xref: /openbmc/linux/net/sctp/socket.c (revision 206a81c1)
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, see
32  * <http://www.gnu.org/licenses/>.
33  *
34  * Please send any bug reports or fixes you make to the
35  * email address(es):
36  *    lksctp developers <linux-sctp@vger.kernel.org>
37  *
38  * Written or modified by:
39  *    La Monte H.P. Yarroll <piggy@acm.org>
40  *    Narasimha Budihal     <narsi@refcode.org>
41  *    Karl Knutson          <karl@athena.chicago.il.us>
42  *    Jon Grimm             <jgrimm@us.ibm.com>
43  *    Xingang Guo           <xingang.guo@intel.com>
44  *    Daisy Chang           <daisyc@us.ibm.com>
45  *    Sridhar Samudrala     <samudrala@us.ibm.com>
46  *    Inaky Perez-Gonzalez  <inaky.gonzalez@intel.com>
47  *    Ardelle Fan	    <ardelle.fan@intel.com>
48  *    Ryan Layer	    <rmlayer@us.ibm.com>
49  *    Anup Pemmaiah         <pemmaiah@cc.usu.edu>
50  *    Kevin Gao             <kevin.gao@intel.com>
51  */
52 
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
54 
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
59 #include <linux/ip.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
68 
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
75 
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
81 
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 				size_t msg_len);
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 					union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 			    struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 			      struct sctp_association *, sctp_socket_type_t);
104 
105 extern struct kmem_cache *sctp_bucket_cachep;
106 extern long sysctl_sctp_mem[3];
107 extern int sysctl_sctp_rmem[3];
108 extern int sysctl_sctp_wmem[3];
109 
110 static int sctp_memory_pressure;
111 static atomic_long_t sctp_memory_allocated;
112 struct percpu_counter sctp_sockets_allocated;
113 
114 static void sctp_enter_memory_pressure(struct sock *sk)
115 {
116 	sctp_memory_pressure = 1;
117 }
118 
119 
120 /* Get the sndbuf space available at the time on the association.  */
121 static inline int sctp_wspace(struct sctp_association *asoc)
122 {
123 	int amt;
124 
125 	if (asoc->ep->sndbuf_policy)
126 		amt = asoc->sndbuf_used;
127 	else
128 		amt = sk_wmem_alloc_get(asoc->base.sk);
129 
130 	if (amt >= asoc->base.sk->sk_sndbuf) {
131 		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 			amt = 0;
133 		else {
134 			amt = sk_stream_wspace(asoc->base.sk);
135 			if (amt < 0)
136 				amt = 0;
137 		}
138 	} else {
139 		amt = asoc->base.sk->sk_sndbuf - amt;
140 	}
141 	return amt;
142 }
143 
144 /* Increment the used sndbuf space count of the corresponding association by
145  * the size of the outgoing data chunk.
146  * Also, set the skb destructor for sndbuf accounting later.
147  *
148  * Since it is always 1-1 between chunk and skb, and also a new skb is always
149  * allocated for chunk bundling in sctp_packet_transmit(), we can use the
150  * destructor in the data chunk skb for the purpose of the sndbuf space
151  * tracking.
152  */
153 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
154 {
155 	struct sctp_association *asoc = chunk->asoc;
156 	struct sock *sk = asoc->base.sk;
157 
158 	/* The sndbuf space is tracked per association.  */
159 	sctp_association_hold(asoc);
160 
161 	skb_set_owner_w(chunk->skb, sk);
162 
163 	chunk->skb->destructor = sctp_wfree;
164 	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
165 	*((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
166 
167 	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
168 				sizeof(struct sk_buff) +
169 				sizeof(struct sctp_chunk);
170 
171 	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
172 	sk->sk_wmem_queued += chunk->skb->truesize;
173 	sk_mem_charge(sk, chunk->skb->truesize);
174 }
175 
176 /* Verify that this is a valid address. */
177 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
178 				   int len)
179 {
180 	struct sctp_af *af;
181 
182 	/* Verify basic sockaddr. */
183 	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
184 	if (!af)
185 		return -EINVAL;
186 
187 	/* Is this a valid SCTP address?  */
188 	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
189 		return -EINVAL;
190 
191 	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 		return -EINVAL;
193 
194 	return 0;
195 }
196 
197 /* Look up the association by its id.  If this is not a UDP-style
198  * socket, the ID field is always ignored.
199  */
200 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
201 {
202 	struct sctp_association *asoc = NULL;
203 
204 	/* If this is not a UDP-style socket, assoc id should be ignored. */
205 	if (!sctp_style(sk, UDP)) {
206 		/* Return NULL if the socket state is not ESTABLISHED. It
207 		 * could be a TCP-style listening socket or a socket which
208 		 * hasn't yet called connect() to establish an association.
209 		 */
210 		if (!sctp_sstate(sk, ESTABLISHED))
211 			return NULL;
212 
213 		/* Get the first and the only association from the list. */
214 		if (!list_empty(&sctp_sk(sk)->ep->asocs))
215 			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
216 					  struct sctp_association, asocs);
217 		return asoc;
218 	}
219 
220 	/* Otherwise this is a UDP-style socket. */
221 	if (!id || (id == (sctp_assoc_t)-1))
222 		return NULL;
223 
224 	spin_lock_bh(&sctp_assocs_id_lock);
225 	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
226 	spin_unlock_bh(&sctp_assocs_id_lock);
227 
228 	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 		return NULL;
230 
231 	return asoc;
232 }
233 
234 /* Look up the transport from an address and an assoc id. If both address and
235  * id are specified, the associations matching the address and the id should be
236  * the same.
237  */
238 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
239 					      struct sockaddr_storage *addr,
240 					      sctp_assoc_t id)
241 {
242 	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
243 	struct sctp_transport *transport;
244 	union sctp_addr *laddr = (union sctp_addr *)addr;
245 
246 	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
247 					       laddr,
248 					       &transport);
249 
250 	if (!addr_asoc)
251 		return NULL;
252 
253 	id_asoc = sctp_id2assoc(sk, id);
254 	if (id_asoc && (id_asoc != addr_asoc))
255 		return NULL;
256 
257 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
258 						(union sctp_addr *)addr);
259 
260 	return transport;
261 }
262 
263 /* API 3.1.2 bind() - UDP Style Syntax
264  * The syntax of bind() is,
265  *
266  *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
267  *
268  *   sd      - the socket descriptor returned by socket().
269  *   addr    - the address structure (struct sockaddr_in or struct
270  *             sockaddr_in6 [RFC 2553]),
271  *   addr_len - the size of the address structure.
272  */
273 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
274 {
275 	int retval = 0;
276 
277 	lock_sock(sk);
278 
279 	pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
280 		 addr, addr_len);
281 
282 	/* Disallow binding twice. */
283 	if (!sctp_sk(sk)->ep->base.bind_addr.port)
284 		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
285 				      addr_len);
286 	else
287 		retval = -EINVAL;
288 
289 	release_sock(sk);
290 
291 	return retval;
292 }
293 
294 static long sctp_get_port_local(struct sock *, union sctp_addr *);
295 
296 /* Verify this is a valid sockaddr. */
297 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
298 					union sctp_addr *addr, int len)
299 {
300 	struct sctp_af *af;
301 
302 	/* Check minimum size.  */
303 	if (len < sizeof (struct sockaddr))
304 		return NULL;
305 
306 	/* V4 mapped address are really of AF_INET family */
307 	if (addr->sa.sa_family == AF_INET6 &&
308 	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
309 		if (!opt->pf->af_supported(AF_INET, opt))
310 			return NULL;
311 	} else {
312 		/* Does this PF support this AF? */
313 		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
314 			return NULL;
315 	}
316 
317 	/* If we get this far, af is valid. */
318 	af = sctp_get_af_specific(addr->sa.sa_family);
319 
320 	if (len < af->sockaddr_len)
321 		return NULL;
322 
323 	return af;
324 }
325 
326 /* Bind a local address either to an endpoint or to an association.  */
327 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
328 {
329 	struct net *net = sock_net(sk);
330 	struct sctp_sock *sp = sctp_sk(sk);
331 	struct sctp_endpoint *ep = sp->ep;
332 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
333 	struct sctp_af *af;
334 	unsigned short snum;
335 	int ret = 0;
336 
337 	/* Common sockaddr verification. */
338 	af = sctp_sockaddr_af(sp, addr, len);
339 	if (!af) {
340 		pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
341 			 __func__, sk, addr, len);
342 		return -EINVAL;
343 	}
344 
345 	snum = ntohs(addr->v4.sin_port);
346 
347 	pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
348 		 __func__, sk, &addr->sa, bp->port, snum, len);
349 
350 	/* PF specific bind() address verification. */
351 	if (!sp->pf->bind_verify(sp, addr))
352 		return -EADDRNOTAVAIL;
353 
354 	/* We must either be unbound, or bind to the same port.
355 	 * It's OK to allow 0 ports if we are already bound.
356 	 * We'll just inhert an already bound port in this case
357 	 */
358 	if (bp->port) {
359 		if (!snum)
360 			snum = bp->port;
361 		else if (snum != bp->port) {
362 			pr_debug("%s: new port %d doesn't match existing port "
363 				 "%d\n", __func__, snum, bp->port);
364 			return -EINVAL;
365 		}
366 	}
367 
368 	if (snum && snum < PROT_SOCK &&
369 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
370 		return -EACCES;
371 
372 	/* See if the address matches any of the addresses we may have
373 	 * already bound before checking against other endpoints.
374 	 */
375 	if (sctp_bind_addr_match(bp, addr, sp))
376 		return -EINVAL;
377 
378 	/* Make sure we are allowed to bind here.
379 	 * The function sctp_get_port_local() does duplicate address
380 	 * detection.
381 	 */
382 	addr->v4.sin_port = htons(snum);
383 	if ((ret = sctp_get_port_local(sk, addr))) {
384 		return -EADDRINUSE;
385 	}
386 
387 	/* Refresh ephemeral port.  */
388 	if (!bp->port)
389 		bp->port = inet_sk(sk)->inet_num;
390 
391 	/* Add the address to the bind address list.
392 	 * Use GFP_ATOMIC since BHs will be disabled.
393 	 */
394 	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
395 
396 	/* Copy back into socket for getsockname() use. */
397 	if (!ret) {
398 		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
399 		af->to_sk_saddr(addr, sk);
400 	}
401 
402 	return ret;
403 }
404 
405  /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
406  *
407  * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408  * at any one time.  If a sender, after sending an ASCONF chunk, decides
409  * it needs to transfer another ASCONF Chunk, it MUST wait until the
410  * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411  * subsequent ASCONF. Note this restriction binds each side, so at any
412  * time two ASCONF may be in-transit on any given association (one sent
413  * from each endpoint).
414  */
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 			    struct sctp_chunk *chunk)
417 {
418 	struct net 	*net = sock_net(asoc->base.sk);
419 	int		retval = 0;
420 
421 	/* If there is an outstanding ASCONF chunk, queue it for later
422 	 * transmission.
423 	 */
424 	if (asoc->addip_last_asconf) {
425 		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
426 		goto out;
427 	}
428 
429 	/* Hold the chunk until an ASCONF_ACK is received. */
430 	sctp_chunk_hold(chunk);
431 	retval = sctp_primitive_ASCONF(net, asoc, chunk);
432 	if (retval)
433 		sctp_chunk_free(chunk);
434 	else
435 		asoc->addip_last_asconf = chunk;
436 
437 out:
438 	return retval;
439 }
440 
441 /* Add a list of addresses as bind addresses to local endpoint or
442  * association.
443  *
444  * Basically run through each address specified in the addrs/addrcnt
445  * array/length pair, determine if it is IPv6 or IPv4 and call
446  * sctp_do_bind() on it.
447  *
448  * If any of them fails, then the operation will be reversed and the
449  * ones that were added will be removed.
450  *
451  * Only sctp_setsockopt_bindx() is supposed to call this function.
452  */
453 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
454 {
455 	int cnt;
456 	int retval = 0;
457 	void *addr_buf;
458 	struct sockaddr *sa_addr;
459 	struct sctp_af *af;
460 
461 	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
462 		 addrs, addrcnt);
463 
464 	addr_buf = addrs;
465 	for (cnt = 0; cnt < addrcnt; cnt++) {
466 		/* The list may contain either IPv4 or IPv6 address;
467 		 * determine the address length for walking thru the list.
468 		 */
469 		sa_addr = addr_buf;
470 		af = sctp_get_af_specific(sa_addr->sa_family);
471 		if (!af) {
472 			retval = -EINVAL;
473 			goto err_bindx_add;
474 		}
475 
476 		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
477 				      af->sockaddr_len);
478 
479 		addr_buf += af->sockaddr_len;
480 
481 err_bindx_add:
482 		if (retval < 0) {
483 			/* Failed. Cleanup the ones that have been added */
484 			if (cnt > 0)
485 				sctp_bindx_rem(sk, addrs, cnt);
486 			return retval;
487 		}
488 	}
489 
490 	return retval;
491 }
492 
493 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
494  * associations that are part of the endpoint indicating that a list of local
495  * addresses are added to the endpoint.
496  *
497  * If any of the addresses is already in the bind address list of the
498  * association, we do not send the chunk for that association.  But it will not
499  * affect other associations.
500  *
501  * Only sctp_setsockopt_bindx() is supposed to call this function.
502  */
503 static int sctp_send_asconf_add_ip(struct sock		*sk,
504 				   struct sockaddr	*addrs,
505 				   int 			addrcnt)
506 {
507 	struct net *net = sock_net(sk);
508 	struct sctp_sock		*sp;
509 	struct sctp_endpoint		*ep;
510 	struct sctp_association		*asoc;
511 	struct sctp_bind_addr		*bp;
512 	struct sctp_chunk		*chunk;
513 	struct sctp_sockaddr_entry	*laddr;
514 	union sctp_addr			*addr;
515 	union sctp_addr			saveaddr;
516 	void				*addr_buf;
517 	struct sctp_af			*af;
518 	struct list_head		*p;
519 	int 				i;
520 	int 				retval = 0;
521 
522 	if (!net->sctp.addip_enable)
523 		return retval;
524 
525 	sp = sctp_sk(sk);
526 	ep = sp->ep;
527 
528 	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
529 		 __func__, sk, addrs, addrcnt);
530 
531 	list_for_each_entry(asoc, &ep->asocs, asocs) {
532 		if (!asoc->peer.asconf_capable)
533 			continue;
534 
535 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
536 			continue;
537 
538 		if (!sctp_state(asoc, ESTABLISHED))
539 			continue;
540 
541 		/* Check if any address in the packed array of addresses is
542 		 * in the bind address list of the association. If so,
543 		 * do not send the asconf chunk to its peer, but continue with
544 		 * other associations.
545 		 */
546 		addr_buf = addrs;
547 		for (i = 0; i < addrcnt; i++) {
548 			addr = addr_buf;
549 			af = sctp_get_af_specific(addr->v4.sin_family);
550 			if (!af) {
551 				retval = -EINVAL;
552 				goto out;
553 			}
554 
555 			if (sctp_assoc_lookup_laddr(asoc, addr))
556 				break;
557 
558 			addr_buf += af->sockaddr_len;
559 		}
560 		if (i < addrcnt)
561 			continue;
562 
563 		/* Use the first valid address in bind addr list of
564 		 * association as Address Parameter of ASCONF CHUNK.
565 		 */
566 		bp = &asoc->base.bind_addr;
567 		p = bp->address_list.next;
568 		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
569 		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
570 						   addrcnt, SCTP_PARAM_ADD_IP);
571 		if (!chunk) {
572 			retval = -ENOMEM;
573 			goto out;
574 		}
575 
576 		/* Add the new addresses to the bind address list with
577 		 * use_as_src set to 0.
578 		 */
579 		addr_buf = addrs;
580 		for (i = 0; i < addrcnt; i++) {
581 			addr = addr_buf;
582 			af = sctp_get_af_specific(addr->v4.sin_family);
583 			memcpy(&saveaddr, addr, af->sockaddr_len);
584 			retval = sctp_add_bind_addr(bp, &saveaddr,
585 						    SCTP_ADDR_NEW, GFP_ATOMIC);
586 			addr_buf += af->sockaddr_len;
587 		}
588 		if (asoc->src_out_of_asoc_ok) {
589 			struct sctp_transport *trans;
590 
591 			list_for_each_entry(trans,
592 			    &asoc->peer.transport_addr_list, transports) {
593 				/* Clear the source and route cache */
594 				dst_release(trans->dst);
595 				trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
596 				    2*asoc->pathmtu, 4380));
597 				trans->ssthresh = asoc->peer.i.a_rwnd;
598 				trans->rto = asoc->rto_initial;
599 				sctp_max_rto(asoc, trans);
600 				trans->rtt = trans->srtt = trans->rttvar = 0;
601 				sctp_transport_route(trans, NULL,
602 				    sctp_sk(asoc->base.sk));
603 			}
604 		}
605 		retval = sctp_send_asconf(asoc, chunk);
606 	}
607 
608 out:
609 	return retval;
610 }
611 
612 /* Remove a list of addresses from bind addresses list.  Do not remove the
613  * last address.
614  *
615  * Basically run through each address specified in the addrs/addrcnt
616  * array/length pair, determine if it is IPv6 or IPv4 and call
617  * sctp_del_bind() on it.
618  *
619  * If any of them fails, then the operation will be reversed and the
620  * ones that were removed will be added back.
621  *
622  * At least one address has to be left; if only one address is
623  * available, the operation will return -EBUSY.
624  *
625  * Only sctp_setsockopt_bindx() is supposed to call this function.
626  */
627 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
628 {
629 	struct sctp_sock *sp = sctp_sk(sk);
630 	struct sctp_endpoint *ep = sp->ep;
631 	int cnt;
632 	struct sctp_bind_addr *bp = &ep->base.bind_addr;
633 	int retval = 0;
634 	void *addr_buf;
635 	union sctp_addr *sa_addr;
636 	struct sctp_af *af;
637 
638 	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
639 		 __func__, sk, addrs, addrcnt);
640 
641 	addr_buf = addrs;
642 	for (cnt = 0; cnt < addrcnt; cnt++) {
643 		/* If the bind address list is empty or if there is only one
644 		 * bind address, there is nothing more to be removed (we need
645 		 * at least one address here).
646 		 */
647 		if (list_empty(&bp->address_list) ||
648 		    (sctp_list_single_entry(&bp->address_list))) {
649 			retval = -EBUSY;
650 			goto err_bindx_rem;
651 		}
652 
653 		sa_addr = addr_buf;
654 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
655 		if (!af) {
656 			retval = -EINVAL;
657 			goto err_bindx_rem;
658 		}
659 
660 		if (!af->addr_valid(sa_addr, sp, NULL)) {
661 			retval = -EADDRNOTAVAIL;
662 			goto err_bindx_rem;
663 		}
664 
665 		if (sa_addr->v4.sin_port &&
666 		    sa_addr->v4.sin_port != htons(bp->port)) {
667 			retval = -EINVAL;
668 			goto err_bindx_rem;
669 		}
670 
671 		if (!sa_addr->v4.sin_port)
672 			sa_addr->v4.sin_port = htons(bp->port);
673 
674 		/* FIXME - There is probably a need to check if sk->sk_saddr and
675 		 * sk->sk_rcv_addr are currently set to one of the addresses to
676 		 * be removed. This is something which needs to be looked into
677 		 * when we are fixing the outstanding issues with multi-homing
678 		 * socket routing and failover schemes. Refer to comments in
679 		 * sctp_do_bind(). -daisy
680 		 */
681 		retval = sctp_del_bind_addr(bp, sa_addr);
682 
683 		addr_buf += af->sockaddr_len;
684 err_bindx_rem:
685 		if (retval < 0) {
686 			/* Failed. Add the ones that has been removed back */
687 			if (cnt > 0)
688 				sctp_bindx_add(sk, addrs, cnt);
689 			return retval;
690 		}
691 	}
692 
693 	return retval;
694 }
695 
696 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
697  * the associations that are part of the endpoint indicating that a list of
698  * local addresses are removed from the endpoint.
699  *
700  * If any of the addresses is already in the bind address list of the
701  * association, we do not send the chunk for that association.  But it will not
702  * affect other associations.
703  *
704  * Only sctp_setsockopt_bindx() is supposed to call this function.
705  */
706 static int sctp_send_asconf_del_ip(struct sock		*sk,
707 				   struct sockaddr	*addrs,
708 				   int			addrcnt)
709 {
710 	struct net *net = sock_net(sk);
711 	struct sctp_sock	*sp;
712 	struct sctp_endpoint	*ep;
713 	struct sctp_association	*asoc;
714 	struct sctp_transport	*transport;
715 	struct sctp_bind_addr	*bp;
716 	struct sctp_chunk	*chunk;
717 	union sctp_addr		*laddr;
718 	void			*addr_buf;
719 	struct sctp_af		*af;
720 	struct sctp_sockaddr_entry *saddr;
721 	int 			i;
722 	int 			retval = 0;
723 	int			stored = 0;
724 
725 	chunk = NULL;
726 	if (!net->sctp.addip_enable)
727 		return retval;
728 
729 	sp = sctp_sk(sk);
730 	ep = sp->ep;
731 
732 	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
733 		 __func__, sk, addrs, addrcnt);
734 
735 	list_for_each_entry(asoc, &ep->asocs, asocs) {
736 
737 		if (!asoc->peer.asconf_capable)
738 			continue;
739 
740 		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
741 			continue;
742 
743 		if (!sctp_state(asoc, ESTABLISHED))
744 			continue;
745 
746 		/* Check if any address in the packed array of addresses is
747 		 * not present in the bind address list of the association.
748 		 * If so, do not send the asconf chunk to its peer, but
749 		 * continue with other associations.
750 		 */
751 		addr_buf = addrs;
752 		for (i = 0; i < addrcnt; i++) {
753 			laddr = addr_buf;
754 			af = sctp_get_af_specific(laddr->v4.sin_family);
755 			if (!af) {
756 				retval = -EINVAL;
757 				goto out;
758 			}
759 
760 			if (!sctp_assoc_lookup_laddr(asoc, laddr))
761 				break;
762 
763 			addr_buf += af->sockaddr_len;
764 		}
765 		if (i < addrcnt)
766 			continue;
767 
768 		/* Find one address in the association's bind address list
769 		 * that is not in the packed array of addresses. This is to
770 		 * make sure that we do not delete all the addresses in the
771 		 * association.
772 		 */
773 		bp = &asoc->base.bind_addr;
774 		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
775 					       addrcnt, sp);
776 		if ((laddr == NULL) && (addrcnt == 1)) {
777 			if (asoc->asconf_addr_del_pending)
778 				continue;
779 			asoc->asconf_addr_del_pending =
780 			    kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
781 			if (asoc->asconf_addr_del_pending == NULL) {
782 				retval = -ENOMEM;
783 				goto out;
784 			}
785 			asoc->asconf_addr_del_pending->sa.sa_family =
786 				    addrs->sa_family;
787 			asoc->asconf_addr_del_pending->v4.sin_port =
788 				    htons(bp->port);
789 			if (addrs->sa_family == AF_INET) {
790 				struct sockaddr_in *sin;
791 
792 				sin = (struct sockaddr_in *)addrs;
793 				asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
794 			} else if (addrs->sa_family == AF_INET6) {
795 				struct sockaddr_in6 *sin6;
796 
797 				sin6 = (struct sockaddr_in6 *)addrs;
798 				asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
799 			}
800 
801 			pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
802 				 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
803 				 asoc->asconf_addr_del_pending);
804 
805 			asoc->src_out_of_asoc_ok = 1;
806 			stored = 1;
807 			goto skip_mkasconf;
808 		}
809 
810 		if (laddr == NULL)
811 			return -EINVAL;
812 
813 		/* We do not need RCU protection throughout this loop
814 		 * because this is done under a socket lock from the
815 		 * setsockopt call.
816 		 */
817 		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
818 						   SCTP_PARAM_DEL_IP);
819 		if (!chunk) {
820 			retval = -ENOMEM;
821 			goto out;
822 		}
823 
824 skip_mkasconf:
825 		/* Reset use_as_src flag for the addresses in the bind address
826 		 * list that are to be deleted.
827 		 */
828 		addr_buf = addrs;
829 		for (i = 0; i < addrcnt; i++) {
830 			laddr = addr_buf;
831 			af = sctp_get_af_specific(laddr->v4.sin_family);
832 			list_for_each_entry(saddr, &bp->address_list, list) {
833 				if (sctp_cmp_addr_exact(&saddr->a, laddr))
834 					saddr->state = SCTP_ADDR_DEL;
835 			}
836 			addr_buf += af->sockaddr_len;
837 		}
838 
839 		/* Update the route and saddr entries for all the transports
840 		 * as some of the addresses in the bind address list are
841 		 * about to be deleted and cannot be used as source addresses.
842 		 */
843 		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
844 					transports) {
845 			dst_release(transport->dst);
846 			sctp_transport_route(transport, NULL,
847 					     sctp_sk(asoc->base.sk));
848 		}
849 
850 		if (stored)
851 			/* We don't need to transmit ASCONF */
852 			continue;
853 		retval = sctp_send_asconf(asoc, chunk);
854 	}
855 out:
856 	return retval;
857 }
858 
859 /* set addr events to assocs in the endpoint.  ep and addr_wq must be locked */
860 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
861 {
862 	struct sock *sk = sctp_opt2sk(sp);
863 	union sctp_addr *addr;
864 	struct sctp_af *af;
865 
866 	/* It is safe to write port space in caller. */
867 	addr = &addrw->a;
868 	addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
869 	af = sctp_get_af_specific(addr->sa.sa_family);
870 	if (!af)
871 		return -EINVAL;
872 	if (sctp_verify_addr(sk, addr, af->sockaddr_len))
873 		return -EINVAL;
874 
875 	if (addrw->state == SCTP_ADDR_NEW)
876 		return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
877 	else
878 		return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
879 }
880 
881 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
882  *
883  * API 8.1
884  * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
885  *                int flags);
886  *
887  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
888  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
889  * or IPv6 addresses.
890  *
891  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
892  * Section 3.1.2 for this usage.
893  *
894  * addrs is a pointer to an array of one or more socket addresses. Each
895  * address is contained in its appropriate structure (i.e. struct
896  * sockaddr_in or struct sockaddr_in6) the family of the address type
897  * must be used to distinguish the address length (note that this
898  * representation is termed a "packed array" of addresses). The caller
899  * specifies the number of addresses in the array with addrcnt.
900  *
901  * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
902  * -1, and sets errno to the appropriate error code.
903  *
904  * For SCTP, the port given in each socket address must be the same, or
905  * sctp_bindx() will fail, setting errno to EINVAL.
906  *
907  * The flags parameter is formed from the bitwise OR of zero or more of
908  * the following currently defined flags:
909  *
910  * SCTP_BINDX_ADD_ADDR
911  *
912  * SCTP_BINDX_REM_ADDR
913  *
914  * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
915  * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
916  * addresses from the association. The two flags are mutually exclusive;
917  * if both are given, sctp_bindx() will fail with EINVAL. A caller may
918  * not remove all addresses from an association; sctp_bindx() will
919  * reject such an attempt with EINVAL.
920  *
921  * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
922  * additional addresses with an endpoint after calling bind().  Or use
923  * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
924  * socket is associated with so that no new association accepted will be
925  * associated with those addresses. If the endpoint supports dynamic
926  * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
927  * endpoint to send the appropriate message to the peer to change the
928  * peers address lists.
929  *
930  * Adding and removing addresses from a connected association is
931  * optional functionality. Implementations that do not support this
932  * functionality should return EOPNOTSUPP.
933  *
934  * Basically do nothing but copying the addresses from user to kernel
935  * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
936  * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
937  * from userspace.
938  *
939  * We don't use copy_from_user() for optimization: we first do the
940  * sanity checks (buffer size -fast- and access check-healthy
941  * pointer); if all of those succeed, then we can alloc the memory
942  * (expensive operation) needed to copy the data to kernel. Then we do
943  * the copying without checking the user space area
944  * (__copy_from_user()).
945  *
946  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
947  * it.
948  *
949  * sk        The sk of the socket
950  * addrs     The pointer to the addresses in user land
951  * addrssize Size of the addrs buffer
952  * op        Operation to perform (add or remove, see the flags of
953  *           sctp_bindx)
954  *
955  * Returns 0 if ok, <0 errno code on error.
