2110  *     uint32_t                spp_pathmtu;
2111  *     uint32_t                spp_sackdelay;
2112  *     uint32_t                spp_flags;
2113  * };
2114  *
2115  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2116  *                     application, and identifies the association for
2117  *                     this query.
2118  *   spp_address     - This specifies which address is of interest.
2119  *   spp_hbinterval  - This contains the value of the heartbeat interval,
2120  *                     in milliseconds.  If a  value of zero
2121  *                     is present in this field then no changes are to
2122  *                     be made to this parameter.
2123  *   spp_pathmaxrxt  - This contains the maximum number of
2124  *                     retransmissions before this address shall be
2125  *                     considered unreachable. If a  value of zero
2126  *                     is present in this field then no changes are to
2127  *                     be made to this parameter.
2128  *   spp_pathmtu     - When Path MTU discovery is disabled the value
2129  *                     specified here will be the "fixed" path mtu.
2130  *                     Note that if the spp_address field is empty
2131  *                     then all associations on this address will
2132  *                     have this fixed path mtu set upon them.
2133  *
2134  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2135  *                     the number of milliseconds that sacks will be delayed
2136  *                     for. This value will apply to all addresses of an
2137  *                     association if the spp_address field is empty. Note
2138  *                     also, that if delayed sack is enabled and this
2139  *                     value is set to 0, no change is made to the last
2140  *                     recorded delayed sack timer value.
2141  *
2142  *   spp_flags       - These flags are used to control various features
2143  *                     on an association. The flag field may contain
2144  *                     zero or more of the following options.
2145  *
2146  *                     SPP_HB_ENABLE  - Enable heartbeats on the
2147  *                     specified address. Note that if the address
2148  *                     field is empty all addresses for the association
2149  *                     have heartbeats enabled upon them.
2150  *
2151  *                     SPP_HB_DISABLE - Disable heartbeats on the
2152  *                     speicifed address. Note that if the address
2153  *                     field is empty all addresses for the association
2154  *                     will have their heartbeats disabled. Note also
2155  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2156  *                     mutually exclusive, only one of these two should
2157  *                     be specified. Enabling both fields will have
2158  *                     undetermined results.
2159  *
2160  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2161  *                     to be made immediately.
2162  *
2163  *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2164  *                     heartbeat delayis to be set to the value of 0
2165  *                     milliseconds.
2166  *
2167  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2168  *                     discovery upon the specified address. Note that
2169  *                     if the address feild is empty then all addresses
2170  *                     on the association are effected.
2171  *
2172  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2173  *                     discovery upon the specified address. Note that
2174  *                     if the address feild is empty then all addresses
2175  *                     on the association are effected. Not also that
2176  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2177  *                     exclusive. Enabling both will have undetermined
2178  *                     results.
2179  *
2180  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2181  *                     on delayed sack. The time specified in spp_sackdelay
2182  *                     is used to specify the sack delay for this address. Note
2183  *                     that if spp_address is empty then all addresses will
2184  *                     enable delayed sack and take on the sack delay
2185  *                     value specified in spp_sackdelay.
2186  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2187  *                     off delayed sack. If the spp_address field is blank then
2188  *                     delayed sack is disabled for the entire association. Note
2189  *                     also that this field is mutually exclusive to
2190  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2191  *                     results.
2192  */
2193 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2194 				       struct sctp_transport   *trans,
2195 				       struct sctp_association *asoc,
2196 				       struct sctp_sock        *sp,
2197 				       int                      hb_change,
2198 				       int                      pmtud_change,
2199 				       int                      sackdelay_change)
2200 {
2201 	int error;
2202 
2203 	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2204 		error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2205 		if (error)
2206 			return error;
2207 	}
2208 
2209 	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2210 	 * this field is ignored.  Note also that a value of zero indicates
2211 	 * the current setting should be left unchanged.
2212 	 */
2213 	if (params->spp_flags & SPP_HB_ENABLE) {
2214 
2215 		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2216 		 * set.  This lets us use 0 value when this flag
2217 		 * is set.
2218 		 */
2219 		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2220 			params->spp_hbinterval = 0;
2221 
2222 		if (params->spp_hbinterval ||
2223 		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2224 			if (trans) {
2225 				trans->hbinterval =
2226 				    msecs_to_jiffies(params->spp_hbinterval);
2227 			} else if (asoc) {
2228 				asoc->hbinterval =
2229 				    msecs_to_jiffies(params->spp_hbinterval);
2230 			} else {
2231 				sp->hbinterval = params->spp_hbinterval;
2232 			}
2233 		}
2234 	}
2235 
2236 	if (hb_change) {
2237 		if (trans) {
2238 			trans->param_flags =
2239 				(trans->param_flags & ~SPP_HB) | hb_change;
2240 		} else if (asoc) {
2241 			asoc->param_flags =
2242 				(asoc->param_flags & ~SPP_HB) | hb_change;
2243 		} else {
2244 			sp->param_flags =
2245 				(sp->param_flags & ~SPP_HB) | hb_change;
2246 		}
2247 	}
2248 
2249 	/* When Path MTU discovery is disabled the value specified here will
2250 	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2251 	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2252 	 * effect).
2253 	 */
2254 	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2255 		if (trans) {
2256 			trans->pathmtu = params->spp_pathmtu;
2257 			sctp_assoc_sync_pmtu(asoc);
2258 		} else if (asoc) {
2259 			asoc->pathmtu = params->spp_pathmtu;
2260 			sctp_frag_point(asoc, params->spp_pathmtu);
2261 		} else {
2262 			sp->pathmtu = params->spp_pathmtu;
2263 		}
2264 	}
2265 
2266 	if (pmtud_change) {
2267 		if (trans) {
2268 			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2269 				(params->spp_flags & SPP_PMTUD_ENABLE);
2270 			trans->param_flags =
2271 				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2272 			if (update) {
2273 				sctp_transport_pmtu(trans);
2274 				sctp_assoc_sync_pmtu(asoc);
2275 			}
2276 		} else if (asoc) {
2277 			asoc->param_flags =
2278 				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2279 		} else {
2280 			sp->param_flags =
2281 				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2282 		}
2283 	}
2284 
2285 	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2286 	 * value of this field is ignored.  Note also that a value of zero
2287 	 * indicates the current setting should be left unchanged.
2288 	 */
2289 	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2290 		if (trans) {
2291 			trans->sackdelay =
2292 				msecs_to_jiffies(params->spp_sackdelay);
2293 		} else if (asoc) {
2294 			asoc->sackdelay =
2295 				msecs_to_jiffies(params->spp_sackdelay);
2296 		} else {
2297 			sp->sackdelay = params->spp_sackdelay;
2298 		}
2299 	}
2300 
2301 	if (sackdelay_change) {
2302 		if (trans) {
2303 			trans->param_flags =
2304 				(trans->param_flags & ~SPP_SACKDELAY) |
2305 				sackdelay_change;
2306 		} else if (asoc) {
2307 			asoc->param_flags =
2308 				(asoc->param_flags & ~SPP_SACKDELAY) |
2309 				sackdelay_change;
2310 		} else {
2311 			sp->param_flags =
2312 				(sp->param_flags & ~SPP_SACKDELAY) |
2313 				sackdelay_change;
2314 		}
2315 	}
2316 
2317 	/* Note that a value of zero indicates the current setting should be
2318 	   left unchanged.
2319 	 */
2320 	if (params->spp_pathmaxrxt) {
2321 		if (trans) {
2322 			trans->pathmaxrxt = params->spp_pathmaxrxt;
2323 		} else if (asoc) {
2324 			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2325 		} else {
2326 			sp->pathmaxrxt = params->spp_pathmaxrxt;
2327 		}
2328 	}
2329 
2330 	return 0;
2331 }
2332 
2333 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2334 					    char __user *optval,
2335 					    unsigned int optlen)
2336 {
2337 	struct sctp_paddrparams  params;
2338 	struct sctp_transport   *trans = NULL;
2339 	struct sctp_association *asoc = NULL;
2340 	struct sctp_sock        *sp = sctp_sk(sk);
2341 	int error;
2342 	int hb_change, pmtud_change, sackdelay_change;
2343 
2344 	if (optlen != sizeof(struct sctp_paddrparams))
2345 		return - EINVAL;
2346 
2347 	if (copy_from_user(&params, optval, optlen))
2348 		return -EFAULT;
2349 
2350 	/* Validate flags and value parameters. */
2351 	hb_change        = params.spp_flags & SPP_HB;
2352 	pmtud_change     = params.spp_flags & SPP_PMTUD;
2353 	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2354 
2355 	if (hb_change        == SPP_HB ||
2356 	    pmtud_change     == SPP_PMTUD ||
2357 	    sackdelay_change == SPP_SACKDELAY ||
2358 	    params.spp_sackdelay > 500 ||
2359 	    (params.spp_pathmtu &&
2360 	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2361 		return -EINVAL;
2362 
2363 	/* If an address other than INADDR_ANY is specified, and
2364 	 * no transport is found, then the request is invalid.
2365 	 */
2366 	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2367 		trans = sctp_addr_id2transport(sk, &params.spp_address,
2368 					       params.spp_assoc_id);
2369 		if (!trans)
2370 			return -EINVAL;
2371 	}
2372 
2373 	/* Get association, if assoc_id != 0 and the socket is a one
2374 	 * to many style socket, and an association was not found, then
2375 	 * the id was invalid.
2376 	 */
2377 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2378 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2379 		return -EINVAL;
2380 
2381 	/* Heartbeat demand can only be sent on a transport or
2382 	 * association, but not a socket.
2383 	 */
2384 	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2385 		return -EINVAL;
2386 
2387 	/* Process parameters. */
2388 	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2389 					    hb_change, pmtud_change,
2390 					    sackdelay_change);
2391 
2392 	if (error)
2393 		return error;
2394 
2395 	/* If changes are for association, also apply parameters to each
2396 	 * transport.
2397 	 */
2398 	if (!trans && asoc) {
2399 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2400 				transports) {
2401 			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2402 						    hb_change, pmtud_change,
2403 						    sackdelay_change);
2404 		}
2405 	}
2406 
2407 	return 0;
2408 }
2409 
2410 /*
2411  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2412  *
2413  * This option will effect the way delayed acks are performed.  This
2414  * option allows you to get or set the delayed ack time, in
2415  * milliseconds.  It also allows changing the delayed ack frequency.
2416  * Changing the frequency to 1 disables the delayed sack algorithm.  If
2417  * the assoc_id is 0, then this sets or gets the endpoints default
2418  * values.  If the assoc_id field is non-zero, then the set or get
2419  * effects the specified association for the one to many model (the
2420  * assoc_id field is ignored by the one to one model).  Note that if
2421  * sack_delay or sack_freq are 0 when setting this option, then the
2422  * current values will remain unchanged.
2423  *
2424  * struct sctp_sack_info {
2425  *     sctp_assoc_t            sack_assoc_id;
2426  *     uint32_t                sack_delay;
2427  *     uint32_t                sack_freq;
2428  * };
2429  *
2430  * sack_assoc_id -  This parameter, indicates which association the user
2431  *    is performing an action upon.  Note that if this field's value is
2432  *    zero then the endpoints default value is changed (effecting future
2433  *    associations only).
2434  *
2435  * sack_delay -  This parameter contains the number of milliseconds that
2436  *    the user is requesting the delayed ACK timer be set to.  Note that
2437  *    this value is defined in the standard to be between 200 and 500
2438  *    milliseconds.
2439  *
2440  * sack_freq -  This parameter contains the number of packets that must
2441  *    be received before a sack is sent without waiting for the delay
2442  *    timer to expire.  The default value for this is 2, setting this
2443  *    value to 1 will disable the delayed sack algorithm.
2444  */
2445 
2446 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2447 				       char __user *optval, unsigned int optlen)
2448 {
2449 	struct sctp_sack_info    params;
2450 	struct sctp_transport   *trans = NULL;
2451 	struct sctp_association *asoc = NULL;
2452 	struct sctp_sock        *sp = sctp_sk(sk);
2453 
2454 	if (optlen == sizeof(struct sctp_sack_info)) {
2455 		if (copy_from_user(&params, optval, optlen))
2456 			return -EFAULT;
2457 
2458 		if (params.sack_delay == 0 && params.sack_freq == 0)
2459 			return 0;
2460 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2461 		printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2462 		       "in delayed_ack socket option deprecated\n");
2463 		printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2464 		if (copy_from_user(&params, optval, optlen))
2465 			return -EFAULT;
2466 
2467 		if (params.sack_delay == 0)
2468 			params.sack_freq = 1;
2469 		else
2470 			params.sack_freq = 0;
2471 	} else
2472 		return - EINVAL;
2473 
2474 	/* Validate value parameter. */
2475 	if (params.sack_delay > 500)
2476 		return -EINVAL;
2477 
2478 	/* Get association, if sack_assoc_id != 0 and the socket is a one
2479 	 * to many style socket, and an association was not found, then
2480 	 * the id was invalid.
2481 	 */
2482 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2483 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2484 		return -EINVAL;
2485 
2486 	if (params.sack_delay) {
2487 		if (asoc) {
2488 			asoc->sackdelay =
2489 				msecs_to_jiffies(params.sack_delay);
2490 			asoc->param_flags =
2491 				(asoc->param_flags & ~SPP_SACKDELAY) |
2492 				SPP_SACKDELAY_ENABLE;
2493 		} else {
2494 			sp->sackdelay = params.sack_delay;
2495 			sp->param_flags =
2496 				(sp->param_flags & ~SPP_SACKDELAY) |
2497 				SPP_SACKDELAY_ENABLE;
2498 		}
2499 	}
2500 
2501 	if (params.sack_freq == 1) {
2502 		if (asoc) {
2503 			asoc->param_flags =
2504 				(asoc->param_flags & ~SPP_SACKDELAY) |
2505 				SPP_SACKDELAY_DISABLE;
2506 		} else {
2507 			sp->param_flags =
2508 				(sp->param_flags & ~SPP_SACKDELAY) |
2509 				SPP_SACKDELAY_DISABLE;
2510 		}
2511 	} else if (params.sack_freq > 1) {
2512 		if (asoc) {
2513 			asoc->sackfreq = params.sack_freq;
2514 			asoc->param_flags =
2515 				(asoc->param_flags & ~SPP_SACKDELAY) |
2516 				SPP_SACKDELAY_ENABLE;
2517 		} else {
2518 			sp->sackfreq = params.sack_freq;
2519 			sp->param_flags =
2520 				(sp->param_flags & ~SPP_SACKDELAY) |
2521 				SPP_SACKDELAY_ENABLE;
2522 		}
2523 	}
2524 
2525 	/* If change is for association, also apply to each transport. */
2526 	if (asoc) {
2527 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2528 				transports) {
2529 			if (params.sack_delay) {
2530 				trans->sackdelay =
2531 					msecs_to_jiffies(params.sack_delay);
2532 				trans->param_flags =
2533 					(trans->param_flags & ~SPP_SACKDELAY) |
2534 					SPP_SACKDELAY_ENABLE;
2535 			}
2536 			if (params.sack_freq == 1) {
2537 				trans->param_flags =
2538 					(trans->param_flags & ~SPP_SACKDELAY) |
2539 					SPP_SACKDELAY_DISABLE;
2540 			} else if (params.sack_freq > 1) {
2541 				trans->sackfreq = params.sack_freq;
2542 				trans->param_flags =
2543 					(trans->param_flags & ~SPP_SACKDELAY) |
2544 					SPP_SACKDELAY_ENABLE;
2545 			}
2546 		}
2547 	}
2548 
2549 	return 0;
2550 }
2551 
2552 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2553  *
2554  * Applications can specify protocol parameters for the default association
2555  * initialization.  The option name argument to setsockopt() and getsockopt()
2556  * is SCTP_INITMSG.
2557  *
2558  * Setting initialization parameters is effective only on an unconnected
2559  * socket (for UDP-style sockets only future associations are effected
2560  * by the change).  With TCP-style sockets, this option is inherited by
2561  * sockets derived from a listener socket.
2562  */
2563 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2564 {
2565 	struct sctp_initmsg sinit;
2566 	struct sctp_sock *sp = sctp_sk(sk);
2567 
2568 	if (optlen != sizeof(struct sctp_initmsg))
2569 		return -EINVAL;
2570 	if (copy_from_user(&sinit, optval, optlen))
2571 		return -EFAULT;
2572 
2573 	if (sinit.sinit_num_ostreams)
2574 		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2575 	if (sinit.sinit_max_instreams)
2576 		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2577 	if (sinit.sinit_max_attempts)
2578 		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2579 	if (sinit.sinit_max_init_timeo)
2580 		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2581 
2582 	return 0;
2583 }
2584 
2585 /*
2586  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2587  *
2588  *   Applications that wish to use the sendto() system call may wish to
2589  *   specify a default set of parameters that would normally be supplied
2590  *   through the inclusion of ancillary data.  This socket option allows
2591  *   such an application to set the default sctp_sndrcvinfo structure.
2592  *   The application that wishes to use this socket option simply passes
2593  *   in to this call the sctp_sndrcvinfo structure defined in Section
2594  *   5.2.2) The input parameters accepted by this call include
2595  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2596  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2597  *   to this call if the caller is using the UDP model.
2598  */
2599 static int sctp_setsockopt_default_send_param(struct sock *sk,
2600 					      char __user *optval,
2601 					      unsigned int optlen)
2602 {
2603 	struct sctp_sndrcvinfo info;
2604 	struct sctp_association *asoc;
2605 	struct sctp_sock *sp = sctp_sk(sk);
2606 
2607 	if (optlen != sizeof(struct sctp_sndrcvinfo))
2608 		return -EINVAL;
2609 	if (copy_from_user(&info, optval, optlen))
2610 		return -EFAULT;
2611 
2612 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2613 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2614 		return -EINVAL;
2615 
2616 	if (asoc) {
2617 		asoc->default_stream = info.sinfo_stream;
2618 		asoc->default_flags = info.sinfo_flags;
2619 		asoc->default_ppid = info.sinfo_ppid;
2620 		asoc->default_context = info.sinfo_context;
2621 		asoc->default_timetolive = info.sinfo_timetolive;
2622 	} else {
2623 		sp->default_stream = info.sinfo_stream;
2624 		sp->default_flags = info.sinfo_flags;
2625 		sp->default_ppid = info.sinfo_ppid;
2626 		sp->default_context = info.sinfo_context;
2627 		sp->default_timetolive = info.sinfo_timetolive;
2628 	}
2629 
2630 	return 0;
2631 }
2632 
2633 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2634  *
2635  * Requests that the local SCTP stack use the enclosed peer address as
2636  * the association primary.  The enclosed address must be one of the
2637  * association peer's addresses.