956  */
957 static int sctp_setsockopt_bindx(struct sock *sk,
958 				 struct sockaddr __user *addrs,
959 				 int addrs_size, int op)
960 {
961 	struct sockaddr *kaddrs;
962 	int err;
963 	int addrcnt = 0;
964 	int walk_size = 0;
965 	struct sockaddr *sa_addr;
966 	void *addr_buf;
967 	struct sctp_af *af;
968 
969 	pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
970 		 __func__, sk, addrs, addrs_size, op);
971 
972 	if (unlikely(addrs_size <= 0))
973 		return -EINVAL;
974 
975 	/* Check the user passed a healthy pointer.  */
976 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
977 		return -EFAULT;
978 
979 	/* Alloc space for the address array in kernel memory.  */
980 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
981 	if (unlikely(!kaddrs))
982 		return -ENOMEM;
983 
984 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
985 		kfree(kaddrs);
986 		return -EFAULT;
987 	}
988 
989 	/* Walk through the addrs buffer and count the number of addresses. */
990 	addr_buf = kaddrs;
991 	while (walk_size < addrs_size) {
992 		if (walk_size + sizeof(sa_family_t) > addrs_size) {
993 			kfree(kaddrs);
994 			return -EINVAL;
995 		}
996 
997 		sa_addr = addr_buf;
998 		af = sctp_get_af_specific(sa_addr->sa_family);
999 
1000 		/* If the address family is not supported or if this address
1001 		 * causes the address buffer to overflow return EINVAL.
1002 		 */
1003 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1004 			kfree(kaddrs);
1005 			return -EINVAL;
1006 		}
1007 		addrcnt++;
1008 		addr_buf += af->sockaddr_len;
1009 		walk_size += af->sockaddr_len;
1010 	}
1011 
1012 	/* Do the work. */
1013 	switch (op) {
1014 	case SCTP_BINDX_ADD_ADDR:
1015 		err = sctp_bindx_add(sk, kaddrs, addrcnt);
1016 		if (err)
1017 			goto out;
1018 		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1019 		break;
1020 
1021 	case SCTP_BINDX_REM_ADDR:
1022 		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1023 		if (err)
1024 			goto out;
1025 		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1026 		break;
1027 
1028 	default:
1029 		err = -EINVAL;
1030 		break;
1031 	}
1032 
1033 out:
1034 	kfree(kaddrs);
1035 
1036 	return err;
1037 }
1038 
1039 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1040  *
1041  * Common routine for handling connect() and sctp_connectx().
1042  * Connect will come in with just a single address.
1043  */
1044 static int __sctp_connect(struct sock *sk,
1045 			  struct sockaddr *kaddrs,
1046 			  int addrs_size,
1047 			  sctp_assoc_t *assoc_id)
1048 {
1049 	struct net *net = sock_net(sk);
1050 	struct sctp_sock *sp;
1051 	struct sctp_endpoint *ep;
1052 	struct sctp_association *asoc = NULL;
1053 	struct sctp_association *asoc2;
1054 	struct sctp_transport *transport;
1055 	union sctp_addr to;
1056 	struct sctp_af *af;
1057 	sctp_scope_t scope;
1058 	long timeo;
1059 	int err = 0;
1060 	int addrcnt = 0;
1061 	int walk_size = 0;
1062 	union sctp_addr *sa_addr = NULL;
1063 	void *addr_buf;
1064 	unsigned short port;
1065 	unsigned int f_flags = 0;
1066 
1067 	sp = sctp_sk(sk);
1068 	ep = sp->ep;
1069 
1070 	/* connect() cannot be done on a socket that is already in ESTABLISHED
1071 	 * state - UDP-style peeled off socket or a TCP-style socket that
1072 	 * is already connected.
1073 	 * It cannot be done even on a TCP-style listening socket.
1074 	 */
1075 	if (sctp_sstate(sk, ESTABLISHED) ||
1076 	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1077 		err = -EISCONN;
1078 		goto out_free;
1079 	}
1080 
1081 	/* Walk through the addrs buffer and count the number of addresses. */
1082 	addr_buf = kaddrs;
1083 	while (walk_size < addrs_size) {
1084 		if (walk_size + sizeof(sa_family_t) > addrs_size) {
1085 			err = -EINVAL;
1086 			goto out_free;
1087 		}
1088 
1089 		sa_addr = addr_buf;
1090 		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1091 
1092 		/* If the address family is not supported or if this address
1093 		 * causes the address buffer to overflow return EINVAL.
1094 		 */
1095 		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1096 			err = -EINVAL;
1097 			goto out_free;
1098 		}
1099 
1100 		port = ntohs(sa_addr->v4.sin_port);
1101 
1102 		/* Save current address so we can work with it */
1103 		memcpy(&to, sa_addr, af->sockaddr_len);
1104 
1105 		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1106 		if (err)
1107 			goto out_free;
1108 
1109 		/* Make sure the destination port is correctly set
1110 		 * in all addresses.
1111 		 */
1112 		if (asoc && asoc->peer.port && asoc->peer.port != port) {
1113 			err = -EINVAL;
1114 			goto out_free;
1115 		}
1116 
1117 		/* Check if there already is a matching association on the
1118 		 * endpoint (other than the one created here).
1119 		 */
1120 		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1121 		if (asoc2 && asoc2 != asoc) {
1122 			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1123 				err = -EISCONN;
1124 			else
1125 				err = -EALREADY;
1126 			goto out_free;
1127 		}
1128 
1129 		/* If we could not find a matching association on the endpoint,
1130 		 * make sure that there is no peeled-off association matching
1131 		 * the peer address even on another socket.
1132 		 */
1133 		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1134 			err = -EADDRNOTAVAIL;
1135 			goto out_free;
1136 		}
1137 
1138 		if (!asoc) {
1139 			/* If a bind() or sctp_bindx() is not called prior to
1140 			 * an sctp_connectx() call, the system picks an
1141 			 * ephemeral port and will choose an address set
1142 			 * equivalent to binding with a wildcard address.
1143 			 */
1144 			if (!ep->base.bind_addr.port) {
1145 				if (sctp_autobind(sk)) {
1146 					err = -EAGAIN;
1147 					goto out_free;
1148 				}
1149 			} else {
1150 				/*
1151 				 * If an unprivileged user inherits a 1-many
1152 				 * style socket with open associations on a
1153 				 * privileged port, it MAY be permitted to
1154 				 * accept new associations, but it SHOULD NOT
1155 				 * be permitted to open new associations.
1156 				 */
1157 				if (ep->base.bind_addr.port < PROT_SOCK &&
1158 				    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1159 					err = -EACCES;
1160 					goto out_free;
1161 				}
1162 			}
1163 
1164 			scope = sctp_scope(&to);
1165 			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1166 			if (!asoc) {
1167 				err = -ENOMEM;
1168 				goto out_free;
1169 			}
1170 
1171 			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1172 							      GFP_KERNEL);
1173 			if (err < 0) {
1174 				goto out_free;
1175 			}
1176 
1177 		}
1178 
1179 		/* Prime the peer's transport structures.  */
1180 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1181 						SCTP_UNKNOWN);
1182 		if (!transport) {
1183 			err = -ENOMEM;
1184 			goto out_free;
1185 		}
1186 
1187 		addrcnt++;
1188 		addr_buf += af->sockaddr_len;
1189 		walk_size += af->sockaddr_len;
1190 	}
1191 
1192 	/* In case the user of sctp_connectx() wants an association
1193 	 * id back, assign one now.
1194 	 */
1195 	if (assoc_id) {
1196 		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1197 		if (err < 0)
1198 			goto out_free;
1199 	}
1200 
1201 	err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1202 	if (err < 0) {
1203 		goto out_free;
1204 	}
1205 
1206 	/* Initialize sk's dport and daddr for getpeername() */
1207 	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1208 	af = sctp_get_af_specific(sa_addr->sa.sa_family);
1209 	af->to_sk_daddr(sa_addr, sk);
1210 	sk->sk_err = 0;
1211 
1212 	/* in-kernel sockets don't generally have a file allocated to them
1213 	 * if all they do is call sock_create_kern().
1214 	 */
1215 	if (sk->sk_socket->file)
1216 		f_flags = sk->sk_socket->file->f_flags;
1217 
1218 	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1219 
1220 	err = sctp_wait_for_connect(asoc, &timeo);
1221 	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1222 		*assoc_id = asoc->assoc_id;
1223 
1224 	/* Don't free association on exit. */
1225 	asoc = NULL;
1226 
1227 out_free:
1228 	pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1229 		 __func__, asoc, kaddrs, err);
1230 
1231 	if (asoc) {
1232 		/* sctp_primitive_ASSOCIATE may have added this association
1233 		 * To the hash table, try to unhash it, just in case, its a noop
1234 		 * if it wasn't hashed so we're safe
1235 		 */
1236 		sctp_unhash_established(asoc);
1237 		sctp_association_free(asoc);
1238 	}
1239 	return err;
1240 }
1241 
1242 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1243  *
1244  * API 8.9
1245  * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1246  * 			sctp_assoc_t *asoc);
1247  *
1248  * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1249  * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1250  * or IPv6 addresses.
1251  *
1252  * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1253  * Section 3.1.2 for this usage.
1254  *
1255  * addrs is a pointer to an array of one or more socket addresses. Each
1256  * address is contained in its appropriate structure (i.e. struct
1257  * sockaddr_in or struct sockaddr_in6) the family of the address type
1258  * must be used to distengish the address length (note that this
1259  * representation is termed a "packed array" of addresses). The caller
1260  * specifies the number of addresses in the array with addrcnt.
1261  *
1262  * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1263  * the association id of the new association.  On failure, sctp_connectx()
1264  * returns -1, and sets errno to the appropriate error code.  The assoc_id
1265  * is not touched by the kernel.
1266  *
1267  * For SCTP, the port given in each socket address must be the same, or
1268  * sctp_connectx() will fail, setting errno to EINVAL.
1269  *
1270  * An application can use sctp_connectx to initiate an association with
1271  * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1272  * allows a caller to specify multiple addresses at which a peer can be
1273  * reached.  The way the SCTP stack uses the list of addresses to set up
1274  * the association is implementation dependent.  This function only
1275  * specifies that the stack will try to make use of all the addresses in
1276  * the list when needed.
1277  *
1278  * Note that the list of addresses passed in is only used for setting up
1279  * the association.  It does not necessarily equal the set of addresses
1280  * the peer uses for the resulting association.  If the caller wants to
1281  * find out the set of peer addresses, it must use sctp_getpaddrs() to
1282  * retrieve them after the association has been set up.
1283  *
1284  * Basically do nothing but copying the addresses from user to kernel
1285  * land and invoking either sctp_connectx(). This is used for tunneling
1286  * the sctp_connectx() request through sctp_setsockopt() from userspace.
1287  *
1288  * We don't use copy_from_user() for optimization: we first do the
1289  * sanity checks (buffer size -fast- and access check-healthy
1290  * pointer); if all of those succeed, then we can alloc the memory
1291  * (expensive operation) needed to copy the data to kernel. Then we do
1292  * the copying without checking the user space area
1293  * (__copy_from_user()).
1294  *
1295  * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1296  * it.
1297  *
1298  * sk        The sk of the socket
1299  * addrs     The pointer to the addresses in user land
1300  * addrssize Size of the addrs buffer
1301  *
1302  * Returns >=0 if ok, <0 errno code on error.
1303  */
1304 static int __sctp_setsockopt_connectx(struct sock *sk,
1305 				      struct sockaddr __user *addrs,
1306 				      int addrs_size,
1307 				      sctp_assoc_t *assoc_id)
1308 {
1309 	int err = 0;
1310 	struct sockaddr *kaddrs;
1311 
1312 	pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1313 		 __func__, sk, addrs, addrs_size);
1314 
1315 	if (unlikely(addrs_size <= 0))
1316 		return -EINVAL;
1317 
1318 	/* Check the user passed a healthy pointer.  */
1319 	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1320 		return -EFAULT;
1321 
1322 	/* Alloc space for the address array in kernel memory.  */
1323 	kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1324 	if (unlikely(!kaddrs))
1325 		return -ENOMEM;
1326 
1327 	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1328 		err = -EFAULT;
1329 	} else {
1330 		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1331 	}
1332 
1333 	kfree(kaddrs);
1334 
1335 	return err;
1336 }
1337 
1338 /*
1339  * This is an older interface.  It's kept for backward compatibility
1340  * to the option that doesn't provide association id.
1341  */
1342 static int sctp_setsockopt_connectx_old(struct sock *sk,
1343 					struct sockaddr __user *addrs,
1344 					int addrs_size)
1345 {
1346 	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1347 }
1348 
1349 /*
1350  * New interface for the API.  The since the API is done with a socket
1351  * option, to make it simple we feed back the association id is as a return
1352  * indication to the call.  Error is always negative and association id is
1353  * always positive.
1354  */
1355 static int sctp_setsockopt_connectx(struct sock *sk,
1356 				    struct sockaddr __user *addrs,
1357 				    int addrs_size)
1358 {
1359 	sctp_assoc_t assoc_id = 0;
1360 	int err = 0;
1361 
1362 	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1363 
1364 	if (err)
1365 		return err;
1366 	else
1367 		return assoc_id;
1368 }
1369 
1370 /*
1371  * New (hopefully final) interface for the API.
1372  * We use the sctp_getaddrs_old structure so that use-space library
1373  * can avoid any unnecessary allocations. The only different part
1374  * is that we store the actual length of the address buffer into the
1375  * addrs_num structure member. That way we can re-use the existing
1376  * code.
1377  */
1378 #ifdef CONFIG_COMPAT
1379 struct compat_sctp_getaddrs_old {
1380 	sctp_assoc_t	assoc_id;
1381 	s32		addr_num;
1382 	compat_uptr_t	addrs;		/* struct sockaddr * */
1383 };
1384 #endif
1385 
1386 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1387 				     char __user *optval,
1388 				     int __user *optlen)
1389 {
1390 	struct sctp_getaddrs_old param;
1391 	sctp_assoc_t assoc_id = 0;
1392 	int err = 0;
1393 
1394 #ifdef CONFIG_COMPAT
1395 	if (is_compat_task()) {
1396 		struct compat_sctp_getaddrs_old param32;
1397 
1398 		if (len < sizeof(param32))
1399 			return -EINVAL;
1400 		if (copy_from_user(&param32, optval, sizeof(param32)))
1401 			return -EFAULT;
1402 
1403 		param.assoc_id = param32.assoc_id;
1404 		param.addr_num = param32.addr_num;
1405 		param.addrs = compat_ptr(param32.addrs);
1406 	} else
1407 #endif
1408 	{
1409 		if (len < sizeof(param))
1410 			return -EINVAL;
1411 		if (copy_from_user(&param, optval, sizeof(param)))
1412 			return -EFAULT;
1413 	}
1414 
1415 	err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1416 					 param.addrs, param.addr_num,
1417 					 &assoc_id);
1418 	if (err == 0 || err == -EINPROGRESS) {
1419 		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 			return -EFAULT;
1421 		if (put_user(sizeof(assoc_id), optlen))
1422 			return -EFAULT;
1423 	}
1424 
1425 	return err;
1426 }
1427 
1428 /* API 3.1.4 close() - UDP Style Syntax
1429  * Applications use close() to perform graceful shutdown (as described in
1430  * Section 10.1 of [SCTP]) on ALL the associations currently represented
1431  * by a UDP-style socket.
1432  *
1433  * The syntax is
1434  *
1435  *   ret = close(int sd);
1436  *
1437  *   sd      - the socket descriptor of the associations to be closed.
1438  *
1439  * To gracefully shutdown a specific association represented by the
1440  * UDP-style socket, an application should use the sendmsg() call,
1441  * passing no user data, but including the appropriate flag in the
1442  * ancillary data (see Section xxxx).
1443  *
1444  * If sd in the close() call is a branched-off socket representing only
1445  * one association, the shutdown is performed on that association only.
1446  *
1447  * 4.1.6 close() - TCP Style Syntax
1448  *
1449  * Applications use close() to gracefully close down an association.
1450  *
1451  * The syntax is:
1452  *
1453  *    int close(int sd);
1454  *
1455  *      sd      - the socket descriptor of the association to be closed.
1456  *
1457  * After an application calls close() on a socket descriptor, no further
1458  * socket operations will succeed on that descriptor.
1459  *
1460  * API 7.1.4 SO_LINGER
1461  *
1462  * An application using the TCP-style socket can use this option to
1463  * perform the SCTP ABORT primitive.  The linger option structure is:
1464  *
1465  *  struct  linger {
1466  *     int     l_onoff;                // option on/off
1467  *     int     l_linger;               // linger time
1468  * };
1469  *
1470  * To enable the option, set l_onoff to 1.  If the l_linger value is set
1471  * to 0, calling close() is the same as the ABORT primitive.  If the
1472  * value is set to a negative value, the setsockopt() call will return
1473  * an error.  If the value is set to a positive value linger_time, the
1474  * close() can be blocked for at most linger_time ms.  If the graceful
1475  * shutdown phase does not finish during this period, close() will
1476  * return but the graceful shutdown phase continues in the system.
1477  */
1478 static void sctp_close(struct sock *sk, long timeout)
1479 {
1480 	struct net *net = sock_net(sk);
1481 	struct sctp_endpoint *ep;
1482 	struct sctp_association *asoc;
1483 	struct list_head *pos, *temp;
1484 	unsigned int data_was_unread;
1485 
1486 	pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1487 
1488 	lock_sock(sk);
1489 	sk->sk_shutdown = SHUTDOWN_MASK;
1490 	sk->sk_state = SCTP_SS_CLOSING;
1491 
1492 	ep = sctp_sk(sk)->ep;
1493 
1494 	/* Clean up any skbs sitting on the receive queue.  */
1495 	data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1496 	data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1497 
1498 	/* Walk all associations on an endpoint.  */
1499 	list_for_each_safe(pos, temp, &ep->asocs) {
1500 		asoc = list_entry(pos, struct sctp_association, asocs);
1501 
1502 		if (sctp_style(sk, TCP)) {
1503 			/* A closed association can still be in the list if
1504 			 * it belongs to a TCP-style listening socket that is
1505 			 * not yet accepted. If so, free it. If not, send an
1506 			 * ABORT or SHUTDOWN based on the linger options.
1507 			 */
1508 			if (sctp_state(asoc, CLOSED)) {
1509 				sctp_unhash_established(asoc);
1510 				sctp_association_free(asoc);
1511 				continue;
1512 			}
1513 		}
1514 
1515 		if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1516 		    !skb_queue_empty(&asoc->ulpq.reasm) ||
1517 		    (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1518 			struct sctp_chunk *chunk;
1519 
1520 			chunk = sctp_make_abort_user(asoc, NULL, 0);
1521 			if (chunk)
1522 				sctp_primitive_ABORT(net, asoc, chunk);
1523 		} else
1524 			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1525 	}
1526 
1527 	/* On a TCP-style socket, block for at most linger_time if set. */
1528 	if (sctp_style(sk, TCP) && timeout)
1529 		sctp_wait_for_close(sk, timeout);
1530 
1531 	/* This will run the backlog queue.  */
1532 	release_sock(sk);
1533 
1534 	/* Supposedly, no process has access to the socket, but
1535 	 * the net layers still may.
1536 	 */
1537 	local_bh_disable();
1538 	bh_lock_sock(sk);
1539 
1540 	/* Hold the sock, since sk_common_release() will put sock_put()
1541 	 * and we have just a little more cleanup.
1542 	 */
1543 	sock_hold(sk);
1544 	sk_common_release(sk);
1545 
1546 	bh_unlock_sock(sk);
1547 	local_bh_enable();
1548 
1549 	sock_put(sk);
1550 
1551 	SCTP_DBG_OBJCNT_DEC(sock);
1552 }
1553 
1554 /* Handle EPIPE error. */
1555 static int sctp_error(struct sock *sk, int flags, int err)
1556 {
1557 	if (err == -EPIPE)
1558 		err = sock_error(sk) ? : -EPIPE;
1559 	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1560 		send_sig(SIGPIPE, current, 0);
1561 	return err;
1562 }
1563 
1564 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565  *
1566  * An application uses sendmsg() and recvmsg() calls to transmit data to
1567  * and receive data from its peer.
1568  *
1569  *  ssize_t sendmsg(int socket, const struct msghdr *message,
1570  *                  int flags);
1571  *
1572  *  socket  - the socket descriptor of the endpoint.
1573  *  message - pointer to the msghdr structure which contains a single
1574  *            user message and possibly some ancillary data.
1575  *
1576  *            See Section 5 for complete description of the data
1577  *            structures.
1578  *
1579  *  flags   - flags sent or received with the user message, see Section
1580  *            5 for complete description of the flags.
1581  *
1582  * Note:  This function could use a rewrite especially when explicit
1583  * connect support comes in.
1584  */
1585 /* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1586 
1587 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1588 
1589 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1590 			struct msghdr *msg, size_t msg_len)
1591 {
1592 	struct net *net = sock_net(sk);
1593 	struct sctp_sock *sp;
1594 	struct sctp_endpoint *ep;
1595 	struct sctp_association *new_asoc = NULL, *asoc = NULL;
1596 	struct sctp_transport *transport, *chunk_tp;
1597 	struct sctp_chunk *chunk;
1598 	union sctp_addr to;
1599 	struct sockaddr *msg_name = NULL;
1600 	struct sctp_sndrcvinfo default_sinfo;
1601 	struct sctp_sndrcvinfo *sinfo;
1602 	struct sctp_initmsg *sinit;
1603 	sctp_assoc_t associd = 0;
1604 	sctp_cmsgs_t cmsgs = { NULL };
1605 	int err;
1606 	sctp_scope_t scope;
1607 	long timeo;
1608 	__u16 sinfo_flags = 0;
1609 	struct sctp_datamsg *datamsg;
1610 	int msg_flags = msg->msg_flags;
1611 
1612 	err = 0;
1613 	sp = sctp_sk(sk);
1614 	ep = sp->ep;
1615 
1616 	pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1617 		 msg, msg_len, ep);
1618 
1619 	/* We cannot send a message over a TCP-style listening socket. */
1620 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1621 		err = -EPIPE;
1622 		goto out_nounlock;
1623 	}
1624 
1625 	/* Parse out the SCTP CMSGs.  */
1626 	err = sctp_msghdr_parse(msg, &cmsgs);
1627 	if (err) {
1628 		pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1629 		goto out_nounlock;
1630 	}
1631 
1632 	/* Fetch the destination address for this packet.  This
1633 	 * address only selects the association--it is not necessarily
1634 	 * the address we will send to.
1635 	 * For a peeled-off socket, msg_name is ignored.
1636 	 */
1637 	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1638 		int msg_namelen = msg->msg_namelen;
1639 
1640 		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1641 				       msg_namelen);
1642 		if (err)
1643 			return err;
1644 
1645 		if (msg_namelen > sizeof(to))
1646 			msg_namelen = sizeof(to);
1647 		memcpy(&to, msg->msg_name, msg_namelen);
1648 		msg_name = msg->msg_name;
1649 	}
1650 
1651 	sinfo = cmsgs.info;
1652 	sinit = cmsgs.init;
1653 
1654 	/* Did the user specify SNDRCVINFO?  */
1655 	if (sinfo) {
1656 		sinfo_flags = sinfo->sinfo_flags;
1657 		associd = sinfo->sinfo_assoc_id;
1658 	}
1659 
1660 	pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1661 		 msg_len, sinfo_flags);
1662 
1663 	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1664 	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1665 		err = -EINVAL;
1666 		goto out_nounlock;
1667 	}
1668 
1669 	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1670 	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1671 	 * If SCTP_ABORT is set, the message length could be non zero with
1672 	 * the msg_iov set to the user abort reason.
1673 	 */
1674 	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1675 	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1676 		err = -EINVAL;
1677 		goto out_nounlock;
1678 	}
1679 
1680 	/* If SCTP_ADDR_OVER is set, there must be an address
1681 	 * specified in msg_name.
1682 	 */
1683 	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1684 		err = -EINVAL;
1685 		goto out_nounlock;
1686 	}
1687 
1688 	transport = NULL;
1689 
1690 	pr_debug("%s: about to look up association\n", __func__);
1691 
1692 	lock_sock(sk);
1693 
1694 	/* If a msg_name has been specified, assume this is to be used.  */
1695 	if (msg_name) {
1696 		/* Look for a matching association on the endpoint. */
1697 		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1698 		if (!asoc) {
1699 			/* If we could not find a matching association on the
1700 			 * endpoint, make sure that it is not a TCP-style
1701 			 * socket that already has an association or there is
1702 			 * no peeled-off association on another socket.
1703 			 */
1704 			if ((sctp_style(sk, TCP) &&
1705 			     sctp_sstate(sk, ESTABLISHED)) ||
1706 			    sctp_endpoint_is_peeled_off(ep, &to)) {
1707 				err = -EADDRNOTAVAIL;
1708 				goto out_unlock;
1709 			}
1710 		}
1711 	} else {
1712 		asoc = sctp_id2assoc(sk, associd);
1713 		if (!asoc) {
1714 			err = -EPIPE;
1715 			goto out_unlock;
1716 		}
1717 	}
1718 
1719 	if (asoc) {
1720 		pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1721 
1722 		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1723 		 * socket that has an association in CLOSED state. This can
1724 		 * happen when an accepted socket has an association that is
1725 		 * already CLOSED.
1726 		 */
1727 		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1728 			err = -EPIPE;
1729 			goto out_unlock;
1730 		}
1731 
1732 		if (sinfo_flags & SCTP_EOF) {
1733 			pr_debug("%s: shutting down association:%p\n",
1734 				 __func__, asoc);
1735 
1736 			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1737 			err = 0;
1738 			goto out_unlock;
1739 		}
1740 		if (sinfo_flags & SCTP_ABORT) {
1741 
1742 			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1743 			if (!chunk) {
1744 				err = -ENOMEM;
1745 				goto out_unlock;
1746 			}
1747 
1748 			pr_debug("%s: aborting association:%p\n",
1749 				 __func__, asoc);
1750 
1751 			sctp_primitive_ABORT(net, asoc, chunk);
1752 			err = 0;
1753 			goto out_unlock;
1754 		}
1755 	}
1756 
1757 	/* Do we need to create the association?  */
1758 	if (!asoc) {
1759 		pr_debug("%s: there is no association yet\n", __func__);
1760 
1761 		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1762 			err = -EINVAL;
1763 			goto out_unlock;
1764 		}
1765 
1766 		/* Check for invalid stream against the stream counts,
1767 		 * either the default or the user specified stream counts.
1768 		 */
1769 		if (sinfo) {
1770 			if (!sinit || !sinit->sinit_num_ostreams) {
1771 				/* Check against the defaults. */
1772 				if (sinfo->sinfo_stream >=
1773 				    sp->initmsg.sinit_num_ostreams) {
1774 					err = -EINVAL;
1775 					goto out_unlock;
1776 				}
1777 			} else {
1778 				/* Check against the requested.  */
1779 				if (sinfo->sinfo_stream >=
1780 				    sinit->sinit_num_ostreams) {
1781 					err = -EINVAL;
1782 					goto out_unlock;
1783 				}
1784 			}
1785 		}
1786 
1787 		/*
1788 		 * API 3.1.2 bind() - UDP Style Syntax
1789 		 * If a bind() or sctp_bindx() is not called prior to a
1790 		 * sendmsg() call that initiates a new association, the
1791 		 * system picks an ephemeral port and will choose an address
1792 		 * set equivalent to binding with a wildcard address.
1793 		 */
1794 		if (!ep->base.bind_addr.port) {
1795 			if (sctp_autobind(sk)) {
1796 				err = -EAGAIN;
1797 				goto out_unlock;
1798 			}
1799 		} else {
1800 			/*
1801 			 * If an unprivileged user inherits a one-to-many
1802 			 * style socket with open associations on a privileged
1803 			 * port, it MAY be permitted to accept new associations,
1804 			 * but it SHOULD NOT be permitted to open new
1805 			 * associations.
1806 			 */
1807 			if (ep->base.bind_addr.port < PROT_SOCK &&
1808 			    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1809 				err = -EACCES;
1810 				goto out_unlock;
1811 			}
1812 		}
1813 
1814 		scope = sctp_scope(&to);
1815 		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1816 		if (!new_asoc) {
1817 			err = -ENOMEM;
1818 			goto out_unlock;
1819 		}
1820 		asoc = new_asoc;
1821 		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1822 		if (err < 0) {
1823 			err = -ENOMEM;
1824 			goto out_free;
1825 		}
1826 
1827 		/* If the SCTP_INIT ancillary data is specified, set all
1828 		 * the association init values accordingly.
1829 		 */
1830 		if (sinit) {
1831 			if (sinit->sinit_num_ostreams) {
1832 				asoc->c.sinit_num_ostreams =
1833 					sinit->sinit_num_ostreams;
1834 			}
1835 			if (sinit->sinit_max_instreams) {
1836 				asoc->c.sinit_max_instreams =
1837 					sinit->sinit_max_instreams;
1838 			}
1839 			if (sinit->sinit_max_attempts) {
1840 				asoc->max_init_attempts
1841 					= sinit->sinit_max_attempts;
1842 			}
1843 			if (sinit->sinit_max_init_timeo) {
1844 				asoc->max_init_timeo =
1845 				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1846 			}
1847 		}
1848 
1849 		/* Prime the peer's transport structures.  */
1850 		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1851 		if (!transport) {
1852 			err = -ENOMEM;
1853 			goto out_free;
1854 		}
1855 	}
1856 
1857 	/* ASSERT: we have a valid association at this point.  */
1858 	pr_debug("%s: we have a valid association\n", __func__);
1859 
1860 	if (!sinfo) {
1861 		/* If the user didn't specify SNDRCVINFO, make up one with
1862 		 * some defaults.
1863 		 */
1864 		memset(&default_sinfo, 0, sizeof(default_sinfo));
1865 		default_sinfo.sinfo_stream = asoc->default_stream;
1866 		default_sinfo.sinfo_flags = asoc->default_flags;
1867 		default_sinfo.sinfo_ppid = asoc->default_ppid;
1868 		default_sinfo.sinfo_context = asoc->default_context;
1869 		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1870 		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1871 		sinfo = &default_sinfo;
1872 	}
1873 
1874 	/* API 7.1.7, the sndbuf size per association bounds the
1875 	 * maximum size of data that can be sent in a single send call.
1876 	 */
1877 	if (msg_len > sk->sk_sndbuf) {
1878 		err = -EMSGSIZE;
1879 		goto out_free;
1880 	}
1881 
1882 	if (asoc->pmtu_pending)
1883 		sctp_assoc_pending_pmtu(sk, asoc);
1884 
1885 	/* If fragmentation is disabled and the message length exceeds the
1886 	 * association fragmentation point, return EMSGSIZE.  The I-D
1887 	 * does not specify what this error is, but this looks like
1888 	 * a great fit.
1889 	 */
1890 	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1891 		err = -EMSGSIZE;
1892 		goto out_free;
1893 	}
1894 
1895 	/* Check for invalid stream. */
1896 	if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1897 		err = -EINVAL;
1898 		goto out_free;
1899 	}
1900 
1901 	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1902 	if (!sctp_wspace(asoc)) {
1903 		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1904 		if (err)
1905 			goto out_free;
1906 	}
1907 
1908 	/* If an address is passed with the sendto/sendmsg call, it is used
1909 	 * to override the primary destination address in the TCP model, or
1910 	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1911 	 */
1912 	if ((sctp_style(sk, TCP) && msg_name) ||
1913 	    (sinfo_flags & SCTP_ADDR_OVER)) {
1914 		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1915 		if (!chunk_tp) {
1916 			err = -EINVAL;
1917 			goto out_free;
1918 		}
1919 	} else
1920 		chunk_tp = NULL;
1921 
1922 	/* Auto-connect, if we aren't connected already. */
1923 	if (sctp_state(asoc, CLOSED)) {
1924 		err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1925 		if (err < 0)
1926 			goto out_free;
1927 
1928 		pr_debug("%s: we associated primitively\n", __func__);
1929 	}
1930 
1931 	/* Break the message into multiple chunks of maximum size. */
1932 	datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1933 	if (IS_ERR(datamsg)) {
1934 		err = PTR_ERR(datamsg);
1935 		goto out_free;
1936 	}
1937 
1938 	/* Now send the (possibly) fragmented message. */
1939 	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1940 		sctp_chunk_hold(chunk);
1941 
1942 		/* Do accounting for the write space.  */
1943 		sctp_set_owner_w(chunk);
1944 
1945 		chunk->transport = chunk_tp;
1946 	}
1947 
1948 	/* Send it to the lower layers.  Note:  all chunks
1949 	 * must either fail or succeed.   The lower layer
1950 	 * works that way today.  Keep it that way or this
1951 	 * breaks.
1952 	 */
1953 	err = sctp_primitive_SEND(net, asoc, datamsg);
1954 	/* Did the lower layer accept the chunk? */
1955 	if (err) {
1956 		sctp_datamsg_free(datamsg);
1957 		goto out_free;
1958 	}
1959 
1960 	pr_debug("%s: we sent primitively\n", __func__);
1961 
1962 	sctp_datamsg_put(datamsg);
1963 	err = msg_len;
1964 
1965 	/* If we are already past ASSOCIATE, the lower
1966 	 * layers are responsible for association cleanup.
1967 	 */
1968 	goto out_unlock;
1969 
1970 out_free:
1971 	if (new_asoc) {
1972 		sctp_unhash_established(asoc);
1973 		sctp_association_free(asoc);
1974 	}
1975 out_unlock:
1976 	release_sock(sk);
1977 
1978 out_nounlock:
1979 	return sctp_error(sk, msg_flags, err);
1980 
1981 #if 0
1982 do_sock_err:
1983 	if (msg_len)
1984 		err = msg_len;
1985 	else
1986 		err = sock_error(sk);
1987 	goto out;
1988 
1989 do_interrupted:
1990 	if (msg_len)
1991 		err = msg_len;
1992 	goto out;
1993 #endif /* 0 */
1994 }
1995 
1996 /* This is an extended version of skb_pull() that removes the data from the
1997  * start of a skb even when data is spread across the list of skb's in the
1998  * frag_list. len specifies the total amount of data that needs to be removed.
1999  * when 'len' bytes could be removed from the skb, it returns 0.
2000  * If 'len' exceeds the total skb length,  it returns the no. of bytes that
2001  * could not be removed.
2002  */
2003 static int sctp_skb_pull(struct sk_buff *skb, int len)
2004 {
2005 	struct sk_buff *list;
2006 	int skb_len = skb_headlen(skb);
2007 	int rlen;
2008 
2009 	if (len <= skb_len) {
2010 		__skb_pull(skb, len);
2011 		return 0;
2012 	}
2013 	len -= skb_len;
2014 	__skb_pull(skb, skb_len);
2015 
2016 	skb_walk_frags(skb, list) {
2017 		rlen = sctp_skb_pull(list, len);
2018 		skb->len -= (len-rlen);
2019 		skb->data_len -= (len-rlen);
2020 
2021 		if (!rlen)
2022 			return 0;
2023 
2024 		len = rlen;
2025 	}
2026 
2027 	return len;
2028 }
2029 
2030 /* API 3.1.3  recvmsg() - UDP Style Syntax
2031  *
2032  *  ssize_t recvmsg(int socket, struct msghdr *message,
2033  *                    int flags);
2034  *
2035  *  socket  - the socket descriptor of the endpoint.
2036  *  message - pointer to the msghdr structure which contains a single
2037  *            user message and possibly some ancillary data.
2038  *
2039  *            See Section 5 for complete description of the data
2040  *            structures.
2041  *
2042  *  flags   - flags sent or received with the user message, see Section
2043  *            5 for complete description of the flags.
2044  */
2045 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2046 
2047 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2048 			struct msghdr *msg, size_t len, int noblock,
2049 			int flags, int *addr_len)
2050 {
2051 	struct sctp_ulpevent *event = NULL;
2052 	struct sctp_sock *sp = sctp_sk(sk);
2053 	struct sk_buff *skb;
2054 	int copied;
2055 	int err = 0;
2056 	int skb_len;
2057 
2058 	pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2059 		 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2060 		 addr_len);
2061 
2062 	lock_sock(sk);
2063 
2064 	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2065 		err = -ENOTCONN;
2066 		goto out;
2067 	}
2068 
2069 	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2070 	if (!skb)
2071 		goto out;
2072 
2073 	/* Get the total length of the skb including any skb's in the
2074 	 * frag_list.
2075 	 */
2076 	skb_len = skb->len;
2077 
2078 	copied = skb_len;
2079 	if (copied > len)
2080 		copied = len;
2081 
2082 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2083 
2084 	event = sctp_skb2event(skb);
2085 
2086 	if (err)
2087 		goto out_free;
2088 
2089 	sock_recv_ts_and_drops(msg, sk, skb);
2090 	if (sctp_ulpevent_is_notification(event)) {
2091 		msg->msg_flags |= MSG_NOTIFICATION;
2092 		sp->pf->event_msgname(event, msg->msg_name, addr_len);
2093 	} else {
2094 		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2095 	}
2096 
2097 	/* Check if we allow SCTP_SNDRCVINFO. */
2098 	if (sp->subscribe.sctp_data_io_event)
2099 		sctp_ulpevent_read_sndrcvinfo(event, msg);
2100 #if 0
2101 	/* FIXME: we should be calling IP/IPv6 layers.  */
2102 	if (sk->sk_protinfo.af_inet.cmsg_flags)
2103 		ip_cmsg_recv(msg, skb);
2104 #endif
2105 
2106 	err = copied;
2107 
2108 	/* If skb's length exceeds the user's buffer, update the skb and
2109 	 * push it back to the receive_queue so that the next call to
2110 	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2111 	 */
2112 	if (skb_len > copied) {
2113 		msg->msg_flags &= ~MSG_EOR;
2114 		if (flags & MSG_PEEK)
2115 			goto out_free;
2116 		sctp_skb_pull(skb, copied);
2117 		skb_queue_head(&sk->sk_receive_queue, skb);
2118 
2119 		/* When only partial message is copied to the user, increase
2120 		 * rwnd by that amount. If all the data in the skb is read,
2121 		 * rwnd is updated when the event is freed.
2122 		 */
2123 		if (!sctp_ulpevent_is_notification(event))
2124 			sctp_assoc_rwnd_increase(event->asoc, copied);
2125 		goto out;
2126 	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2127 		   (event->msg_flags & MSG_EOR))
2128 		msg->msg_flags |= MSG_EOR;
2129 	else
2130 		msg->msg_flags &= ~MSG_EOR;
2131 
2132 out_free:
2133 	if (flags & MSG_PEEK) {
2134 		/* Release the skb reference acquired after peeking the skb in
2135 		 * sctp_skb_recv_datagram().
2136 		 */
2137 		kfree_skb(skb);
2138 	} else {
2139 		/* Free the event which includes releasing the reference to
2140 		 * the owner of the skb, freeing the skb and updating the
2141 		 * rwnd.
2142 		 */
2143 		sctp_ulpevent_free(event);
2144 	}
2145 out:
2146 	release_sock(sk);
2147 	return err;
2148 }
2149 
2150 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2151  *
2152  * This option is a on/off flag.  If enabled no SCTP message
2153  * fragmentation will be performed.  Instead if a message being sent
2154  * exceeds the current PMTU size, the message will NOT be sent and
2155  * instead a error will be indicated to the user.
2156  */
2157 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2158 					     char __user *optval,
2159 					     unsigned int optlen)
2160 {
2161 	int val;
2162 
2163 	if (optlen < sizeof(int))
2164 		return -EINVAL;
2165 
2166 	if (get_user(val, (int __user *)optval))
2167 		return -EFAULT;
2168 
2169 	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2170 
2171 	return 0;
2172 }
2173 
2174 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2175 				  unsigned int optlen)
2176 {
2177 	struct sctp_association *asoc;
2178 	struct sctp_ulpevent *event;
2179 
2180 	if (optlen > sizeof(struct sctp_event_subscribe))
2181 		return -EINVAL;
2182 	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2183 		return -EFAULT;
2184 
2185 	/*
2186 	 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2187 	 * if there is no data to be sent or retransmit, the stack will
2188 	 * immediately send up this notification.
2189 	 */
2190 	if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2191 				       &sctp_sk(sk)->subscribe)) {
2192 		asoc = sctp_id2assoc(sk, 0);
2193 
2194 		if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2195 			event = sctp_ulpevent_make_sender_dry_event(asoc,
2196 					GFP_ATOMIC);
2197 			if (!event)
2198 				return -ENOMEM;
2199 
2200 			sctp_ulpq_tail_event(&asoc->ulpq, event);
2201 		}
2202 	}
2203 
2204 	return 0;
2205 }
2206 
2207 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2208  *
2209  * This socket option is applicable to the UDP-style socket only.  When
2210  * set it will cause associations that are idle for more than the
2211  * specified number of seconds to automatically close.  An association
2212  * being idle is defined an association that has NOT sent or received
2213  * user data.  The special value of '0' indicates that no automatic
2214  * close of any associations should be performed.  The option expects an
2215  * integer defining the number of seconds of idle time before an
2216  * association is closed.
2217  */
2218 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2219 				     unsigned int optlen)
2220 {
2221 	struct sctp_sock *sp = sctp_sk(sk);
2222 	struct net *net = sock_net(sk);
2223 
2224 	/* Applicable to UDP-style socket only */
2225 	if (sctp_style(sk, TCP))
2226 		return -EOPNOTSUPP;
2227 	if (optlen != sizeof(int))
2228 		return -EINVAL;
2229 	if (copy_from_user(&sp->autoclose, optval, optlen))
2230 		return -EFAULT;
2231 
2232 	if (sp->autoclose > net->sctp.max_autoclose)
2233 		sp->autoclose = net->sctp.max_autoclose;
2234 
2235 	return 0;
2236 }
2237 
2238 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2239  *
2240  * Applications can enable or disable heartbeats for any peer address of
2241  * an association, modify an address's heartbeat interval, force a
2242  * heartbeat to be sent immediately, and adjust the address's maximum
2243  * number of retransmissions sent before an address is considered
2244  * unreachable.  The following structure is used to access and modify an
2245  * address's parameters:
2246  *
2247  *  struct sctp_paddrparams {
2248  *     sctp_assoc_t            spp_assoc_id;
2249  *     struct sockaddr_storage spp_address;
2250  *     uint32_t                spp_hbinterval;
2251  *     uint16_t                spp_pathmaxrxt;
2252  *     uint32_t                spp_pathmtu;
2253  *     uint32_t                spp_sackdelay;
2254  *     uint32_t                spp_flags;
2255  * };
2256  *
2257  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2258  *                     application, and identifies the association for
2259  *                     this query.
2260  *   spp_address     - This specifies which address is of interest.
2261  *   spp_hbinterval  - This contains the value of the heartbeat interval,
2262  *                     in milliseconds.  If a  value of zero
2263  *                     is present in this field then no changes are to
2264  *                     be made to this parameter.
2265  *   spp_pathmaxrxt  - This contains the maximum number of
2266  *                     retransmissions before this address shall be
2267  *                     considered unreachable. If a  value of zero
2268  *                     is present in this field then no changes are to
2269  *                     be made to this parameter.
2270  *   spp_pathmtu     - When Path MTU discovery is disabled the value
2271  *                     specified here will be the "fixed" path mtu.
2272  *                     Note that if the spp_address field is empty
2273  *                     then all associations on this address will
2274  *                     have this fixed path mtu set upon them.
2275  *
2276  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2277  *                     the number of milliseconds that sacks will be delayed
2278  *                     for. This value will apply to all addresses of an
2279  *                     association if the spp_address field is empty. Note
2280  *                     also, that if delayed sack is enabled and this
2281  *                     value is set to 0, no change is made to the last
2282  *                     recorded delayed sack timer value.
2283  *
2284  *   spp_flags       - These flags are used to control various features
2285  *                     on an association. The flag field may contain
2286  *                     zero or more of the following options.
2287  *
2288  *                     SPP_HB_ENABLE  - Enable heartbeats on the
2289  *                     specified address. Note that if the address
2290  *                     field is empty all addresses for the association
2291  *                     have heartbeats enabled upon them.
2292  *
2293  *                     SPP_HB_DISABLE - Disable heartbeats on the
2294  *                     speicifed address. Note that if the address
2295  *                     field is empty all addresses for the association
2296  *                     will have their heartbeats disabled. Note also
2297  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2298  *                     mutually exclusive, only one of these two should
2299  *                     be specified. Enabling both fields will have
2300  *                     undetermined results.
2301  *
2302  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2303  *                     to be made immediately.
2304  *
2305  *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2306  *                     heartbeat delayis to be set to the value of 0
2307  *                     milliseconds.
2308  *
2309  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2310  *                     discovery upon the specified address. Note that
2311  *                     if the address feild is empty then all addresses
2312  *                     on the association are effected.
2313  *
2314  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2315  *                     discovery upon the specified address. Note that
2316  *                     if the address feild is empty then all addresses
2317  *                     on the association are effected. Not also that
2318  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2319  *                     exclusive. Enabling both will have undetermined
2320  *                     results.
2321  *
2322  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2323  *                     on delayed sack. The time specified in spp_sackdelay
2324  *                     is used to specify the sack delay for this address. Note
2325  *                     that if spp_address is empty then all addresses will
2326  *                     enable delayed sack and take on the sack delay
2327  *                     value specified in spp_sackdelay.
2328  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2329  *                     off delayed sack. If the spp_address field is blank then
2330  *                     delayed sack is disabled for the entire association. Note
2331  *                     also that this field is mutually exclusive to
2332  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2333  *                     results.
2334  */
2335 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2336 				       struct sctp_transport   *trans,
2337 				       struct sctp_association *asoc,
2338 				       struct sctp_sock        *sp,
2339 				       int                      hb_change,
2340 				       int                      pmtud_change,
2341 				       int                      sackdelay_change)
2342 {
2343 	int error;
2344 
2345 	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2346 		struct net *net = sock_net(trans->asoc->base.sk);
2347 
2348 		error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2349 		if (error)
2350 			return error;
2351 	}
2352 
2353 	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2354 	 * this field is ignored.  Note also that a value of zero indicates
2355 	 * the current setting should be left unchanged.
2356 	 */
2357 	if (params->spp_flags & SPP_HB_ENABLE) {
2358 
2359 		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2360 		 * set.  This lets us use 0 value when this flag
2361 		 * is set.
2362 		 */
2363 		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2364 			params->spp_hbinterval = 0;
2365 
2366 		if (params->spp_hbinterval ||
2367 		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2368 			if (trans) {
2369 				trans->hbinterval =
2370 				    msecs_to_jiffies(params->spp_hbinterval);
2371 			} else if (asoc) {
2372 				asoc->hbinterval =
2373 				    msecs_to_jiffies(params->spp_hbinterval);
2374 			} else {
2375 				sp->hbinterval = params->spp_hbinterval;
2376 			}
2377 		}
2378 	}
2379 
2380 	if (hb_change) {
2381 		if (trans) {
2382 			trans->param_flags =
2383 				(trans->param_flags & ~SPP_HB) | hb_change;
2384 		} else if (asoc) {
2385 			asoc->param_flags =
2386 				(asoc->param_flags & ~SPP_HB) | hb_change;
2387 		} else {
2388 			sp->param_flags =
2389 				(sp->param_flags & ~SPP_HB) | hb_change;
2390 		}
2391 	}
2392 
2393 	/* When Path MTU discovery is disabled the value specified here will
2394 	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2395 	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2396 	 * effect).
2397 	 */
2398 	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2399 		if (trans) {
2400 			trans->pathmtu = params->spp_pathmtu;
2401 			sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2402 		} else if (asoc) {
2403 			asoc->pathmtu = params->spp_pathmtu;
2404 			sctp_frag_point(asoc, params->spp_pathmtu);
2405 		} else {
2406 			sp->pathmtu = params->spp_pathmtu;
2407 		}
2408 	}
2409 
2410 	if (pmtud_change) {
2411 		if (trans) {
2412 			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2413 				(params->spp_flags & SPP_PMTUD_ENABLE);
2414 			trans->param_flags =
2415 				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2416 			if (update) {
2417 				sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2418 				sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2419 			}
2420 		} else if (asoc) {
2421 			asoc->param_flags =
2422 				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2423 		} else {
2424 			sp->param_flags =
2425 				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2426 		}
2427 	}
2428 
2429 	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2430 	 * value of this field is ignored.  Note also that a value of zero
2431 	 * indicates the current setting should be left unchanged.
2432 	 */
2433 	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2434 		if (trans) {
2435 			trans->sackdelay =
2436 				msecs_to_jiffies(params->spp_sackdelay);
2437 		} else if (asoc) {
2438 			asoc->sackdelay =
2439 				msecs_to_jiffies(params->spp_sackdelay);
2440 		} else {
2441 			sp->sackdelay = params->spp_sackdelay;
2442 		}
2443 	}
2444 
2445 	if (sackdelay_change) {
2446 		if (trans) {
2447 			trans->param_flags =
2448 				(trans->param_flags & ~SPP_SACKDELAY) |
2449 				sackdelay_change;
2450 		} else if (asoc) {
2451 			asoc->param_flags =
2452 				(asoc->param_flags & ~SPP_SACKDELAY) |
2453 				sackdelay_change;
2454 		} else {
2455 			sp->param_flags =
2456 				(sp->param_flags & ~SPP_SACKDELAY) |
2457 				sackdelay_change;
2458 		}
2459 	}
2460 
2461 	/* Note that a value of zero indicates the current setting should be
2462 	   left unchanged.
2463 	 */
2464 	if (params->spp_pathmaxrxt) {
2465 		if (trans) {
2466 			trans->pathmaxrxt = params->spp_pathmaxrxt;
2467 		} else if (asoc) {
2468 			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2469 		} else {
2470 			sp->pathmaxrxt = params->spp_pathmaxrxt;
2471 		}
2472 	}
2473 
2474 	return 0;
2475 }
2476 
2477 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2478 					    char __user *optval,
2479 					    unsigned int optlen)
2480 {
2481 	struct sctp_paddrparams  params;
2482 	struct sctp_transport   *trans = NULL;
2483 	struct sctp_association *asoc = NULL;
2484 	struct sctp_sock        *sp = sctp_sk(sk);
2485 	int error;
2486 	int hb_change, pmtud_change, sackdelay_change;
2487 
2488 	if (optlen != sizeof(struct sctp_paddrparams))
2489 		return -EINVAL;
2490 
2491 	if (copy_from_user(&params, optval, optlen))
2492 		return -EFAULT;
2493 
2494 	/* Validate flags and value parameters. */
2495 	hb_change        = params.spp_flags & SPP_HB;
2496 	pmtud_change     = params.spp_flags & SPP_PMTUD;
2497 	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2498 
2499 	if (hb_change        == SPP_HB ||
2500 	    pmtud_change     == SPP_PMTUD ||
2501 	    sackdelay_change == SPP_SACKDELAY ||
2502 	    params.spp_sackdelay > 500 ||
2503 	    (params.spp_pathmtu &&
2504 	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2505 		return -EINVAL;
2506 
2507 	/* If an address other than INADDR_ANY is specified, and
2508 	 * no transport is found, then the request is invalid.
2509 	 */
2510 	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2511 		trans = sctp_addr_id2transport(sk, &params.spp_address,
2512 					       params.spp_assoc_id);
2513 		if (!trans)
2514 			return -EINVAL;
2515 	}
2516 
2517 	/* Get association, if assoc_id != 0 and the socket is a one
2518 	 * to many style socket, and an association was not found, then
2519 	 * the id was invalid.
2520 	 */
2521 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2522 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2523 		return -EINVAL;
2524 
2525 	/* Heartbeat demand can only be sent on a transport or
2526 	 * association, but not a socket.
2527 	 */
2528 	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2529 		return -EINVAL;
2530 
2531 	/* Process parameters. */
2532 	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2533 					    hb_change, pmtud_change,
2534 					    sackdelay_change);
2535 
2536 	if (error)
2537 		return error;
2538 
2539 	/* If changes are for association, also apply parameters to each
2540 	 * transport.
2541 	 */
2542 	if (!trans && asoc) {
2543 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2544 				transports) {
2545 			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2546 						    hb_change, pmtud_change,
2547 						    sackdelay_change);
2548 		}
2549 	}
2550 
2551 	return 0;
2552 }
2553 
2554 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2555 {
2556 	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2557 }
2558 
2559 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2560 {
2561 	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2562 }
2563 
2564 /*
2565  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2566  *
2567  * This option will effect the way delayed acks are performed.  This
2568  * option allows you to get or set the delayed ack time, in
2569  * milliseconds.  It also allows changing the delayed ack frequency.
2570  * Changing the frequency to 1 disables the delayed sack algorithm.  If
2571  * the assoc_id is 0, then this sets or gets the endpoints default
2572  * values.  If the assoc_id field is non-zero, then the set or get
2573  * effects the specified association for the one to many model (the
2574  * assoc_id field is ignored by the one to one model).  Note that if
2575  * sack_delay or sack_freq are 0 when setting this option, then the
2576  * current values will remain unchanged.
2577  *
2578  * struct sctp_sack_info {
2579  *     sctp_assoc_t            sack_assoc_id;
2580  *     uint32_t                sack_delay;
2581  *     uint32_t                sack_freq;
2582  * };
2583  *
2584  * sack_assoc_id -  This parameter, indicates which association the user
2585  *    is performing an action upon.  Note that if this field's value is
2586  *    zero then the endpoints default value is changed (effecting future
2587  *    associations only).
2588  *
2589  * sack_delay -  This parameter contains the number of milliseconds that
2590  *    the user is requesting the delayed ACK timer be set to.  Note that
2591  *    this value is defined in the standard to be between 200 and 500
2592  *    milliseconds.
2593  *
2594  * sack_freq -  This parameter contains the number of packets that must
2595  *    be received before a sack is sent without waiting for the delay
2596  *    timer to expire.  The default value for this is 2, setting this
2597  *    value to 1 will disable the delayed sack algorithm.
2598  */
2599 
2600 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2601 				       char __user *optval, unsigned int optlen)
2602 {
2603 	struct sctp_sack_info    params;
2604 	struct sctp_transport   *trans = NULL;
2605 	struct sctp_association *asoc = NULL;
2606 	struct sctp_sock        *sp = sctp_sk(sk);
2607 
2608 	if (optlen == sizeof(struct sctp_sack_info)) {
2609 		if (copy_from_user(&params, optval, optlen))
2610 			return -EFAULT;
2611 
2612 		if (params.sack_delay == 0 && params.sack_freq == 0)
2613 			return 0;
2614 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2615 		pr_warn_ratelimited(DEPRECATED
2616 				    "%s (pid %d) "
2617 				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2618 				    "Use struct sctp_sack_info instead\n",
2619 				    current->comm, task_pid_nr(current));
2620 		if (copy_from_user(&params, optval, optlen))
2621 			return -EFAULT;
2622 
2623 		if (params.sack_delay == 0)
2624 			params.sack_freq = 1;
2625 		else
2626 			params.sack_freq = 0;
2627 	} else
2628 		return -EINVAL;
2629 
2630 	/* Validate value parameter. */
2631 	if (params.sack_delay > 500)
2632 		return -EINVAL;
2633 
2634 	/* Get association, if sack_assoc_id != 0 and the socket is a one
2635 	 * to many style socket, and an association was not found, then
2636 	 * the id was invalid.
2637 	 */
2638 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2639 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2640 		return -EINVAL;
2641 
2642 	if (params.sack_delay) {
2643 		if (asoc) {
2644 			asoc->sackdelay =
2645 				msecs_to_jiffies(params.sack_delay);
2646 			asoc->param_flags =
2647 				sctp_spp_sackdelay_enable(asoc->param_flags);
2648 		} else {
2649 			sp->sackdelay = params.sack_delay;
2650 			sp->param_flags =
2651 				sctp_spp_sackdelay_enable(sp->param_flags);
2652 		}
2653 	}
2654 
2655 	if (params.sack_freq == 1) {
2656 		if (asoc) {
2657 			asoc->param_flags =
2658 				sctp_spp_sackdelay_disable(asoc->param_flags);
2659 		} else {
2660 			sp->param_flags =
2661 				sctp_spp_sackdelay_disable(sp->param_flags);
2662 		}
2663 	} else if (params.sack_freq > 1) {
2664 		if (asoc) {
2665 			asoc->sackfreq = params.sack_freq;
2666 			asoc->param_flags =
2667 				sctp_spp_sackdelay_enable(asoc->param_flags);
2668 		} else {
2669 			sp->sackfreq = params.sack_freq;
2670 			sp->param_flags =
2671 				sctp_spp_sackdelay_enable(sp->param_flags);
2672 		}
2673 	}
2674 
2675 	/* If change is for association, also apply to each transport. */
2676 	if (asoc) {
2677 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2678 				transports) {
2679 			if (params.sack_delay) {
2680 				trans->sackdelay =
2681 					msecs_to_jiffies(params.sack_delay);
2682 				trans->param_flags =
2683 					sctp_spp_sackdelay_enable(trans->param_flags);
2684 			}
2685 			if (params.sack_freq == 1) {
2686 				trans->param_flags =
2687 					sctp_spp_sackdelay_disable(trans->param_flags);
2688 			} else if (params.sack_freq > 1) {
2689 				trans->sackfreq = params.sack_freq;
2690 				trans->param_flags =
2691 					sctp_spp_sackdelay_enable(trans->param_flags);
2692 			}
2693 		}
2694 	}
2695 
2696 	return 0;
2697 }
2698 
2699 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2700  *
2701  * Applications can specify protocol parameters for the default association
2702  * initialization.  The option name argument to setsockopt() and getsockopt()
2703  * is SCTP_INITMSG.