2638  */
2639 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2640 					unsigned int optlen)
2641 {
2642 	struct sctp_prim prim;
2643 	struct sctp_transport *trans;
2644 
2645 	if (optlen != sizeof(struct sctp_prim))
2646 		return -EINVAL;
2647 
2648 	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2649 		return -EFAULT;
2650 
2651 	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2652 	if (!trans)
2653 		return -EINVAL;
2654 
2655 	sctp_assoc_set_primary(trans->asoc, trans);
2656 
2657 	return 0;
2658 }
2659 
2660 /*
2661  * 7.1.5 SCTP_NODELAY
2662  *
2663  * Turn on/off any Nagle-like algorithm.  This means that packets are
2664  * generally sent as soon as possible and no unnecessary delays are
2665  * introduced, at the cost of more packets in the network.  Expects an
2666  *  integer boolean flag.
2667  */
2668 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2669 				   unsigned int optlen)
2670 {
2671 	int val;
2672 
2673 	if (optlen < sizeof(int))
2674 		return -EINVAL;
2675 	if (get_user(val, (int __user *)optval))
2676 		return -EFAULT;
2677 
2678 	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2679 	return 0;
2680 }
2681 
2682 /*
2683  *
2684  * 7.1.1 SCTP_RTOINFO
2685  *
2686  * The protocol parameters used to initialize and bound retransmission
2687  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2688  * and modify these parameters.
2689  * All parameters are time values, in milliseconds.  A value of 0, when
2690  * modifying the parameters, indicates that the current value should not
2691  * be changed.
2692  *
2693  */
2694 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2695 {
2696 	struct sctp_rtoinfo rtoinfo;
2697 	struct sctp_association *asoc;
2698 
2699 	if (optlen != sizeof (struct sctp_rtoinfo))
2700 		return -EINVAL;
2701 
2702 	if (copy_from_user(&rtoinfo, optval, optlen))
2703 		return -EFAULT;
2704 
2705 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2706 
2707 	/* Set the values to the specific association */
2708 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2709 		return -EINVAL;
2710 
2711 	if (asoc) {
2712 		if (rtoinfo.srto_initial != 0)
2713 			asoc->rto_initial =
2714 				msecs_to_jiffies(rtoinfo.srto_initial);
2715 		if (rtoinfo.srto_max != 0)
2716 			asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2717 		if (rtoinfo.srto_min != 0)
2718 			asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2719 	} else {
2720 		/* If there is no association or the association-id = 0
2721 		 * set the values to the endpoint.
2722 		 */
2723 		struct sctp_sock *sp = sctp_sk(sk);
2724 
2725 		if (rtoinfo.srto_initial != 0)
2726 			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2727 		if (rtoinfo.srto_max != 0)
2728 			sp->rtoinfo.srto_max = rtoinfo.srto_max;
2729 		if (rtoinfo.srto_min != 0)
2730 			sp->rtoinfo.srto_min = rtoinfo.srto_min;
2731 	}
2732 
2733 	return 0;
2734 }
2735 
2736 /*
2737  *
2738  * 7.1.2 SCTP_ASSOCINFO
2739  *
2740  * This option is used to tune the maximum retransmission attempts
2741  * of the association.
2742  * Returns an error if the new association retransmission value is
2743  * greater than the sum of the retransmission value  of the peer.
2744  * See [SCTP] for more information.
2745  *
2746  */
2747 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2748 {
2749 
2750 	struct sctp_assocparams assocparams;
2751 	struct sctp_association *asoc;
2752 
2753 	if (optlen != sizeof(struct sctp_assocparams))
2754 		return -EINVAL;
2755 	if (copy_from_user(&assocparams, optval, optlen))
2756 		return -EFAULT;
2757 
2758 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2759 
2760 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2761 		return -EINVAL;
2762 
2763 	/* Set the values to the specific association */
2764 	if (asoc) {
2765 		if (assocparams.sasoc_asocmaxrxt != 0) {
2766 			__u32 path_sum = 0;
2767 			int   paths = 0;
2768 			struct sctp_transport *peer_addr;
2769 
2770 			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2771 					transports) {
2772 				path_sum += peer_addr->pathmaxrxt;
2773 				paths++;
2774 			}
2775 
2776 			/* Only validate asocmaxrxt if we have more than
2777 			 * one path/transport.  We do this because path
2778 			 * retransmissions are only counted when we have more
2779 			 * then one path.
2780 			 */
2781 			if (paths > 1 &&
2782 			    assocparams.sasoc_asocmaxrxt > path_sum)
2783 				return -EINVAL;
2784 
2785 			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2786 		}
2787 
2788 		if (assocparams.sasoc_cookie_life != 0) {
2789 			asoc->cookie_life.tv_sec =
2790 					assocparams.sasoc_cookie_life / 1000;
2791 			asoc->cookie_life.tv_usec =
2792 					(assocparams.sasoc_cookie_life % 1000)
2793 					* 1000;
2794 		}
2795 	} else {
2796 		/* Set the values to the endpoint */
2797 		struct sctp_sock *sp = sctp_sk(sk);
2798 
2799 		if (assocparams.sasoc_asocmaxrxt != 0)
2800 			sp->assocparams.sasoc_asocmaxrxt =
2801 						assocparams.sasoc_asocmaxrxt;
2802 		if (assocparams.sasoc_cookie_life != 0)
2803 			sp->assocparams.sasoc_cookie_life =
2804 						assocparams.sasoc_cookie_life;
2805 	}
2806 	return 0;
2807 }
2808 
2809 /*
2810  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2811  *
2812  * This socket option is a boolean flag which turns on or off mapped V4
2813  * addresses.  If this option is turned on and the socket is type
2814  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2815  * If this option is turned off, then no mapping will be done of V4
2816  * addresses and a user will receive both PF_INET6 and PF_INET type
2817  * addresses on the socket.
2818  */
2819 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2820 {
2821 	int val;
2822 	struct sctp_sock *sp = sctp_sk(sk);
2823 
2824 	if (optlen < sizeof(int))
2825 		return -EINVAL;
2826 	if (get_user(val, (int __user *)optval))
2827 		return -EFAULT;
2828 	if (val)
2829 		sp->v4mapped = 1;
2830 	else
2831 		sp->v4mapped = 0;
2832 
2833 	return 0;
2834 }
2835 
2836 /*
2837  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2838  * This option will get or set the maximum size to put in any outgoing
2839  * SCTP DATA chunk.  If a message is larger than this size it will be
2840  * fragmented by SCTP into the specified size.  Note that the underlying
2841  * SCTP implementation may fragment into smaller sized chunks when the
2842  * PMTU of the underlying association is smaller than the value set by
2843  * the user.  The default value for this option is '0' which indicates
2844  * the user is NOT limiting fragmentation and only the PMTU will effect
2845  * SCTP's choice of DATA chunk size.  Note also that values set larger
2846  * than the maximum size of an IP datagram will effectively let SCTP
2847  * control fragmentation (i.e. the same as setting this option to 0).
2848  *
2849  * The following structure is used to access and modify this parameter:
2850  *
2851  * struct sctp_assoc_value {
2852  *   sctp_assoc_t assoc_id;
2853  *   uint32_t assoc_value;
2854  * };
2855  *
2856  * assoc_id:  This parameter is ignored for one-to-one style sockets.
2857  *    For one-to-many style sockets this parameter indicates which
2858  *    association the user is performing an action upon.  Note that if
2859  *    this field's value is zero then the endpoints default value is
2860  *    changed (effecting future associations only).
2861  * assoc_value:  This parameter specifies the maximum size in bytes.
2862  */
2863 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2864 {
2865 	struct sctp_assoc_value params;
2866 	struct sctp_association *asoc;
2867 	struct sctp_sock *sp = sctp_sk(sk);
2868 	int val;
2869 
2870 	if (optlen == sizeof(int)) {
2871 		printk(KERN_WARNING
2872 		   "SCTP: Use of int in maxseg socket option deprecated\n");
2873 		printk(KERN_WARNING
2874 		   "SCTP: Use struct sctp_assoc_value instead\n");
2875 		if (copy_from_user(&val, optval, optlen))
2876 			return -EFAULT;
2877 		params.assoc_id = 0;
2878 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2879 		if (copy_from_user(&params, optval, optlen))
2880 			return -EFAULT;
2881 		val = params.assoc_value;
2882 	} else
2883 		return -EINVAL;
2884 
2885 	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2886 		return -EINVAL;
2887 
2888 	asoc = sctp_id2assoc(sk, params.assoc_id);
2889 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2890 		return -EINVAL;
2891 
2892 	if (asoc) {
2893 		if (val == 0) {
2894 			val = asoc->pathmtu;
2895 			val -= sp->pf->af->net_header_len;
2896 			val -= sizeof(struct sctphdr) +
2897 					sizeof(struct sctp_data_chunk);
2898 		}
2899 		asoc->user_frag = val;
2900 		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2901 	} else {
2902 		sp->user_frag = val;
2903 	}
2904 
2905 	return 0;
2906 }
2907 
2908 
2909 /*
2910  *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2911  *
2912  *   Requests that the peer mark the enclosed address as the association
2913  *   primary. The enclosed address must be one of the association's
2914  *   locally bound addresses. The following structure is used to make a
2915  *   set primary request:
2916  */
2917 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2918 					     unsigned int optlen)
2919 {
2920 	struct sctp_sock	*sp;
2921 	struct sctp_endpoint	*ep;
2922 	struct sctp_association	*asoc = NULL;
2923 	struct sctp_setpeerprim	prim;
2924 	struct sctp_chunk	*chunk;
2925 	int 			err;
2926 
2927 	sp = sctp_sk(sk);
2928 	ep = sp->ep;
2929 
2930 	if (!sctp_addip_enable)
2931 		return -EPERM;
2932 
2933 	if (optlen != sizeof(struct sctp_setpeerprim))
2934 		return -EINVAL;
2935 
2936 	if (copy_from_user(&prim, optval, optlen))
2937 		return -EFAULT;
2938 
2939 	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2940 	if (!asoc)
2941 		return -EINVAL;
2942 
2943 	if (!asoc->peer.asconf_capable)
2944 		return -EPERM;
2945 
2946 	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2947 		return -EPERM;
2948 
2949 	if (!sctp_state(asoc, ESTABLISHED))
2950 		return -ENOTCONN;
2951 
2952 	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2953 		return -EADDRNOTAVAIL;
2954 
2955 	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
2956 	chunk = sctp_make_asconf_set_prim(asoc,
2957 					  (union sctp_addr *)&prim.sspp_addr);
2958 	if (!chunk)
2959 		return -ENOMEM;
2960 
2961 	err = sctp_send_asconf(asoc, chunk);
2962 
2963 	SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2964 
2965 	return err;
2966 }
2967 
2968 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2969 					    unsigned int optlen)
2970 {
2971 	struct sctp_setadaptation adaptation;
2972 
2973 	if (optlen != sizeof(struct sctp_setadaptation))
2974 		return -EINVAL;
2975 	if (copy_from_user(&adaptation, optval, optlen))
2976 		return -EFAULT;
2977 
2978 	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2979 
2980 	return 0;
2981 }
2982 
2983 /*
2984  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
2985  *
2986  * The context field in the sctp_sndrcvinfo structure is normally only
2987  * used when a failed message is retrieved holding the value that was
2988  * sent down on the actual send call.  This option allows the setting of
2989  * a default context on an association basis that will be received on
2990  * reading messages from the peer.  This is especially helpful in the
2991  * one-2-many model for an application to keep some reference to an
2992  * internal state machine that is processing messages on the
2993  * association.  Note that the setting of this value only effects
2994  * received messages from the peer and does not effect the value that is
2995  * saved with outbound messages.
2996  */
2997 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2998 				   unsigned int optlen)
2999 {
3000 	struct sctp_assoc_value params;
3001 	struct sctp_sock *sp;
3002 	struct sctp_association *asoc;
3003 
3004 	if (optlen != sizeof(struct sctp_assoc_value))
3005 		return -EINVAL;
3006 	if (copy_from_user(&params, optval, optlen))
3007 		return -EFAULT;
3008 
3009 	sp = sctp_sk(sk);
3010 
3011 	if (params.assoc_id != 0) {
3012 		asoc = sctp_id2assoc(sk, params.assoc_id);
3013 		if (!asoc)
3014 			return -EINVAL;
3015 		asoc->default_rcv_context = params.assoc_value;
3016 	} else {
3017 		sp->default_rcv_context = params.assoc_value;
3018 	}
3019 
3020 	return 0;
3021 }
3022 
3023 /*
3024  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3025  *
3026  * This options will at a minimum specify if the implementation is doing
3027  * fragmented interleave.  Fragmented interleave, for a one to many
3028  * socket, is when subsequent calls to receive a message may return
3029  * parts of messages from different associations.  Some implementations
3030  * may allow you to turn this value on or off.  If so, when turned off,
3031  * no fragment interleave will occur (which will cause a head of line
3032  * blocking amongst multiple associations sharing the same one to many
3033  * socket).  When this option is turned on, then each receive call may
3034  * come from a different association (thus the user must receive data
3035  * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3036  * association each receive belongs to.
3037  *
3038  * This option takes a boolean value.  A non-zero value indicates that
3039  * fragmented interleave is on.  A value of zero indicates that
3040  * fragmented interleave is off.
3041  *
3042  * Note that it is important that an implementation that allows this
3043  * option to be turned on, have it off by default.  Otherwise an unaware
3044  * application using the one to many model may become confused and act
3045  * incorrectly.
3046  */
3047 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3048 					       char __user *optval,
3049 					       unsigned int optlen)
3050 {
3051 	int val;
3052 
3053 	if (optlen != sizeof(int))
3054 		return -EINVAL;
3055 	if (get_user(val, (int __user *)optval))
3056 		return -EFAULT;
3057 
3058 	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3059 
3060 	return 0;
3061 }
3062 
3063 /*
3064  * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3065  *       (SCTP_PARTIAL_DELIVERY_POINT)
3066  *
3067  * This option will set or get the SCTP partial delivery point.  This
3068  * point is the size of a message where the partial delivery API will be
3069  * invoked to help free up rwnd space for the peer.  Setting this to a
3070  * lower value will cause partial deliveries to happen more often.  The
3071  * calls argument is an integer that sets or gets the partial delivery
3072  * point.  Note also that the call will fail if the user attempts to set
3073  * this value larger than the socket receive buffer size.
3074  *
3075  * Note that any single message having a length smaller than or equal to
3076  * the SCTP partial delivery point will be delivered in one single read
3077  * call as long as the user provided buffer is large enough to hold the
3078  * message.
3079  */
3080 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3081 						  char __user *optval,
3082 						  unsigned int optlen)
3083 {
3084 	u32 val;
3085 
3086 	if (optlen != sizeof(u32))
3087 		return -EINVAL;
3088 	if (get_user(val, (int __user *)optval))
3089 		return -EFAULT;
3090 
3091 	/* Note: We double the receive buffer from what the user sets
3092 	 * it to be, also initial rwnd is based on rcvbuf/2.
3093 	 */
3094 	if (val > (sk->sk_rcvbuf >> 1))
3095 		return -EINVAL;
3096 
3097 	sctp_sk(sk)->pd_point = val;
3098 
3099 	return 0; /* is this the right error code? */
3100 }
3101 
3102 /*
3103  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3104  *
3105  * This option will allow a user to change the maximum burst of packets
3106  * that can be emitted by this association.  Note that the default value
3107  * is 4, and some implementations may restrict this setting so that it
3108  * can only be lowered.
3109  *
3110  * NOTE: This text doesn't seem right.  Do this on a socket basis with
3111  * future associations inheriting the socket value.
3112  */
3113 static int sctp_setsockopt_maxburst(struct sock *sk,
3114 				    char __user *optval,
3115 				    unsigned int optlen)
3116 {
3117 	struct sctp_assoc_value params;
3118 	struct sctp_sock *sp;
3119 	struct sctp_association *asoc;
3120 	int val;
3121 	int assoc_id = 0;
3122 
3123 	if (optlen == sizeof(int)) {
3124 		printk(KERN_WARNING
3125 		   "SCTP: Use of int in max_burst socket option deprecated\n");
3126 		printk(KERN_WARNING
3127 		   "SCTP: Use struct sctp_assoc_value instead\n");
3128 		if (copy_from_user(&val, optval, optlen))
3129 			return -EFAULT;
3130 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3131 		if (copy_from_user(&params, optval, optlen))
3132 			return -EFAULT;
3133 		val = params.assoc_value;
3134 		assoc_id = params.assoc_id;
3135 	} else
3136 		return -EINVAL;
3137 
3138 	sp = sctp_sk(sk);
3139 
3140 	if (assoc_id != 0) {
3141 		asoc = sctp_id2assoc(sk, assoc_id);
3142 		if (!asoc)
3143 			return -EINVAL;
3144 		asoc->max_burst = val;
3145 	} else
3146 		sp->max_burst = val;
3147 
3148 	return 0;
3149 }
3150 
3151 /*
3152  * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3153  *
3154  * This set option adds a chunk type that the user is requesting to be
3155  * received only in an authenticated way.  Changes to the list of chunks
3156  * will only effect future associations on the socket.
3157  */
3158 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3159 				      char __user *optval,
3160 				      unsigned int optlen)
3161 {
3162 	struct sctp_authchunk val;
3163 
3164 	if (!sctp_auth_enable)
3165 		return -EACCES;
3166 
3167 	if (optlen != sizeof(struct sctp_authchunk))
3168 		return -EINVAL;
3169 	if (copy_from_user(&val, optval, optlen))
3170 		return -EFAULT;
3171 
3172 	switch (val.sauth_chunk) {
3173 		case SCTP_CID_INIT:
3174 		case SCTP_CID_INIT_ACK:
3175 		case SCTP_CID_SHUTDOWN_COMPLETE:
3176 		case SCTP_CID_AUTH:
3177 			return -EINVAL;
3178 	}
3179 
3180 	/* add this chunk id to the endpoint */
3181 	return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3182 }
3183 
3184 /*
3185  * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3186  *
3187  * This option gets or sets the list of HMAC algorithms that the local
3188  * endpoint requires the peer to use.