2704  *
2705  * Setting initialization parameters is effective only on an unconnected
2706  * socket (for UDP-style sockets only future associations are effected
2707  * by the change).  With TCP-style sockets, this option is inherited by
2708  * sockets derived from a listener socket.
2709  */
2710 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2711 {
2712 	struct sctp_initmsg sinit;
2713 	struct sctp_sock *sp = sctp_sk(sk);
2714 
2715 	if (optlen != sizeof(struct sctp_initmsg))
2716 		return -EINVAL;
2717 	if (copy_from_user(&sinit, optval, optlen))
2718 		return -EFAULT;
2719 
2720 	if (sinit.sinit_num_ostreams)
2721 		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2722 	if (sinit.sinit_max_instreams)
2723 		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2724 	if (sinit.sinit_max_attempts)
2725 		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2726 	if (sinit.sinit_max_init_timeo)
2727 		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2728 
2729 	return 0;
2730 }
2731 
2732 /*
2733  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2734  *
2735  *   Applications that wish to use the sendto() system call may wish to
2736  *   specify a default set of parameters that would normally be supplied
2737  *   through the inclusion of ancillary data.  This socket option allows
2738  *   such an application to set the default sctp_sndrcvinfo structure.
2739  *   The application that wishes to use this socket option simply passes
2740  *   in to this call the sctp_sndrcvinfo structure defined in Section
2741  *   5.2.2) The input parameters accepted by this call include
2742  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2743  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2744  *   to this call if the caller is using the UDP model.
2745  */
2746 static int sctp_setsockopt_default_send_param(struct sock *sk,
2747 					      char __user *optval,
2748 					      unsigned int optlen)
2749 {
2750 	struct sctp_sndrcvinfo info;
2751 	struct sctp_association *asoc;
2752 	struct sctp_sock *sp = sctp_sk(sk);
2753 
2754 	if (optlen != sizeof(struct sctp_sndrcvinfo))
2755 		return -EINVAL;
2756 	if (copy_from_user(&info, optval, optlen))
2757 		return -EFAULT;
2758 
2759 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2760 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2761 		return -EINVAL;
2762 
2763 	if (asoc) {
2764 		asoc->default_stream = info.sinfo_stream;
2765 		asoc->default_flags = info.sinfo_flags;
2766 		asoc->default_ppid = info.sinfo_ppid;
2767 		asoc->default_context = info.sinfo_context;
2768 		asoc->default_timetolive = info.sinfo_timetolive;
2769 	} else {
2770 		sp->default_stream = info.sinfo_stream;
2771 		sp->default_flags = info.sinfo_flags;
2772 		sp->default_ppid = info.sinfo_ppid;
2773 		sp->default_context = info.sinfo_context;
2774 		sp->default_timetolive = info.sinfo_timetolive;
2775 	}
2776 
2777 	return 0;
2778 }
2779 
2780 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2781  *
2782  * Requests that the local SCTP stack use the enclosed peer address as
2783  * the association primary.  The enclosed address must be one of the
2784  * association peer's addresses.
2785  */
2786 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2787 					unsigned int optlen)
2788 {
2789 	struct sctp_prim prim;
2790 	struct sctp_transport *trans;
2791 
2792 	if (optlen != sizeof(struct sctp_prim))
2793 		return -EINVAL;
2794 
2795 	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2796 		return -EFAULT;
2797 
2798 	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2799 	if (!trans)
2800 		return -EINVAL;
2801 
2802 	sctp_assoc_set_primary(trans->asoc, trans);
2803 
2804 	return 0;
2805 }
2806 
2807 /*
2808  * 7.1.5 SCTP_NODELAY
2809  *
2810  * Turn on/off any Nagle-like algorithm.  This means that packets are
2811  * generally sent as soon as possible and no unnecessary delays are
2812  * introduced, at the cost of more packets in the network.  Expects an
2813  *  integer boolean flag.
2814  */
2815 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2816 				   unsigned int optlen)
2817 {
2818 	int val;
2819 
2820 	if (optlen < sizeof(int))
2821 		return -EINVAL;
2822 	if (get_user(val, (int __user *)optval))
2823 		return -EFAULT;
2824 
2825 	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2826 	return 0;
2827 }
2828 
2829 /*
2830  *
2831  * 7.1.1 SCTP_RTOINFO
2832  *
2833  * The protocol parameters used to initialize and bound retransmission
2834  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2835  * and modify these parameters.
2836  * All parameters are time values, in milliseconds.  A value of 0, when
2837  * modifying the parameters, indicates that the current value should not
2838  * be changed.
2839  *
2840  */
2841 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2842 {
2843 	struct sctp_rtoinfo rtoinfo;
2844 	struct sctp_association *asoc;
2845 	unsigned long rto_min, rto_max;
2846 	struct sctp_sock *sp = sctp_sk(sk);
2847 
2848 	if (optlen != sizeof (struct sctp_rtoinfo))
2849 		return -EINVAL;
2850 
2851 	if (copy_from_user(&rtoinfo, optval, optlen))
2852 		return -EFAULT;
2853 
2854 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2855 
2856 	/* Set the values to the specific association */
2857 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2858 		return -EINVAL;
2859 
2860 	rto_max = rtoinfo.srto_max;
2861 	rto_min = rtoinfo.srto_min;
2862 
2863 	if (rto_max)
2864 		rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2865 	else
2866 		rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2867 
2868 	if (rto_min)
2869 		rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2870 	else
2871 		rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2872 
2873 	if (rto_min > rto_max)
2874 		return -EINVAL;
2875 
2876 	if (asoc) {
2877 		if (rtoinfo.srto_initial != 0)
2878 			asoc->rto_initial =
2879 				msecs_to_jiffies(rtoinfo.srto_initial);
2880 		asoc->rto_max = rto_max;
2881 		asoc->rto_min = rto_min;
2882 	} else {
2883 		/* If there is no association or the association-id = 0
2884 		 * set the values to the endpoint.
2885 		 */
2886 		if (rtoinfo.srto_initial != 0)
2887 			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2888 		sp->rtoinfo.srto_max = rto_max;
2889 		sp->rtoinfo.srto_min = rto_min;
2890 	}
2891 
2892 	return 0;
2893 }
2894 
2895 /*
2896  *
2897  * 7.1.2 SCTP_ASSOCINFO
2898  *
2899  * This option is used to tune the maximum retransmission attempts
2900  * of the association.
2901  * Returns an error if the new association retransmission value is
2902  * greater than the sum of the retransmission value  of the peer.
2903  * See [SCTP] for more information.
2904  *
2905  */
2906 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2907 {
2908 
2909 	struct sctp_assocparams assocparams;
2910 	struct sctp_association *asoc;
2911 
2912 	if (optlen != sizeof(struct sctp_assocparams))
2913 		return -EINVAL;
2914 	if (copy_from_user(&assocparams, optval, optlen))
2915 		return -EFAULT;
2916 
2917 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2918 
2919 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2920 		return -EINVAL;
2921 
2922 	/* Set the values to the specific association */
2923 	if (asoc) {
2924 		if (assocparams.sasoc_asocmaxrxt != 0) {
2925 			__u32 path_sum = 0;
2926 			int   paths = 0;
2927 			struct sctp_transport *peer_addr;
2928 
2929 			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2930 					transports) {
2931 				path_sum += peer_addr->pathmaxrxt;
2932 				paths++;
2933 			}
2934 
2935 			/* Only validate asocmaxrxt if we have more than
2936 			 * one path/transport.  We do this because path
2937 			 * retransmissions are only counted when we have more
2938 			 * then one path.
2939 			 */
2940 			if (paths > 1 &&
2941 			    assocparams.sasoc_asocmaxrxt > path_sum)
2942 				return -EINVAL;
2943 
2944 			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2945 		}
2946 
2947 		if (assocparams.sasoc_cookie_life != 0)
2948 			asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2949 	} else {
2950 		/* Set the values to the endpoint */
2951 		struct sctp_sock *sp = sctp_sk(sk);
2952 
2953 		if (assocparams.sasoc_asocmaxrxt != 0)
2954 			sp->assocparams.sasoc_asocmaxrxt =
2955 						assocparams.sasoc_asocmaxrxt;
2956 		if (assocparams.sasoc_cookie_life != 0)
2957 			sp->assocparams.sasoc_cookie_life =
2958 						assocparams.sasoc_cookie_life;
2959 	}
2960 	return 0;
2961 }
2962 
2963 /*
2964  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2965  *
2966  * This socket option is a boolean flag which turns on or off mapped V4
2967  * addresses.  If this option is turned on and the socket is type
2968  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2969  * If this option is turned off, then no mapping will be done of V4
2970  * addresses and a user will receive both PF_INET6 and PF_INET type
2971  * addresses on the socket.
2972  */
2973 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2974 {
2975 	int val;
2976 	struct sctp_sock *sp = sctp_sk(sk);
2977 
2978 	if (optlen < sizeof(int))
2979 		return -EINVAL;
2980 	if (get_user(val, (int __user *)optval))
2981 		return -EFAULT;
2982 	if (val)
2983 		sp->v4mapped = 1;
2984 	else
2985 		sp->v4mapped = 0;
2986 
2987 	return 0;
2988 }
2989 
2990 /*
2991  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2992  * This option will get or set the maximum size to put in any outgoing
2993  * SCTP DATA chunk.  If a message is larger than this size it will be
2994  * fragmented by SCTP into the specified size.  Note that the underlying
2995  * SCTP implementation may fragment into smaller sized chunks when the
2996  * PMTU of the underlying association is smaller than the value set by
2997  * the user.  The default value for this option is '0' which indicates
2998  * the user is NOT limiting fragmentation and only the PMTU will effect
2999  * SCTP's choice of DATA chunk size.  Note also that values set larger
3000  * than the maximum size of an IP datagram will effectively let SCTP
3001  * control fragmentation (i.e. the same as setting this option to 0).
3002  *
3003  * The following structure is used to access and modify this parameter:
3004  *
3005  * struct sctp_assoc_value {
3006  *   sctp_assoc_t assoc_id;
3007  *   uint32_t assoc_value;
3008  * };
3009  *
3010  * assoc_id:  This parameter is ignored for one-to-one style sockets.
3011  *    For one-to-many style sockets this parameter indicates which
3012  *    association the user is performing an action upon.  Note that if
3013  *    this field's value is zero then the endpoints default value is
3014  *    changed (effecting future associations only).
3015  * assoc_value:  This parameter specifies the maximum size in bytes.
3016  */
3017 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3018 {
3019 	struct sctp_assoc_value params;
3020 	struct sctp_association *asoc;
3021 	struct sctp_sock *sp = sctp_sk(sk);
3022 	int val;
3023 
3024 	if (optlen == sizeof(int)) {
3025 		pr_warn_ratelimited(DEPRECATED
3026 				    "%s (pid %d) "
3027 				    "Use of int in maxseg socket option.\n"
3028 				    "Use struct sctp_assoc_value instead\n",
3029 				    current->comm, task_pid_nr(current));
3030 		if (copy_from_user(&val, optval, optlen))
3031 			return -EFAULT;
3032 		params.assoc_id = 0;
3033 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3034 		if (copy_from_user(&params, optval, optlen))
3035 			return -EFAULT;
3036 		val = params.assoc_value;
3037 	} else
3038 		return -EINVAL;
3039 
3040 	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3041 		return -EINVAL;
3042 
3043 	asoc = sctp_id2assoc(sk, params.assoc_id);
3044 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3045 		return -EINVAL;
3046 
3047 	if (asoc) {
3048 		if (val == 0) {
3049 			val = asoc->pathmtu;
3050 			val -= sp->pf->af->net_header_len;
3051 			val -= sizeof(struct sctphdr) +
3052 					sizeof(struct sctp_data_chunk);
3053 		}
3054 		asoc->user_frag = val;
3055 		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3056 	} else {
3057 		sp->user_frag = val;
3058 	}
3059 
3060 	return 0;
3061 }
3062 
3063 
3064 /*
3065  *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3066  *
3067  *   Requests that the peer mark the enclosed address as the association
3068  *   primary. The enclosed address must be one of the association's
3069  *   locally bound addresses. The following structure is used to make a
3070  *   set primary request:
3071  */
3072 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3073 					     unsigned int optlen)
3074 {
3075 	struct net *net = sock_net(sk);
3076 	struct sctp_sock	*sp;
3077 	struct sctp_association	*asoc = NULL;
3078 	struct sctp_setpeerprim	prim;
3079 	struct sctp_chunk	*chunk;
3080 	struct sctp_af		*af;
3081 	int 			err;
3082 
3083 	sp = sctp_sk(sk);
3084 
3085 	if (!net->sctp.addip_enable)
3086 		return -EPERM;
3087 
3088 	if (optlen != sizeof(struct sctp_setpeerprim))
3089 		return -EINVAL;
3090 
3091 	if (copy_from_user(&prim, optval, optlen))
3092 		return -EFAULT;
3093 
3094 	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3095 	if (!asoc)
3096 		return -EINVAL;
3097 
3098 	if (!asoc->peer.asconf_capable)
3099 		return -EPERM;
3100 
3101 	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3102 		return -EPERM;
3103 
3104 	if (!sctp_state(asoc, ESTABLISHED))
3105 		return -ENOTCONN;
3106 
3107 	af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3108 	if (!af)
3109 		return -EINVAL;
3110 
3111 	if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3112 		return -EADDRNOTAVAIL;
3113 
3114 	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3115 		return -EADDRNOTAVAIL;
3116 
3117 	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
3118 	chunk = sctp_make_asconf_set_prim(asoc,
3119 					  (union sctp_addr *)&prim.sspp_addr);
3120 	if (!chunk)
3121 		return -ENOMEM;
3122 
3123 	err = sctp_send_asconf(asoc, chunk);
3124 
3125 	pr_debug("%s: we set peer primary addr primitively\n", __func__);
3126 
3127 	return err;
3128 }
3129 
3130 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3131 					    unsigned int optlen)
3132 {
3133 	struct sctp_setadaptation adaptation;
3134 
3135 	if (optlen != sizeof(struct sctp_setadaptation))
3136 		return -EINVAL;
3137 	if (copy_from_user(&adaptation, optval, optlen))
3138 		return -EFAULT;
3139 
3140 	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3141 
3142 	return 0;
3143 }
3144 
3145 /*
3146  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
3147  *
3148  * The context field in the sctp_sndrcvinfo structure is normally only
3149  * used when a failed message is retrieved holding the value that was
3150  * sent down on the actual send call.  This option allows the setting of
3151  * a default context on an association basis that will be received on
3152  * reading messages from the peer.  This is especially helpful in the
3153  * one-2-many model for an application to keep some reference to an
3154  * internal state machine that is processing messages on the
3155  * association.  Note that the setting of this value only effects
3156  * received messages from the peer and does not effect the value that is
3157  * saved with outbound messages.
3158  */
3159 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3160 				   unsigned int optlen)
3161 {
3162 	struct sctp_assoc_value params;
3163 	struct sctp_sock *sp;
3164 	struct sctp_association *asoc;
3165 
3166 	if (optlen != sizeof(struct sctp_assoc_value))
3167 		return -EINVAL;
3168 	if (copy_from_user(&params, optval, optlen))
3169 		return -EFAULT;
3170 
3171 	sp = sctp_sk(sk);
3172 
3173 	if (params.assoc_id != 0) {
3174 		asoc = sctp_id2assoc(sk, params.assoc_id);
3175 		if (!asoc)
3176 			return -EINVAL;
3177 		asoc->default_rcv_context = params.assoc_value;
3178 	} else {
3179 		sp->default_rcv_context = params.assoc_value;
3180 	}
3181 
3182 	return 0;
3183 }
3184 
3185 /*
3186  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3187  *
3188  * This options will at a minimum specify if the implementation is doing
3189  * fragmented interleave.  Fragmented interleave, for a one to many
3190  * socket, is when subsequent calls to receive a message may return
3191  * parts of messages from different associations.  Some implementations
3192  * may allow you to turn this value on or off.  If so, when turned off,
3193  * no fragment interleave will occur (which will cause a head of line
3194  * blocking amongst multiple associations sharing the same one to many
3195  * socket).  When this option is turned on, then each receive call may
3196  * come from a different association (thus the user must receive data
3197  * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3198  * association each receive belongs to.
3199  *
3200  * This option takes a boolean value.  A non-zero value indicates that
3201  * fragmented interleave is on.  A value of zero indicates that
3202  * fragmented interleave is off.
3203  *
3204  * Note that it is important that an implementation that allows this
3205  * option to be turned on, have it off by default.  Otherwise an unaware
3206  * application using the one to many model may become confused and act
3207  * incorrectly.
3208  */
3209 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3210 					       char __user *optval,
3211 					       unsigned int optlen)
3212 {
3213 	int val;
3214 
3215 	if (optlen != sizeof(int))
3216 		return -EINVAL;
3217 	if (get_user(val, (int __user *)optval))
3218 		return -EFAULT;
3219 
3220 	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3221 
3222 	return 0;
3223 }
3224 
3225 /*
3226  * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3227  *       (SCTP_PARTIAL_DELIVERY_POINT)
3228  *
3229  * This option will set or get the SCTP partial delivery point.  This
3230  * point is the size of a message where the partial delivery API will be
3231  * invoked to help free up rwnd space for the peer.  Setting this to a
3232  * lower value will cause partial deliveries to happen more often.  The
3233  * calls argument is an integer that sets or gets the partial delivery
3234  * point.  Note also that the call will fail if the user attempts to set
3235  * this value larger than the socket receive buffer size.
3236  *
3237  * Note that any single message having a length smaller than or equal to
3238  * the SCTP partial delivery point will be delivered in one single read
3239  * call as long as the user provided buffer is large enough to hold the
3240  * message.
3241  */
3242 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3243 						  char __user *optval,
3244 						  unsigned int optlen)
3245 {
3246 	u32 val;
3247 
3248 	if (optlen != sizeof(u32))
3249 		return -EINVAL;
3250 	if (get_user(val, (int __user *)optval))
3251 		return -EFAULT;
3252 
3253 	/* Note: We double the receive buffer from what the user sets
3254 	 * it to be, also initial rwnd is based on rcvbuf/2.
3255 	 */
3256 	if (val > (sk->sk_rcvbuf >> 1))
3257 		return -EINVAL;
3258 
3259 	sctp_sk(sk)->pd_point = val;
3260 
3261 	return 0; /* is this the right error code? */
3262 }
3263 
3264 /*
3265  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3266  *
3267  * This option will allow a user to change the maximum burst of packets
3268  * that can be emitted by this association.  Note that the default value
3269  * is 4, and some implementations may restrict this setting so that it
3270  * can only be lowered.
3271  *
3272  * NOTE: This text doesn't seem right.  Do this on a socket basis with
3273  * future associations inheriting the socket value.
3274  */
3275 static int sctp_setsockopt_maxburst(struct sock *sk,
3276 				    char __user *optval,
3277 				    unsigned int optlen)
3278 {
3279 	struct sctp_assoc_value params;
3280 	struct sctp_sock *sp;
3281 	struct sctp_association *asoc;
3282 	int val;
3283 	int assoc_id = 0;
3284 
3285 	if (optlen == sizeof(int)) {
3286 		pr_warn_ratelimited(DEPRECATED
3287 				    "%s (pid %d) "
3288 				    "Use of int in max_burst socket option deprecated.\n"
3289 				    "Use struct sctp_assoc_value instead\n",
3290 				    current->comm, task_pid_nr(current));
3291 		if (copy_from_user(&val, optval, optlen))
3292 			return -EFAULT;
3293 	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3294 		if (copy_from_user(&params, optval, optlen))
3295 			return -EFAULT;
3296 		val = params.assoc_value;
3297 		assoc_id = params.assoc_id;
3298 	} else
3299 		return -EINVAL;
3300 
3301 	sp = sctp_sk(sk);
3302 
3303 	if (assoc_id != 0) {
3304 		asoc = sctp_id2assoc(sk, assoc_id);
3305 		if (!asoc)
3306 			return -EINVAL;
3307 		asoc->max_burst = val;
3308 	} else
3309 		sp->max_burst = val;
3310 
3311 	return 0;
3312 }
3313 
3314 /*
3315  * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3316  *
3317  * This set option adds a chunk type that the user is requesting to be
3318  * received only in an authenticated way.  Changes to the list of chunks
3319  * will only effect future associations on the socket.
3320  */
3321 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3322 				      char __user *optval,
3323 				      unsigned int optlen)
3324 {
3325 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3326 	struct sctp_authchunk val;
3327 
3328 	if (!ep->auth_enable)
3329 		return -EACCES;
3330 
3331 	if (optlen != sizeof(struct sctp_authchunk))
3332 		return -EINVAL;
3333 	if (copy_from_user(&val, optval, optlen))
3334 		return -EFAULT;
3335 
3336 	switch (val.sauth_chunk) {
3337 	case SCTP_CID_INIT:
3338 	case SCTP_CID_INIT_ACK:
3339 	case SCTP_CID_SHUTDOWN_COMPLETE:
3340 	case SCTP_CID_AUTH:
3341 		return -EINVAL;
3342 	}
3343 
3344 	/* add this chunk id to the endpoint */
3345 	return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3346 }
3347 
3348 /*
3349  * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3350  *
3351  * This option gets or sets the list of HMAC algorithms that the local
3352  * endpoint requires the peer to use.
3353  */
3354 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3355 				      char __user *optval,
3356 				      unsigned int optlen)
3357 {
3358 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3359 	struct sctp_hmacalgo *hmacs;
3360 	u32 idents;
3361 	int err;
3362 
3363 	if (!ep->auth_enable)
3364 		return -EACCES;
3365 
3366 	if (optlen < sizeof(struct sctp_hmacalgo))
3367 		return -EINVAL;
3368 
3369 	hmacs = memdup_user(optval, optlen);
3370 	if (IS_ERR(hmacs))
3371 		return PTR_ERR(hmacs);
3372 
3373 	idents = hmacs->shmac_num_idents;
3374 	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3375 	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3376 		err = -EINVAL;
3377 		goto out;
3378 	}
3379 
3380 	err = sctp_auth_ep_set_hmacs(ep, hmacs);
3381 out:
3382 	kfree(hmacs);
3383 	return err;
3384 }
3385 
3386 /*
3387  * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3388  *
3389  * This option will set a shared secret key which is used to build an
3390  * association shared key.
3391  */
3392 static int sctp_setsockopt_auth_key(struct sock *sk,
3393 				    char __user *optval,
3394 				    unsigned int optlen)
3395 {
3396 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3397 	struct sctp_authkey *authkey;
3398 	struct sctp_association *asoc;
3399 	int ret;
3400 
3401 	if (!ep->auth_enable)
3402 		return -EACCES;
3403 
3404 	if (optlen <= sizeof(struct sctp_authkey))
3405 		return -EINVAL;
3406 
3407 	authkey = memdup_user(optval, optlen);
3408 	if (IS_ERR(authkey))
3409 		return PTR_ERR(authkey);
3410 
3411 	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3412 		ret = -EINVAL;
3413 		goto out;
3414 	}
3415 
3416 	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3417 	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3418 		ret = -EINVAL;
3419 		goto out;
3420 	}
3421 
3422 	ret = sctp_auth_set_key(ep, asoc, authkey);
3423 out:
3424 	kzfree(authkey);
3425 	return ret;
3426 }
3427 
3428 /*
3429  * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3430  *
3431  * This option will get or set the active shared key to be used to build
3432  * the association shared key.
3433  */
3434 static int sctp_setsockopt_active_key(struct sock *sk,
3435 				      char __user *optval,
3436 				      unsigned int optlen)
3437 {
3438 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3439 	struct sctp_authkeyid val;
3440 	struct sctp_association *asoc;
3441 
3442 	if (!ep->auth_enable)
3443 		return -EACCES;
3444 
3445 	if (optlen != sizeof(struct sctp_authkeyid))
3446 		return -EINVAL;
3447 	if (copy_from_user(&val, optval, optlen))
3448 		return -EFAULT;
3449 
3450 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3451 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3452 		return -EINVAL;
3453 
3454 	return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3455 }
3456 
3457 /*
3458  * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3459  *
3460  * This set option will delete a shared secret key from use.
3461  */
3462 static int sctp_setsockopt_del_key(struct sock *sk,
3463 				   char __user *optval,
3464 				   unsigned int optlen)
3465 {
3466 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3467 	struct sctp_authkeyid val;
3468 	struct sctp_association *asoc;
3469 
3470 	if (!ep->auth_enable)
3471 		return -EACCES;
3472 
3473 	if (optlen != sizeof(struct sctp_authkeyid))
3474 		return -EINVAL;
3475 	if (copy_from_user(&val, optval, optlen))
3476 		return -EFAULT;
3477 
3478 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3479 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3480 		return -EINVAL;
3481 
3482 	return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3483 
3484 }
3485 
3486 /*
3487  * 8.1.23 SCTP_AUTO_ASCONF
3488  *
3489  * This option will enable or disable the use of the automatic generation of
3490  * ASCONF chunks to add and delete addresses to an existing association.  Note
3491  * that this option has two caveats namely: a) it only affects sockets that
3492  * are bound to all addresses available to the SCTP stack, and b) the system
3493  * administrator may have an overriding control that turns the ASCONF feature
3494  * off no matter what setting the socket option may have.
3495  * This option expects an integer boolean flag, where a non-zero value turns on
3496  * the option, and a zero value turns off the option.
3497  * Note. In this implementation, socket operation overrides default parameter
3498  * being set by sysctl as well as FreeBSD implementation
3499  */
3500 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3501 					unsigned int optlen)
3502 {
3503 	int val;
3504 	struct sctp_sock *sp = sctp_sk(sk);
3505 
3506 	if (optlen < sizeof(int))
3507 		return -EINVAL;
3508 	if (get_user(val, (int __user *)optval))
3509 		return -EFAULT;
3510 	if (!sctp_is_ep_boundall(sk) && val)
3511 		return -EINVAL;
3512 	if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3513 		return 0;
3514 
3515 	if (val == 0 && sp->do_auto_asconf) {
3516 		list_del(&sp->auto_asconf_list);
3517 		sp->do_auto_asconf = 0;
3518 	} else if (val && !sp->do_auto_asconf) {
3519 		list_add_tail(&sp->auto_asconf_list,
3520 		    &sock_net(sk)->sctp.auto_asconf_splist);
3521 		sp->do_auto_asconf = 1;
3522 	}
3523 	return 0;
3524 }
3525 
3526 
3527 /*
3528  * SCTP_PEER_ADDR_THLDS
3529  *
3530  * This option allows us to alter the partially failed threshold for one or all
3531  * transports in an association.  See Section 6.1 of:
3532  * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3533  */
3534 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3535 					    char __user *optval,
3536 					    unsigned int optlen)
3537 {
3538 	struct sctp_paddrthlds val;
3539 	struct sctp_transport *trans;
3540 	struct sctp_association *asoc;
3541 
3542 	if (optlen < sizeof(struct sctp_paddrthlds))
3543 		return -EINVAL;
3544 	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3545 			   sizeof(struct sctp_paddrthlds)))
3546 		return -EFAULT;
3547 
3548 
3549 	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3550 		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3551 		if (!asoc)
3552 			return -ENOENT;
3553 		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3554 				    transports) {
3555 			if (val.spt_pathmaxrxt)
3556 				trans->pathmaxrxt = val.spt_pathmaxrxt;
3557 			trans->pf_retrans = val.spt_pathpfthld;
3558 		}
3559 
3560 		if (val.spt_pathmaxrxt)
3561 			asoc->pathmaxrxt = val.spt_pathmaxrxt;
3562 		asoc->pf_retrans = val.spt_pathpfthld;
3563 	} else {
3564 		trans = sctp_addr_id2transport(sk, &val.spt_address,
3565 					       val.spt_assoc_id);
3566 		if (!trans)
3567 			return -ENOENT;
3568 
3569 		if (val.spt_pathmaxrxt)
3570 			trans->pathmaxrxt = val.spt_pathmaxrxt;
3571 		trans->pf_retrans = val.spt_pathpfthld;
3572 	}
3573 
3574 	return 0;
3575 }
3576 
3577 /* API 6.2 setsockopt(), getsockopt()
3578  *
3579  * Applications use setsockopt() and getsockopt() to set or retrieve
3580  * socket options.  Socket options are used to change the default
3581  * behavior of sockets calls.  They are described in Section 7.
3582  *
3583  * The syntax is:
3584  *
3585  *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3586  *                    int __user *optlen);
3587  *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3588  *                    int optlen);
3589  *
3590  *   sd      - the socket descript.
3591  *   level   - set to IPPROTO_SCTP for all SCTP options.
3592  *   optname - the option name.
3593  *   optval  - the buffer to store the value of the option.
3594  *   optlen  - the size of the buffer.
3595  */
3596 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3597 			   char __user *optval, unsigned int optlen)
3598 {
3599 	int retval = 0;
3600 
3601 	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3602 
3603 	/* I can hardly begin to describe how wrong this is.  This is
3604 	 * so broken as to be worse than useless.  The API draft
3605 	 * REALLY is NOT helpful here...  I am not convinced that the
3606 	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3607 	 * are at all well-founded.
3608 	 */
3609 	if (level != SOL_SCTP) {
3610 		struct sctp_af *af = sctp_sk(sk)->pf->af;
3611 		retval = af->setsockopt(sk, level, optname, optval, optlen);
3612 		goto out_nounlock;
3613 	}
3614 
3615 	lock_sock(sk);
3616 
3617 	switch (optname) {
3618 	case SCTP_SOCKOPT_BINDX_ADD:
3619 		/* 'optlen' is the size of the addresses buffer. */
3620 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3621 					       optlen, SCTP_BINDX_ADD_ADDR);
3622 		break;
3623 
3624 	case SCTP_SOCKOPT_BINDX_REM:
3625 		/* 'optlen' is the size of the addresses buffer. */
3626 		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3627 					       optlen, SCTP_BINDX_REM_ADDR);
3628 		break;
3629 
3630 	case SCTP_SOCKOPT_CONNECTX_OLD:
3631 		/* 'optlen' is the size of the addresses buffer. */
3632 		retval = sctp_setsockopt_connectx_old(sk,
3633 					    (struct sockaddr __user *)optval,
3634 					    optlen);
3635 		break;
3636 
3637 	case SCTP_SOCKOPT_CONNECTX:
3638 		/* 'optlen' is the size of the addresses buffer. */
3639 		retval = sctp_setsockopt_connectx(sk,
3640 					    (struct sockaddr __user *)optval,
3641 					    optlen);
3642 		break;
3643 
3644 	case SCTP_DISABLE_FRAGMENTS:
3645 		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3646 		break;
3647 
3648 	case SCTP_EVENTS:
3649 		retval = sctp_setsockopt_events(sk, optval, optlen);
3650 		break;
3651 
3652 	case SCTP_AUTOCLOSE:
3653 		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3654 		break;
3655 
3656 	case SCTP_PEER_ADDR_PARAMS:
3657 		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3658 		break;
3659 
3660 	case SCTP_DELAYED_SACK:
3661 		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3662 		break;
3663 	case SCTP_PARTIAL_DELIVERY_POINT:
3664 		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3665 		break;
3666 
3667 	case SCTP_INITMSG:
3668 		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3669 		break;
3670 	case SCTP_DEFAULT_SEND_PARAM:
3671 		retval = sctp_setsockopt_default_send_param(sk, optval,
3672 							    optlen);
3673 		break;
3674 	case SCTP_PRIMARY_ADDR:
3675 		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3676 		break;
3677 	case SCTP_SET_PEER_PRIMARY_ADDR:
3678 		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3679 		break;
3680 	case SCTP_NODELAY:
3681 		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3682 		break;
3683 	case SCTP_RTOINFO:
3684 		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3685 		break;
3686 	case SCTP_ASSOCINFO:
3687 		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3688 		break;
3689 	case SCTP_I_WANT_MAPPED_V4_ADDR:
3690 		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3691 		break;
3692 	case SCTP_MAXSEG:
3693 		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3694 		break;
3695 	case SCTP_ADAPTATION_LAYER:
3696 		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3697 		break;
3698 	case SCTP_CONTEXT:
3699 		retval = sctp_setsockopt_context(sk, optval, optlen);
3700 		break;
3701 	case SCTP_FRAGMENT_INTERLEAVE:
3702 		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3703 		break;
3704 	case SCTP_MAX_BURST:
3705 		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3706 		break;
3707 	case SCTP_AUTH_CHUNK:
3708 		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3709 		break;
3710 	case SCTP_HMAC_IDENT:
3711 		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3712 		break;
3713 	case SCTP_AUTH_KEY:
3714 		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3715 		break;
3716 	case SCTP_AUTH_ACTIVE_KEY:
3717 		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3718 		break;
3719 	case SCTP_AUTH_DELETE_KEY:
3720 		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3721 		break;
3722 	case SCTP_AUTO_ASCONF:
3723 		retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3724 		break;
3725 	case SCTP_PEER_ADDR_THLDS:
3726 		retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3727 		break;
3728 	default:
3729 		retval = -ENOPROTOOPT;
3730 		break;
3731 	}
3732 
3733 	release_sock(sk);
3734 
3735 out_nounlock:
3736 	return retval;
3737 }
3738 
3739 /* API 3.1.6 connect() - UDP Style Syntax
3740  *
3741  * An application may use the connect() call in the UDP model to initiate an
3742  * association without sending data.
3743  *
3744  * The syntax is:
3745  *
3746  * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3747  *
3748  * sd: the socket descriptor to have a new association added to.
3749  *
3750  * nam: the address structure (either struct sockaddr_in or struct
3751  *    sockaddr_in6 defined in RFC2553 [7]).
3752  *
3753  * len: the size of the address.
3754  */
3755 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3756 			int addr_len)
3757 {
3758 	int err = 0;
3759 	struct sctp_af *af;
3760 
3761 	lock_sock(sk);
3762 
3763 	pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3764 		 addr, addr_len);
3765 
3766 	/* Validate addr_len before calling common connect/connectx routine. */
3767 	af = sctp_get_af_specific(addr->sa_family);
3768 	if (!af || addr_len < af->sockaddr_len) {
3769 		err = -EINVAL;
3770 	} else {
3771 		/* Pass correct addr len to common routine (so it knows there
3772 		 * is only one address being passed.
3773 		 */
3774 		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3775 	}
3776 
3777 	release_sock(sk);
3778 	return err;
3779 }
3780 
3781 /* FIXME: Write comments. */
3782 static int sctp_disconnect(struct sock *sk, int flags)
3783 {
3784 	return -EOPNOTSUPP; /* STUB */
3785 }
3786 
3787 /* 4.1.4 accept() - TCP Style Syntax
3788  *
3789  * Applications use accept() call to remove an established SCTP
3790  * association from the accept queue of the endpoint.  A new socket
3791  * descriptor will be returned from accept() to represent the newly
3792  * formed association.
3793  */
3794 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3795 {
3796 	struct sctp_sock *sp;
3797 	struct sctp_endpoint *ep;
3798 	struct sock *newsk = NULL;
3799 	struct sctp_association *asoc;
3800 	long timeo;
3801 	int error = 0;
3802 
3803 	lock_sock(sk);
3804 
3805 	sp = sctp_sk(sk);
3806 	ep = sp->ep;
3807 
3808 	if (!sctp_style(sk, TCP)) {
3809 		error = -EOPNOTSUPP;
3810 		goto out;
3811 	}
3812 
3813 	if (!sctp_sstate(sk, LISTENING)) {
3814 		error = -EINVAL;
3815 		goto out;
3816 	}
3817 
3818 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3819 
3820 	error = sctp_wait_for_accept(sk, timeo);
3821 	if (error)
3822 		goto out;
3823 
3824 	/* We treat the list of associations on the endpoint as the accept
3825 	 * queue and pick the first association on the list.
3826 	 */
3827 	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3828 
3829 	newsk = sp->pf->create_accept_sk(sk, asoc);
3830 	if (!newsk) {
3831 		error = -ENOMEM;
3832 		goto out;
3833 	}
3834 
3835 	/* Populate the fields of the newsk from the oldsk and migrate the
3836 	 * asoc to the newsk.
3837 	 */
3838 	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3839 
3840 out:
3841 	release_sock(sk);
3842 	*err = error;
3843 	return newsk;
3844 }
3845 
3846 /* The SCTP ioctl handler. */
3847 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3848 {
3849 	int rc = -ENOTCONN;
3850 
3851 	lock_sock(sk);
3852 
3853 	/*
3854 	 * SEQPACKET-style sockets in LISTENING state are valid, for
3855 	 * SCTP, so only discard TCP-style sockets in LISTENING state.
3856 	 */
3857 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3858 		goto out;
3859 
3860 	switch (cmd) {
3861 	case SIOCINQ: {
3862 		struct sk_buff *skb;
3863 		unsigned int amount = 0;
3864 
3865 		skb = skb_peek(&sk->sk_receive_queue);
3866 		if (skb != NULL) {
3867 			/*
3868 			 * We will only return the amount of this packet since
3869 			 * that is all that will be read.
3870 			 */
3871 			amount = skb->len;
3872 		}
3873 		rc = put_user(amount, (int __user *)arg);
3874 		break;
3875 	}
3876 	default:
3877 		rc = -ENOIOCTLCMD;
3878 		break;
3879 	}
3880 out:
3881 	release_sock(sk);
3882 	return rc;
3883 }
3884 
3885 /* This is the function which gets called during socket creation to
3886  * initialized the SCTP-specific portion of the sock.
3887  * The sock structure should already be zero-filled memory.
3888  */
3889 static int sctp_init_sock(struct sock *sk)
3890 {
3891 	struct net *net = sock_net(sk);
3892 	struct sctp_sock *sp;
3893 
3894 	pr_debug("%s: sk:%p\n", __func__, sk);
3895 
3896 	sp = sctp_sk(sk);
3897 
3898 	/* Initialize the SCTP per socket area.  */
3899 	switch (sk->sk_type) {
3900 	case SOCK_SEQPACKET:
3901 		sp->type = SCTP_SOCKET_UDP;
3902 		break;
3903 	case SOCK_STREAM:
3904 		sp->type = SCTP_SOCKET_TCP;
3905 		break;
3906 	default:
3907 		return -ESOCKTNOSUPPORT;
3908 	}
3909 
3910 	/* Initialize default send parameters. These parameters can be
3911 	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3912 	 */
3913 	sp->default_stream = 0;
3914 	sp->default_ppid = 0;
3915 	sp->default_flags = 0;
3916 	sp->default_context = 0;
3917 	sp->default_timetolive = 0;
3918 
3919 	sp->default_rcv_context = 0;
3920 	sp->max_burst = net->sctp.max_burst;
3921 
3922 	sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3923 
3924 	/* Initialize default setup parameters. These parameters
3925 	 * can be modified with the SCTP_INITMSG socket option or
3926 	 * overridden by the SCTP_INIT CMSG.
3927 	 */
3928 	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
3929 	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
3930 	sp->initmsg.sinit_max_attempts   = net->sctp.max_retrans_init;
3931 	sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3932 
3933 	/* Initialize default RTO related parameters.  These parameters can
3934 	 * be modified for with the SCTP_RTOINFO socket option.
3935 	 */
3936 	sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3937 	sp->rtoinfo.srto_max     = net->sctp.rto_max;
3938 	sp->rtoinfo.srto_min     = net->sctp.rto_min;
3939 
3940 	/* Initialize default association related parameters. These parameters
3941 	 * can be modified with the SCTP_ASSOCINFO socket option.
3942 	 */
3943 	sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3944 	sp->assocparams.sasoc_number_peer_destinations = 0;
3945 	sp->assocparams.sasoc_peer_rwnd = 0;
3946 	sp->assocparams.sasoc_local_rwnd = 0;
3947 	sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3948 
3949 	/* Initialize default event subscriptions. By default, all the
3950 	 * options are off.
3951 	 */
3952 	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3953 
3954 	/* Default Peer Address Parameters.  These defaults can
3955 	 * be modified via SCTP_PEER_ADDR_PARAMS
3956 	 */
3957 	sp->hbinterval  = net->sctp.hb_interval;
3958 	sp->pathmaxrxt  = net->sctp.max_retrans_path;
3959 	sp->pathmtu     = 0; /* allow default discovery */
3960 	sp->sackdelay   = net->sctp.sack_timeout;
3961 	sp->sackfreq	= 2;
3962 	sp->param_flags = SPP_HB_ENABLE |
3963 			  SPP_PMTUD_ENABLE |
3964 			  SPP_SACKDELAY_ENABLE;
3965 
3966 	/* If enabled no SCTP message fragmentation will be performed.
3967 	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3968 	 */
3969 	sp->disable_fragments = 0;
3970 
3971 	/* Enable Nagle algorithm by default.  */
3972 	sp->nodelay           = 0;
3973 
3974 	/* Enable by default. */
3975 	sp->v4mapped          = 1;
3976 
3977 	/* Auto-close idle associations after the configured
3978 	 * number of seconds.  A value of 0 disables this
3979 	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
3980 	 * for UDP-style sockets only.
3981 	 */
3982 	sp->autoclose         = 0;
3983 
3984 	/* User specified fragmentation limit. */
3985 	sp->user_frag         = 0;
3986 
3987 	sp->adaptation_ind = 0;
3988 
3989 	sp->pf = sctp_get_pf_specific(sk->sk_family);
3990 
3991 	/* Control variables for partial data delivery. */
3992 	atomic_set(&sp->pd_mode, 0);
3993 	skb_queue_head_init(&sp->pd_lobby);
3994 	sp->frag_interleave = 0;
3995 
3996 	/* Create a per socket endpoint structure.  Even if we
3997 	 * change the data structure relationships, this may still
3998 	 * be useful for storing pre-connect address information.
3999 	 */
4000 	sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4001 	if (!sp->ep)
4002 		return -ENOMEM;
4003 
4004 	sp->hmac = NULL;
4005 
4006 	sk->sk_destruct = sctp_destruct_sock;
4007 
4008 	SCTP_DBG_OBJCNT_INC(sock);
4009 
4010 	local_bh_disable();
4011 	percpu_counter_inc(&sctp_sockets_allocated);
4012 	sock_prot_inuse_add(net, sk->sk_prot, 1);
4013 	if (net->sctp.default_auto_asconf) {
4014 		list_add_tail(&sp->auto_asconf_list,
4015 		    &net->sctp.auto_asconf_splist);
4016 		sp->do_auto_asconf = 1;
4017 	} else
4018 		sp->do_auto_asconf = 0;
4019 	local_bh_enable();
4020 
4021 	return 0;
4022 }
4023 
4024 /* Cleanup any SCTP per socket resources.  */
4025 static void sctp_destroy_sock(struct sock *sk)
4026 {
4027 	struct sctp_sock *sp;
4028 
4029 	pr_debug("%s: sk:%p\n", __func__, sk);
4030 
4031 	/* Release our hold on the endpoint. */
4032 	sp = sctp_sk(sk);
4033 	/* This could happen during socket init, thus we bail out
4034 	 * early, since the rest of the below is not setup either.
4035 	 */
4036 	if (sp->ep == NULL)
4037 		return;
4038 
4039 	if (sp->do_auto_asconf) {
4040 		sp->do_auto_asconf = 0;
4041 		list_del(&sp->auto_asconf_list);
4042 	}
4043 	sctp_endpoint_free(sp->ep);
4044 	local_bh_disable();
4045 	percpu_counter_dec(&sctp_sockets_allocated);
4046 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4047 	local_bh_enable();
4048 }
4049 
4050 /* Triggered when there are no references on the socket anymore */
4051 static void sctp_destruct_sock(struct sock *sk)
4052 {
4053 	struct sctp_sock *sp = sctp_sk(sk);
4054 
4055 	/* Free up the HMAC transform. */
4056 	crypto_free_hash(sp->hmac);
4057 
4058 	inet_sock_destruct(sk);
4059 }
4060 
4061 /* API 4.1.7 shutdown() - TCP Style Syntax
4062  *     int shutdown(int socket, int how);
4063  *
4064  *     sd      - the socket descriptor of the association to be closed.
4065  *     how     - Specifies the type of shutdown.  The  values  are
4066  *               as follows:
4067  *               SHUT_RD
4068  *                     Disables further receive operations. No SCTP
4069  *                     protocol action is taken.
4070  *               SHUT_WR
4071  *                     Disables further send operations, and initiates
4072  *                     the SCTP shutdown sequence.
4073  *               SHUT_RDWR
4074  *                     Disables further send  and  receive  operations
4075  *                     and initiates the SCTP shutdown sequence.
4076  */
4077 static void sctp_shutdown(struct sock *sk, int how)
4078 {
4079 	struct net *net = sock_net(sk);
4080 	struct sctp_endpoint *ep;
4081 	struct sctp_association *asoc;
4082 
4083 	if (!sctp_style(sk, TCP))
4084 		return;
4085 
4086 	if (how & SEND_SHUTDOWN) {
4087 		ep = sctp_sk(sk)->ep;
4088 		if (!list_empty(&ep->asocs)) {
4089 			asoc = list_entry(ep->asocs.next,
4090 					  struct sctp_association, asocs);
4091 			sctp_primitive_SHUTDOWN(net, asoc, NULL);
4092 		}
4093 	}
4094 }
4095 
4096 /* 7.2.1 Association Status (SCTP_STATUS)
4097 
4098  * Applications can retrieve current status information about an
4099  * association, including association state, peer receiver window size,
4100  * number of unacked data chunks, and number of data chunks pending
4101  * receipt.  This information is read-only.
4102  */
4103 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4104 				       char __user *optval,
4105 				       int __user *optlen)
4106 {
4107 	struct sctp_status status;
4108 	struct sctp_association *asoc = NULL;
4109 	struct sctp_transport *transport;
4110 	sctp_assoc_t associd;
4111 	int retval = 0;
4112 
4113 	if (len < sizeof(status)) {
4114 		retval = -EINVAL;
4115 		goto out;
4116 	}
4117 
4118 	len = sizeof(status);
4119 	if (copy_from_user(&status, optval, len)) {
4120 		retval = -EFAULT;
4121 		goto out;
4122 	}
4123 
4124 	associd = status.sstat_assoc_id;
4125 	asoc = sctp_id2assoc(sk, associd);
4126 	if (!asoc) {
4127 		retval = -EINVAL;
4128 		goto out;
4129 	}
4130 
4131 	transport = asoc->peer.primary_path;
4132 
4133 	status.sstat_assoc_id = sctp_assoc2id(asoc);
4134 	status.sstat_state = asoc->state;
4135 	status.sstat_rwnd =  asoc->peer.rwnd;
4136 	status.sstat_unackdata = asoc->unack_data;
4137 
4138 	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4139 	status.sstat_instrms = asoc->c.sinit_max_instreams;
4140 	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4141 	status.sstat_fragmentation_point = asoc->frag_point;
4142 	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4143 	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4144 			transport->af_specific->sockaddr_len);
4145 	/* Map ipv4 address into v4-mapped-on-v6 address.  */
4146 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4147 		(union sctp_addr *)&status.sstat_primary.spinfo_address);
4148 	status.sstat_primary.spinfo_state = transport->state;
4149 	status.sstat_primary.spinfo_cwnd = transport->cwnd;
4150 	status.sstat_primary.spinfo_srtt = transport->srtt;
4151 	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4152 	status.sstat_primary.spinfo_mtu = transport->pathmtu;
4153 
4154 	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4155 		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4156 
4157 	if (put_user(len, optlen)) {
4158 		retval = -EFAULT;
4159 		goto out;
4160 	}
4161 
4162 	pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4163 		 __func__, len, status.sstat_state, status.sstat_rwnd,
4164 		 status.sstat_assoc_id);
4165 
4166 	if (copy_to_user(optval, &status, len)) {
4167 		retval = -EFAULT;
4168 		goto out;
4169 	}
4170 
4171 out:
4172 	return retval;
4173 }
4174 
4175 
4176 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4177  *
4178  * Applications can retrieve information about a specific peer address
4179  * of an association, including its reachability state, congestion
4180  * window, and retransmission timer values.  This information is
4181  * read-only.
4182  */
4183 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4184 					  char __user *optval,
4185 					  int __user *optlen)
4186 {
4187 	struct sctp_paddrinfo pinfo;
4188 	struct sctp_transport *transport;
4189 	int retval = 0;
4190 
4191 	if (len < sizeof(pinfo)) {
4192 		retval = -EINVAL;
4193 		goto out;
4194 	}
4195 
4196 	len = sizeof(pinfo);
4197 	if (copy_from_user(&pinfo, optval, len)) {
4198 		retval = -EFAULT;
4199 		goto out;
4200 	}
4201 
4202 	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4203 					   pinfo.spinfo_assoc_id);
4204 	if (!transport)
4205 		return -EINVAL;
4206 
4207 	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4208 	pinfo.spinfo_state = transport->state;
4209 	pinfo.spinfo_cwnd = transport->cwnd;
4210 	pinfo.spinfo_srtt = transport->srtt;
4211 	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4212 	pinfo.spinfo_mtu = transport->pathmtu;
4213 
4214 	if (pinfo.spinfo_state == SCTP_UNKNOWN)
4215 		pinfo.spinfo_state = SCTP_ACTIVE;
4216 
4217 	if (put_user(len, optlen)) {
4218 		retval = -EFAULT;
4219 		goto out;
4220 	}
4221 
4222 	if (copy_to_user(optval, &pinfo, len)) {
4223 		retval = -EFAULT;
4224 		goto out;
4225 	}
4226 
4227 out:
4228 	return retval;
4229 }
4230 
4231 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4232  *
4233  * This option is a on/off flag.  If enabled no SCTP message
4234  * fragmentation will be performed.  Instead if a message being sent
4235  * exceeds the current PMTU size, the message will NOT be sent and
4236  * instead a error will be indicated to the user.
4237  */
4238 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4239 					char __user *optval, int __user *optlen)
4240 {
4241 	int val;
4242 
4243 	if (len < sizeof(int))
4244 		return -EINVAL;
4245 
4246 	len = sizeof(int);
4247 	val = (sctp_sk(sk)->disable_fragments == 1);
4248 	if (put_user(len, optlen))
4249 		return -EFAULT;
4250 	if (copy_to_user(optval, &val, len))
4251 		return -EFAULT;
4252 	return 0;
4253 }
4254 
4255 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4256  *
4257  * This socket option is used to specify various notifications and
4258  * ancillary data the user wishes to receive.
4259  */
4260 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4261 				  int __user *optlen)
4262 {
4263 	if (len <= 0)
4264 		return -EINVAL;
4265 	if (len > sizeof(struct sctp_event_subscribe))
4266 		len = sizeof(struct sctp_event_subscribe);
4267 	if (put_user(len, optlen))
4268 		return -EFAULT;
4269 	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4270 		return -EFAULT;
4271 	return 0;
4272 }
4273 
4274 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4275  *
4276  * This socket option is applicable to the UDP-style socket only.  When
4277  * set it will cause associations that are idle for more than the
4278  * specified number of seconds to automatically close.  An association
4279  * being idle is defined an association that has NOT sent or received
4280  * user data.  The special value of '0' indicates that no automatic
4281  * close of any associations should be performed.  The option expects an
4282  * integer defining the number of seconds of idle time before an
4283  * association is closed.
4284  */
4285 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4286 {
4287 	/* Applicable to UDP-style socket only */
4288 	if (sctp_style(sk, TCP))
4289 		return -EOPNOTSUPP;
4290 	if (len < sizeof(int))
4291 		return -EINVAL;
4292 	len = sizeof(int);
4293 	if (put_user(len, optlen))
4294 		return -EFAULT;
4295 	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4296 		return -EFAULT;
4297 	return 0;
4298 }
4299 
4300 /* Helper routine to branch off an association to a new socket.  */
4301 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4302 {
4303 	struct sctp_association *asoc = sctp_id2assoc(sk, id);
4304 	struct socket *sock;
4305 	struct sctp_af *af;
4306 	int err = 0;
4307 
4308 	if (!asoc)
4309 		return -EINVAL;
4310 
4311 	/* An association cannot be branched off from an already peeled-off
4312 	 * socket, nor is this supported for tcp style sockets.
4313 	 */
4314 	if (!sctp_style(sk, UDP))
4315 		return -EINVAL;
4316 
4317 	/* Create a new socket.  */
4318 	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4319 	if (err < 0)
4320 		return err;
4321 
4322 	sctp_copy_sock(sock->sk, sk, asoc);
4323 
4324 	/* Make peeled-off sockets more like 1-1 accepted sockets.
4325 	 * Set the daddr and initialize id to something more random
4326 	 */
4327 	af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4328 	af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4329 
4330 	/* Populate the fields of the newsk from the oldsk and migrate the
4331 	 * asoc to the newsk.
4332 	 */
4333 	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4334 
4335 	*sockp = sock;
4336 
4337 	return err;
4338 }
4339 EXPORT_SYMBOL(sctp_do_peeloff);
4340 
4341 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4342 {
4343 	sctp_peeloff_arg_t peeloff;
4344 	struct socket *newsock;
4345 	struct file *newfile;
4346 	int retval = 0;
4347 
4348 	if (len < sizeof(sctp_peeloff_arg_t))
4349 		return -EINVAL;
4350 	len = sizeof(sctp_peeloff_arg_t);
4351 	if (copy_from_user(&peeloff, optval, len))
4352 		return -EFAULT;
4353 
4354 	retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4355 	if (retval < 0)
4356 		goto out;
4357 
4358 	/* Map the socket to an unused fd that can be returned to the user.  */
4359 	retval = get_unused_fd_flags(0);
4360 	if (retval < 0) {
4361 		sock_release(newsock);
4362 		goto out;
4363 	}
4364 
4365 	newfile = sock_alloc_file(newsock, 0, NULL);
4366 	if (unlikely(IS_ERR(newfile))) {
4367 		put_unused_fd(retval);
4368 		sock_release(newsock);
4369 		return PTR_ERR(newfile);
4370 	}
4371 
4372 	pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4373 		 retval);
4374 
4375 	/* Return the fd mapped to the new socket.  */
4376 	if (put_user(len, optlen)) {
4377 		fput(newfile);
4378 		put_unused_fd(retval);
4379 		return -EFAULT;
4380 	}
4381 	peeloff.sd = retval;
4382 	if (copy_to_user(optval, &peeloff, len)) {
4383 		fput(newfile);
4384 		put_unused_fd(retval);
4385 		return -EFAULT;
4386 	}
4387 	fd_install(retval, newfile);
4388 out:
4389 	return retval;
4390 }
4391 
4392 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4393  *
4394  * Applications can enable or disable heartbeats for any peer address of
4395  * an association, modify an address's heartbeat interval, force a
4396  * heartbeat to be sent immediately, and adjust the address's maximum
4397  * number of retransmissions sent before an address is considered
4398  * unreachable.  The following structure is used to access and modify an
4399  * address's parameters:
4400  *
4401  *  struct sctp_paddrparams {
4402  *     sctp_assoc_t            spp_assoc_id;
4403  *     struct sockaddr_storage spp_address;
4404  *     uint32_t                spp_hbinterval;
4405  *     uint16_t                spp_pathmaxrxt;
4406  *     uint32_t                spp_pathmtu;
4407  *     uint32_t                spp_sackdelay;
4408  *     uint32_t                spp_flags;
4409  * };
4410  *
4411  *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4412  *                     application, and identifies the association for
4413  *                     this query.
4414  *   spp_address     - This specifies which address is of interest.
4415  *   spp_hbinterval  - This contains the value of the heartbeat interval,
4416  *                     in milliseconds.  If a  value of zero
4417  *                     is present in this field then no changes are to
4418  *                     be made to this parameter.