3189  */
3190 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3191 				      char __user *optval,
3192 				      unsigned int optlen)
3193 {
3194 	struct sctp_hmacalgo *hmacs;
3195 	u32 idents;
3196 	int err;
3197 
3198 	if (!sctp_auth_enable)
3199 		return -EACCES;
3200 
3201 	if (optlen < sizeof(struct sctp_hmacalgo))
3202 		return -EINVAL;
3203 
3204 	hmacs = kmalloc(optlen, GFP_KERNEL);
3205 	if (!hmacs)
3206 		return -ENOMEM;
3207 
3208 	if (copy_from_user(hmacs, optval, optlen)) {
3209 		err = -EFAULT;
3210 		goto out;
3211 	}
3212 
3213 	idents = hmacs->shmac_num_idents;
3214 	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3215 	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3216 		err = -EINVAL;
3217 		goto out;
3218 	}
3219 
3220 	err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3221 out:
3222 	kfree(hmacs);
3223 	return err;
3224 }
3225 
3226 /*
3227  * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3228  *
3229  * This option will set a shared secret key which is used to build an
3230  * association shared key.
3231  */
3232 static int sctp_setsockopt_auth_key(struct sock *sk,
3233 				    char __user *optval,
3234 				    unsigned int optlen)
3235 {
3236 	struct sctp_authkey *authkey;
3237 	struct sctp_association *asoc;
3238 	int ret;
3239 
3240 	if (!sctp_auth_enable)
3241 		return -EACCES;
3242 
3243 	if (optlen <= sizeof(struct sctp_authkey))
3244 		return -EINVAL;
3245 
3246 	authkey = kmalloc(optlen, GFP_KERNEL);
3247 	if (!authkey)
3248 		return -ENOMEM;
3249 
3250 	if (copy_from_user(authkey, optval, optlen)) {
3251 		ret = -EFAULT;
3252 		goto out;
3253 	}
3254 
3255 	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3256 		ret = -EINVAL;
3257 		goto out;
3258 	}
3259 
3260 	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3261 	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3262 		ret = -EINVAL;
3263 		goto out;
3264 	}
3265 
3266 	ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3267 out:
3268 	kfree(authkey);
3269 	return ret;
3270 }
3271 
3272 /*
3273  * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3274  *
3275  * This option will get or set the active shared key to be used to build
3276  * the association shared key.
3277  */
3278 static int sctp_setsockopt_active_key(struct sock *sk,
3279 				      char __user *optval,
3280 				      unsigned int optlen)
3281 {
3282 	struct sctp_authkeyid val;
3283 	struct sctp_association *asoc;
3284 
3285 	if (!sctp_auth_enable)
3286 		return -EACCES;
3287 
3288 	if (optlen != sizeof(struct sctp_authkeyid))
3289 		return -EINVAL;
3290 	if (copy_from_user(&val, optval, optlen))
3291 		return -EFAULT;
3292 
3293 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3294 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3295 		return -EINVAL;
3296 
3297 	return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3298 					val.scact_keynumber);
3299 }
3300 
3301 /*
3302  * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3303  *
3304  * This set option will delete a shared secret key from use.
3305  */
3306 static int sctp_setsockopt_del_key(struct sock *sk,
3307 				   char __user *optval,
3308 				   unsigned int optlen)
3309 {
3310 	struct sctp_authkeyid val;
3311 	struct sctp_association *asoc;
3312 
3313 	if (!sctp_auth_enable)
3314 		return -EACCES;
3315 
3316 	if (optlen != sizeof(struct sctp_authkeyid))
3317 		return -EINVAL;
3318 	if (copy_from_user(&val, optval, optlen))
3319 		return -EFAULT;
3320 
3321 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3322 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3323 		return -EINVAL;
3324 
3325 	return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3326 				    val.scact_keynumber);
3327 
3328 }
3329 
3330 
3331 /* API 6.2 setsockopt(), getsockopt()
3332  *
3333  * Applications use setsockopt() and getsockopt() to set or retrieve
3334  * socket options.  Socket options are used to change the default
3335  * behavior of sockets calls.  They are described in Section 7.
3336  *
3337  * The syntax is:
3338  *
3339  *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3340  *                    int __user *optlen);
3341  *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3342  *                    int optlen);
3343  *
3344  *   sd      - the socket descript.
3345  *   level   - set to IPPROTO_SCTP for all SCTP options.
3346  *   optname - the option name.
3347  *   optval  - the buffer to store the value of the option.
3348  *   optlen  - the size of the buffer.
3349  */
3350 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3351 				char __user *optval, unsigned int optlen)
3352 {
3353 	int retval = 0;
3354 
3355 	SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3356 			  sk, optname);
3357 
3358 	/* I can hardly begin to describe how wrong this is.  This is
3359 	 * so broken as to be worse than useless.  The API draft
3360 	 * REALLY is NOT helpful here...  I am not convinced that the
3361 	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3362 	 * are at all well-founded.
3363 	 */
3364 	if (level != SOL_SCTP) {
3365 		struct sctp_af *af = sctp_sk(sk)->pf->af;
3366 		retval = af->setsockopt(sk, level, optname, optval, optlen);
3367 		goto out_nounlock;
3368 	}
3369 
3370 	sctp_lock_sock(sk);
3371 
3372 	switch (optname) {
3373 	case SCTP_SOCKOPT_BINDX_ADD:
3374 		/* 'optlen' is the size of the addresses buffer. */
3375 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3376 					       optlen, SCTP_BINDX_ADD_ADDR);
3377 		break;
3378 
3379 	case SCTP_SOCKOPT_BINDX_REM:
3380 		/* 'optlen' is the size of the addresses buffer. */
3381 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3382 					       optlen, SCTP_BINDX_REM_ADDR);
3383 		break;
3384 
3385 	case SCTP_SOCKOPT_CONNECTX_OLD:
3386 		/* 'optlen' is the size of the addresses buffer. */
3387 		retval = sctp_setsockopt_connectx_old(sk,
3388 					    (struct sockaddr __user *)optval,
3389 					    optlen);
3390 		break;
3391 
3392 	case SCTP_SOCKOPT_CONNECTX:
3393 		/* 'optlen' is the size of the addresses buffer. */
3394 		retval = sctp_setsockopt_connectx(sk,
3395 					    (struct sockaddr __user *)optval,
3396 					    optlen);
3397 		break;
3398 
3399 	case SCTP_DISABLE_FRAGMENTS:
3400 		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3401 		break;
3402 
3403 	case SCTP_EVENTS:
3404 		retval = sctp_setsockopt_events(sk, optval, optlen);
3405 		break;
3406 
3407 	case SCTP_AUTOCLOSE:
3408 		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3409 		break;
3410 
3411 	case SCTP_PEER_ADDR_PARAMS:
3412 		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3413 		break;
3414 
3415 	case SCTP_DELAYED_ACK:
3416 		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3417 		break;
3418 	case SCTP_PARTIAL_DELIVERY_POINT:
3419 		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3420 		break;
3421 
3422 	case SCTP_INITMSG:
3423 		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3424 		break;
3425 	case SCTP_DEFAULT_SEND_PARAM:
3426 		retval = sctp_setsockopt_default_send_param(sk, optval,
3427 							    optlen);
3428 		break;
3429 	case SCTP_PRIMARY_ADDR:
3430 		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3431 		break;
3432 	case SCTP_SET_PEER_PRIMARY_ADDR:
3433 		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3434 		break;
3435 	case SCTP_NODELAY:
3436 		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3437 		break;
3438 	case SCTP_RTOINFO:
3439 		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3440 		break;
3441 	case SCTP_ASSOCINFO:
3442 		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3443 		break;
3444 	case SCTP_I_WANT_MAPPED_V4_ADDR:
3445 		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3446 		break;
3447 	case SCTP_MAXSEG:
3448 		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3449 		break;
3450 	case SCTP_ADAPTATION_LAYER:
3451 		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3452 		break;
3453 	case SCTP_CONTEXT:
3454 		retval = sctp_setsockopt_context(sk, optval, optlen);
3455 		break;
3456 	case SCTP_FRAGMENT_INTERLEAVE:
3457 		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3458 		break;
3459 	case SCTP_MAX_BURST:
3460 		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3461 		break;
3462 	case SCTP_AUTH_CHUNK:
3463 		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3464 		break;
3465 	case SCTP_HMAC_IDENT:
3466 		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3467 		break;
3468 	case SCTP_AUTH_KEY:
3469 		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3470 		break;
3471 	case SCTP_AUTH_ACTIVE_KEY:
3472 		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3473 		break;
3474 	case SCTP_AUTH_DELETE_KEY:
3475 		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3476 		break;
3477 	default:
3478 		retval = -ENOPROTOOPT;
3479 		break;
3480 	}
3481 
3482 	sctp_release_sock(sk);
3483 
3484 out_nounlock:
3485 	return retval;
3486 }
3487 
3488 /* API 3.1.6 connect() - UDP Style Syntax
3489  *
3490  * An application may use the connect() call in the UDP model to initiate an
3491  * association without sending data.
3492  *
3493  * The syntax is:
3494  *
3495  * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3496  *
3497  * sd: the socket descriptor to have a new association added to.
3498  *
3499  * nam: the address structure (either struct sockaddr_in or struct
3500  *    sockaddr_in6 defined in RFC2553 [7]).
3501  *
3502  * len: the size of the address.
3503  */
3504 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3505 			     int addr_len)
3506 {
3507 	int err = 0;
3508 	struct sctp_af *af;
3509 
3510 	sctp_lock_sock(sk);
3511 
3512 	SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3513 			  __func__, sk, addr, addr_len);
3514 
3515 	/* Validate addr_len before calling common connect/connectx routine. */
3516 	af = sctp_get_af_specific(addr->sa_family);
3517 	if (!af || addr_len < af->sockaddr_len) {
3518 		err = -EINVAL;
3519 	} else {
3520 		/* Pass correct addr len to common routine (so it knows there
3521 		 * is only one address being passed.
3522 		 */
3523 		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3524 	}
3525 
3526 	sctp_release_sock(sk);
3527 	return err;
3528 }
3529 
3530 /* FIXME: Write comments. */
3531 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3532 {
3533 	return -EOPNOTSUPP; /* STUB */
3534 }
3535 
3536 /* 4.1.4 accept() - TCP Style Syntax
3537  *
3538  * Applications use accept() call to remove an established SCTP
3539  * association from the accept queue of the endpoint.  A new socket
3540  * descriptor will be returned from accept() to represent the newly
3541  * formed association.
3542  */
3543 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3544 {
3545 	struct sctp_sock *sp;
3546 	struct sctp_endpoint *ep;
3547 	struct sock *newsk = NULL;
3548 	struct sctp_association *asoc;
3549 	long timeo;
3550 	int error = 0;
3551 
3552 	sctp_lock_sock(sk);
3553 
3554 	sp = sctp_sk(sk);
3555 	ep = sp->ep;
3556 
3557 	if (!sctp_style(sk, TCP)) {
3558 		error = -EOPNOTSUPP;
3559 		goto out;
3560 	}
3561 
3562 	if (!sctp_sstate(sk, LISTENING)) {
3563 		error = -EINVAL;
3564 		goto out;
3565 	}
3566 
3567 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3568 
3569 	error = sctp_wait_for_accept(sk, timeo);
3570 	if (error)
3571 		goto out;
3572 
3573 	/* We treat the list of associations on the endpoint as the accept
3574 	 * queue and pick the first association on the list.
3575 	 */
3576 	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3577 
3578 	newsk = sp->pf->create_accept_sk(sk, asoc);
3579 	if (!newsk) {
3580 		error = -ENOMEM;
3581 		goto out;
3582 	}
3583 
3584 	/* Populate the fields of the newsk from the oldsk and migrate the
3585 	 * asoc to the newsk.
3586 	 */
3587 	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3588 
3589 out:
3590 	sctp_release_sock(sk);
3591 	*err = error;
3592 	return newsk;
3593 }
3594 
3595 /* The SCTP ioctl handler. */
3596 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3597 {
3598 	return -ENOIOCTLCMD;
3599 }
3600 
3601 /* This is the function which gets called during socket creation to
3602  * initialized the SCTP-specific portion of the sock.
3603  * The sock structure should already be zero-filled memory.
3604  */
3605 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3606 {
3607 	struct sctp_endpoint *ep;
3608 	struct sctp_sock *sp;
3609 
3610 	SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3611 
3612 	sp = sctp_sk(sk);
3613 
3614 	/* Initialize the SCTP per socket area.  */
3615 	switch (sk->sk_type) {
3616 	case SOCK_SEQPACKET:
3617 		sp->type = SCTP_SOCKET_UDP;
3618 		break;
3619 	case SOCK_STREAM:
3620 		sp->type = SCTP_SOCKET_TCP;
3621 		break;
3622 	default:
3623 		return -ESOCKTNOSUPPORT;
3624 	}
3625 
3626 	/* Initialize default send parameters. These parameters can be
3627 	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3628 	 */
3629 	sp->default_stream = 0;
3630 	sp->default_ppid = 0;
3631 	sp->default_flags = 0;
3632 	sp->default_context = 0;
3633 	sp->default_timetolive = 0;
3634 
3635 	sp->default_rcv_context = 0;
3636 	sp->max_burst = sctp_max_burst;
3637 
3638 	/* Initialize default setup parameters. These parameters
3639 	 * can be modified with the SCTP_INITMSG socket option or
3640 	 * overridden by the SCTP_INIT CMSG.
3641 	 */
3642 	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3643 	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3644 	sp->initmsg.sinit_max_attempts   = sctp_max_retrans_init;
3645 	sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3646 
3647 	/* Initialize default RTO related parameters.  These parameters can
3648 	 * be modified for with the SCTP_RTOINFO socket option.
3649 	 */
3650 	sp->rtoinfo.srto_initial = sctp_rto_initial;
3651 	sp->rtoinfo.srto_max     = sctp_rto_max;
3652 	sp->rtoinfo.srto_min     = sctp_rto_min;
3653 
3654 	/* Initialize default association related parameters. These parameters
3655 	 * can be modified with the SCTP_ASSOCINFO socket option.
3656 	 */
3657 	sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3658 	sp->assocparams.sasoc_number_peer_destinations = 0;
3659 	sp->assocparams.sasoc_peer_rwnd = 0;
3660 	sp->assocparams.sasoc_local_rwnd = 0;
3661 	sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3662 
3663 	/* Initialize default event subscriptions. By default, all the
3664 	 * options are off.
3665 	 */
3666 	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3667 
3668 	/* Default Peer Address Parameters.  These defaults can
3669 	 * be modified via SCTP_PEER_ADDR_PARAMS
3670 	 */
3671 	sp->hbinterval  = sctp_hb_interval;
3672 	sp->pathmaxrxt  = sctp_max_retrans_path;
3673 	sp->pathmtu     = 0; // allow default discovery
3674 	sp->sackdelay   = sctp_sack_timeout;
3675 	sp->sackfreq	= 2;
3676 	sp->param_flags = SPP_HB_ENABLE |
3677 			  SPP_PMTUD_ENABLE |
3678 			  SPP_SACKDELAY_ENABLE;
3679 
3680 	/* If enabled no SCTP message fragmentation will be performed.
3681 	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3682 	 */
3683 	sp->disable_fragments = 0;
3684 
3685 	/* Enable Nagle algorithm by default.  */
3686 	sp->nodelay           = 0;
3687 
3688 	/* Enable by default. */
3689 	sp->v4mapped          = 1;
3690 
3691 	/* Auto-close idle associations after the configured
3692 	 * number of seconds.  A value of 0 disables this
3693 	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3694 	 * for UDP-style sockets only.
3695 	 */
3696 	sp->autoclose         = 0;
3697 
3698 	/* User specified fragmentation limit. */
3699 	sp->user_frag         = 0;
3700 
3701 	sp->adaptation_ind = 0;
3702 
3703 	sp->pf = sctp_get_pf_specific(sk->sk_family);
3704 
3705 	/* Control variables for partial data delivery. */
3706 	atomic_set(&sp->pd_mode, 0);
3707 	skb_queue_head_init(&sp->pd_lobby);
3708 	sp->frag_interleave = 0;
3709 
3710 	/* Create a per socket endpoint structure.  Even if we
3711 	 * change the data structure relationships, this may still
3712 	 * be useful for storing pre-connect address information.
3713 	 */
3714 	ep = sctp_endpoint_new(sk, GFP_KERNEL);
3715 	if (!ep)
3716 		return -ENOMEM;
3717 
3718 	sp->ep = ep;
3719 	sp->hmac = NULL;
3720 
3721 	SCTP_DBG_OBJCNT_INC(sock);
3722 
3723 	local_bh_disable();
3724 	percpu_counter_inc(&sctp_sockets_allocated);
3725 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3726 	local_bh_enable();
3727 
3728 	return 0;
3729 }
3730 
3731 /* Cleanup any SCTP per socket resources.  */
3732 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3733 {
3734 	struct sctp_endpoint *ep;
3735 
3736 	SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3737 
3738 	/* Release our hold on the endpoint. */
3739 	ep = sctp_sk(sk)->ep;
3740 	sctp_endpoint_free(ep);
3741 	local_bh_disable();
3742 	percpu_counter_dec(&sctp_sockets_allocated);
3743 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3744 	local_bh_enable();
3745 }
3746 
3747 /* API 4.1.7 shutdown() - TCP Style Syntax
3748  *     int shutdown(int socket, int how);
3749  *
3750  *     sd      - the socket descriptor of the association to be closed.
3751  *     how     - Specifies the type of shutdown.  The  values  are
3752  *               as follows:
3753  *               SHUT_RD
3754  *                     Disables further receive operations. No SCTP
3755  *                     protocol action is taken.
3756  *               SHUT_WR
3757  *                     Disables further send operations, and initiates
3758  *                     the SCTP shutdown sequence.
3759  *               SHUT_RDWR
3760  *                     Disables further send  and  receive  operations
3761  *                     and initiates the SCTP shutdown sequence.
3762  */
3763 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3764 {
3765 	struct sctp_endpoint *ep;
3766 	struct sctp_association *asoc;
3767 
3768 	if (!sctp_style(sk, TCP))
3769 		return;
3770 
3771 	if (how & SEND_SHUTDOWN) {
3772 		ep = sctp_sk(sk)->ep;
3773 		if (!list_empty(&ep->asocs)) {
3774 			asoc = list_entry(ep->asocs.next,
3775 					  struct sctp_association, asocs);
3776 			sctp_primitive_SHUTDOWN(asoc, NULL);
3777 		}
3778 	}
3779 }
3780 
3781 /* 7.2.1 Association Status (SCTP_STATUS)
3782 
3783  * Applications can retrieve current status information about an
3784  * association, including association state, peer receiver window size,
3785  * number of unacked data chunks, and number of data chunks pending
3786  * receipt.  This information is read-only.