4419  *   spp_pathmaxrxt  - This contains the maximum number of
4420  *                     retransmissions before this address shall be
4421  *                     considered unreachable. If a  value of zero
4422  *                     is present in this field then no changes are to
4423  *                     be made to this parameter.
4424  *   spp_pathmtu     - When Path MTU discovery is disabled the value
4425  *                     specified here will be the "fixed" path mtu.
4426  *                     Note that if the spp_address field is empty
4427  *                     then all associations on this address will
4428  *                     have this fixed path mtu set upon them.
4429  *
4430  *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4431  *                     the number of milliseconds that sacks will be delayed
4432  *                     for. This value will apply to all addresses of an
4433  *                     association if the spp_address field is empty. Note
4434  *                     also, that if delayed sack is enabled and this
4435  *                     value is set to 0, no change is made to the last
4436  *                     recorded delayed sack timer value.
4437  *
4438  *   spp_flags       - These flags are used to control various features
4439  *                     on an association. The flag field may contain
4440  *                     zero or more of the following options.
4441  *
4442  *                     SPP_HB_ENABLE  - Enable heartbeats on the
4443  *                     specified address. Note that if the address
4444  *                     field is empty all addresses for the association
4445  *                     have heartbeats enabled upon them.
4446  *
4447  *                     SPP_HB_DISABLE - Disable heartbeats on the
4448  *                     speicifed address. Note that if the address
4449  *                     field is empty all addresses for the association
4450  *                     will have their heartbeats disabled. Note also
4451  *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4452  *                     mutually exclusive, only one of these two should
4453  *                     be specified. Enabling both fields will have
4454  *                     undetermined results.
4455  *
4456  *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4457  *                     to be made immediately.
4458  *
4459  *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4460  *                     discovery upon the specified address. Note that
4461  *                     if the address feild is empty then all addresses
4462  *                     on the association are effected.
4463  *
4464  *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4465  *                     discovery upon the specified address. Note that
4466  *                     if the address feild is empty then all addresses
4467  *                     on the association are effected. Not also that
4468  *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4469  *                     exclusive. Enabling both will have undetermined
4470  *                     results.
4471  *
4472  *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4473  *                     on delayed sack. The time specified in spp_sackdelay
4474  *                     is used to specify the sack delay for this address. Note
4475  *                     that if spp_address is empty then all addresses will
4476  *                     enable delayed sack and take on the sack delay
4477  *                     value specified in spp_sackdelay.
4478  *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4479  *                     off delayed sack. If the spp_address field is blank then
4480  *                     delayed sack is disabled for the entire association. Note
4481  *                     also that this field is mutually exclusive to
4482  *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4483  *                     results.
4484  */
4485 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4486 					    char __user *optval, int __user *optlen)
4487 {
4488 	struct sctp_paddrparams  params;
4489 	struct sctp_transport   *trans = NULL;
4490 	struct sctp_association *asoc = NULL;
4491 	struct sctp_sock        *sp = sctp_sk(sk);
4492 
4493 	if (len < sizeof(struct sctp_paddrparams))
4494 		return -EINVAL;
4495 	len = sizeof(struct sctp_paddrparams);
4496 	if (copy_from_user(&params, optval, len))
4497 		return -EFAULT;
4498 
4499 	/* If an address other than INADDR_ANY is specified, and
4500 	 * no transport is found, then the request is invalid.
4501 	 */
4502 	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4503 		trans = sctp_addr_id2transport(sk, &params.spp_address,
4504 					       params.spp_assoc_id);
4505 		if (!trans) {
4506 			pr_debug("%s: failed no transport\n", __func__);
4507 			return -EINVAL;
4508 		}
4509 	}
4510 
4511 	/* Get association, if assoc_id != 0 and the socket is a one
4512 	 * to many style socket, and an association was not found, then
4513 	 * the id was invalid.
4514 	 */
4515 	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4516 	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4517 		pr_debug("%s: failed no association\n", __func__);
4518 		return -EINVAL;
4519 	}
4520 
4521 	if (trans) {
4522 		/* Fetch transport values. */
4523 		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4524 		params.spp_pathmtu    = trans->pathmtu;
4525 		params.spp_pathmaxrxt = trans->pathmaxrxt;
4526 		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4527 
4528 		/*draft-11 doesn't say what to return in spp_flags*/
4529 		params.spp_flags      = trans->param_flags;
4530 	} else if (asoc) {
4531 		/* Fetch association values. */
4532 		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4533 		params.spp_pathmtu    = asoc->pathmtu;
4534 		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4535 		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4536 
4537 		/*draft-11 doesn't say what to return in spp_flags*/
4538 		params.spp_flags      = asoc->param_flags;
4539 	} else {
4540 		/* Fetch socket values. */
4541 		params.spp_hbinterval = sp->hbinterval;
4542 		params.spp_pathmtu    = sp->pathmtu;
4543 		params.spp_sackdelay  = sp->sackdelay;
4544 		params.spp_pathmaxrxt = sp->pathmaxrxt;
4545 
4546 		/*draft-11 doesn't say what to return in spp_flags*/
4547 		params.spp_flags      = sp->param_flags;
4548 	}
4549 
4550 	if (copy_to_user(optval, &params, len))
4551 		return -EFAULT;
4552 
4553 	if (put_user(len, optlen))
4554 		return -EFAULT;
4555 
4556 	return 0;
4557 }
4558 
4559 /*
4560  * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4561  *
4562  * This option will effect the way delayed acks are performed.  This
4563  * option allows you to get or set the delayed ack time, in
4564  * milliseconds.  It also allows changing the delayed ack frequency.
4565  * Changing the frequency to 1 disables the delayed sack algorithm.  If
4566  * the assoc_id is 0, then this sets or gets the endpoints default
4567  * values.  If the assoc_id field is non-zero, then the set or get
4568  * effects the specified association for the one to many model (the
4569  * assoc_id field is ignored by the one to one model).  Note that if
4570  * sack_delay or sack_freq are 0 when setting this option, then the
4571  * current values will remain unchanged.
4572  *
4573  * struct sctp_sack_info {
4574  *     sctp_assoc_t            sack_assoc_id;
4575  *     uint32_t                sack_delay;
4576  *     uint32_t                sack_freq;
4577  * };
4578  *
4579  * sack_assoc_id -  This parameter, indicates which association the user
4580  *    is performing an action upon.  Note that if this field's value is
4581  *    zero then the endpoints default value is changed (effecting future
4582  *    associations only).
4583  *
4584  * sack_delay -  This parameter contains the number of milliseconds that
4585  *    the user is requesting the delayed ACK timer be set to.  Note that
4586  *    this value is defined in the standard to be between 200 and 500
4587  *    milliseconds.
4588  *
4589  * sack_freq -  This parameter contains the number of packets that must
4590  *    be received before a sack is sent without waiting for the delay
4591  *    timer to expire.  The default value for this is 2, setting this
4592  *    value to 1 will disable the delayed sack algorithm.
4593  */
4594 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4595 					    char __user *optval,
4596 					    int __user *optlen)
4597 {
4598 	struct sctp_sack_info    params;
4599 	struct sctp_association *asoc = NULL;
4600 	struct sctp_sock        *sp = sctp_sk(sk);
4601 
4602 	if (len >= sizeof(struct sctp_sack_info)) {
4603 		len = sizeof(struct sctp_sack_info);
4604 
4605 		if (copy_from_user(&params, optval, len))
4606 			return -EFAULT;
4607 	} else if (len == sizeof(struct sctp_assoc_value)) {
4608 		pr_warn_ratelimited(DEPRECATED
4609 				    "%s (pid %d) "
4610 				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4611 				    "Use struct sctp_sack_info instead\n",
4612 				    current->comm, task_pid_nr(current));
4613 		if (copy_from_user(&params, optval, len))
4614 			return -EFAULT;
4615 	} else
4616 		return -EINVAL;
4617 
4618 	/* Get association, if sack_assoc_id != 0 and the socket is a one
4619 	 * to many style socket, and an association was not found, then
4620 	 * the id was invalid.
4621 	 */
4622 	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4623 	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4624 		return -EINVAL;
4625 
4626 	if (asoc) {
4627 		/* Fetch association values. */
4628 		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4629 			params.sack_delay = jiffies_to_msecs(
4630 				asoc->sackdelay);
4631 			params.sack_freq = asoc->sackfreq;
4632 
4633 		} else {
4634 			params.sack_delay = 0;
4635 			params.sack_freq = 1;
4636 		}
4637 	} else {
4638 		/* Fetch socket values. */
4639 		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4640 			params.sack_delay  = sp->sackdelay;
4641 			params.sack_freq = sp->sackfreq;
4642 		} else {
4643 			params.sack_delay  = 0;
4644 			params.sack_freq = 1;
4645 		}
4646 	}
4647 
4648 	if (copy_to_user(optval, &params, len))
4649 		return -EFAULT;
4650 
4651 	if (put_user(len, optlen))
4652 		return -EFAULT;
4653 
4654 	return 0;
4655 }
4656 
4657 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4658  *
4659  * Applications can specify protocol parameters for the default association
4660  * initialization.  The option name argument to setsockopt() and getsockopt()
4661  * is SCTP_INITMSG.
4662  *
4663  * Setting initialization parameters is effective only on an unconnected
4664  * socket (for UDP-style sockets only future associations are effected
4665  * by the change).  With TCP-style sockets, this option is inherited by
4666  * sockets derived from a listener socket.
4667  */
4668 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4669 {
4670 	if (len < sizeof(struct sctp_initmsg))
4671 		return -EINVAL;
4672 	len = sizeof(struct sctp_initmsg);
4673 	if (put_user(len, optlen))
4674 		return -EFAULT;
4675 	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4676 		return -EFAULT;
4677 	return 0;
4678 }
4679 
4680 
4681 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4682 				      char __user *optval, int __user *optlen)
4683 {
4684 	struct sctp_association *asoc;
4685 	int cnt = 0;
4686 	struct sctp_getaddrs getaddrs;
4687 	struct sctp_transport *from;
4688 	void __user *to;
4689 	union sctp_addr temp;
4690 	struct sctp_sock *sp = sctp_sk(sk);
4691 	int addrlen;
4692 	size_t space_left;
4693 	int bytes_copied;
4694 
4695 	if (len < sizeof(struct sctp_getaddrs))
4696 		return -EINVAL;
4697 
4698 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4699 		return -EFAULT;
4700 
4701 	/* For UDP-style sockets, id specifies the association to query.  */
4702 	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4703 	if (!asoc)
4704 		return -EINVAL;
4705 
4706 	to = optval + offsetof(struct sctp_getaddrs, addrs);
4707 	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4708 
4709 	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4710 				transports) {
4711 		memcpy(&temp, &from->ipaddr, sizeof(temp));
4712 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4713 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4714 		if (space_left < addrlen)
4715 			return -ENOMEM;
4716 		if (copy_to_user(to, &temp, addrlen))
4717 			return -EFAULT;
4718 		to += addrlen;
4719 		cnt++;
4720 		space_left -= addrlen;
4721 	}
4722 
4723 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4724 		return -EFAULT;
4725 	bytes_copied = ((char __user *)to) - optval;
4726 	if (put_user(bytes_copied, optlen))
4727 		return -EFAULT;
4728 
4729 	return 0;
4730 }
4731 
4732 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4733 			    size_t space_left, int *bytes_copied)
4734 {
4735 	struct sctp_sockaddr_entry *addr;
4736 	union sctp_addr temp;
4737 	int cnt = 0;
4738 	int addrlen;
4739 	struct net *net = sock_net(sk);
4740 
4741 	rcu_read_lock();
4742 	list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4743 		if (!addr->valid)
4744 			continue;
4745 
4746 		if ((PF_INET == sk->sk_family) &&
4747 		    (AF_INET6 == addr->a.sa.sa_family))
4748 			continue;
4749 		if ((PF_INET6 == sk->sk_family) &&
4750 		    inet_v6_ipv6only(sk) &&
4751 		    (AF_INET == addr->a.sa.sa_family))
4752 			continue;
4753 		memcpy(&temp, &addr->a, sizeof(temp));
4754 		if (!temp.v4.sin_port)
4755 			temp.v4.sin_port = htons(port);
4756 
4757 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4758 								&temp);
4759 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4760 		if (space_left < addrlen) {
4761 			cnt =  -ENOMEM;
4762 			break;
4763 		}
4764 		memcpy(to, &temp, addrlen);
4765 
4766 		to += addrlen;
4767 		cnt++;
4768 		space_left -= addrlen;
4769 		*bytes_copied += addrlen;
4770 	}
4771 	rcu_read_unlock();
4772 
4773 	return cnt;
4774 }
4775 
4776 
4777 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4778 				       char __user *optval, int __user *optlen)
4779 {
4780 	struct sctp_bind_addr *bp;
4781 	struct sctp_association *asoc;
4782 	int cnt = 0;
4783 	struct sctp_getaddrs getaddrs;
4784 	struct sctp_sockaddr_entry *addr;
4785 	void __user *to;
4786 	union sctp_addr temp;
4787 	struct sctp_sock *sp = sctp_sk(sk);
4788 	int addrlen;
4789 	int err = 0;
4790 	size_t space_left;
4791 	int bytes_copied = 0;
4792 	void *addrs;
4793 	void *buf;
4794 
4795 	if (len < sizeof(struct sctp_getaddrs))
4796 		return -EINVAL;
4797 
4798 	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4799 		return -EFAULT;
4800 
4801 	/*
4802 	 *  For UDP-style sockets, id specifies the association to query.
4803 	 *  If the id field is set to the value '0' then the locally bound
4804 	 *  addresses are returned without regard to any particular
4805 	 *  association.
4806 	 */
4807 	if (0 == getaddrs.assoc_id) {
4808 		bp = &sctp_sk(sk)->ep->base.bind_addr;
4809 	} else {
4810 		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4811 		if (!asoc)
4812 			return -EINVAL;
4813 		bp = &asoc->base.bind_addr;
4814 	}
4815 
4816 	to = optval + offsetof(struct sctp_getaddrs, addrs);
4817 	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4818 
4819 	addrs = kmalloc(space_left, GFP_KERNEL);
4820 	if (!addrs)
4821 		return -ENOMEM;
4822 
4823 	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4824 	 * addresses from the global local address list.
4825 	 */
4826 	if (sctp_list_single_entry(&bp->address_list)) {
4827 		addr = list_entry(bp->address_list.next,
4828 				  struct sctp_sockaddr_entry, list);
4829 		if (sctp_is_any(sk, &addr->a)) {
4830 			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4831 						space_left, &bytes_copied);
4832 			if (cnt < 0) {
4833 				err = cnt;
4834 				goto out;
4835 			}
4836 			goto copy_getaddrs;
4837 		}
4838 	}
4839 
4840 	buf = addrs;
4841 	/* Protection on the bound address list is not needed since
4842 	 * in the socket option context we hold a socket lock and
4843 	 * thus the bound address list can't change.
4844 	 */
4845 	list_for_each_entry(addr, &bp->address_list, list) {
4846 		memcpy(&temp, &addr->a, sizeof(temp));
4847 		sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4848 		addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4849 		if (space_left < addrlen) {
4850 			err =  -ENOMEM; /*fixme: right error?*/
4851 			goto out;
4852 		}
4853 		memcpy(buf, &temp, addrlen);
4854 		buf += addrlen;
4855 		bytes_copied += addrlen;
4856 		cnt++;
4857 		space_left -= addrlen;
4858 	}
4859 
4860 copy_getaddrs:
4861 	if (copy_to_user(to, addrs, bytes_copied)) {
4862 		err = -EFAULT;
4863 		goto out;
4864 	}
4865 	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4866 		err = -EFAULT;
4867 		goto out;
4868 	}
4869 	if (put_user(bytes_copied, optlen))
4870 		err = -EFAULT;
4871 out:
4872 	kfree(addrs);
4873 	return err;
4874 }
4875 
4876 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4877  *
4878  * Requests that the local SCTP stack use the enclosed peer address as
4879  * the association primary.  The enclosed address must be one of the
4880  * association peer's addresses.
4881  */
4882 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4883 					char __user *optval, int __user *optlen)
4884 {
4885 	struct sctp_prim prim;
4886 	struct sctp_association *asoc;
4887 	struct sctp_sock *sp = sctp_sk(sk);
4888 
4889 	if (len < sizeof(struct sctp_prim))
4890 		return -EINVAL;
4891 
4892 	len = sizeof(struct sctp_prim);
4893 
4894 	if (copy_from_user(&prim, optval, len))
4895 		return -EFAULT;
4896 
4897 	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4898 	if (!asoc)
4899 		return -EINVAL;
4900 
4901 	if (!asoc->peer.primary_path)
4902 		return -ENOTCONN;
4903 
4904 	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4905 		asoc->peer.primary_path->af_specific->sockaddr_len);
4906 
4907 	sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4908 			(union sctp_addr *)&prim.ssp_addr);
4909 
4910 	if (put_user(len, optlen))
4911 		return -EFAULT;
4912 	if (copy_to_user(optval, &prim, len))
4913 		return -EFAULT;
4914 
4915 	return 0;
4916 }
4917 
4918 /*
4919  * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4920  *
4921  * Requests that the local endpoint set the specified Adaptation Layer
4922  * Indication parameter for all future INIT and INIT-ACK exchanges.
4923  */
4924 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4925 				  char __user *optval, int __user *optlen)
4926 {
4927 	struct sctp_setadaptation adaptation;
4928 
4929 	if (len < sizeof(struct sctp_setadaptation))
4930 		return -EINVAL;
4931 
4932 	len = sizeof(struct sctp_setadaptation);
4933 
4934 	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4935 
4936 	if (put_user(len, optlen))
4937 		return -EFAULT;
4938 	if (copy_to_user(optval, &adaptation, len))
4939 		return -EFAULT;
4940 
4941 	return 0;
4942 }
4943 
4944 /*
4945  *
4946  * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4947  *
4948  *   Applications that wish to use the sendto() system call may wish to
4949  *   specify a default set of parameters that would normally be supplied
4950  *   through the inclusion of ancillary data.  This socket option allows
4951  *   such an application to set the default sctp_sndrcvinfo structure.
4952 
4953 
4954  *   The application that wishes to use this socket option simply passes
4955  *   in to this call the sctp_sndrcvinfo structure defined in Section
4956  *   5.2.2) The input parameters accepted by this call include
4957  *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4958  *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
4959  *   to this call if the caller is using the UDP model.
4960  *
4961  *   For getsockopt, it get the default sctp_sndrcvinfo structure.
4962  */
4963 static int sctp_getsockopt_default_send_param(struct sock *sk,
4964 					int len, char __user *optval,
4965 					int __user *optlen)
4966 {
4967 	struct sctp_sndrcvinfo info;
4968 	struct sctp_association *asoc;
4969 	struct sctp_sock *sp = sctp_sk(sk);
4970 
4971 	if (len < sizeof(struct sctp_sndrcvinfo))
4972 		return -EINVAL;
4973 
4974 	len = sizeof(struct sctp_sndrcvinfo);
4975 
4976 	if (copy_from_user(&info, optval, len))
4977 		return -EFAULT;
4978 
4979 	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4980 	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4981 		return -EINVAL;
4982 
4983 	if (asoc) {
4984 		info.sinfo_stream = asoc->default_stream;
4985 		info.sinfo_flags = asoc->default_flags;
4986 		info.sinfo_ppid = asoc->default_ppid;
4987 		info.sinfo_context = asoc->default_context;
4988 		info.sinfo_timetolive = asoc->default_timetolive;
4989 	} else {
4990 		info.sinfo_stream = sp->default_stream;
4991 		info.sinfo_flags = sp->default_flags;
4992 		info.sinfo_ppid = sp->default_ppid;
4993 		info.sinfo_context = sp->default_context;
4994 		info.sinfo_timetolive = sp->default_timetolive;
4995 	}
4996 
4997 	if (put_user(len, optlen))
4998 		return -EFAULT;
4999 	if (copy_to_user(optval, &info, len))
5000 		return -EFAULT;
5001 
5002 	return 0;
5003 }
5004 
5005 /*
5006  *
5007  * 7.1.5 SCTP_NODELAY
5008  *
5009  * Turn on/off any Nagle-like algorithm.  This means that packets are
5010  * generally sent as soon as possible and no unnecessary delays are
5011  * introduced, at the cost of more packets in the network.  Expects an
5012  * integer boolean flag.
5013  */
5014 
5015 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5016 				   char __user *optval, int __user *optlen)
5017 {
5018 	int val;
5019 
5020 	if (len < sizeof(int))
5021 		return -EINVAL;
5022 
5023 	len = sizeof(int);
5024 	val = (sctp_sk(sk)->nodelay == 1);
5025 	if (put_user(len, optlen))
5026 		return -EFAULT;
5027 	if (copy_to_user(optval, &val, len))
5028 		return -EFAULT;
5029 	return 0;
5030 }
5031 
5032 /*
5033  *
5034  * 7.1.1 SCTP_RTOINFO
5035  *
5036  * The protocol parameters used to initialize and bound retransmission
5037  * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5038  * and modify these parameters.
5039  * All parameters are time values, in milliseconds.  A value of 0, when
5040  * modifying the parameters, indicates that the current value should not
5041  * be changed.
5042  *
5043  */
5044 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5045 				char __user *optval,
5046 				int __user *optlen) {
5047 	struct sctp_rtoinfo rtoinfo;
5048 	struct sctp_association *asoc;
5049 
5050 	if (len < sizeof (struct sctp_rtoinfo))
5051 		return -EINVAL;
5052 
5053 	len = sizeof(struct sctp_rtoinfo);
5054 
5055 	if (copy_from_user(&rtoinfo, optval, len))
5056 		return -EFAULT;
5057 
5058 	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5059 
5060 	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5061 		return -EINVAL;
5062 
5063 	/* Values corresponding to the specific association. */
5064 	if (asoc) {
5065 		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5066 		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5067 		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5068 	} else {
5069 		/* Values corresponding to the endpoint. */
5070 		struct sctp_sock *sp = sctp_sk(sk);
5071 
5072 		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5073 		rtoinfo.srto_max = sp->rtoinfo.srto_max;
5074 		rtoinfo.srto_min = sp->rtoinfo.srto_min;
5075 	}
5076 
5077 	if (put_user(len, optlen))
5078 		return -EFAULT;
5079 
5080 	if (copy_to_user(optval, &rtoinfo, len))
5081 		return -EFAULT;
5082 
5083 	return 0;
5084 }
5085 
5086 /*
5087  *
5088  * 7.1.2 SCTP_ASSOCINFO
5089  *
5090  * This option is used to tune the maximum retransmission attempts
5091  * of the association.
5092  * Returns an error if the new association retransmission value is
5093  * greater than the sum of the retransmission value  of the peer.
5094  * See [SCTP] for more information.
5095  *
5096  */
5097 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5098 				     char __user *optval,
5099 				     int __user *optlen)
5100 {
5101 
5102 	struct sctp_assocparams assocparams;
5103 	struct sctp_association *asoc;
5104 	struct list_head *pos;
5105 	int cnt = 0;
5106 
5107 	if (len < sizeof (struct sctp_assocparams))
5108 		return -EINVAL;
5109 
5110 	len = sizeof(struct sctp_assocparams);
5111 
5112 	if (copy_from_user(&assocparams, optval, len))
5113 		return -EFAULT;
5114 
5115 	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5116 
5117 	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5118 		return -EINVAL;
5119 
5120 	/* Values correspoinding to the specific association */
5121 	if (asoc) {
5122 		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5123 		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5124 		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5125 		assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5126 
5127 		list_for_each(pos, &asoc->peer.transport_addr_list) {
5128 			cnt++;
5129 		}
5130 
5131 		assocparams.sasoc_number_peer_destinations = cnt;
5132 	} else {
5133 		/* Values corresponding to the endpoint */
5134 		struct sctp_sock *sp = sctp_sk(sk);
5135 
5136 		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5137 		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5138 		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5139 		assocparams.sasoc_cookie_life =
5140 					sp->assocparams.sasoc_cookie_life;
5141 		assocparams.sasoc_number_peer_destinations =
5142 					sp->assocparams.
5143 					sasoc_number_peer_destinations;
5144 	}
5145 
5146 	if (put_user(len, optlen))
5147 		return -EFAULT;
5148 
5149 	if (copy_to_user(optval, &assocparams, len))
5150 		return -EFAULT;
5151 
5152 	return 0;
5153 }
5154 
5155 /*
5156  * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5157  *
5158  * This socket option is a boolean flag which turns on or off mapped V4
5159  * addresses.  If this option is turned on and the socket is type
5160  * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5161  * If this option is turned off, then no mapping will be done of V4
5162  * addresses and a user will receive both PF_INET6 and PF_INET type
5163  * addresses on the socket.
5164  */
5165 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5166 				    char __user *optval, int __user *optlen)
5167 {
5168 	int val;
5169 	struct sctp_sock *sp = sctp_sk(sk);
5170 
5171 	if (len < sizeof(int))
5172 		return -EINVAL;
5173 
5174 	len = sizeof(int);
5175 	val = sp->v4mapped;
5176 	if (put_user(len, optlen))
5177 		return -EFAULT;
5178 	if (copy_to_user(optval, &val, len))
5179 		return -EFAULT;
5180 
5181 	return 0;
5182 }
5183 
5184 /*
5185  * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
5186  * (chapter and verse is quoted at sctp_setsockopt_context())
5187  */
5188 static int sctp_getsockopt_context(struct sock *sk, int len,
5189 				   char __user *optval, int __user *optlen)
5190 {
5191 	struct sctp_assoc_value params;
5192 	struct sctp_sock *sp;
5193 	struct sctp_association *asoc;
5194 
5195 	if (len < sizeof(struct sctp_assoc_value))
5196 		return -EINVAL;
5197 
5198 	len = sizeof(struct sctp_assoc_value);
5199 
5200 	if (copy_from_user(&params, optval, len))
5201 		return -EFAULT;
5202 
5203 	sp = sctp_sk(sk);
5204 
5205 	if (params.assoc_id != 0) {
5206 		asoc = sctp_id2assoc(sk, params.assoc_id);
5207 		if (!asoc)
5208 			return -EINVAL;
5209 		params.assoc_value = asoc->default_rcv_context;
5210 	} else {
5211 		params.assoc_value = sp->default_rcv_context;
5212 	}
5213 
5214 	if (put_user(len, optlen))
5215 		return -EFAULT;
5216 	if (copy_to_user(optval, &params, len))
5217 		return -EFAULT;
5218 
5219 	return 0;
5220 }
5221 
5222 /*
5223  * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5224  * This option will get or set the maximum size to put in any outgoing
5225  * SCTP DATA chunk.  If a message is larger than this size it will be
5226  * fragmented by SCTP into the specified size.  Note that the underlying
5227  * SCTP implementation may fragment into smaller sized chunks when the
5228  * PMTU of the underlying association is smaller than the value set by
5229  * the user.  The default value for this option is '0' which indicates
5230  * the user is NOT limiting fragmentation and only the PMTU will effect
5231  * SCTP's choice of DATA chunk size.  Note also that values set larger
5232  * than the maximum size of an IP datagram will effectively let SCTP
5233  * control fragmentation (i.e. the same as setting this option to 0).
5234  *
5235  * The following structure is used to access and modify this parameter:
5236  *
5237  * struct sctp_assoc_value {
5238  *   sctp_assoc_t assoc_id;
5239  *   uint32_t assoc_value;
5240  * };
5241  *
5242  * assoc_id:  This parameter is ignored for one-to-one style sockets.
5243  *    For one-to-many style sockets this parameter indicates which
5244  *    association the user is performing an action upon.  Note that if
5245  *    this field's value is zero then the endpoints default value is
5246  *    changed (effecting future associations only).
5247  * assoc_value:  This parameter specifies the maximum size in bytes.