3787  */
3788 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3789 				       char __user *optval,
3790 				       int __user *optlen)
3791 {
3792 	struct sctp_status status;
3793 	struct sctp_association *asoc = NULL;
3794 	struct sctp_transport *transport;
3795 	sctp_assoc_t associd;
3796 	int retval = 0;
3797 
3798 	if (len < sizeof(status)) {
3799 		retval = -EINVAL;
3800 		goto out;
3801 	}
3802 
3803 	len = sizeof(status);
3804 	if (copy_from_user(&status, optval, len)) {
3805 		retval = -EFAULT;
3806 		goto out;
3807 	}
3808 
3809 	associd = status.sstat_assoc_id;
3810 	asoc = sctp_id2assoc(sk, associd);
3811 	if (!asoc) {
3812 		retval = -EINVAL;
3813 		goto out;
3814 	}
3815 
3816 	transport = asoc->peer.primary_path;
3817 
3818 	status.sstat_assoc_id = sctp_assoc2id(asoc);
3819 	status.sstat_state = asoc->state;
3820 	status.sstat_rwnd =  asoc->peer.rwnd;
3821 	status.sstat_unackdata = asoc->unack_data;
3822 
3823 	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3824 	status.sstat_instrms = asoc->c.sinit_max_instreams;
3825 	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3826 	status.sstat_fragmentation_point = asoc->frag_point;
3827 	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3828 	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3829 			transport->af_specific->sockaddr_len);
3830 	/* Map ipv4 address into v4-mapped-on-v6 address.  */
3831 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3832 		(union sctp_addr *)&status.sstat_primary.spinfo_address);
3833 	status.sstat_primary.spinfo_state = transport->state;
3834 	status.sstat_primary.spinfo_cwnd = transport->cwnd;
3835 	status.sstat_primary.spinfo_srtt = transport->srtt;
3836 	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3837 	status.sstat_primary.spinfo_mtu = transport->pathmtu;
3838 
3839 	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3840 		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3841 
3842 	if (put_user(len, optlen)) {
3843 		retval = -EFAULT;
3844 		goto out;
3845 	}
3846 
3847 	SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3848 			  len, status.sstat_state, status.sstat_rwnd,
3849 			  status.sstat_assoc_id);
3850 
3851 	if (copy_to_user(optval, &status, len)) {
3852 		retval = -EFAULT;
3853 		goto out;
3854 	}
3855 
3856 out:
3857 	return (retval);
3858 }
3859 
3860 
3861 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3862  *
3863  * Applications can retrieve information about a specific peer address
3864  * of an association, including its reachability state, congestion
3865  * window, and retransmission timer values.  This information is
3866  * read-only.
3867  */
3868 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3869 					  char __user *optval,
3870 					  int __user *optlen)
3871 {
3872 	struct sctp_paddrinfo pinfo;
3873 	struct sctp_transport *transport;
3874 	int retval = 0;
3875 
3876 	if (len < sizeof(pinfo)) {
3877 		retval = -EINVAL;
3878 		goto out;
3879 	}
3880 
3881 	len = sizeof(pinfo);
3882 	if (copy_from_user(&pinfo, optval, len)) {
3883 		retval = -EFAULT;
3884 		goto out;
3885 	}
3886 
3887 	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3888 					   pinfo.spinfo_assoc_id);
3889 	if (!transport)
3890 		return -EINVAL;
3891 
3892 	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3893 	pinfo.spinfo_state = transport->state;
3894 	pinfo.spinfo_cwnd = transport->cwnd;
3895 	pinfo.spinfo_srtt = transport->srtt;
3896 	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3897 	pinfo.spinfo_mtu = transport->pathmtu;
3898 
3899 	if (pinfo.spinfo_state == SCTP_UNKNOWN)
3900 		pinfo.spinfo_state = SCTP_ACTIVE;
3901 
3902 	if (put_user(len, optlen)) {
3903 		retval = -EFAULT;
3904 		goto out;
3905 	}
3906 
3907 	if (copy_to_user(optval, &pinfo, len)) {
3908 		retval = -EFAULT;
3909 		goto out;
3910 	}
3911 
3912 out:
3913 	return (retval);
3914 }
3915 
3916 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3917  *
3918  * This option is a on/off flag.  If enabled no SCTP message
3919  * fragmentation will be performed.  Instead if a message being sent
3920  * exceeds the current PMTU size, the message will NOT be sent and
3921  * instead a error will be indicated to the user.
3922  */
3923 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3924 					char __user *optval, int __user *optlen)
3925 {
3926 	int val;
3927 
3928 	if (len < sizeof(int))
3929 		return -EINVAL;
3930 
3931 	len = sizeof(int);
3932 	val = (sctp_sk(sk)->disable_fragments == 1);
3933 	if (put_user(len, optlen))
3934 		return -EFAULT;
3935 	if (copy_to_user(optval, &val, len))
3936 		return -EFAULT;
3937 	return 0;
3938 }
3939 
3940 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3941  *
3942  * This socket option is used to specify various notifications and
3943  * ancillary data the user wishes to receive.
3944  */
3945 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3946 				  int __user *optlen)
3947 {
3948 	if (len < sizeof(struct sctp_event_subscribe))
3949 		return -EINVAL;
3950 	len = sizeof(struct sctp_event_subscribe);
3951 	if (put_user(len, optlen))
3952 		return -EFAULT;
3953 	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3954 		return -EFAULT;
3955 	return 0;
3956 }
3957 
3958 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3959  *
3960  * This socket option is applicable to the UDP-style socket only.  When
3961  * set it will cause associations that are idle for more than the
3962  * specified number of seconds to automatically close.  An association
3963  * being idle is defined an association that has NOT sent or received
3964  * user data.  The special value of '0' indicates that no automatic
3965  * close of any associations should be performed.  The option expects an
3966  * integer defining the number of seconds of idle time before an
3967  * association is closed.
3968  */
3969 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3970 {
3971 	/* Applicable to UDP-style socket only */
3972 	if (sctp_style(sk, TCP))
3973 		return -EOPNOTSUPP;
3974 	if (len < sizeof(int))
3975 		return -EINVAL;
3976 	len = sizeof(int);
3977 	if (put_user(len, optlen))
3978 		return -EFAULT;
3979 	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3980 		return -EFAULT;
3981 	return 0;
3982 }
3983 
3984 /* Helper routine to branch off an association to a new socket.  */
3985 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3986 				struct socket **sockp)
3987 {
3988 	struct sock *sk = asoc->base.sk;
3989 	struct socket *sock;
3990 	struct sctp_af *af;
3991 	int err = 0;
3992 
3993 	/* An association cannot be branched off from an already peeled-off
3994 	 * socket, nor is this supported for tcp style sockets.
3995 	 */
3996 	if (!sctp_style(sk, UDP))
3997 		return -EINVAL;
3998 
3999 	/* Create a new socket.  */
4000 	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4001 	if (err < 0)
4002 		return err;
4003 
4004 	sctp_copy_sock(sock->sk, sk, asoc);
4005 
4006 	/* Make peeled-off sockets more like 1-1 accepted sockets.
4007 	 * Set the daddr and initialize id to something more random
4008 	 */
4009 	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4010 	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4011 
4012 	/* Populate the fields of the newsk from the oldsk and migrate the
4013 	 * asoc to the newsk.
4014 	 */
4015 	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4016 
4017 	*sockp = sock;
4018 
4019 	return err;
4020 }
4021 
4022 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4023 {
4024 	sctp_peeloff_arg_t peeloff;
4025 	struct socket *newsock;
4026 	int retval = 0;
4027 	struct sctp_association *asoc;
4028 
4029 	if (len < sizeof(sctp_peeloff_arg_t))
4030 		return -EINVAL;
4031 	len = sizeof(sctp_peeloff_arg_t);
4032 	if (copy_from_user(&peeloff, optval, len))
4033 		return -EFAULT;
4034 
4035 	asoc = sctp_id2assoc(sk, peeloff.associd);
4036 	if (!asoc) {
4037 		retval = -EINVAL;
4038 		goto out;
4039 	}
4040 
4041 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4042 
4043 	retval = sctp_do_peeloff(asoc, &newsock);
4044 	if (retval < 0)
4045 		goto out;
4046 
4047 	/* Map the socket to an unused fd that can be returned to the user.  */
4048 	retval = sock_map_fd(newsock, 0);
4049 	if (retval < 0) {
4050 		sock_release(newsock);
4051 		goto out;
4052 	}
4053 
4054 	SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4055 			  __func__, sk, asoc, newsock->sk, retval);
4056 
4057 	/* Return the fd mapped to the new socket.  */
4058 	peeloff.sd = retval;
4059 	if (put_user(len, optlen))
4060 		return -EFAULT;
4061 	if (copy_to_user(optval, &peeloff, len))
4062 		retval = -EFAULT;
4063 
4064 out:
4065 	return retval;
4066 }
4067 
4068 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4069  *
4070  * Applications can enable or disable heartbeats for any peer address of
4071  * an association, modify an address's heartbeat interval, force a
4072  * heartbeat to be sent immediately, and adjust the address's maximum
4073  * number of retransmissions sent before an address is considered
4074  * unreachable.  The following structure is used to access and modify an
4075  * address's parameters:
4076  *
4077  *  struct sctp_paddrparams {
4078  *     sctp_assoc_t            spp_assoc_id;
4079  *     struct sockaddr_storage spp_address;
4080  *     uint32_t                spp_hbinterval;
4081  *     uint16_t                spp_pathmaxrxt;
4082  *     uint32_t                spp_pathmtu;
4083  *     uint32_t                spp_sackdelay;
4084  *     uint32_t                spp_flags;
4085  * };
4086  *
4087  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4088  *                     application, and identifies the association for
4089  *                     this query.
4090  *   spp_address     - This specifies which address is of interest.
4091  *   spp_hbinterval  - This contains the value of the heartbeat interval,
4092  *                     in milliseconds.  If a  value of zero
4093  *                     is present in this field then no changes are to
4094  *                     be made to this parameter.
4095  *   spp_pathmaxrxt  - This contains the maximum number of
4096  *                     retransmissions before this address shall be
4097  *                     considered unreachable. If a  value of zero
4098  *                     is present in this field then no changes are to
4099  *                     be made to this parameter.
4100  *   spp_pathmtu     - When Path MTU discovery is disabled the value
4101  *                     specified here will be the "fixed" path mtu.
4102  *                     Note that if the spp_address field is empty
4103  *                     then all associations on this address will
4104  *                     have this fixed path mtu set upon them.
4105  *
4106  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4107  *                     the number of milliseconds that sacks will be delayed
4108  *                     for. This value will apply to all addresses of an
4109  *                     association if the spp_address field is empty. Note
4110  *                     also, that if delayed sack is enabled and this
4111  *                     value is set to 0, no change is made to the last
4112  *                     recorded delayed sack timer value.
4113  *
4114  *   spp_flags       - These flags are used to control various features
4115  *                     on an association. The flag field may contain
4116  *                     zero or more of the following options.
4117  *
4118  *                     SPP_HB_ENABLE  - Enable heartbeats on the
4119  *                     specified address. Note that if the address
4120  *                     field is empty all addresses for the association
4121  *                     have heartbeats enabled upon them.
4122  *
4123  *                     SPP_HB_DISABLE - Disable heartbeats on the
4124  *                     speicifed address. Note that if the address
4125  *                     field is empty all addresses for the association
4126  *                     will have their heartbeats disabled. Note also
4127  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4128  *                     mutually exclusive, only one of these two should
4129  *                     be specified. Enabling both fields will have
4130  *                     undetermined results.
4131  *
4132  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4133  *                     to be made immediately.
4134  *
4135  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4136  *                     discovery upon the specified address. Note that
4137  *                     if the address feild is empty then all addresses
4138  *                     on the association are effected.
4139  *
4140  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4141  *                     discovery upon the specified address. Note that
4142  *                     if the address feild is empty then all addresses
4143  *                     on the association are effected. Not also that
4144  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4145  *                     exclusive. Enabling both will have undetermined
4146  *                     results.
4147  *
4148  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4149  *                     on delayed sack. The time specified in spp_sackdelay
4150  *                     is used to specify the sack delay for this address. Note
4151  *                     that if spp_address is empty then all addresses will
4152  *                     enable delayed sack and take on the sack delay
4153  *                     value specified in spp_sackdelay.
4154  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4155  *                     off delayed sack. If the spp_address field is blank then
4156  *                     delayed sack is disabled for the entire association. Note
4157  *                     also that this field is mutually exclusive to
4158  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4159  *                     results.
4160  */
4161 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4162 					    char __user *optval, int __user *optlen)
4163 {
4164 	struct sctp_paddrparams  params;
4165 	struct sctp_transport   *trans = NULL;
4166 	struct sctp_association *asoc = NULL;
4167 	struct sctp_sock        *sp = sctp_sk(sk);
4168 
4169 	if (len < sizeof(struct sctp_paddrparams))
4170 		return -EINVAL;
4171 	len = sizeof(struct sctp_paddrparams);
4172 	if (copy_from_user(&params, optval, len))
4173 		return -EFAULT;
4174 
4175 	/* If an address other than INADDR_ANY is specified, and
4176 	 * no transport is found, then the request is invalid.
4177 	 */
4178 	if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4179 		trans = sctp_addr_id2transport(sk, &params.spp_address,
4180 					       params.spp_assoc_id);
4181 		if (!trans) {
4182 			SCTP_DEBUG_PRINTK("Failed no transport\n");
4183 			return -EINVAL;
4184 		}
4185 	}
4186 
4187 	/* Get association, if assoc_id != 0 and the socket is a one
4188 	 * to many style socket, and an association was not found, then
4189 	 * the id was invalid.
4190 	 */
4191 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4192 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4193 		SCTP_DEBUG_PRINTK("Failed no association\n");
4194 		return -EINVAL;
4195 	}
4196 
4197 	if (trans) {
4198 		/* Fetch transport values. */
4199 		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4200 		params.spp_pathmtu    = trans->pathmtu;
4201 		params.spp_pathmaxrxt = trans->pathmaxrxt;
4202 		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4203 
4204 		/*draft-11 doesn't say what to return in spp_flags*/
4205 		params.spp_flags      = trans->param_flags;
4206 	} else if (asoc) {
4207 		/* Fetch association values. */
4208 		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4209 		params.spp_pathmtu    = asoc->pathmtu;
4210 		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4211 		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4212 
4213 		/*draft-11 doesn't say what to return in spp_flags*/
4214 		params.spp_flags      = asoc->param_flags;
4215 	} else {
4216 		/* Fetch socket values. */
4217 		params.spp_hbinterval = sp->hbinterval;
4218 		params.spp_pathmtu    = sp->pathmtu;
4219 		params.spp_sackdelay  = sp->sackdelay;
4220 		params.spp_pathmaxrxt = sp->pathmaxrxt;
4221 
4222 		/*draft-11 doesn't say what to return in spp_flags*/
4223 		params.spp_flags      = sp->param_flags;
4224 	}
4225 
4226 	if (copy_to_user(optval, &params, len))
4227 		return -EFAULT;
4228 
4229 	if (put_user(len, optlen))
4230 		return -EFAULT;
4231 
4232 	return 0;
4233 }
4234 
4235 /*
4236  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4237  *
4238  * This option will effect the way delayed acks are performed.  This
4239  * option allows you to get or set the delayed ack time, in
4240  * milliseconds.  It also allows changing the delayed ack frequency.
4241  * Changing the frequency to 1 disables the delayed sack algorithm.  If
4242  * the assoc_id is 0, then this sets or gets the endpoints default
4243  * values.  If the assoc_id field is non-zero, then the set or get
4244  * effects the specified association for the one to many model (the
4245  * assoc_id field is ignored by the one to one model).  Note that if
4246  * sack_delay or sack_freq are 0 when setting this option, then the
4247  * current values will remain unchanged.
4248  *
4249  * struct sctp_sack_info {
4250  *     sctp_assoc_t            sack_assoc_id;
4251  *     uint32_t                sack_delay;
4252  *     uint32_t                sack_freq;
4253  * };
4254  *
4255  * sack_assoc_id -  This parameter, indicates which association the user
4256  *    is performing an action upon.  Note that if this field's value is
4257  *    zero then the endpoints default value is changed (effecting future
4258  *    associations only).
4259  *
4260  * sack_delay -  This parameter contains the number of milliseconds that
4261  *    the user is requesting the delayed ACK timer be set to.  Note that
4262  *    this value is defined in the standard to be between 200 and 500
4263  *    milliseconds.
4264  *
4265  * sack_freq -  This parameter contains the number of packets that must
4266  *    be received before a sack is sent without waiting for the delay
4267  *    timer to expire.  The default value for this is 2, setting this
4268  *    value to 1 will disable the delayed sack algorithm.
4269  */
4270 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4271 					    char __user *optval,
4272 					    int __user *optlen)
4273 {
4274 	struct sctp_sack_info    params;
4275 	struct sctp_association *asoc = NULL;
4276 	struct sctp_sock        *sp = sctp_sk(sk);
4277 
4278 	if (len >= sizeof(struct sctp_sack_info)) {
4279 		len = sizeof(struct sctp_sack_info);
4280 
4281 		if (copy_from_user(&params, optval, len))
4282 			return -EFAULT;
4283 	} else if (len == sizeof(struct sctp_assoc_value)) {
4284 		printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4285 		       "in delayed_ack socket option deprecated\n");
4286 		printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4287 		if (copy_from_user(&params, optval, len))
4288 			return -EFAULT;
4289 	} else
4290 		return - EINVAL;
4291 
4292 	/* Get association, if sack_assoc_id != 0 and the socket is a one
4293 	 * to many style socket, and an association was not found, then
4294 	 * the id was invalid.
4295 	 */
4296 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4297 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4298 		return -EINVAL;
4299 
4300 	if (asoc) {
4301 		/* Fetch association values. */
4302 		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4303 			params.sack_delay = jiffies_to_msecs(
4304 				asoc->sackdelay);
4305 			params.sack_freq = asoc->sackfreq;
4306 
4307 		} else {
4308 			params.sack_delay = 0;
4309 			params.sack_freq = 1;
4310 		}
4311 	} else {
4312 		/* Fetch socket values. */
4313 		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4314 			params.sack_delay  = sp->sackdelay;
4315 			params.sack_freq = sp->sackfreq;
4316 		} else {
4317 			params.sack_delay  = 0;
4318 			params.sack_freq = 1;
4319 		}
4320 	}
4321 
4322 	if (copy_to_user(optval, &params, len))
4323 		return -EFAULT;
4324 
4325 	if (put_user(len, optlen))
4326 		return -EFAULT;
4327 
4328 	return 0;
4329 }
4330 
4331 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4332  *
4333  * Applications can specify protocol parameters for the default association
4334  * initialization.  The option name argument to setsockopt() and getsockopt()
4335  * is SCTP_INITMSG.
4336  *
4337  * Setting initialization parameters is effective only on an unconnected
4338  * socket (for UDP-style sockets only future associations are effected
4339  * by the change).  With TCP-style sockets, this option is inherited by
4340  * sockets derived from a listener socket.
4341  */
4342 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4343 {
4344 	if (len < sizeof(struct sctp_initmsg))
4345 		return -EINVAL;
4346 	len = sizeof(struct sctp_initmsg);
4347 	if (put_user(len, optlen))
4348 		return -EFAULT;
4349 	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4350 		return -EFAULT;
4351 	return 0;
4352 }
4353 
4354 
4355 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4356 				      char __user *optval, int __user *optlen)
4357 {
4358 	struct sctp_association *asoc;
4359 	int cnt = 0;
4360 	struct sctp_getaddrs getaddrs;
4361 	struct sctp_transport *from;
4362 	void __user *to;
4363 	union sctp_addr temp;
4364 	struct sctp_sock *sp = sctp_sk(sk);
4365 	int addrlen;
4366 	size_t space_left;
4367 	int bytes_copied;
4368 
4369 	if (len < sizeof(struct sctp_getaddrs))
4370 		return -EINVAL;
4371 
4372 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4373 		return -EFAULT;
4374 
4375 	/* For UDP-style sockets, id specifies the association to query.  */
4376 	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4377 	if (!asoc)
4378 		return -EINVAL;
4379 
4380 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4381 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4382 
4383 	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4384 				transports) {
4385 		memcpy(&temp, &from->ipaddr, sizeof(temp));
4386 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4387 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4388 		if (space_left < addrlen)
4389 			return -ENOMEM;
4390 		if (copy_to_user(to, &temp, addrlen))
4391 			return -EFAULT;
4392 		to += addrlen;
4393 		cnt++;
4394 		space_left -= addrlen;
4395 	}
4396 
4397 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4398 		return -EFAULT;
4399 	bytes_copied = ((char __user *)to) - optval;
4400 	if (put_user(bytes_copied, optlen))
4401 		return -EFAULT;
4402 
4403 	return 0;
4404 }
4405 
4406 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4407 			    size_t space_left, int *bytes_copied)
4408 {
4409 	struct sctp_sockaddr_entry *addr;
4410 	union sctp_addr temp;
4411 	int cnt = 0;
4412 	int addrlen;
4413 
4414 	rcu_read_lock();
4415 	list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4416 		if (!addr->valid)
4417 			continue;
4418 
4419 		if ((PF_INET == sk->sk_family) &&
4420 		    (AF_INET6 == addr->a.sa.sa_family))
4421 			continue;
4422 		if ((PF_INET6 == sk->sk_family) &&
4423 		    inet_v6_ipv6only(sk) &&
4424 		    (AF_INET == addr->a.sa.sa_family))
4425 			continue;
4426 		memcpy(&temp, &addr->a, sizeof(temp));
4427 		if (!temp.v4.sin_port)
4428 			temp.v4.sin_port = htons(port);
4429 
4430 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4431 								&temp);
4432 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4433 		if (space_left < addrlen) {
4434 			cnt =  -ENOMEM;
4435 			break;
4436 		}
4437 		memcpy(to, &temp, addrlen);
4438 
4439 		to += addrlen;
4440 		cnt ++;
4441 		space_left -= addrlen;
4442 		*bytes_copied += addrlen;
4443 	}
4444 	rcu_read_unlock();
4445 
4446 	return cnt;
4447 }
4448 
4449 
4450 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4451 				       char __user *optval, int __user *optlen)
4452 {
4453 	struct sctp_bind_addr *bp;
4454 	struct sctp_association *asoc;
4455 	int cnt = 0;
4456 	struct sctp_getaddrs getaddrs;
4457 	struct sctp_sockaddr_entry *addr;
4458 	void __user *to;
4459 	union sctp_addr temp;
4460 	struct sctp_sock *sp = sctp_sk(sk);
4461 	int addrlen;
4462 	int err = 0;
4463 	size_t space_left;
4464 	int bytes_copied = 0;
4465 	void *addrs;
4466 	void *buf;
4467 
4468 	if (len < sizeof(struct sctp_getaddrs))
4469 		return -EINVAL;
4470 
4471 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4472 		return -EFAULT;
4473 
4474 	/*
4475 	 *  For UDP-style sockets, id specifies the association to query.
4476 	 *  If the id field is set to the value '0' then the locally bound
4477 	 *  addresses are returned without regard to any particular
4478 	 *  association.
4479 	 */
4480 	if (0 == getaddrs.assoc_id) {
4481 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4482 	} else {
4483 		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4484 		if (!asoc)
4485 			return -EINVAL;
4486 		bp = &asoc->base.bind_addr;
4487 	}
4488 
4489 	to = optval + offsetof(struct sctp_getaddrs,addrs);
4490 	space_left = len - offsetof(struct sctp_getaddrs,addrs);
4491 
4492 	addrs = kmalloc(space_left, GFP_KERNEL);
4493 	if (!addrs)
4494 		return -ENOMEM;
4495 
4496 	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4497 	 * addresses from the global local address list.
4498 	 */
4499 	if (sctp_list_single_entry(&bp->address_list)) {
4500 		addr = list_entry(bp->address_list.next,
4501 				  struct sctp_sockaddr_entry, list);
4502 		if (sctp_is_any(sk, &addr->a)) {
4503 			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4504 						space_left, &bytes_copied);
4505 			if (cnt < 0) {
4506 				err = cnt;
4507 				goto out;
4508 			}
4509 			goto copy_getaddrs;
4510 		}
4511 	}
4512 
4513 	buf = addrs;
4514 	/* Protection on the bound address list is not needed since
4515 	 * in the socket option context we hold a socket lock and
4516 	 * thus the bound address list can't change.
4517 	 */
4518 	list_for_each_entry(addr, &bp->address_list, list) {
4519 		memcpy(&temp, &addr->a, sizeof(temp));
4520 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4521 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4522 		if (space_left < addrlen) {
4523 			err =  -ENOMEM; /*fixme: right error?*/
4524 			goto out;
4525 		}
4526 		memcpy(buf, &temp, addrlen);
4527 		buf += addrlen;
4528 		bytes_copied += addrlen;
4529 		cnt ++;
4530 		space_left -= addrlen;
4531 	}
4532 
4533 copy_getaddrs:
4534 	if (copy_to_user(to, addrs, bytes_copied)) {
4535 		err = -EFAULT;
4536 		goto out;
4537 	}
4538 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4539 		err = -EFAULT;
4540 		goto out;
4541 	}
4542 	if (put_user(bytes_copied, optlen))
4543 		err = -EFAULT;
4544 out:
4545 	kfree(addrs);
4546 	return err;
4547 }
4548 
4549 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4550  *
4551  * Requests that the local SCTP stack use the enclosed peer address as
4552  * the association primary.  The enclosed address must be one of the
4553  * association peer's addresses.
4554  */
4555 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4556 					char __user *optval, int __user *optlen)
4557 {
4558 	struct sctp_prim prim;
4559 	struct sctp_association *asoc;
4560 	struct sctp_sock *sp = sctp_sk(sk);
4561 
4562 	if (len < sizeof(struct sctp_prim))
4563 		return -EINVAL;
4564 
4565 	len = sizeof(struct sctp_prim);
4566 
4567 	if (copy_from_user(&prim, optval, len))
4568 		return -EFAULT;
4569 
4570 	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4571 	if (!asoc)
4572 		return -EINVAL;
4573 
4574 	if (!asoc->peer.primary_path)
4575 		return -ENOTCONN;
4576 
4577 	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4578 		asoc->peer.primary_path->af_specific->sockaddr_len);
4579 
4580 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4581 			(union sctp_addr *)&prim.ssp_addr);
4582 
4583 	if (put_user(len, optlen))
4584 		return -EFAULT;
4585 	if (copy_to_user(optval, &prim, len))
4586 		return -EFAULT;
4587 
4588 	return 0;
4589 }
4590 
4591 /*
4592  * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4593  *
4594  * Requests that the local endpoint set the specified Adaptation Layer
4595  * Indication parameter for all future INIT and INIT-ACK exchanges.
4596  */
4597 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4598 				  char __user *optval, int __user *optlen)
4599 {
4600 	struct sctp_setadaptation adaptation;
4601 
4602 	if (len < sizeof(struct sctp_setadaptation))
4603 		return -EINVAL;
4604 
4605 	len = sizeof(struct sctp_setadaptation);
4606 
4607 	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4608 
4609 	if (put_user(len, optlen))
4610 		return -EFAULT;
4611 	if (copy_to_user(optval, &adaptation, len))
4612 		return -EFAULT;
4613 
4614 	return 0;
4615 }
4616 
4617 /*
4618  *
4619  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4620  *
4621  *   Applications that wish to use the sendto() system call may wish to
4622  *   specify a default set of parameters that would normally be supplied
4623  *   through the inclusion of ancillary data.  This socket option allows
4624  *   such an application to set the default sctp_sndrcvinfo structure.
4625 
4626 
4627  *   The application that wishes to use this socket option simply passes
4628  *   in to this call the sctp_sndrcvinfo structure defined in Section
4629  *   5.2.2) The input parameters accepted by this call include
4630  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4631  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4632  *   to this call if the caller is using the UDP model.
4633  *
4634  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4635  */
4636 static int sctp_getsockopt_default_send_param(struct sock *sk,
4637 					int len, char __user *optval,
4638 					int __user *optlen)
4639 {
4640 	struct sctp_sndrcvinfo info;
4641 	struct sctp_association *asoc;
4642 	struct sctp_sock *sp = sctp_sk(sk);
4643 
4644 	if (len < sizeof(struct sctp_sndrcvinfo))
4645 		return -EINVAL;
4646 
4647 	len = sizeof(struct sctp_sndrcvinfo);
4648 
4649 	if (copy_from_user(&info, optval, len))
4650 		return -EFAULT;
4651 
4652 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4653 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4654 		return -EINVAL;
4655 
4656 	if (asoc) {
4657 		info.sinfo_stream = asoc->default_stream;
4658 		info.sinfo_flags = asoc->default_flags;
4659 		info.sinfo_ppid = asoc->default_ppid;
4660 		info.sinfo_context = asoc->default_context;
4661 		info.sinfo_timetolive = asoc->default_timetolive;
4662 	} else {
4663 		info.sinfo_stream = sp->default_stream;
4664 		info.sinfo_flags = sp->default_flags;
4665 		info.sinfo_ppid = sp->default_ppid;
4666 		info.sinfo_context = sp->default_context;
4667 		info.sinfo_timetolive = sp->default_timetolive;
4668 	}
4669 
4670 	if (put_user(len, optlen))
4671 		return -EFAULT;
4672 	if (copy_to_user(optval, &info, len))
4673 		return -EFAULT;
4674 
4675 	return 0;
4676 }
4677 
4678 /*
4679  *
4680  * 7.1.5 SCTP_NODELAY
4681  *
4682  * Turn on/off any Nagle-like algorithm.  This means that packets are
4683  * generally sent as soon as possible and no unnecessary delays are
4684  * introduced, at the cost of more packets in the network.  Expects an
4685  * integer boolean flag.
4686  */
4687 
4688 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4689 				   char __user *optval, int __user *optlen)
4690 {
4691 	int val;
4692 
4693 	if (len < sizeof(int))
4694 		return -EINVAL;
4695 
4696 	len = sizeof(int);
4697 	val = (sctp_sk(sk)->nodelay == 1);
4698 	if (put_user(len, optlen))
4699 		return -EFAULT;
4700 	if (copy_to_user(optval, &val, len))
4701 		return -EFAULT;
4702 	return 0;
4703 }
4704 
4705 /*
4706  *
4707  * 7.1.1 SCTP_RTOINFO
4708  *
4709  * The protocol parameters used to initialize and bound retransmission
4710  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4711  * and modify these parameters.
4712  * All parameters are time values, in milliseconds.  A value of 0, when
4713  * modifying the parameters, indicates that the current value should not
4714  * be changed.
4715  *
4716  */
4717 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4718 				char __user *optval,
4719 				int __user *optlen) {
4720 	struct sctp_rtoinfo rtoinfo;
4721 	struct sctp_association *asoc;
4722 
4723 	if (len < sizeof (struct sctp_rtoinfo))
4724 		return -EINVAL;
4725 
4726 	len = sizeof(struct sctp_rtoinfo);
4727 
4728 	if (copy_from_user(&rtoinfo, optval, len))
4729 		return -EFAULT;
4730 
4731 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4732 
4733 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4734 		return -EINVAL;
4735 
4736 	/* Values corresponding to the specific association. */
4737 	if (asoc) {
4738 		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4739 		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4740 		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4741 	} else {
4742 		/* Values corresponding to the endpoint. */
4743 		struct sctp_sock *sp = sctp_sk(sk);
4744 
4745 		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4746 		rtoinfo.srto_max = sp->rtoinfo.srto_max;
4747 		rtoinfo.srto_min = sp->rtoinfo.srto_min;
4748 	}
4749 
4750 	if (put_user(len, optlen))
4751 		return -EFAULT;
4752 
4753 	if (copy_to_user(optval, &rtoinfo, len))
4754 		return -EFAULT;
4755 
4756 	return 0;
4757 }
4758 
4759 /*
4760  *
4761  * 7.1.2 SCTP_ASSOCINFO
4762  *
4763  * This option is used to tune the maximum retransmission attempts
4764  * of the association.
4765  * Returns an error if the new association retransmission value is
4766  * greater than the sum of the retransmission value  of the peer.
4767  * See [SCTP] for more information.
4768  *
4769  */
4770 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4771 				     char __user *optval,
4772 				     int __user *optlen)
4773 {
4774 
4775 	struct sctp_assocparams assocparams;
4776 	struct sctp_association *asoc;
4777 	struct list_head *pos;
4778 	int cnt = 0;
4779 
4780 	if (len < sizeof (struct sctp_assocparams))
4781 		return -EINVAL;
4782 
4783 	len = sizeof(struct sctp_assocparams);
4784 
4785 	if (copy_from_user(&assocparams, optval, len))
4786 		return -EFAULT;
4787 
4788 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4789 
4790 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4791 		return -EINVAL;
4792 
4793 	/* Values correspoinding to the specific association */
4794 	if (asoc) {
4795 		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4796 		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4797 		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4798 		assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4799 						* 1000) +
4800 						(asoc->cookie_life.tv_usec
4801 						/ 1000);
4802 
4803 		list_for_each(pos, &asoc->peer.transport_addr_list) {
4804 			cnt ++;
4805 		}
4806 
4807 		assocparams.sasoc_number_peer_destinations = cnt;
4808 	} else {
4809 		/* Values corresponding to the endpoint */
4810 		struct sctp_sock *sp = sctp_sk(sk);
4811 
4812 		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4813 		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4814 		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4815 		assocparams.sasoc_cookie_life =
4816 					sp->assocparams.sasoc_cookie_life;
4817 		assocparams.sasoc_number_peer_destinations =
4818 					sp->assocparams.
4819 					sasoc_number_peer_destinations;
4820 	}
4821 
4822 	if (put_user(len, optlen))
4823 		return -EFAULT;
4824 
4825 	if (copy_to_user(optval, &assocparams, len))
4826 		return -EFAULT;
4827 
4828 	return 0;
4829 }
4830 
4831 /*
4832  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4833  *
4834  * This socket option is a boolean flag which turns on or off mapped V4
4835  * addresses.  If this option is turned on and the socket is type
4836  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4837  * If this option is turned off, then no mapping will be done of V4
4838  * addresses and a user will receive both PF_INET6 and PF_INET type
4839  * addresses on the socket.
4840  */
4841 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4842 				    char __user *optval, int __user *optlen)
4843 {
4844 	int val;
4845 	struct sctp_sock *sp = sctp_sk(sk);
4846 
4847 	if (len < sizeof(int))
4848 		return -EINVAL;
4849 
4850 	len = sizeof(int);
4851 	val = sp->v4mapped;
4852 	if (put_user(len, optlen))
4853 		return -EFAULT;
4854 	if (copy_to_user(optval, &val, len))
4855 		return -EFAULT;
4856 
4857 	return 0;
4858 }
4859 
4860 /*
4861  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
4862  * (chapter and verse is quoted at sctp_setsockopt_context())
4863  */
4864 static int sctp_getsockopt_context(struct sock *sk, int len,
4865 				   char __user *optval, int __user *optlen)
4866 {
4867 	struct sctp_assoc_value params;
4868 	struct sctp_sock *sp;
4869 	struct sctp_association *asoc;
4870 
4871 	if (len < sizeof(struct sctp_assoc_value))
4872 		return -EINVAL;
4873 
4874 	len = sizeof(struct sctp_assoc_value);
4875 
4876 	if (copy_from_user(&params, optval, len))
4877 		return -EFAULT;
4878 
4879 	sp = sctp_sk(sk);
4880 
4881 	if (params.assoc_id != 0) {
4882 		asoc = sctp_id2assoc(sk, params.assoc_id);
4883 		if (!asoc)
4884 			return -EINVAL;
4885 		params.assoc_value = asoc->default_rcv_context;
4886 	} else {
4887 		params.assoc_value = sp->default_rcv_context;
4888 	}
4889 
4890 	if (put_user(len, optlen))
4891 		return -EFAULT;
4892 	if (copy_to_user(optval, &params, len))
4893 		return -EFAULT;
4894 
4895 	return 0;
4896 }
4897 
4898 /*
4899  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4900  * This option will get or set the maximum size to put in any outgoing
4901  * SCTP DATA chunk.  If a message is larger than this size it will be
4902  * fragmented by SCTP into the specified size.  Note that the underlying
4903  * SCTP implementation may fragment into smaller sized chunks when the
4904  * PMTU of the underlying association is smaller than the value set by
4905  * the user.  The default value for this option is '0' which indicates
4906  * the user is NOT limiting fragmentation and only the PMTU will effect
4907  * SCTP's choice of DATA chunk size.  Note also that values set larger
4908  * than the maximum size of an IP datagram will effectively let SCTP
4909  * control fragmentation (i.e. the same as setting this option to 0).
4910  *
4911  * The following structure is used to access and modify this parameter:
4912  *
4913  * struct sctp_assoc_value {
4914  *   sctp_assoc_t assoc_id;
4915  *   uint32_t assoc_value;
4916  * };
4917  *
4918  * assoc_id:  This parameter is ignored for one-to-one style sockets.
4919  *    For one-to-many style sockets this parameter indicates which
4920  *    association the user is performing an action upon.  Note that if
4921  *    this field's value is zero then the endpoints default value is
4922  *    changed (effecting future associations only).
4923  * assoc_value:  This parameter specifies the maximum size in bytes.