5248  */
5249 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5250 				  char __user *optval, int __user *optlen)
5251 {
5252 	struct sctp_assoc_value params;
5253 	struct sctp_association *asoc;
5254 
5255 	if (len == sizeof(int)) {
5256 		pr_warn_ratelimited(DEPRECATED
5257 				    "%s (pid %d) "
5258 				    "Use of int in maxseg socket option.\n"
5259 				    "Use struct sctp_assoc_value instead\n",
5260 				    current->comm, task_pid_nr(current));
5261 		params.assoc_id = 0;
5262 	} else if (len >= sizeof(struct sctp_assoc_value)) {
5263 		len = sizeof(struct sctp_assoc_value);
5264 		if (copy_from_user(&params, optval, sizeof(params)))
5265 			return -EFAULT;
5266 	} else
5267 		return -EINVAL;
5268 
5269 	asoc = sctp_id2assoc(sk, params.assoc_id);
5270 	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5271 		return -EINVAL;
5272 
5273 	if (asoc)
5274 		params.assoc_value = asoc->frag_point;
5275 	else
5276 		params.assoc_value = sctp_sk(sk)->user_frag;
5277 
5278 	if (put_user(len, optlen))
5279 		return -EFAULT;
5280 	if (len == sizeof(int)) {
5281 		if (copy_to_user(optval, &params.assoc_value, len))
5282 			return -EFAULT;
5283 	} else {
5284 		if (copy_to_user(optval, &params, len))
5285 			return -EFAULT;
5286 	}
5287 
5288 	return 0;
5289 }
5290 
5291 /*
5292  * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5293  * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5294  */
5295 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5296 					       char __user *optval, int __user *optlen)
5297 {
5298 	int val;
5299 
5300 	if (len < sizeof(int))
5301 		return -EINVAL;
5302 
5303 	len = sizeof(int);
5304 
5305 	val = sctp_sk(sk)->frag_interleave;
5306 	if (put_user(len, optlen))
5307 		return -EFAULT;
5308 	if (copy_to_user(optval, &val, len))
5309 		return -EFAULT;
5310 
5311 	return 0;
5312 }
5313 
5314 /*
5315  * 7.1.25.  Set or Get the sctp partial delivery point
5316  * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5317  */
5318 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5319 						  char __user *optval,
5320 						  int __user *optlen)
5321 {
5322 	u32 val;
5323 
5324 	if (len < sizeof(u32))
5325 		return -EINVAL;
5326 
5327 	len = sizeof(u32);
5328 
5329 	val = sctp_sk(sk)->pd_point;
5330 	if (put_user(len, optlen))
5331 		return -EFAULT;
5332 	if (copy_to_user(optval, &val, len))
5333 		return -EFAULT;
5334 
5335 	return 0;
5336 }
5337 
5338 /*
5339  * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5340  * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5341  */
5342 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5343 				    char __user *optval,
5344 				    int __user *optlen)
5345 {
5346 	struct sctp_assoc_value params;
5347 	struct sctp_sock *sp;
5348 	struct sctp_association *asoc;
5349 
5350 	if (len == sizeof(int)) {
5351 		pr_warn_ratelimited(DEPRECATED
5352 				    "%s (pid %d) "
5353 				    "Use of int in max_burst socket option.\n"
5354 				    "Use struct sctp_assoc_value instead\n",
5355 				    current->comm, task_pid_nr(current));
5356 		params.assoc_id = 0;
5357 	} else if (len >= sizeof(struct sctp_assoc_value)) {
5358 		len = sizeof(struct sctp_assoc_value);
5359 		if (copy_from_user(&params, optval, len))
5360 			return -EFAULT;
5361 	} else
5362 		return -EINVAL;
5363 
5364 	sp = sctp_sk(sk);
5365 
5366 	if (params.assoc_id != 0) {
5367 		asoc = sctp_id2assoc(sk, params.assoc_id);
5368 		if (!asoc)
5369 			return -EINVAL;
5370 		params.assoc_value = asoc->max_burst;
5371 	} else
5372 		params.assoc_value = sp->max_burst;
5373 
5374 	if (len == sizeof(int)) {
5375 		if (copy_to_user(optval, &params.assoc_value, len))
5376 			return -EFAULT;
5377 	} else {
5378 		if (copy_to_user(optval, &params, len))
5379 			return -EFAULT;
5380 	}
5381 
5382 	return 0;
5383 
5384 }
5385 
5386 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5387 				    char __user *optval, int __user *optlen)
5388 {
5389 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5390 	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5391 	struct sctp_hmac_algo_param *hmacs;
5392 	__u16 data_len = 0;
5393 	u32 num_idents;
5394 
5395 	if (!ep->auth_enable)
5396 		return -EACCES;
5397 
5398 	hmacs = ep->auth_hmacs_list;
5399 	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5400 
5401 	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5402 		return -EINVAL;
5403 
5404 	len = sizeof(struct sctp_hmacalgo) + data_len;
5405 	num_idents = data_len / sizeof(u16);
5406 
5407 	if (put_user(len, optlen))
5408 		return -EFAULT;
5409 	if (put_user(num_idents, &p->shmac_num_idents))
5410 		return -EFAULT;
5411 	if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5412 		return -EFAULT;
5413 	return 0;
5414 }
5415 
5416 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5417 				    char __user *optval, int __user *optlen)
5418 {
5419 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5420 	struct sctp_authkeyid val;
5421 	struct sctp_association *asoc;
5422 
5423 	if (!ep->auth_enable)
5424 		return -EACCES;
5425 
5426 	if (len < sizeof(struct sctp_authkeyid))
5427 		return -EINVAL;
5428 	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5429 		return -EFAULT;
5430 
5431 	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5432 	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5433 		return -EINVAL;
5434 
5435 	if (asoc)
5436 		val.scact_keynumber = asoc->active_key_id;
5437 	else
5438 		val.scact_keynumber = ep->active_key_id;
5439 
5440 	len = sizeof(struct sctp_authkeyid);
5441 	if (put_user(len, optlen))
5442 		return -EFAULT;
5443 	if (copy_to_user(optval, &val, len))
5444 		return -EFAULT;
5445 
5446 	return 0;
5447 }
5448 
5449 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5450 				    char __user *optval, int __user *optlen)
5451 {
5452 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5453 	struct sctp_authchunks __user *p = (void __user *)optval;
5454 	struct sctp_authchunks val;
5455 	struct sctp_association *asoc;
5456 	struct sctp_chunks_param *ch;
5457 	u32    num_chunks = 0;
5458 	char __user *to;
5459 
5460 	if (!ep->auth_enable)
5461 		return -EACCES;
5462 
5463 	if (len < sizeof(struct sctp_authchunks))
5464 		return -EINVAL;
5465 
5466 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5467 		return -EFAULT;
5468 
5469 	to = p->gauth_chunks;
5470 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5471 	if (!asoc)
5472 		return -EINVAL;
5473 
5474 	ch = asoc->peer.peer_chunks;
5475 	if (!ch)
5476 		goto num;
5477 
5478 	/* See if the user provided enough room for all the data */
5479 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5480 	if (len < num_chunks)
5481 		return -EINVAL;
5482 
5483 	if (copy_to_user(to, ch->chunks, num_chunks))
5484 		return -EFAULT;
5485 num:
5486 	len = sizeof(struct sctp_authchunks) + num_chunks;
5487 	if (put_user(len, optlen))
5488 		return -EFAULT;
5489 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5490 		return -EFAULT;
5491 	return 0;
5492 }
5493 
5494 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5495 				    char __user *optval, int __user *optlen)
5496 {
5497 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5498 	struct sctp_authchunks __user *p = (void __user *)optval;
5499 	struct sctp_authchunks val;
5500 	struct sctp_association *asoc;
5501 	struct sctp_chunks_param *ch;
5502 	u32    num_chunks = 0;
5503 	char __user *to;
5504 
5505 	if (!ep->auth_enable)
5506 		return -EACCES;
5507 
5508 	if (len < sizeof(struct sctp_authchunks))
5509 		return -EINVAL;
5510 
5511 	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5512 		return -EFAULT;
5513 
5514 	to = p->gauth_chunks;
5515 	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5516 	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5517 		return -EINVAL;
5518 
5519 	if (asoc)
5520 		ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5521 	else
5522 		ch = ep->auth_chunk_list;
5523 
5524 	if (!ch)
5525 		goto num;
5526 
5527 	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5528 	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5529 		return -EINVAL;
5530 
5531 	if (copy_to_user(to, ch->chunks, num_chunks))
5532 		return -EFAULT;
5533 num:
5534 	len = sizeof(struct sctp_authchunks) + num_chunks;
5535 	if (put_user(len, optlen))
5536 		return -EFAULT;
5537 	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5538 		return -EFAULT;
5539 
5540 	return 0;
5541 }
5542 
5543 /*
5544  * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5545  * This option gets the current number of associations that are attached
5546  * to a one-to-many style socket.  The option value is an uint32_t.
5547  */
5548 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5549 				    char __user *optval, int __user *optlen)
5550 {
5551 	struct sctp_sock *sp = sctp_sk(sk);
5552 	struct sctp_association *asoc;
5553 	u32 val = 0;
5554 
5555 	if (sctp_style(sk, TCP))
5556 		return -EOPNOTSUPP;
5557 
5558 	if (len < sizeof(u32))
5559 		return -EINVAL;
5560 
5561 	len = sizeof(u32);
5562 
5563 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5564 		val++;
5565 	}
5566 
5567 	if (put_user(len, optlen))
5568 		return -EFAULT;
5569 	if (copy_to_user(optval, &val, len))
5570 		return -EFAULT;
5571 
5572 	return 0;
5573 }
5574 
5575 /*
5576  * 8.1.23 SCTP_AUTO_ASCONF
5577  * See the corresponding setsockopt entry as description
5578  */
5579 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5580 				   char __user *optval, int __user *optlen)
5581 {
5582 	int val = 0;
5583 
5584 	if (len < sizeof(int))
5585 		return -EINVAL;
5586 
5587 	len = sizeof(int);
5588 	if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5589 		val = 1;
5590 	if (put_user(len, optlen))
5591 		return -EFAULT;
5592 	if (copy_to_user(optval, &val, len))
5593 		return -EFAULT;
5594 	return 0;
5595 }
5596 
5597 /*
5598  * 8.2.6. Get the Current Identifiers of Associations
5599  *        (SCTP_GET_ASSOC_ID_LIST)
5600  *
5601  * This option gets the current list of SCTP association identifiers of
5602  * the SCTP associations handled by a one-to-many style socket.
5603  */
5604 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5605 				    char __user *optval, int __user *optlen)
5606 {
5607 	struct sctp_sock *sp = sctp_sk(sk);
5608 	struct sctp_association *asoc;
5609 	struct sctp_assoc_ids *ids;
5610 	u32 num = 0;
5611 
5612 	if (sctp_style(sk, TCP))
5613 		return -EOPNOTSUPP;
5614 
5615 	if (len < sizeof(struct sctp_assoc_ids))
5616 		return -EINVAL;
5617 
5618 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5619 		num++;
5620 	}
5621 
5622 	if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5623 		return -EINVAL;
5624 
5625 	len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5626 
5627 	ids = kmalloc(len, GFP_KERNEL);
5628 	if (unlikely(!ids))
5629 		return -ENOMEM;
5630 
5631 	ids->gaids_number_of_ids = num;
5632 	num = 0;
5633 	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5634 		ids->gaids_assoc_id[num++] = asoc->assoc_id;
5635 	}
5636 
5637 	if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5638 		kfree(ids);
5639 		return -EFAULT;
5640 	}
5641 
5642 	kfree(ids);
5643 	return 0;
5644 }
5645 
5646 /*
5647  * SCTP_PEER_ADDR_THLDS
5648  *
5649  * This option allows us to fetch the partially failed threshold for one or all
5650  * transports in an association.  See Section 6.1 of:
5651  * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5652  */
5653 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5654 					    char __user *optval,
5655 					    int len,
5656 					    int __user *optlen)
5657 {
5658 	struct sctp_paddrthlds val;
5659 	struct sctp_transport *trans;
5660 	struct sctp_association *asoc;
5661 
5662 	if (len < sizeof(struct sctp_paddrthlds))
5663 		return -EINVAL;
5664 	len = sizeof(struct sctp_paddrthlds);
5665 	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5666 		return -EFAULT;
5667 
5668 	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5669 		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5670 		if (!asoc)
5671 			return -ENOENT;
5672 
5673 		val.spt_pathpfthld = asoc->pf_retrans;
5674 		val.spt_pathmaxrxt = asoc->pathmaxrxt;
5675 	} else {
5676 		trans = sctp_addr_id2transport(sk, &val.spt_address,
5677 					       val.spt_assoc_id);
5678 		if (!trans)
5679 			return -ENOENT;
5680 
5681 		val.spt_pathmaxrxt = trans->pathmaxrxt;
5682 		val.spt_pathpfthld = trans->pf_retrans;
5683 	}
5684 
5685 	if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5686 		return -EFAULT;
5687 
5688 	return 0;
5689 }
5690 
5691 /*
5692  * SCTP_GET_ASSOC_STATS
5693  *
5694  * This option retrieves local per endpoint statistics. It is modeled
5695  * after OpenSolaris' implementation
5696  */
5697 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5698 				       char __user *optval,
5699 				       int __user *optlen)
5700 {
5701 	struct sctp_assoc_stats sas;
5702 	struct sctp_association *asoc = NULL;
5703 
5704 	/* User must provide at least the assoc id */
5705 	if (len < sizeof(sctp_assoc_t))
5706 		return -EINVAL;
5707 
5708 	/* Allow the struct to grow and fill in as much as possible */
5709 	len = min_t(size_t, len, sizeof(sas));
5710 
5711 	if (copy_from_user(&sas, optval, len))
5712 		return -EFAULT;
5713 
5714 	asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5715 	if (!asoc)
5716 		return -EINVAL;
5717 
5718 	sas.sas_rtxchunks = asoc->stats.rtxchunks;
5719 	sas.sas_gapcnt = asoc->stats.gapcnt;
5720 	sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5721 	sas.sas_osacks = asoc->stats.osacks;
5722 	sas.sas_isacks = asoc->stats.isacks;
5723 	sas.sas_octrlchunks = asoc->stats.octrlchunks;
5724 	sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5725 	sas.sas_oodchunks = asoc->stats.oodchunks;
5726 	sas.sas_iodchunks = asoc->stats.iodchunks;
5727 	sas.sas_ouodchunks = asoc->stats.ouodchunks;
5728 	sas.sas_iuodchunks = asoc->stats.iuodchunks;
5729 	sas.sas_idupchunks = asoc->stats.idupchunks;
5730 	sas.sas_opackets = asoc->stats.opackets;
5731 	sas.sas_ipackets = asoc->stats.ipackets;
5732 
5733 	/* New high max rto observed, will return 0 if not a single
5734 	 * RTO update took place. obs_rto_ipaddr will be bogus
5735 	 * in such a case
5736 	 */
5737 	sas.sas_maxrto = asoc->stats.max_obs_rto;
5738 	memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5739 		sizeof(struct sockaddr_storage));
5740 
5741 	/* Mark beginning of a new observation period */
5742 	asoc->stats.max_obs_rto = asoc->rto_min;
5743 
5744 	if (put_user(len, optlen))
5745 		return -EFAULT;
5746 
5747 	pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5748 
5749 	if (copy_to_user(optval, &sas, len))
5750 		return -EFAULT;
5751 
5752 	return 0;
5753 }
5754 
5755 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5756 			   char __user *optval, int __user *optlen)
5757 {
5758 	int retval = 0;
5759 	int len;
5760 
5761 	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5762 
5763 	/* I can hardly begin to describe how wrong this is.  This is
5764 	 * so broken as to be worse than useless.  The API draft
5765 	 * REALLY is NOT helpful here...  I am not convinced that the
5766 	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5767 	 * are at all well-founded.
5768 	 */
5769 	if (level != SOL_SCTP) {
5770 		struct sctp_af *af = sctp_sk(sk)->pf->af;
5771 
5772 		retval = af->getsockopt(sk, level, optname, optval, optlen);
5773 		return retval;
5774 	}
5775 
5776 	if (get_user(len, optlen))
5777 		return -EFAULT;
5778 
5779 	lock_sock(sk);
5780 
5781 	switch (optname) {
5782 	case SCTP_STATUS:
5783 		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5784 		break;
5785 	case SCTP_DISABLE_FRAGMENTS:
5786 		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5787 							   optlen);
5788 		break;
5789 	case SCTP_EVENTS:
5790 		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5791 		break;
5792 	case SCTP_AUTOCLOSE:
5793 		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5794 		break;
5795 	case SCTP_SOCKOPT_PEELOFF:
5796 		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5797 		break;
5798 	case SCTP_PEER_ADDR_PARAMS:
5799 		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5800 							  optlen);
5801 		break;
5802 	case SCTP_DELAYED_SACK:
5803 		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5804 							  optlen);
5805 		break;
5806 	case SCTP_INITMSG:
5807 		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5808 		break;
5809 	case SCTP_GET_PEER_ADDRS:
5810 		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5811 						    optlen);
5812 		break;
5813 	case SCTP_GET_LOCAL_ADDRS:
5814 		retval = sctp_getsockopt_local_addrs(sk, len, optval,
5815 						     optlen);
5816 		break;
5817 	case SCTP_SOCKOPT_CONNECTX3:
5818 		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5819 		break;
5820 	case SCTP_DEFAULT_SEND_PARAM:
5821 		retval = sctp_getsockopt_default_send_param(sk, len,
5822 							    optval, optlen);
5823 		break;
5824 	case SCTP_PRIMARY_ADDR:
5825 		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5826 		break;
5827 	case SCTP_NODELAY:
5828 		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5829 		break;
5830 	case SCTP_RTOINFO:
5831 		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5832 		break;
5833 	case SCTP_ASSOCINFO:
5834 		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5835 		break;
5836 	case SCTP_I_WANT_MAPPED_V4_ADDR:
5837 		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5838 		break;
5839 	case SCTP_MAXSEG:
5840 		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5841 		break;
5842 	case SCTP_GET_PEER_ADDR_INFO:
5843 		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5844 							optlen);
5845 		break;
5846 	case SCTP_ADAPTATION_LAYER:
5847 		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5848 							optlen);
5849 		break;
5850 	case SCTP_CONTEXT:
5851 		retval = sctp_getsockopt_context(sk, len, optval, optlen);
5852 		break;
5853 	case SCTP_FRAGMENT_INTERLEAVE:
5854 		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5855 							     optlen);
5856 		break;
5857 	case SCTP_PARTIAL_DELIVERY_POINT:
5858 		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5859 								optlen);
5860 		break;
5861 	case SCTP_MAX_BURST:
5862 		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5863 		break;
5864 	case SCTP_AUTH_KEY:
5865 	case SCTP_AUTH_CHUNK:
5866 	case SCTP_AUTH_DELETE_KEY:
5867 		retval = -EOPNOTSUPP;
5868 		break;
5869 	case SCTP_HMAC_IDENT:
5870 		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5871 		break;
5872 	case SCTP_AUTH_ACTIVE_KEY:
5873 		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5874 		break;
5875 	case SCTP_PEER_AUTH_CHUNKS:
5876 		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5877 							optlen);
5878 		break;
5879 	case SCTP_LOCAL_AUTH_CHUNKS:
5880 		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5881 							optlen);
5882 		break;
5883 	case SCTP_GET_ASSOC_NUMBER:
5884 		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5885 		break;
5886 	case SCTP_GET_ASSOC_ID_LIST:
5887 		retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5888 		break;
5889 	case SCTP_AUTO_ASCONF:
5890 		retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5891 		break;
5892 	case SCTP_PEER_ADDR_THLDS:
5893 		retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5894 		break;
5895 	case SCTP_GET_ASSOC_STATS:
5896 		retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5897 		break;
5898 	default:
5899 		retval = -ENOPROTOOPT;
5900 		break;
5901 	}
5902 
5903 	release_sock(sk);
5904 	return retval;
5905 }
5906 
5907 static void sctp_hash(struct sock *sk)
5908 {
5909 	/* STUB */
5910 }
5911 
5912 static void sctp_unhash(struct sock *sk)
5913 {
5914 	/* STUB */
5915 }
5916 
5917 /* Check if port is acceptable.  Possibly find first available port.
5918  *
5919  * The port hash table (contained in the 'global' SCTP protocol storage
5920  * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5921  * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5922  * list (the list number is the port number hashed out, so as you
5923  * would expect from a hash function, all the ports in a given list have
5924  * such a number that hashes out to the same list number; you were
5925  * expecting that, right?); so each list has a set of ports, with a
5926  * link to the socket (struct sock) that uses it, the port number and
5927  * a fastreuse flag (FIXME: NPI ipg).
5928  */
5929 static struct sctp_bind_bucket *sctp_bucket_create(
5930 	struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5931 
5932 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5933 {
5934 	struct sctp_bind_hashbucket *head; /* hash list */
5935 	struct sctp_bind_bucket *pp;
5936 	unsigned short snum;
5937 	int ret;
5938 
5939 	snum = ntohs(addr->v4.sin_port);
5940 
5941 	pr_debug("%s: begins, snum:%d\n", __func__, snum);
5942 
5943 	local_bh_disable();
5944 
5945 	if (snum == 0) {
5946 		/* Search for an available port. */
5947 		int low, high, remaining, index;
5948 		unsigned int rover;
5949 		struct net *net = sock_net(sk);
5950 
5951 		inet_get_local_port_range(net, &low, &high);
5952 		remaining = (high - low) + 1;
5953 		rover = prandom_u32() % remaining + low;
5954 
5955 		do {
5956 			rover++;
5957 			if ((rover < low) || (rover > high))
5958 				rover = low;
5959 			if (inet_is_local_reserved_port(net, rover))
5960 				continue;
5961 			index = sctp_phashfn(sock_net(sk), rover);
5962 			head = &sctp_port_hashtable[index];
5963 			spin_lock(&head->lock);
5964 			sctp_for_each_hentry(pp, &head->chain)
5965 				if ((pp->port == rover) &&
5966 				    net_eq(sock_net(sk), pp->net))
5967 					goto next;
5968 			break;
5969 		next:
5970 			spin_unlock(&head->lock);
5971 		} while (--remaining > 0);
5972 
5973 		/* Exhausted local port range during search? */
5974 		ret = 1;
5975 		if (remaining <= 0)
5976 			goto fail;
5977 
5978 		/* OK, here is the one we will use.  HEAD (the port
5979 		 * hash table list entry) is non-NULL and we hold it's
5980 		 * mutex.
5981 		 */
5982 		snum = rover;
5983 	} else {
5984 		/* We are given an specific port number; we verify
5985 		 * that it is not being used. If it is used, we will
5986 		 * exahust the search in the hash list corresponding
5987 		 * to the port number (snum) - we detect that with the
5988 		 * port iterator, pp being NULL.
5989 		 */
5990 		head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5991 		spin_lock(&head->lock);
5992 		sctp_for_each_hentry(pp, &head->chain) {
5993 			if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5994 				goto pp_found;
5995 		}
5996 	}
5997 	pp = NULL;
5998 	goto pp_not_found;
5999 pp_found:
6000 	if (!hlist_empty(&pp->owner)) {
6001 		/* We had a port hash table hit - there is an
6002 		 * available port (pp != NULL) and it is being
6003 		 * used by other socket (pp->owner not empty); that other
6004 		 * socket is going to be sk2.
6005 		 */
6006 		int reuse = sk->sk_reuse;
6007 		struct sock *sk2;
6008 
6009 		pr_debug("%s: found a possible match\n", __func__);
6010 
6011 		if (pp->fastreuse && sk->sk_reuse &&
6012 			sk->sk_state != SCTP_SS_LISTENING)
6013 			goto success;
6014 
6015 		/* Run through the list of sockets bound to the port
6016 		 * (pp->port) [via the pointers bind_next and
6017 		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6018 		 * we get the endpoint they describe and run through
6019 		 * the endpoint's list of IP (v4 or v6) addresses,
6020 		 * comparing each of the addresses with the address of
6021 		 * the socket sk. If we find a match, then that means
6022 		 * that this port/socket (sk) combination are already
6023 		 * in an endpoint.
6024 		 */
6025 		sk_for_each_bound(sk2, &pp->owner) {
6026 			struct sctp_endpoint *ep2;
6027 			ep2 = sctp_sk(sk2)->ep;
6028 
6029 			if (sk == sk2 ||
6030 			    (reuse && sk2->sk_reuse &&
6031 			     sk2->sk_state != SCTP_SS_LISTENING))
6032 				continue;
6033 
6034 			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6035 						 sctp_sk(sk2), sctp_sk(sk))) {
6036 				ret = (long)sk2;
6037 				goto fail_unlock;
6038 			}
6039 		}
6040 
6041 		pr_debug("%s: found a match\n", __func__);
6042 	}
6043 pp_not_found:
6044 	/* If there was a hash table miss, create a new port.  */
6045 	ret = 1;
6046 	if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6047 		goto fail_unlock;
6048 
6049 	/* In either case (hit or miss), make sure fastreuse is 1 only
6050 	 * if sk->sk_reuse is too (that is, if the caller requested
6051 	 * SO_REUSEADDR on this socket -sk-).
6052 	 */
6053 	if (hlist_empty(&pp->owner)) {
6054 		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6055 			pp->fastreuse = 1;
6056 		else
6057 			pp->fastreuse = 0;
6058 	} else if (pp->fastreuse &&
6059 		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6060 		pp->fastreuse = 0;
6061 
6062 	/* We are set, so fill up all the data in the hash table
6063 	 * entry, tie the socket list information with the rest of the
6064 	 * sockets FIXME: Blurry, NPI (ipg).
6065 	 */
6066 success:
6067 	if (!sctp_sk(sk)->bind_hash) {
6068 		inet_sk(sk)->inet_num = snum;
6069 		sk_add_bind_node(sk, &pp->owner);
6070 		sctp_sk(sk)->bind_hash = pp;
6071 	}
6072 	ret = 0;
6073 
6074 fail_unlock:
6075 	spin_unlock(&head->lock);
6076 
6077 fail:
6078 	local_bh_enable();
6079 	return ret;
6080 }
6081 
6082 /* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
6083  * port is requested.
6084  */
6085 static int sctp_get_port(struct sock *sk, unsigned short snum)
6086 {
6087 	union sctp_addr addr;
6088 	struct sctp_af *af = sctp_sk(sk)->pf->af;
6089 
6090 	/* Set up a dummy address struct from the sk. */
6091 	af->from_sk(&addr, sk);
6092 	addr.v4.sin_port = htons(snum);
6093 
6094 	/* Note: sk->sk_num gets filled in if ephemeral port request. */
6095 	return !!sctp_get_port_local(sk, &addr);
6096 }
6097 
6098 /*
6099  *  Move a socket to LISTENING state.
6100  */
6101 static int sctp_listen_start(struct sock *sk, int backlog)
6102 {
6103 	struct sctp_sock *sp = sctp_sk(sk);
6104 	struct sctp_endpoint *ep = sp->ep;
6105 	struct crypto_hash *tfm = NULL;
6106 	char alg[32];
6107 
6108 	/* Allocate HMAC for generating cookie. */
6109 	if (!sp->hmac && sp->sctp_hmac_alg) {
6110 		sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6111 		tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6112 		if (IS_ERR(tfm)) {
6113 			net_info_ratelimited("failed to load transform for %s: %ld\n",
6114 					     sp->sctp_hmac_alg, PTR_ERR(tfm));
6115 			return -ENOSYS;
6116 		}
6117 		sctp_sk(sk)->hmac = tfm;
6118 	}
6119 
6120 	/*
6121 	 * If a bind() or sctp_bindx() is not called prior to a listen()
6122 	 * call that allows new associations to be accepted, the system
6123 	 * picks an ephemeral port and will choose an address set equivalent
6124 	 * to binding with a wildcard address.
6125 	 *
6126 	 * This is not currently spelled out in the SCTP sockets
6127 	 * extensions draft, but follows the practice as seen in TCP
6128 	 * sockets.
6129 	 *
6130 	 */
6131 	sk->sk_state = SCTP_SS_LISTENING;
6132 	if (!ep->base.bind_addr.port) {
6133 		if (sctp_autobind(sk))
6134 			return -EAGAIN;
6135 	} else {
6136 		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6137 			sk->sk_state = SCTP_SS_CLOSED;
6138 			return -EADDRINUSE;
6139 		}
6140 	}
6141 
6142 	sk->sk_max_ack_backlog = backlog;
6143 	sctp_hash_endpoint(ep);
6144 	return 0;
6145 }
6146 
6147 /*
6148  * 4.1.3 / 5.1.3 listen()
6149  *
6150  *   By default, new associations are not accepted for UDP style sockets.
6151  *   An application uses listen() to mark a socket as being able to
6152  *   accept new associations.
6153  *
6154  *   On TCP style sockets, applications use listen() to ready the SCTP
6155  *   endpoint for accepting inbound associations.