4924  */
4925 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4926 				  char __user *optval, int __user *optlen)
4927 {
4928 	struct sctp_assoc_value params;
4929 	struct sctp_association *asoc;
4930 
4931 	if (len == sizeof(int)) {
4932 		printk(KERN_WARNING
4933 		   "SCTP: Use of int in maxseg socket option deprecated\n");
4934 		printk(KERN_WARNING
4935 		   "SCTP: Use struct sctp_assoc_value instead\n");
4936 		params.assoc_id = 0;
4937 	} else if (len >= sizeof(struct sctp_assoc_value)) {
4938 		len = sizeof(struct sctp_assoc_value);
4939 		if (copy_from_user(&params, optval, sizeof(params)))
4940 			return -EFAULT;
4941 	} else
4942 		return -EINVAL;
4943 
4944 	asoc = sctp_id2assoc(sk, params.assoc_id);
4945 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4946 		return -EINVAL;
4947 
4948 	if (asoc)
4949 		params.assoc_value = asoc->frag_point;
4950 	else
4951 		params.assoc_value = sctp_sk(sk)->user_frag;
4952 
4953 	if (put_user(len, optlen))
4954 		return -EFAULT;
4955 	if (len == sizeof(int)) {
4956 		if (copy_to_user(optval, &params.assoc_value, len))
4957 			return -EFAULT;
4958 	} else {
4959 		if (copy_to_user(optval, &params, len))
4960 			return -EFAULT;
4961 	}
4962 
4963 	return 0;
4964 }
4965 
4966 /*
4967  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4968  * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4969  */
4970 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4971 					       char __user *optval, int __user *optlen)
4972 {
4973 	int val;
4974 
4975 	if (len < sizeof(int))
4976 		return -EINVAL;
4977 
4978 	len = sizeof(int);
4979 
4980 	val = sctp_sk(sk)->frag_interleave;
4981 	if (put_user(len, optlen))
4982 		return -EFAULT;
4983 	if (copy_to_user(optval, &val, len))
4984 		return -EFAULT;
4985 
4986 	return 0;
4987 }
4988 
4989 /*
4990  * 7.1.25.  Set or Get the sctp partial delivery point
4991  * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4992  */
4993 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4994 						  char __user *optval,
4995 						  int __user *optlen)
4996 {
4997 	u32 val;
4998 
4999 	if (len < sizeof(u32))
5000 		return -EINVAL;
5001 
5002 	len = sizeof(u32);
5003 
5004 	val = sctp_sk(sk)->pd_point;
5005 	if (put_user(len, optlen))
5006 		return -EFAULT;
5007 	if (copy_to_user(optval, &val, len))
5008 		return -EFAULT;
5009 
5010 	return -ENOTSUPP;
5011 }
5012 
5013 /*
5014  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5015  * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5016  */
5017 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5018 				    char __user *optval,
5019 				    int __user *optlen)
5020 {
5021 	struct sctp_assoc_value params;
5022 	struct sctp_sock *sp;
5023 	struct sctp_association *asoc;
5024 
5025 	if (len == sizeof(int)) {
5026 		printk(KERN_WARNING
5027 		   "SCTP: Use of int in max_burst socket option deprecated\n");
5028 		printk(KERN_WARNING
5029 		   "SCTP: Use struct sctp_assoc_value instead\n");
5030 		params.assoc_id = 0;
5031 	} else if (len >= sizeof(struct sctp_assoc_value)) {
5032 		len = sizeof(struct sctp_assoc_value);
5033 		if (copy_from_user(&params, optval, len))
5034 			return -EFAULT;
5035 	} else
5036 		return -EINVAL;
5037 
5038 	sp = sctp_sk(sk);
5039 
5040 	if (params.assoc_id != 0) {
5041 		asoc = sctp_id2assoc(sk, params.assoc_id);
5042 		if (!asoc)
5043 			return -EINVAL;
5044 		params.assoc_value = asoc->max_burst;
5045 	} else
5046 		params.assoc_value = sp->max_burst;
5047 
5048 	if (len == sizeof(int)) {
5049 		if (copy_to_user(optval, &params.assoc_value, len))
5050 			return -EFAULT;
5051 	} else {
5052 		if (copy_to_user(optval, &params, len))
5053 			return -EFAULT;
5054 	}
5055 
5056 	return 0;
5057 
5058 }
5059 
5060 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5061 				    char __user *optval, int __user *optlen)
5062 {
5063 	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5064 	struct sctp_hmac_algo_param *hmacs;
5065 	__u16 data_len = 0;
5066 	u32 num_idents;
5067 
5068 	if (!sctp_auth_enable)
5069 		return -EACCES;
5070 
5071 	hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5072 	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5073 
5074 	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5075 		return -EINVAL;
5076 
5077 	len = sizeof(struct sctp_hmacalgo) + data_len;
5078 	num_idents = data_len / sizeof(u16);
5079 
5080 	if (put_user(len, optlen))
5081 		return -EFAULT;
5082 	if (put_user(num_idents, &p->shmac_num_idents))
5083 		return -EFAULT;
5084 	if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5085 		return -EFAULT;
5086 	return 0;
5087 }
5088 
5089 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5090 				    char __user *optval, int __user *optlen)
5091 {
5092 	struct sctp_authkeyid val;
5093 	struct sctp_association *asoc;
5094 
5095 	if (!sctp_auth_enable)
5096 		return -EACCES;
5097 
5098 	if (len < sizeof(struct sctp_authkeyid))
5099 		return -EINVAL;
5100 	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5101 		return -EFAULT;
5102 
5103 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5104 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5105 		return -EINVAL;
5106 
5107 	if (asoc)
5108 		val.scact_keynumber = asoc->active_key_id;
5109 	else
5110 		val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5111 
5112 	len = sizeof(struct sctp_authkeyid);
5113 	if (put_user(len, optlen))
5114 		return -EFAULT;
5115 	if (copy_to_user(optval, &val, len))
5116 		return -EFAULT;
5117 
5118 	return 0;
5119 }
5120 
5121 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5122 				    char __user *optval, int __user *optlen)
5123 {
5124 	struct sctp_authchunks __user *p = (void __user *)optval;
5125 	struct sctp_authchunks val;
5126 	struct sctp_association *asoc;
5127 	struct sctp_chunks_param *ch;
5128 	u32    num_chunks = 0;
5129 	char __user *to;
5130 
5131 	if (!sctp_auth_enable)
5132 		return -EACCES;
5133 
5134 	if (len < sizeof(struct sctp_authchunks))
5135 		return -EINVAL;
5136 
5137 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5138 		return -EFAULT;
5139 
5140 	to = p->gauth_chunks;
5141 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5142 	if (!asoc)
5143 		return -EINVAL;
5144 
5145 	ch = asoc->peer.peer_chunks;
5146 	if (!ch)
5147 		goto num;
5148 
5149 	/* See if the user provided enough room for all the data */
5150 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5151 	if (len < num_chunks)
5152 		return -EINVAL;
5153 
5154 	if (copy_to_user(to, ch->chunks, num_chunks))
5155 		return -EFAULT;
5156 num:
5157 	len = sizeof(struct sctp_authchunks) + num_chunks;
5158 	if (put_user(len, optlen)) return -EFAULT;
5159 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5160 		return -EFAULT;
5161 	return 0;
5162 }
5163 
5164 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5165 				    char __user *optval, int __user *optlen)
5166 {
5167 	struct sctp_authchunks __user *p = (void __user *)optval;
5168 	struct sctp_authchunks val;
5169 	struct sctp_association *asoc;
5170 	struct sctp_chunks_param *ch;
5171 	u32    num_chunks = 0;
5172 	char __user *to;
5173 
5174 	if (!sctp_auth_enable)
5175 		return -EACCES;
5176 
5177 	if (len < sizeof(struct sctp_authchunks))
5178 		return -EINVAL;
5179 
5180 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5181 		return -EFAULT;
5182 
5183 	to = p->gauth_chunks;
5184 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5185 	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5186 		return -EINVAL;
5187 
5188 	if (asoc)
5189 		ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5190 	else
5191 		ch = sctp_sk(sk)->ep->auth_chunk_list;
5192 
5193 	if (!ch)
5194 		goto num;
5195 
5196 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5197 	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5198 		return -EINVAL;
5199 
5200 	if (copy_to_user(to, ch->chunks, num_chunks))
5201 		return -EFAULT;
5202 num:
5203 	len = sizeof(struct sctp_authchunks) + num_chunks;
5204 	if (put_user(len, optlen))
5205 		return -EFAULT;
5206 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5207 		return -EFAULT;
5208 
5209 	return 0;
5210 }
5211 
5212 /*
5213  * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5214  * This option gets the current number of associations that are attached
5215  * to a one-to-many style socket.  The option value is an uint32_t.
5216  */
5217 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5218 				    char __user *optval, int __user *optlen)
5219 {
5220 	struct sctp_sock *sp = sctp_sk(sk);
5221 	struct sctp_association *asoc;
5222 	u32 val = 0;
5223 
5224 	if (sctp_style(sk, TCP))
5225 		return -EOPNOTSUPP;
5226 
5227 	if (len < sizeof(u32))
5228 		return -EINVAL;
5229 
5230 	len = sizeof(u32);
5231 
5232 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5233 		val++;
5234 	}
5235 
5236 	if (put_user(len, optlen))
5237 		return -EFAULT;
5238 	if (copy_to_user(optval, &val, len))
5239 		return -EFAULT;
5240 
5241 	return 0;
5242 }
5243 
5244 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5245 				char __user *optval, int __user *optlen)
5246 {
5247 	int retval = 0;
5248 	int len;
5249 
5250 	SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5251 			  sk, optname);
5252 
5253 	/* I can hardly begin to describe how wrong this is.  This is
5254 	 * so broken as to be worse than useless.  The API draft
5255 	 * REALLY is NOT helpful here...  I am not convinced that the
5256 	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5257 	 * are at all well-founded.
5258 	 */
5259 	if (level != SOL_SCTP) {
5260 		struct sctp_af *af = sctp_sk(sk)->pf->af;
5261 
5262 		retval = af->getsockopt(sk, level, optname, optval, optlen);
5263 		return retval;
5264 	}
5265 
5266 	if (get_user(len, optlen))
5267 		return -EFAULT;
5268 
5269 	sctp_lock_sock(sk);
5270 
5271 	switch (optname) {
5272 	case SCTP_STATUS:
5273 		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5274 		break;
5275 	case SCTP_DISABLE_FRAGMENTS:
5276 		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5277 							   optlen);
5278 		break;
5279 	case SCTP_EVENTS:
5280 		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5281 		break;
5282 	case SCTP_AUTOCLOSE:
5283 		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5284 		break;
5285 	case SCTP_SOCKOPT_PEELOFF:
5286 		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5287 		break;
5288 	case SCTP_PEER_ADDR_PARAMS:
5289 		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5290 							  optlen);
5291 		break;
5292 	case SCTP_DELAYED_ACK:
5293 		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5294 							  optlen);
5295 		break;
5296 	case SCTP_INITMSG:
5297 		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5298 		break;
5299 	case SCTP_GET_PEER_ADDRS:
5300 		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5301 						    optlen);
5302 		break;
5303 	case SCTP_GET_LOCAL_ADDRS:
5304 		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5305 						     optlen);
5306 		break;
5307 	case SCTP_SOCKOPT_CONNECTX3:
5308 		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5309 		break;
5310 	case SCTP_DEFAULT_SEND_PARAM:
5311 		retval = sctp_getsockopt_default_send_param(sk, len,
5312 							    optval, optlen);
5313 		break;
5314 	case SCTP_PRIMARY_ADDR:
5315 		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5316 		break;
5317 	case SCTP_NODELAY:
5318 		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5319 		break;
5320 	case SCTP_RTOINFO:
5321 		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5322 		break;
5323 	case SCTP_ASSOCINFO:
5324 		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5325 		break;
5326 	case SCTP_I_WANT_MAPPED_V4_ADDR:
5327 		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5328 		break;
5329 	case SCTP_MAXSEG:
5330 		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5331 		break;
5332 	case SCTP_GET_PEER_ADDR_INFO:
5333 		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5334 							optlen);
5335 		break;
5336 	case SCTP_ADAPTATION_LAYER:
5337 		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5338 							optlen);
5339 		break;
5340 	case SCTP_CONTEXT:
5341 		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5342 		break;
5343 	case SCTP_FRAGMENT_INTERLEAVE:
5344 		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5345 							     optlen);
5346 		break;
5347 	case SCTP_PARTIAL_DELIVERY_POINT:
5348 		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5349 								optlen);
5350 		break;
5351 	case SCTP_MAX_BURST:
5352 		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5353 		break;
5354 	case SCTP_AUTH_KEY:
5355 	case SCTP_AUTH_CHUNK:
5356 	case SCTP_AUTH_DELETE_KEY:
5357 		retval = -EOPNOTSUPP;
5358 		break;
5359 	case SCTP_HMAC_IDENT:
5360 		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5361 		break;
5362 	case SCTP_AUTH_ACTIVE_KEY:
5363 		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5364 		break;
5365 	case SCTP_PEER_AUTH_CHUNKS:
5366 		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5367 							optlen);
5368 		break;
5369 	case SCTP_LOCAL_AUTH_CHUNKS:
5370 		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5371 							optlen);
5372 		break;
5373 	case SCTP_GET_ASSOC_NUMBER:
5374 		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5375 		break;
5376 	default:
5377 		retval = -ENOPROTOOPT;
5378 		break;
5379 	}
5380 
5381 	sctp_release_sock(sk);
5382 	return retval;
5383 }
5384 
5385 static void sctp_hash(struct sock *sk)
5386 {
5387 	/* STUB */
5388 }
5389 
5390 static void sctp_unhash(struct sock *sk)
5391 {
5392 	/* STUB */
5393 }
5394 
5395 /* Check if port is acceptable.  Possibly find first available port.
5396  *
5397  * The port hash table (contained in the 'global' SCTP protocol storage
5398  * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5399  * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5400  * list (the list number is the port number hashed out, so as you
5401  * would expect from a hash function, all the ports in a given list have
5402  * such a number that hashes out to the same list number; you were
5403  * expecting that, right?); so each list has a set of ports, with a
5404  * link to the socket (struct sock) that uses it, the port number and
5405  * a fastreuse flag (FIXME: NPI ipg).
5406  */
5407 static struct sctp_bind_bucket *sctp_bucket_create(
5408 	struct sctp_bind_hashbucket *head, unsigned short snum);
5409 
5410 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5411 {
5412 	struct sctp_bind_hashbucket *head; /* hash list */
5413 	struct sctp_bind_bucket *pp; /* hash list port iterator */
5414 	struct hlist_node *node;
5415 	unsigned short snum;
5416 	int ret;
5417 
5418 	snum = ntohs(addr->v4.sin_port);
5419 
5420 	SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5421 	sctp_local_bh_disable();
5422 
5423 	if (snum == 0) {
5424 		/* Search for an available port. */
5425 		int low, high, remaining, index;
5426 		unsigned int rover;
5427 
5428 		inet_get_local_port_range(&low, &high);
5429 		remaining = (high - low) + 1;
5430 		rover = net_random() % remaining + low;
5431 
5432 		do {
5433 			rover++;
5434 			if ((rover < low) || (rover > high))
5435 				rover = low;
5436 			if (inet_is_reserved_local_port(rover))
5437 				continue;
5438 			index = sctp_phashfn(rover);
5439 			head = &sctp_port_hashtable[index];
5440 			sctp_spin_lock(&head->lock);
5441 			sctp_for_each_hentry(pp, node, &head->chain)
5442 				if (pp->port == rover)
5443 					goto next;
5444 			break;
5445 		next:
5446 			sctp_spin_unlock(&head->lock);
5447 		} while (--remaining > 0);
5448 
5449 		/* Exhausted local port range during search? */
5450 		ret = 1;
5451 		if (remaining <= 0)
5452 			goto fail;
5453 
5454 		/* OK, here is the one we will use.  HEAD (the port
5455 		 * hash table list entry) is non-NULL and we hold it's
5456 		 * mutex.
5457 		 */
5458 		snum = rover;
5459 	} else {
5460 		/* We are given an specific port number; we verify
5461 		 * that it is not being used. If it is used, we will
5462 		 * exahust the search in the hash list corresponding
5463 		 * to the port number (snum) - we detect that with the
5464 		 * port iterator, pp being NULL.
5465 		 */
5466 		head = &sctp_port_hashtable[sctp_phashfn(snum)];
5467 		sctp_spin_lock(&head->lock);
5468 		sctp_for_each_hentry(pp, node, &head->chain) {
5469 			if (pp->port == snum)
5470 				goto pp_found;
5471 		}
5472 	}
5473 	pp = NULL;
5474 	goto pp_not_found;
5475 pp_found:
5476 	if (!hlist_empty(&pp->owner)) {
5477 		/* We had a port hash table hit - there is an
5478 		 * available port (pp != NULL) and it is being
5479 		 * used by other socket (pp->owner not empty); that other
5480 		 * socket is going to be sk2.
5481 		 */
5482 		int reuse = sk->sk_reuse;
5483 		struct sock *sk2;
5484 
5485 		SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5486 		if (pp->fastreuse && sk->sk_reuse &&
5487 			sk->sk_state != SCTP_SS_LISTENING)
5488 			goto success;
5489 
5490 		/* Run through the list of sockets bound to the port
5491 		 * (pp->port) [via the pointers bind_next and
5492 		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5493 		 * we get the endpoint they describe and run through
5494 		 * the endpoint's list of IP (v4 or v6) addresses,
5495 		 * comparing each of the addresses with the address of
5496 		 * the socket sk. If we find a match, then that means
5497 		 * that this port/socket (sk) combination are already
5498 		 * in an endpoint.
5499 		 */
5500 		sk_for_each_bound(sk2, node, &pp->owner) {
5501 			struct sctp_endpoint *ep2;
5502 			ep2 = sctp_sk(sk2)->ep;
5503 
5504 			if (sk == sk2 ||
5505 			    (reuse && sk2->sk_reuse &&
5506 			     sk2->sk_state != SCTP_SS_LISTENING))
5507 				continue;
5508 
5509 			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5510 						 sctp_sk(sk2), sctp_sk(sk))) {
5511 				ret = (long)sk2;
5512 				goto fail_unlock;
5513 			}
5514 		}
5515 		SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5516 	}
5517 pp_not_found:
5518 	/* If there was a hash table miss, create a new port.  */
5519 	ret = 1;
5520 	if (!pp && !(pp = sctp_bucket_create(head, snum)))
5521 		goto fail_unlock;
5522 
5523 	/* In either case (hit or miss), make sure fastreuse is 1 only
5524 	 * if sk->sk_reuse is too (that is, if the caller requested
5525 	 * SO_REUSEADDR on this socket -sk-).
5526 	 */
5527 	if (hlist_empty(&pp->owner)) {
5528 		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5529 			pp->fastreuse = 1;
5530 		else
5531 			pp->fastreuse = 0;
5532 	} else if (pp->fastreuse &&
5533 		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5534 		pp->fastreuse = 0;
5535 
5536 	/* We are set, so fill up all the data in the hash table
5537 	 * entry, tie the socket list information with the rest of the
5538 	 * sockets FIXME: Blurry, NPI (ipg).
5539 	 */
5540 success:
5541 	if (!sctp_sk(sk)->bind_hash) {
5542 		inet_sk(sk)->inet_num = snum;
5543 		sk_add_bind_node(sk, &pp->owner);
5544 		sctp_sk(sk)->bind_hash = pp;
5545 	}
5546 	ret = 0;
5547 
5548 fail_unlock:
5549 	sctp_spin_unlock(&head->lock);
5550 
5551 fail:
5552 	sctp_local_bh_enable();
5553 	return ret;
5554 }
5555 
5556 /* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
5557  * port is requested.
5558  */
5559 static int sctp_get_port(struct sock *sk, unsigned short snum)
5560 {
5561 	long ret;
5562 	union sctp_addr addr;
5563 	struct sctp_af *af = sctp_sk(sk)->pf->af;
5564 
5565 	/* Set up a dummy address struct from the sk. */
5566 	af->from_sk(&addr, sk);
5567 	addr.v4.sin_port = htons(snum);
5568 
5569 	/* Note: sk->sk_num gets filled in if ephemeral port request. */
5570 	ret = sctp_get_port_local(sk, &addr);
5571 
5572 	return (ret ? 1 : 0);
5573 }
5574 
5575 /*
5576  *  Move a socket to LISTENING state.
5577  */
5578 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5579 {
5580 	struct sctp_sock *sp = sctp_sk(sk);
5581 	struct sctp_endpoint *ep = sp->ep;
5582 	struct crypto_hash *tfm = NULL;
5583 
5584 	/* Allocate HMAC for generating cookie. */
5585 	if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5586 		tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5587 		if (IS_ERR(tfm)) {
5588 			if (net_ratelimit()) {
5589 				printk(KERN_INFO
5590 				       "SCTP: failed to load transform for %s: %ld\n",
5591 					sctp_hmac_alg, PTR_ERR(tfm));
5592 			}
5593 			return -ENOSYS;
5594 		}
5595 		sctp_sk(sk)->hmac = tfm;
5596 	}
5597 
5598 	/*
5599 	 * If a bind() or sctp_bindx() is not called prior to a listen()
5600 	 * call that allows new associations to be accepted, the system
5601 	 * picks an ephemeral port and will choose an address set equivalent
5602 	 * to binding with a wildcard address.
5603 	 *
5604 	 * This is not currently spelled out in the SCTP sockets
5605 	 * extensions draft, but follows the practice as seen in TCP
5606 	 * sockets.
5607 	 *
5608 	 */
5609 	sk->sk_state = SCTP_SS_LISTENING;
5610 	if (!ep->base.bind_addr.port) {
5611 		if (sctp_autobind(sk))
5612 			return -EAGAIN;
5613 	} else {
5614 		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5615 			sk->sk_state = SCTP_SS_CLOSED;
5616 			return -EADDRINUSE;
5617 		}
5618 	}
5619 
5620 	sk->sk_max_ack_backlog = backlog;
5621 	sctp_hash_endpoint(ep);
5622 	return 0;
5623 }
5624 
5625 /*
5626  * 4.1.3 / 5.1.3 listen()
5627  *
5628  *   By default, new associations are not accepted for UDP style sockets.
5629  *   An application uses listen() to mark a socket as being able to
5630  *   accept new associations.
5631  *
5632  *   On TCP style sockets, applications use listen() to ready the SCTP
5633  *   endpoint for accepting inbound associations.
5634  *
5635  *   On both types of endpoints a backlog of '0' disables listening.
5636  *
5637  *  Move a socket to LISTENING state.
5638  */
5639 int sctp_inet_listen(struct socket *sock, int backlog)
5640 {
5641 	struct sock *sk = sock->sk;
5642 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5643 	int err = -EINVAL;
5644 
5645 	if (unlikely(backlog < 0))
5646 		return err;
5647 
5648 	sctp_lock_sock(sk);
5649 
5650 	/* Peeled-off sockets are not allowed to listen().  */
5651 	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5652 		goto out;
5653 
5654 	if (sock->state != SS_UNCONNECTED)
5655 		goto out;
5656 
5657 	/* If backlog is zero, disable listening. */
5658 	if (!backlog) {
5659 		if (sctp_sstate(sk, CLOSED))
5660 			goto out;
5661 
5662 		err = 0;
5663 		sctp_unhash_endpoint(ep);
5664 		sk->sk_state = SCTP_SS_CLOSED;
5665 		if (sk->sk_reuse)
5666 			sctp_sk(sk)->bind_hash->fastreuse = 1;
5667 		goto out;
5668 	}
5669 
5670 	/* If we are already listening, just update the backlog */
5671 	if (sctp_sstate(sk, LISTENING))
5672 		sk->sk_max_ack_backlog = backlog;
5673 	else {
5674 		err = sctp_listen_start(sk, backlog);
5675 		if (err)
5676 			goto out;
5677 	}
5678 
5679 	err = 0;
5680 out:
5681 	sctp_release_sock(sk);
5682 	return err;
5683 }
5684 
5685 /*
5686  * This function is done by modeling the current datagram_poll() and the
5687  * tcp_poll().  Note that, based on these implementations, we don't
5688  * lock the socket in this function, even though it seems that,
5689  * ideally, locking or some other mechanisms can be used to ensure
5690  * the integrity of the counters (sndbuf and wmem_alloc) used
5691  * in this place.  We assume that we don't need locks either until proven
5692  * otherwise.
5693  *
5694  * Another thing to note is that we include the Async I/O support
5695  * here, again, by modeling the current TCP/UDP code.  We don't have
5696  * a good way to test with it yet.
5697  */
5698 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5699 {
5700 	struct sock *sk = sock->sk;
5701 	struct sctp_sock *sp = sctp_sk(sk);
5702 	unsigned int mask;
5703 
5704 	poll_wait(file, sk_sleep(sk), wait);
5705 
5706 	/* A TCP-style listening socket becomes readable when the accept queue
5707 	 * is not empty.
5708 	 */
5709 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5710 		return (!list_empty(&sp->ep->asocs)) ?
5711 			(POLLIN | POLLRDNORM) : 0;
5712 
5713 	mask = 0;
5714 
5715 	/* Is there any exceptional events?  */
5716 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5717 		mask |= POLLERR;
5718 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5719 		mask |= POLLRDHUP;
5720 	if (sk->sk_shutdown == SHUTDOWN_MASK)
5721 		mask |= POLLHUP;
5722 
5723 	/* Is it readable?  Reconsider this code with TCP-style support.  */
5724 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
5725 	    (sk->sk_shutdown & RCV_SHUTDOWN))
5726 		mask |= POLLIN | POLLRDNORM;
5727 
5728 	/* The association is either gone or not ready.  */
5729 	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5730 		return mask;
5731 
5732 	/* Is it writable?  */
5733 	if (sctp_writeable(sk)) {
5734 		mask |= POLLOUT | POLLWRNORM;
5735 	} else {
5736 		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5737 		/*
5738 		 * Since the socket is not locked, the buffer
5739 		 * might be made available after the writeable check and
5740 		 * before the bit is set.  This could cause a lost I/O
5741 		 * signal.  tcp_poll() has a race breaker for this race
5742 		 * condition.  Based on their implementation, we put
5743 		 * in the following code to cover it as well.
5744 		 */
5745 		if (sctp_writeable(sk))
5746 			mask |= POLLOUT | POLLWRNORM;
5747 	}
5748 	return mask;
5749 }
5750 
5751 /********************************************************************
5752  * 2nd Level Abstractions
5753  ********************************************************************/
5754 
5755 static struct sctp_bind_bucket *sctp_bucket_create(
5756 	struct sctp_bind_hashbucket *head, unsigned short snum)
5757 {
5758 	struct sctp_bind_bucket *pp;
5759 
5760 	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5761 	if (pp) {
5762 		SCTP_DBG_OBJCNT_INC(bind_bucket);
5763 		pp->port = snum;
5764 		pp->fastreuse = 0;
5765 		INIT_HLIST_HEAD(&pp->owner);
5766 		hlist_add_head(&pp->node, &head->chain);
5767 	}
5768 	return pp;
5769 }
5770 
5771 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5772 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5773 {
5774 	if (pp && hlist_empty(&pp->owner)) {
5775 		__hlist_del(&pp->node);
5776 		kmem_cache_free(sctp_bucket_cachep, pp);
5777 		SCTP_DBG_OBJCNT_DEC(bind_bucket);
5778 	}
5779 }
5780 
5781 /* Release this socket's reference to a local port.  */
5782 static inline void __sctp_put_port(struct sock *sk)
5783 {
5784 	struct sctp_bind_hashbucket *head =
5785 		&sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5786 	struct sctp_bind_bucket *pp;
5787 
5788 	sctp_spin_lock(&head->lock);
5789 	pp = sctp_sk(sk)->bind_hash;
5790 	__sk_del_bind_node(sk);
5791 	sctp_sk(sk)->bind_hash = NULL;
5792 	inet_sk(sk)->inet_num = 0;
5793 	sctp_bucket_destroy(pp);
5794 	sctp_spin_unlock(&head->lock);
5795 }
5796 
5797 void sctp_put_port(struct sock *sk)
5798 {
5799 	sctp_local_bh_disable();
5800 	__sctp_put_port(sk);
5801 	sctp_local_bh_enable();
5802 }
5803 
5804 /*
5805  * The system picks an ephemeral port and choose an address set equivalent
5806  * to binding with a wildcard address.
5807  * One of those addresses will be the primary address for the association.
5808  * This automatically enables the multihoming capability of SCTP.
5809  */
5810 static int sctp_autobind(struct sock *sk)
5811 {
5812 	union sctp_addr autoaddr;
5813 	struct sctp_af *af;
5814 	__be16 port;
5815 
5816 	/* Initialize a local sockaddr structure to INADDR_ANY. */
5817 	af = sctp_sk(sk)->pf->af;
5818 
5819 	port = htons(inet_sk(sk)->inet_num);
5820 	af->inaddr_any(&autoaddr, port);
5821 
5822 	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5823 }
5824 
5825 /* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
5826  *
5827  * From RFC 2292
5828  * 4.2 The cmsghdr Structure *
5829  *
5830  * When ancillary data is sent or received, any number of ancillary data
5831  * objects can be specified by the msg_control and msg_controllen members of
5832  * the msghdr structure, because each object is preceded by
5833  * a cmsghdr structure defining the object's length (the cmsg_len member).
5834  * Historically Berkeley-derived implementations have passed only one object
5835  * at a time, but this API allows multiple objects to be
5836  * passed in a single call to sendmsg() or recvmsg(). The following example
5837  * shows two ancillary data objects in a control buffer.
5838  *
5839  *   |<--------------------------- msg_controllen -------------------------->|
5840  *   |                                                                       |
5841  *
5842  *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
5843  *
5844  *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5845  *   |                                   |                                   |
5846  *
5847  *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
5848  *
5849  *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
5850  *   |                                |  |                                |  |
5851  *
5852  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5853  *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
5854  *
5855  *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
5856  *
5857  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5858  *    ^
5859  *    |
5860  *
5861  * msg_control
5862  * points here
5863  */
5864 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5865 				  sctp_cmsgs_t *cmsgs)
5866 {
5867 	struct cmsghdr *cmsg;
5868 	struct msghdr *my_msg = (struct msghdr *)msg;
5869 
5870 	for (cmsg = CMSG_FIRSTHDR(msg);
5871 	     cmsg != NULL;
5872 	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5873 		if (!CMSG_OK(my_msg, cmsg))
5874 			return -EINVAL;
5875 
5876 		/* Should we parse this header or ignore?  */
5877 		if (cmsg->cmsg_level != IPPROTO_SCTP)
5878 			continue;
5879 
5880 		/* Strictly check lengths following example in SCM code.  */
5881 		switch (cmsg->cmsg_type) {
5882 		case SCTP_INIT:
5883 			/* SCTP Socket API Extension
5884 			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5885 			 *
5886 			 * This cmsghdr structure provides information for
5887 			 * initializing new SCTP associations with sendmsg().
5888 			 * The SCTP_INITMSG socket option uses this same data
5889 			 * structure.  This structure is not used for
5890 			 * recvmsg().
5891 			 *
5892 			 * cmsg_level    cmsg_type      cmsg_data[]
5893 			 * ------------  ------------   ----------------------
5894 			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
5895 			 */
5896 			if (cmsg->cmsg_len !=
5897 			    CMSG_LEN(sizeof(struct sctp_initmsg)))
5898 				return -EINVAL;
5899 			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5900 			break;
5901 
5902 		case SCTP_SNDRCV:
5903 			/* SCTP Socket API Extension
5904 			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5905 			 *
5906 			 * This cmsghdr structure specifies SCTP options for
5907 			 * sendmsg() and describes SCTP header information
5908 			 * about a received message through recvmsg().
5909 			 *
5910 			 * cmsg_level    cmsg_type      cmsg_data[]
5911 			 * ------------  ------------   ----------------------
5912 			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
5913 			 */
5914 			if (cmsg->cmsg_len !=
5915 			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5916 				return -EINVAL;
5917 
5918 			cmsgs->info =
5919 				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5920 
5921 			/* Minimally, validate the sinfo_flags. */
5922 			if (cmsgs->info->sinfo_flags &
5923 			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5924 			      SCTP_ABORT | SCTP_EOF))
5925 				return -EINVAL;
5926 			break;
5927 
5928 		default:
5929 			return -EINVAL;
5930 		}
5931 	}
5932 	return 0;
5933 }
5934 
5935 /*
5936  * Wait for a packet..
5937  * Note: This function is the same function as in core/datagram.c
5938  * with a few modifications to make lksctp work.
5939  */
5940 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5941 {
5942 	int error;
5943 	DEFINE_WAIT(wait);
5944 
5945 	prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
5946 
5947 	/* Socket errors? */
5948 	error = sock_error(sk);
5949 	if (error)
5950 		goto out;
5951 
5952 	if (!skb_queue_empty(&sk->sk_receive_queue))
5953 		goto ready;
5954 
5955 	/* Socket shut down?  */
5956 	if (sk->sk_shutdown & RCV_SHUTDOWN)
5957 		goto out;
5958 
5959 	/* Sequenced packets can come disconnected.  If so we report the
5960 	 * problem.
5961 	 */
5962 	error = -ENOTCONN;
5963 
5964 	/* Is there a good reason to think that we may receive some data?  */
5965 	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5966 		goto out;
5967 
5968 	/* Handle signals.  */
5969 	if (signal_pending(current))
5970 		goto interrupted;
5971 
5972 	/* Let another process have a go.  Since we are going to sleep
5973 	 * anyway.  Note: This may cause odd behaviors if the message
5974 	 * does not fit in the user's buffer, but this seems to be the
5975 	 * only way to honor MSG_DONTWAIT realistically.
5976 	 */
5977 	sctp_release_sock(sk);
5978 	*timeo_p = schedule_timeout(*timeo_p);
5979 	sctp_lock_sock(sk);
5980 
5981 ready:
5982 	finish_wait(sk_sleep(sk), &wait);
5983 	return 0;
5984 
5985 interrupted:
5986 	error = sock_intr_errno(*timeo_p);
5987 
5988 out:
5989 	finish_wait(sk_sleep(sk), &wait);
5990 	*err = error;
5991 	return error;
5992 }
5993 
5994 /* Receive a datagram.
5995  * Note: This is pretty much the same routine as in core/datagram.c
5996  * with a few changes to make lksctp work.
5997  */
5998 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5999 					      int noblock, int *err)
6000 {
6001 	int error;
6002 	struct sk_buff *skb;
6003 	long timeo;
6004 
6005 	timeo = sock_rcvtimeo(sk, noblock);
6006 
6007 	SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6008 			  timeo, MAX_SCHEDULE_TIMEOUT);
6009 
6010 	do {
6011 		/* Again only user level code calls this function,
6012 		 * so nothing interrupt level
6013 		 * will suddenly eat the receive_queue.
6014 		 *
6015 		 *  Look at current nfs client by the way...
6016 		 *  However, this function was corrent in any case. 8)
6017 		 */
6018 		if (flags & MSG_PEEK) {
6019 			spin_lock_bh(&sk->sk_receive_queue.lock);
6020 			skb = skb_peek(&sk->sk_receive_queue);
6021 			if (skb)
6022 				atomic_inc(&skb->users);
6023 			spin_unlock_bh(&sk->sk_receive_queue.lock);
6024 		} else {
6025 			skb = skb_dequeue(&sk->sk_receive_queue);
6026 		}
6027 
6028 		if (skb)
6029 			return skb;
6030 
6031 		/* Caller is allowed not to check sk->sk_err before calling. */
6032 		error = sock_error(sk);
6033 		if (error)
6034 			goto no_packet;
6035 
6036 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6037 			break;
6038 
6039 		/* User doesn't want to wait.  */
6040 		error = -EAGAIN;
6041 		if (!timeo)
6042 			goto no_packet;
6043 	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6044 
6045 	return NULL;
6046 
6047 no_packet:
6048 	*err = error;
6049 	return NULL;
6050 }
6051 
6052 /* If sndbuf has changed, wake up per association sndbuf waiters.  */
6053 static void __sctp_write_space(struct sctp_association *asoc)
6054 {
6055 	struct sock *sk = asoc->base.sk;
6056 	struct socket *sock = sk->sk_socket;
6057 
6058 	if ((sctp_wspace(asoc) > 0) && sock) {
6059 		if (waitqueue_active(&asoc->wait))
6060 			wake_up_interruptible(&asoc->wait);
6061 
6062 		if (sctp_writeable(sk)) {
6063 			if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
6064 				wake_up_interruptible(sk_sleep(sk));
6065 
6066 			/* Note that we try to include the Async I/O support
6067 			 * here by modeling from the current TCP/UDP code.
6068 			 * We have not tested with it yet.
6069 			 */
6070 			if (sock->wq->fasync_list &&
6071 			    !(sk->sk_shutdown & SEND_SHUTDOWN))
6072 				sock_wake_async(sock,
6073 						SOCK_WAKE_SPACE, POLL_OUT);
6074 		}
6075 	}
6076 }
6077 
6078 /* Do accounting for the sndbuf space.
6079  * Decrement the used sndbuf space of the corresponding association by the
6080  * data size which was just transmitted(freed).
6081  */
6082 static void sctp_wfree(struct sk_buff *skb)
6083 {
6084 	struct sctp_association *asoc;
6085 	struct sctp_chunk *chunk;
6086 	struct sock *sk;
6087 
6088 	/* Get the saved chunk pointer.  */
6089 	chunk = *((struct sctp_chunk **)(skb->cb));
6090 	asoc = chunk->asoc;
6091 	sk = asoc->base.sk;
6092 	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6093 				sizeof(struct sk_buff) +
6094 				sizeof(struct sctp_chunk);
6095 
6096 	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6097 
6098 	/*
6099 	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6100 	 */
6101 	sk->sk_wmem_queued   -= skb->truesize;
6102 	sk_mem_uncharge(sk, skb->truesize);
6103 
6104 	sock_wfree(skb);
6105 	__sctp_write_space(asoc);
6106 
6107 	sctp_association_put(asoc);
6108 }
6109 
6110 /* Do accounting for the receive space on the socket.
6111  * Accounting for the association is done in ulpevent.c
6112  * We set this as a destructor for the cloned data skbs so that
6113  * accounting is done at the correct time.
6114  */
6115 void sctp_sock_rfree(struct sk_buff *skb)
6116 {
6117 	struct sock *sk = skb->sk;
6118 	struct sctp_ulpevent *event = sctp_skb2event(skb);
6119 
6120 	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6121 
6122 	/*
6123 	 * Mimic the behavior of sock_rfree
6124 	 */
6125 	sk_mem_uncharge(sk, event->rmem_len);
6126 }
6127 
6128 
6129 /* Helper function to wait for space in the sndbuf.  */
6130 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6131 				size_t msg_len)
6132 {
6133 	struct sock *sk = asoc->base.sk;
6134 	int err = 0;
6135 	long current_timeo = *timeo_p;
6136 	DEFINE_WAIT(wait);
6137 
6138 	SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6139 			  asoc, (long)(*timeo_p), msg_len);
6140 
6141 	/* Increment the association's refcnt.  */
6142 	sctp_association_hold(asoc);
6143 
6144 	/* Wait on the association specific sndbuf space. */
6145 	for (;;) {
6146 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6147 					  TASK_INTERRUPTIBLE);
6148 		if (!*timeo_p)
6149 			goto do_nonblock;
6150 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6151 		    asoc->base.dead)
6152 			goto do_error;
6153 		if (signal_pending(current))
6154 			goto do_interrupted;
6155 		if (msg_len <= sctp_wspace(asoc))
6156 			break;
6157 
6158 		/* Let another process have a go.  Since we are going
6159 		 * to sleep anyway.
6160 		 */
6161 		sctp_release_sock(sk);
6162 		current_timeo = schedule_timeout(current_timeo);
6163 		BUG_ON(sk != asoc->base.sk);
6164 		sctp_lock_sock(sk);
6165 
6166 		*timeo_p = current_timeo;
6167 	}
6168 
6169 out:
6170 	finish_wait(&asoc->wait, &wait);
6171 
6172 	/* Release the association's refcnt.  */
6173 	sctp_association_put(asoc);
6174 
6175 	return err;
6176 
6177 do_error:
6178 	err = -EPIPE;
6179 	goto out;
6180 
6181 do_interrupted:
6182 	err = sock_intr_errno(*timeo_p);
6183 	goto out;
6184 
6185 do_nonblock:
6186 	err = -EAGAIN;
6187 	goto out;
6188 }
6189 
6190 void sctp_data_ready(struct sock *sk, int len)
6191 {
6192 	struct socket_wq *wq;
6193 
6194 	rcu_read_lock();
6195 	wq = rcu_dereference(sk->sk_wq);
6196 	if (wq_has_sleeper(wq))
6197 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6198 						POLLRDNORM | POLLRDBAND);
6199 	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6200 	rcu_read_unlock();
6201 }
6202 
6203 /* If socket sndbuf has changed, wake up all per association waiters.  */
6204 void sctp_write_space(struct sock *sk)
6205 {
6206 	struct sctp_association *asoc;
6207 
6208 	/* Wake up the tasks in each wait queue.  */
6209 	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6210 		__sctp_write_space(asoc);
6211 	}
6212 }
6213 
6214 /* Is there any sndbuf space available on the socket?
6215  *
6216  * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6217  * associations on the same socket.  For a UDP-style socket with
6218  * multiple associations, it is possible for it to be "unwriteable"
6219  * prematurely.  I assume that this is acceptable because
6220  * a premature "unwriteable" is better than an accidental "writeable" which
6221  * would cause an unwanted block under certain circumstances.  For the 1-1
6222  * UDP-style sockets or TCP-style sockets, this code should work.
6223  *  - Daisy
6224  */
6225 static int sctp_writeable(struct sock *sk)
6226 {
6227 	int amt = 0;
6228 
6229 	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6230 	if (amt < 0)
6231 		amt = 0;
6232 	return amt;
6233 }
6234 
6235 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6236  * returns immediately with EINPROGRESS.
6237  */
6238 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6239 {
6240 	struct sock *sk = asoc->base.sk;
6241 	int err = 0;
6242 	long current_timeo = *timeo_p;
6243 	DEFINE_WAIT(wait);
6244 
6245 	SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6246 			  (long)(*timeo_p));
6247 
6248 	/* Increment the association's refcnt.  */
6249 	sctp_association_hold(asoc);
6250 
6251 	for (;;) {
6252 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6253 					  TASK_INTERRUPTIBLE);
6254 		if (!*timeo_p)
6255 			goto do_nonblock;
6256 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6257 			break;
6258 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6259 		    asoc->base.dead)
6260 			goto do_error;
6261 		if (signal_pending(current))
6262 			goto do_interrupted;
6263 
6264 		if (sctp_state(asoc, ESTABLISHED))
6265 			break;
6266 
6267 		/* Let another process have a go.  Since we are going
6268 		 * to sleep anyway.
6269 		 */
6270 		sctp_release_sock(sk);
6271 		current_timeo = schedule_timeout(current_timeo);
6272 		sctp_lock_sock(sk);
6273 
6274 		*timeo_p = current_timeo;
6275 	}
6276 
6277 out:
6278 	finish_wait(&asoc->wait, &wait);
6279 
6280 	/* Release the association's refcnt.  */
6281 	sctp_association_put(asoc);
6282 
6283 	return err;
6284 
6285 do_error:
6286 	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6287 		err = -ETIMEDOUT;
6288 	else
6289 		err = -ECONNREFUSED;
6290 	goto out;
6291 
6292 do_interrupted:
6293 	err = sock_intr_errno(*timeo_p);
6294 	goto out;
6295 
6296 do_nonblock:
6297 	err = -EINPROGRESS;
6298 	goto out;
6299 }
6300 
6301 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6302 {
6303 	struct sctp_endpoint *ep;
6304 	int err = 0;
6305 	DEFINE_WAIT(wait);
6306 
6307 	ep = sctp_sk(sk)->ep;
6308 
6309 
6310 	for (;;) {
6311 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6312 					  TASK_INTERRUPTIBLE);
6313 
6314 		if (list_empty(&ep->asocs)) {
6315 			sctp_release_sock(sk);
6316 			timeo = schedule_timeout(timeo);
6317 			sctp_lock_sock(sk);
6318 		}
6319 
6320 		err = -EINVAL;
6321 		if (!sctp_sstate(sk, LISTENING))
6322 			break;
6323 
6324 		err = 0;
6325 		if (!list_empty(&ep->asocs))
6326 			break;
6327 
6328 		err = sock_intr_errno(timeo);
6329 		if (signal_pending(current))
6330 			break;
6331 
6332 		err = -EAGAIN;
6333 		if (!timeo)
6334 			break;
6335 	}
6336 
6337 	finish_wait(sk_sleep(sk), &wait);
6338 
6339 	return err;
6340 }
6341 
6342 static void sctp_wait_for_close(struct sock *sk, long timeout)
6343 {
6344 	DEFINE_WAIT(wait);
6345 
6346 	do {
6347 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6348 		if (list_empty(&sctp_sk(sk)->ep->asocs))
6349 			break;
6350 		sctp_release_sock(sk);
6351 		timeout = schedule_timeout(timeout);
6352 		sctp_lock_sock(sk);
6353 	} while (!signal_pending(current) && timeout);
6354 
6355 	finish_wait(sk_sleep(sk), &wait);
6356 }
6357 
6358 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6359 {
6360 	struct sk_buff *frag;
6361 
6362 	if (!skb->data_len)
6363 		goto done;
6364 
6365 	/* Don't forget the fragments. */
6366 	skb_walk_frags(skb, frag)
6367 		sctp_skb_set_owner_r_frag(frag, sk);
6368 
6369 done:
6370 	sctp_skb_set_owner_r(skb, sk);
6371 }
6372 
6373 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6374 		    struct sctp_association *asoc)
6375 {
6376 	struct inet_sock *inet = inet_sk(sk);
6377 	struct inet_sock *newinet;
6378 
6379 	newsk->sk_type = sk->sk_type;
6380 	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6381 	newsk->sk_flags = sk->sk_flags;
6382 	newsk->sk_no_check = sk->sk_no_check;
6383 	newsk->sk_reuse = sk->sk_reuse;
6384 
6385 	newsk->sk_shutdown = sk->sk_shutdown;
6386 	newsk->sk_destruct = inet_sock_destruct;
6387 	newsk->sk_family = sk->sk_family;
6388 	newsk->sk_protocol = IPPROTO_SCTP;
6389 	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6390 	newsk->sk_sndbuf = sk->sk_sndbuf;
6391 	newsk->sk_rcvbuf = sk->sk_rcvbuf;
6392 	newsk->sk_lingertime = sk->sk_lingertime;
6393 	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6394 	newsk->sk_sndtimeo = sk->sk_sndtimeo;
6395 
6396 	newinet = inet_sk(newsk);
6397 
6398 	/* Initialize sk's sport, dport, rcv_saddr and daddr for
6399 	 * getsockname() and getpeername()
6400 	 */
6401 	newinet->inet_sport = inet->inet_sport;
6402 	newinet->inet_saddr = inet->inet_saddr;
6403 	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6404 	newinet->inet_dport = htons(asoc->peer.port);
6405 	newinet->pmtudisc = inet->pmtudisc;
6406 	newinet->inet_id = asoc->next_tsn ^ jiffies;
6407 
6408 	newinet->uc_ttl = inet->uc_ttl;
6409 	newinet->mc_loop = 1;
6410 	newinet->mc_ttl = 1;
6411 	newinet->mc_index = 0;
6412 	newinet->mc_list = NULL;
6413 }
6414 
6415 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6416  * and its messages to the newsk.
6417  */
6418 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6419 			      struct sctp_association *assoc,
6420 			      sctp_socket_type_t type)
6421 {
6422 	struct sctp_sock *oldsp = sctp_sk(oldsk);
6423 	struct sctp_sock *newsp = sctp_sk(newsk);
6424 	struct sctp_bind_bucket *pp; /* hash list port iterator */
6425 	struct sctp_endpoint *newep = newsp->ep;
6426 	struct sk_buff *skb, *tmp;
6427 	struct sctp_ulpevent *event;
6428 	struct sctp_bind_hashbucket *head;
6429 
6430 	/* Migrate socket buffer sizes and all the socket level options to the
6431 	 * new socket.
6432 	 */
6433 	newsk->sk_sndbuf = oldsk->sk_sndbuf;
6434 	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6435 	/* Brute force copy old sctp opt. */
6436 	inet_sk_copy_descendant(newsk, oldsk);
6437 
6438 	/* Restore the ep value that was overwritten with the above structure
6439 	 * copy.
6440 	 */
6441 	newsp->ep = newep;
6442 	newsp->hmac = NULL;
6443 
6444 	/* Hook this new socket in to the bind_hash list. */
6445 	head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6446 	sctp_local_bh_disable();
6447 	sctp_spin_lock(&head->lock);
6448 	pp = sctp_sk(oldsk)->bind_hash;
6449 	sk_add_bind_node(newsk, &pp->owner);
6450 	sctp_sk(newsk)->bind_hash = pp;
6451 	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6452 	sctp_spin_unlock(&head->lock);
6453 	sctp_local_bh_enable();
6454 
6455 	/* Copy the bind_addr list from the original endpoint to the new
6456 	 * endpoint so that we can handle restarts properly
6457 	 */
6458 	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6459 				&oldsp->ep->base.bind_addr, GFP_KERNEL);
6460 
6461 	/* Move any messages in the old socket's receive queue that are for the
6462 	 * peeled off association to the new socket's receive queue.
6463 	 */
6464 	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6465 		event = sctp_skb2event(skb);
6466 		if (event->asoc == assoc) {
6467 			__skb_unlink(skb, &oldsk->sk_receive_queue);
6468 			__skb_queue_tail(&newsk->sk_receive_queue, skb);
6469 			sctp_skb_set_owner_r_frag(skb, newsk);
6470 		}
6471 	}
6472 
6473 	/* Clean up any messages pending delivery due to partial
6474 	 * delivery.   Three cases:
6475 	 * 1) No partial deliver;  no work.
6476 	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6477 	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6478 	 */
6479 	skb_queue_head_init(&newsp->pd_lobby);
6480 	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6481 
6482 	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6483 		struct sk_buff_head *queue;
6484 
6485 		/* Decide which queue to move pd_lobby skbs to. */
6486 		if (assoc->ulpq.pd_mode) {
6487 			queue = &newsp->pd_lobby;
6488 		} else
6489 			queue = &newsk->sk_receive_queue;
6490 
6491 		/* Walk through the pd_lobby, looking for skbs that
6492 		 * need moved to the new socket.
6493 		 */
6494 		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6495 			event = sctp_skb2event(skb);
6496 			if (event->asoc == assoc) {
6497 				__skb_unlink(skb, &oldsp->pd_lobby);
6498 				__skb_queue_tail(queue, skb);
6499 				sctp_skb_set_owner_r_frag(skb, newsk);
6500 			}
6501 		}
6502 
6503 		/* Clear up any skbs waiting for the partial
6504 		 * delivery to finish.
6505 		 */
6506 		if (assoc->ulpq.pd_mode)
6507 			sctp_clear_pd(oldsk, NULL);
6508 
6509 	}
6510 
6511 	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6512 		sctp_skb_set_owner_r_frag(skb, newsk);
6513 
6514 	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6515 		sctp_skb_set_owner_r_frag(skb, newsk);
6516 
6517 	/* Set the type of socket to indicate that it is peeled off from the
6518 	 * original UDP-style socket or created with the accept() call on a
6519 	 * TCP-style socket..
6520 	 */
6521 	newsp->type = type;
6522 
6523 	/* Mark the new socket "in-use" by the user so that any packets
6524 	 * that may arrive on the association after we've moved it are
6525 	 * queued to the backlog.  This prevents a potential race between
6526 	 * backlog processing on the old socket and new-packet processing
6527 	 * on the new socket.
6528 	 *
6529 	 * The caller has just allocated newsk so we can guarantee that other
6530 	 * paths won't try to lock it and then oldsk.
6531 	 */
6532 	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6533 	sctp_assoc_migrate(assoc, newsk);
6534 
6535 	/* If the association on the newsk is already closed before accept()
6536 	 * is called, set RCV_SHUTDOWN flag.
6537 	 */
6538 	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6539 		newsk->sk_shutdown |= RCV_SHUTDOWN;
6540 
6541 	newsk->sk_state = SCTP_SS_ESTABLISHED;
6542 	sctp_release_sock(newsk);
6543 }
6544 
6545 
6546 /* This proto struct describes the ULP interface for SCTP.  */
6547 struct proto sctp_prot = {
6548 	.name        =	"SCTP",
6549 	.owner       =	THIS_MODULE,
6550 	.close       =	sctp_close,
6551 	.connect     =	sctp_connect,
6552 	.disconnect  =	sctp_disconnect,
6553 	.accept      =	sctp_accept,
6554 	.ioctl       =	sctp_ioctl,
6555 	.init        =	sctp_init_sock,
6556 	.destroy     =	sctp_destroy_sock,
6557 	.shutdown    =	sctp_shutdown,
6558 	.setsockopt  =	sctp_setsockopt,
6559 	.getsockopt  =	sctp_getsockopt,
6560 	.sendmsg     =	sctp_sendmsg,
6561 	.recvmsg     =	sctp_recvmsg,
6562 	.bind        =	sctp_bind,
6563 	.backlog_rcv =	sctp_backlog_rcv,
6564 	.hash        =	sctp_hash,
6565 	.unhash      =	sctp_unhash,
6566 	.get_port    =	sctp_get_port,
6567 	.obj_size    =  sizeof(struct sctp_sock),
6568 	.sysctl_mem  =  sysctl_sctp_mem,
6569 	.sysctl_rmem =  sysctl_sctp_rmem,
6570 	.sysctl_wmem =  sysctl_sctp_wmem,
6571 	.memory_pressure = &sctp_memory_pressure,
6572 	.enter_memory_pressure = sctp_enter_memory_pressure,
6573 	.memory_allocated = &sctp_memory_allocated,
6574 	.sockets_allocated = &sctp_sockets_allocated,
6575 };
6576 
6577 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6578 
6579 struct proto sctpv6_prot = {
6580 	.name		= "SCTPv6",
6581 	.owner		= THIS_MODULE,
6582 	.close		= sctp_close,
6583 	.connect	= sctp_connect,
6584 	.disconnect	= sctp_disconnect,
6585 	.accept		= sctp_accept,
6586 	.ioctl		= sctp_ioctl,
6587 	.init		= sctp_init_sock,
6588 	.destroy	= sctp_destroy_sock,
6589 	.shutdown	= sctp_shutdown,
6590 	.setsockopt	= sctp_setsockopt,
6591 	.getsockopt	= sctp_getsockopt,
6592 	.sendmsg	= sctp_sendmsg,
6593 	.recvmsg	= sctp_recvmsg,
6594 	.bind		= sctp_bind,
6595 	.backlog_rcv	= sctp_backlog_rcv,
6596 	.hash		= sctp_hash,
6597 	.unhash		= sctp_unhash,
6598 	.get_port	= sctp_get_port,
6599 	.obj_size	= sizeof(struct sctp6_sock),
6600 	.sysctl_mem	= sysctl_sctp_mem,
6601 	.sysctl_rmem	= sysctl_sctp_rmem,
6602 	.sysctl_wmem	= sysctl_sctp_wmem,
6603 	.memory_pressure = &sctp_memory_pressure,
6604 	.enter_memory_pressure = sctp_enter_memory_pressure,
6605 	.memory_allocated = &sctp_memory_allocated,
6606 	.sockets_allocated = &sctp_sockets_allocated,
6607 };
6608 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
6609