6156  *
6157  *   On both types of endpoints a backlog of '0' disables listening.
6158  *
6159  *  Move a socket to LISTENING state.
6160  */
6161 int sctp_inet_listen(struct socket *sock, int backlog)
6162 {
6163 	struct sock *sk = sock->sk;
6164 	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6165 	int err = -EINVAL;
6166 
6167 	if (unlikely(backlog < 0))
6168 		return err;
6169 
6170 	lock_sock(sk);
6171 
6172 	/* Peeled-off sockets are not allowed to listen().  */
6173 	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6174 		goto out;
6175 
6176 	if (sock->state != SS_UNCONNECTED)
6177 		goto out;
6178 
6179 	/* If backlog is zero, disable listening. */
6180 	if (!backlog) {
6181 		if (sctp_sstate(sk, CLOSED))
6182 			goto out;
6183 
6184 		err = 0;
6185 		sctp_unhash_endpoint(ep);
6186 		sk->sk_state = SCTP_SS_CLOSED;
6187 		if (sk->sk_reuse)
6188 			sctp_sk(sk)->bind_hash->fastreuse = 1;
6189 		goto out;
6190 	}
6191 
6192 	/* If we are already listening, just update the backlog */
6193 	if (sctp_sstate(sk, LISTENING))
6194 		sk->sk_max_ack_backlog = backlog;
6195 	else {
6196 		err = sctp_listen_start(sk, backlog);
6197 		if (err)
6198 			goto out;
6199 	}
6200 
6201 	err = 0;
6202 out:
6203 	release_sock(sk);
6204 	return err;
6205 }
6206 
6207 /*
6208  * This function is done by modeling the current datagram_poll() and the
6209  * tcp_poll().  Note that, based on these implementations, we don't
6210  * lock the socket in this function, even though it seems that,
6211  * ideally, locking or some other mechanisms can be used to ensure
6212  * the integrity of the counters (sndbuf and wmem_alloc) used
6213  * in this place.  We assume that we don't need locks either until proven
6214  * otherwise.
6215  *
6216  * Another thing to note is that we include the Async I/O support
6217  * here, again, by modeling the current TCP/UDP code.  We don't have
6218  * a good way to test with it yet.
6219  */
6220 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6221 {
6222 	struct sock *sk = sock->sk;
6223 	struct sctp_sock *sp = sctp_sk(sk);
6224 	unsigned int mask;
6225 
6226 	poll_wait(file, sk_sleep(sk), wait);
6227 
6228 	/* A TCP-style listening socket becomes readable when the accept queue
6229 	 * is not empty.
6230 	 */
6231 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6232 		return (!list_empty(&sp->ep->asocs)) ?
6233 			(POLLIN | POLLRDNORM) : 0;
6234 
6235 	mask = 0;
6236 
6237 	/* Is there any exceptional events?  */
6238 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6239 		mask |= POLLERR |
6240 			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6241 	if (sk->sk_shutdown & RCV_SHUTDOWN)
6242 		mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6243 	if (sk->sk_shutdown == SHUTDOWN_MASK)
6244 		mask |= POLLHUP;
6245 
6246 	/* Is it readable?  Reconsider this code with TCP-style support.  */
6247 	if (!skb_queue_empty(&sk->sk_receive_queue))
6248 		mask |= POLLIN | POLLRDNORM;
6249 
6250 	/* The association is either gone or not ready.  */
6251 	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6252 		return mask;
6253 
6254 	/* Is it writable?  */
6255 	if (sctp_writeable(sk)) {
6256 		mask |= POLLOUT | POLLWRNORM;
6257 	} else {
6258 		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6259 		/*
6260 		 * Since the socket is not locked, the buffer
6261 		 * might be made available after the writeable check and
6262 		 * before the bit is set.  This could cause a lost I/O
6263 		 * signal.  tcp_poll() has a race breaker for this race
6264 		 * condition.  Based on their implementation, we put
6265 		 * in the following code to cover it as well.
6266 		 */
6267 		if (sctp_writeable(sk))
6268 			mask |= POLLOUT | POLLWRNORM;
6269 	}
6270 	return mask;
6271 }
6272 
6273 /********************************************************************
6274  * 2nd Level Abstractions
6275  ********************************************************************/
6276 
6277 static struct sctp_bind_bucket *sctp_bucket_create(
6278 	struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6279 {
6280 	struct sctp_bind_bucket *pp;
6281 
6282 	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6283 	if (pp) {
6284 		SCTP_DBG_OBJCNT_INC(bind_bucket);
6285 		pp->port = snum;
6286 		pp->fastreuse = 0;
6287 		INIT_HLIST_HEAD(&pp->owner);
6288 		pp->net = net;
6289 		hlist_add_head(&pp->node, &head->chain);
6290 	}
6291 	return pp;
6292 }
6293 
6294 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6295 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6296 {
6297 	if (pp && hlist_empty(&pp->owner)) {
6298 		__hlist_del(&pp->node);
6299 		kmem_cache_free(sctp_bucket_cachep, pp);
6300 		SCTP_DBG_OBJCNT_DEC(bind_bucket);
6301 	}
6302 }
6303 
6304 /* Release this socket's reference to a local port.  */
6305 static inline void __sctp_put_port(struct sock *sk)
6306 {
6307 	struct sctp_bind_hashbucket *head =
6308 		&sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6309 						  inet_sk(sk)->inet_num)];
6310 	struct sctp_bind_bucket *pp;
6311 
6312 	spin_lock(&head->lock);
6313 	pp = sctp_sk(sk)->bind_hash;
6314 	__sk_del_bind_node(sk);
6315 	sctp_sk(sk)->bind_hash = NULL;
6316 	inet_sk(sk)->inet_num = 0;
6317 	sctp_bucket_destroy(pp);
6318 	spin_unlock(&head->lock);
6319 }
6320 
6321 void sctp_put_port(struct sock *sk)
6322 {
6323 	local_bh_disable();
6324 	__sctp_put_port(sk);
6325 	local_bh_enable();
6326 }
6327 
6328 /*
6329  * The system picks an ephemeral port and choose an address set equivalent
6330  * to binding with a wildcard address.
6331  * One of those addresses will be the primary address for the association.
6332  * This automatically enables the multihoming capability of SCTP.
6333  */
6334 static int sctp_autobind(struct sock *sk)
6335 {
6336 	union sctp_addr autoaddr;
6337 	struct sctp_af *af;
6338 	__be16 port;
6339 
6340 	/* Initialize a local sockaddr structure to INADDR_ANY. */
6341 	af = sctp_sk(sk)->pf->af;
6342 
6343 	port = htons(inet_sk(sk)->inet_num);
6344 	af->inaddr_any(&autoaddr, port);
6345 
6346 	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6347 }
6348 
6349 /* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
6350  *
6351  * From RFC 2292
6352  * 4.2 The cmsghdr Structure *
6353  *
6354  * When ancillary data is sent or received, any number of ancillary data
6355  * objects can be specified by the msg_control and msg_controllen members of
6356  * the msghdr structure, because each object is preceded by
6357  * a cmsghdr structure defining the object's length (the cmsg_len member).
6358  * Historically Berkeley-derived implementations have passed only one object
6359  * at a time, but this API allows multiple objects to be
6360  * passed in a single call to sendmsg() or recvmsg(). The following example
6361  * shows two ancillary data objects in a control buffer.
6362  *
6363  *   |<--------------------------- msg_controllen -------------------------->|
6364  *   |                                                                       |
6365  *
6366  *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
6367  *
6368  *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6369  *   |                                   |                                   |
6370  *
6371  *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
6372  *
6373  *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
6374  *   |                                |  |                                |  |
6375  *
6376  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6377  *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
6378  *
6379  *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
6380  *
6381  *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6382  *    ^
6383  *    |
6384  *
6385  * msg_control
6386  * points here
6387  */
6388 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6389 {
6390 	struct cmsghdr *cmsg;
6391 	struct msghdr *my_msg = (struct msghdr *)msg;
6392 
6393 	for (cmsg = CMSG_FIRSTHDR(msg);
6394 	     cmsg != NULL;
6395 	     cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6396 		if (!CMSG_OK(my_msg, cmsg))
6397 			return -EINVAL;
6398 
6399 		/* Should we parse this header or ignore?  */
6400 		if (cmsg->cmsg_level != IPPROTO_SCTP)
6401 			continue;
6402 
6403 		/* Strictly check lengths following example in SCM code.  */
6404 		switch (cmsg->cmsg_type) {
6405 		case SCTP_INIT:
6406 			/* SCTP Socket API Extension
6407 			 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6408 			 *
6409 			 * This cmsghdr structure provides information for
6410 			 * initializing new SCTP associations with sendmsg().
6411 			 * The SCTP_INITMSG socket option uses this same data
6412 			 * structure.  This structure is not used for
6413 			 * recvmsg().
6414 			 *
6415 			 * cmsg_level    cmsg_type      cmsg_data[]
6416 			 * ------------  ------------   ----------------------
6417 			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
6418 			 */
6419 			if (cmsg->cmsg_len !=
6420 			    CMSG_LEN(sizeof(struct sctp_initmsg)))
6421 				return -EINVAL;
6422 			cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6423 			break;
6424 
6425 		case SCTP_SNDRCV:
6426 			/* SCTP Socket API Extension
6427 			 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6428 			 *
6429 			 * This cmsghdr structure specifies SCTP options for
6430 			 * sendmsg() and describes SCTP header information
6431 			 * about a received message through recvmsg().
6432 			 *
6433 			 * cmsg_level    cmsg_type      cmsg_data[]
6434 			 * ------------  ------------   ----------------------
6435 			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
6436 			 */
6437 			if (cmsg->cmsg_len !=
6438 			    CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6439 				return -EINVAL;
6440 
6441 			cmsgs->info =
6442 				(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6443 
6444 			/* Minimally, validate the sinfo_flags. */
6445 			if (cmsgs->info->sinfo_flags &
6446 			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6447 			      SCTP_ABORT | SCTP_EOF))
6448 				return -EINVAL;
6449 			break;
6450 
6451 		default:
6452 			return -EINVAL;
6453 		}
6454 	}
6455 	return 0;
6456 }
6457 
6458 /*
6459  * Wait for a packet..
6460  * Note: This function is the same function as in core/datagram.c
6461  * with a few modifications to make lksctp work.
6462  */
6463 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6464 {
6465 	int error;
6466 	DEFINE_WAIT(wait);
6467 
6468 	prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6469 
6470 	/* Socket errors? */
6471 	error = sock_error(sk);
6472 	if (error)
6473 		goto out;
6474 
6475 	if (!skb_queue_empty(&sk->sk_receive_queue))
6476 		goto ready;
6477 
6478 	/* Socket shut down?  */
6479 	if (sk->sk_shutdown & RCV_SHUTDOWN)
6480 		goto out;
6481 
6482 	/* Sequenced packets can come disconnected.  If so we report the
6483 	 * problem.
6484 	 */
6485 	error = -ENOTCONN;
6486 
6487 	/* Is there a good reason to think that we may receive some data?  */
6488 	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6489 		goto out;
6490 
6491 	/* Handle signals.  */
6492 	if (signal_pending(current))
6493 		goto interrupted;
6494 
6495 	/* Let another process have a go.  Since we are going to sleep
6496 	 * anyway.  Note: This may cause odd behaviors if the message
6497 	 * does not fit in the user's buffer, but this seems to be the
6498 	 * only way to honor MSG_DONTWAIT realistically.
6499 	 */
6500 	release_sock(sk);
6501 	*timeo_p = schedule_timeout(*timeo_p);
6502 	lock_sock(sk);
6503 
6504 ready:
6505 	finish_wait(sk_sleep(sk), &wait);
6506 	return 0;
6507 
6508 interrupted:
6509 	error = sock_intr_errno(*timeo_p);
6510 
6511 out:
6512 	finish_wait(sk_sleep(sk), &wait);
6513 	*err = error;
6514 	return error;
6515 }
6516 
6517 /* Receive a datagram.
6518  * Note: This is pretty much the same routine as in core/datagram.c
6519  * with a few changes to make lksctp work.
6520  */
6521 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6522 					      int noblock, int *err)
6523 {
6524 	int error;
6525 	struct sk_buff *skb;
6526 	long timeo;
6527 
6528 	timeo = sock_rcvtimeo(sk, noblock);
6529 
6530 	pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6531 		 MAX_SCHEDULE_TIMEOUT);
6532 
6533 	do {
6534 		/* Again only user level code calls this function,
6535 		 * so nothing interrupt level
6536 		 * will suddenly eat the receive_queue.
6537 		 *
6538 		 *  Look at current nfs client by the way...
6539 		 *  However, this function was correct in any case. 8)
6540 		 */
6541 		if (flags & MSG_PEEK) {
6542 			spin_lock_bh(&sk->sk_receive_queue.lock);
6543 			skb = skb_peek(&sk->sk_receive_queue);
6544 			if (skb)
6545 				atomic_inc(&skb->users);
6546 			spin_unlock_bh(&sk->sk_receive_queue.lock);
6547 		} else {
6548 			skb = skb_dequeue(&sk->sk_receive_queue);
6549 		}
6550 
6551 		if (skb)
6552 			return skb;
6553 
6554 		/* Caller is allowed not to check sk->sk_err before calling. */
6555 		error = sock_error(sk);
6556 		if (error)
6557 			goto no_packet;
6558 
6559 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6560 			break;
6561 
6562 		if (sk_can_busy_loop(sk) &&
6563 		    sk_busy_loop(sk, noblock))
6564 			continue;
6565 
6566 		/* User doesn't want to wait.  */
6567 		error = -EAGAIN;
6568 		if (!timeo)
6569 			goto no_packet;
6570 	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6571 
6572 	return NULL;
6573 
6574 no_packet:
6575 	*err = error;
6576 	return NULL;
6577 }
6578 
6579 /* If sndbuf has changed, wake up per association sndbuf waiters.  */
6580 static void __sctp_write_space(struct sctp_association *asoc)
6581 {
6582 	struct sock *sk = asoc->base.sk;
6583 	struct socket *sock = sk->sk_socket;
6584 
6585 	if ((sctp_wspace(asoc) > 0) && sock) {
6586 		if (waitqueue_active(&asoc->wait))
6587 			wake_up_interruptible(&asoc->wait);
6588 
6589 		if (sctp_writeable(sk)) {
6590 			wait_queue_head_t *wq = sk_sleep(sk);
6591 
6592 			if (wq && waitqueue_active(wq))
6593 				wake_up_interruptible(wq);
6594 
6595 			/* Note that we try to include the Async I/O support
6596 			 * here by modeling from the current TCP/UDP code.
6597 			 * We have not tested with it yet.
6598 			 */
6599 			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6600 				sock_wake_async(sock,
6601 						SOCK_WAKE_SPACE, POLL_OUT);
6602 		}
6603 	}
6604 }
6605 
6606 static void sctp_wake_up_waiters(struct sock *sk,
6607 				 struct sctp_association *asoc)
6608 {
6609 	struct sctp_association *tmp = asoc;
6610 
6611 	/* We do accounting for the sndbuf space per association,
6612 	 * so we only need to wake our own association.
6613 	 */
6614 	if (asoc->ep->sndbuf_policy)
6615 		return __sctp_write_space(asoc);
6616 
6617 	/* If association goes down and is just flushing its
6618 	 * outq, then just normally notify others.
6619 	 */
6620 	if (asoc->base.dead)
6621 		return sctp_write_space(sk);
6622 
6623 	/* Accounting for the sndbuf space is per socket, so we
6624 	 * need to wake up others, try to be fair and in case of
6625 	 * other associations, let them have a go first instead
6626 	 * of just doing a sctp_write_space() call.
6627 	 *
6628 	 * Note that we reach sctp_wake_up_waiters() only when
6629 	 * associations free up queued chunks, thus we are under
6630 	 * lock and the list of associations on a socket is
6631 	 * guaranteed not to change.
6632 	 */
6633 	for (tmp = list_next_entry(tmp, asocs); 1;
6634 	     tmp = list_next_entry(tmp, asocs)) {
6635 		/* Manually skip the head element. */
6636 		if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6637 			continue;
6638 		/* Wake up association. */
6639 		__sctp_write_space(tmp);
6640 		/* We've reached the end. */
6641 		if (tmp == asoc)
6642 			break;
6643 	}
6644 }
6645 
6646 /* Do accounting for the sndbuf space.
6647  * Decrement the used sndbuf space of the corresponding association by the
6648  * data size which was just transmitted(freed).
6649  */
6650 static void sctp_wfree(struct sk_buff *skb)
6651 {
6652 	struct sctp_association *asoc;
6653 	struct sctp_chunk *chunk;
6654 	struct sock *sk;
6655 
6656 	/* Get the saved chunk pointer.  */
6657 	chunk = *((struct sctp_chunk **)(skb->cb));
6658 	asoc = chunk->asoc;
6659 	sk = asoc->base.sk;
6660 	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6661 				sizeof(struct sk_buff) +
6662 				sizeof(struct sctp_chunk);
6663 
6664 	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6665 
6666 	/*
6667 	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6668 	 */
6669 	sk->sk_wmem_queued   -= skb->truesize;
6670 	sk_mem_uncharge(sk, skb->truesize);
6671 
6672 	sock_wfree(skb);
6673 	sctp_wake_up_waiters(sk, asoc);
6674 
6675 	sctp_association_put(asoc);
6676 }
6677 
6678 /* Do accounting for the receive space on the socket.
6679  * Accounting for the association is done in ulpevent.c
6680  * We set this as a destructor for the cloned data skbs so that
6681  * accounting is done at the correct time.
6682  */
6683 void sctp_sock_rfree(struct sk_buff *skb)
6684 {
6685 	struct sock *sk = skb->sk;
6686 	struct sctp_ulpevent *event = sctp_skb2event(skb);
6687 
6688 	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6689 
6690 	/*
6691 	 * Mimic the behavior of sock_rfree
6692 	 */
6693 	sk_mem_uncharge(sk, event->rmem_len);
6694 }
6695 
6696 
6697 /* Helper function to wait for space in the sndbuf.  */
6698 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6699 				size_t msg_len)
6700 {
6701 	struct sock *sk = asoc->base.sk;
6702 	int err = 0;
6703 	long current_timeo = *timeo_p;
6704 	DEFINE_WAIT(wait);
6705 
6706 	pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6707 		 *timeo_p, msg_len);
6708 
6709 	/* Increment the association's refcnt.  */
6710 	sctp_association_hold(asoc);
6711 
6712 	/* Wait on the association specific sndbuf space. */
6713 	for (;;) {
6714 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6715 					  TASK_INTERRUPTIBLE);
6716 		if (!*timeo_p)
6717 			goto do_nonblock;
6718 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6719 		    asoc->base.dead)
6720 			goto do_error;
6721 		if (signal_pending(current))
6722 			goto do_interrupted;
6723 		if (msg_len <= sctp_wspace(asoc))
6724 			break;
6725 
6726 		/* Let another process have a go.  Since we are going
6727 		 * to sleep anyway.
6728 		 */
6729 		release_sock(sk);
6730 		current_timeo = schedule_timeout(current_timeo);
6731 		BUG_ON(sk != asoc->base.sk);
6732 		lock_sock(sk);
6733 
6734 		*timeo_p = current_timeo;
6735 	}
6736 
6737 out:
6738 	finish_wait(&asoc->wait, &wait);
6739 
6740 	/* Release the association's refcnt.  */
6741 	sctp_association_put(asoc);
6742 
6743 	return err;
6744 
6745 do_error:
6746 	err = -EPIPE;
6747 	goto out;
6748 
6749 do_interrupted:
6750 	err = sock_intr_errno(*timeo_p);
6751 	goto out;
6752 
6753 do_nonblock:
6754 	err = -EAGAIN;
6755 	goto out;
6756 }
6757 
6758 void sctp_data_ready(struct sock *sk)
6759 {
6760 	struct socket_wq *wq;
6761 
6762 	rcu_read_lock();
6763 	wq = rcu_dereference(sk->sk_wq);
6764 	if (wq_has_sleeper(wq))
6765 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6766 						POLLRDNORM | POLLRDBAND);
6767 	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6768 	rcu_read_unlock();
6769 }
6770 
6771 /* If socket sndbuf has changed, wake up all per association waiters.  */
6772 void sctp_write_space(struct sock *sk)
6773 {
6774 	struct sctp_association *asoc;
6775 
6776 	/* Wake up the tasks in each wait queue.  */
6777 	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6778 		__sctp_write_space(asoc);
6779 	}
6780 }
6781 
6782 /* Is there any sndbuf space available on the socket?
6783  *
6784  * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6785  * associations on the same socket.  For a UDP-style socket with
6786  * multiple associations, it is possible for it to be "unwriteable"
6787  * prematurely.  I assume that this is acceptable because
6788  * a premature "unwriteable" is better than an accidental "writeable" which
6789  * would cause an unwanted block under certain circumstances.  For the 1-1
6790  * UDP-style sockets or TCP-style sockets, this code should work.
6791  *  - Daisy
6792  */
6793 static int sctp_writeable(struct sock *sk)
6794 {
6795 	int amt = 0;
6796 
6797 	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6798 	if (amt < 0)
6799 		amt = 0;
6800 	return amt;
6801 }
6802 
6803 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6804  * returns immediately with EINPROGRESS.
6805  */
6806 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6807 {
6808 	struct sock *sk = asoc->base.sk;
6809 	int err = 0;
6810 	long current_timeo = *timeo_p;
6811 	DEFINE_WAIT(wait);
6812 
6813 	pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6814 
6815 	/* Increment the association's refcnt.  */
6816 	sctp_association_hold(asoc);
6817 
6818 	for (;;) {
6819 		prepare_to_wait_exclusive(&asoc->wait, &wait,
6820 					  TASK_INTERRUPTIBLE);
6821 		if (!*timeo_p)
6822 			goto do_nonblock;
6823 		if (sk->sk_shutdown & RCV_SHUTDOWN)
6824 			break;
6825 		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6826 		    asoc->base.dead)
6827 			goto do_error;
6828 		if (signal_pending(current))
6829 			goto do_interrupted;
6830 
6831 		if (sctp_state(asoc, ESTABLISHED))
6832 			break;
6833 
6834 		/* Let another process have a go.  Since we are going
6835 		 * to sleep anyway.
6836 		 */
6837 		release_sock(sk);
6838 		current_timeo = schedule_timeout(current_timeo);
6839 		lock_sock(sk);
6840 
6841 		*timeo_p = current_timeo;
6842 	}
6843 
6844 out:
6845 	finish_wait(&asoc->wait, &wait);
6846 
6847 	/* Release the association's refcnt.  */
6848 	sctp_association_put(asoc);
6849 
6850 	return err;
6851 
6852 do_error:
6853 	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6854 		err = -ETIMEDOUT;
6855 	else
6856 		err = -ECONNREFUSED;
6857 	goto out;
6858 
6859 do_interrupted:
6860 	err = sock_intr_errno(*timeo_p);
6861 	goto out;
6862 
6863 do_nonblock:
6864 	err = -EINPROGRESS;
6865 	goto out;
6866 }
6867 
6868 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6869 {
6870 	struct sctp_endpoint *ep;
6871 	int err = 0;
6872 	DEFINE_WAIT(wait);
6873 
6874 	ep = sctp_sk(sk)->ep;
6875 
6876 
6877 	for (;;) {
6878 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6879 					  TASK_INTERRUPTIBLE);
6880 
6881 		if (list_empty(&ep->asocs)) {
6882 			release_sock(sk);
6883 			timeo = schedule_timeout(timeo);
6884 			lock_sock(sk);
6885 		}
6886 
6887 		err = -EINVAL;
6888 		if (!sctp_sstate(sk, LISTENING))
6889 			break;
6890 
6891 		err = 0;
6892 		if (!list_empty(&ep->asocs))
6893 			break;
6894 
6895 		err = sock_intr_errno(timeo);
6896 		if (signal_pending(current))
6897 			break;
6898 
6899 		err = -EAGAIN;
6900 		if (!timeo)
6901 			break;
6902 	}
6903 
6904 	finish_wait(sk_sleep(sk), &wait);
6905 
6906 	return err;
6907 }
6908 
6909 static void sctp_wait_for_close(struct sock *sk, long timeout)
6910 {
6911 	DEFINE_WAIT(wait);
6912 
6913 	do {
6914 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6915 		if (list_empty(&sctp_sk(sk)->ep->asocs))
6916 			break;
6917 		release_sock(sk);
6918 		timeout = schedule_timeout(timeout);
6919 		lock_sock(sk);
6920 	} while (!signal_pending(current) && timeout);
6921 
6922 	finish_wait(sk_sleep(sk), &wait);
6923 }
6924 
6925 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6926 {
6927 	struct sk_buff *frag;
6928 
6929 	if (!skb->data_len)
6930 		goto done;
6931 
6932 	/* Don't forget the fragments. */
6933 	skb_walk_frags(skb, frag)
6934 		sctp_skb_set_owner_r_frag(frag, sk);
6935 
6936 done:
6937 	sctp_skb_set_owner_r(skb, sk);
6938 }
6939 
6940 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6941 		    struct sctp_association *asoc)
6942 {
6943 	struct inet_sock *inet = inet_sk(sk);
6944 	struct inet_sock *newinet;
6945 
6946 	newsk->sk_type = sk->sk_type;
6947 	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6948 	newsk->sk_flags = sk->sk_flags;
6949 	newsk->sk_no_check_tx = sk->sk_no_check_tx;
6950 	newsk->sk_no_check_rx = sk->sk_no_check_rx;
6951 	newsk->sk_reuse = sk->sk_reuse;
6952 
6953 	newsk->sk_shutdown = sk->sk_shutdown;
6954 	newsk->sk_destruct = sctp_destruct_sock;
6955 	newsk->sk_family = sk->sk_family;
6956 	newsk->sk_protocol = IPPROTO_SCTP;
6957 	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6958 	newsk->sk_sndbuf = sk->sk_sndbuf;
6959 	newsk->sk_rcvbuf = sk->sk_rcvbuf;
6960 	newsk->sk_lingertime = sk->sk_lingertime;
6961 	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6962 	newsk->sk_sndtimeo = sk->sk_sndtimeo;
6963 
6964 	newinet = inet_sk(newsk);
6965 
6966 	/* Initialize sk's sport, dport, rcv_saddr and daddr for
6967 	 * getsockname() and getpeername()
6968 	 */
6969 	newinet->inet_sport = inet->inet_sport;
6970 	newinet->inet_saddr = inet->inet_saddr;
6971 	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6972 	newinet->inet_dport = htons(asoc->peer.port);
6973 	newinet->pmtudisc = inet->pmtudisc;
6974 	newinet->inet_id = asoc->next_tsn ^ jiffies;
6975 
6976 	newinet->uc_ttl = inet->uc_ttl;
6977 	newinet->mc_loop = 1;
6978 	newinet->mc_ttl = 1;
6979 	newinet->mc_index = 0;
6980 	newinet->mc_list = NULL;
6981 }
6982 
6983 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6984  * and its messages to the newsk.
6985  */
6986 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6987 			      struct sctp_association *assoc,
6988 			      sctp_socket_type_t type)
6989 {
6990 	struct sctp_sock *oldsp = sctp_sk(oldsk);
6991 	struct sctp_sock *newsp = sctp_sk(newsk);
6992 	struct sctp_bind_bucket *pp; /* hash list port iterator */
6993 	struct sctp_endpoint *newep = newsp->ep;
6994 	struct sk_buff *skb, *tmp;
6995 	struct sctp_ulpevent *event;
6996 	struct sctp_bind_hashbucket *head;
6997 	struct list_head tmplist;
6998 
6999 	/* Migrate socket buffer sizes and all the socket level options to the
7000 	 * new socket.
7001 	 */
7002 	newsk->sk_sndbuf = oldsk->sk_sndbuf;
7003 	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7004 	/* Brute force copy old sctp opt. */
7005 	if (oldsp->do_auto_asconf) {
7006 		memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7007 		inet_sk_copy_descendant(newsk, oldsk);
7008 		memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7009 	} else
7010 		inet_sk_copy_descendant(newsk, oldsk);
7011 
7012 	/* Restore the ep value that was overwritten with the above structure
7013 	 * copy.
7014 	 */
7015 	newsp->ep = newep;
7016 	newsp->hmac = NULL;
7017 
7018 	/* Hook this new socket in to the bind_hash list. */
7019 	head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7020 						 inet_sk(oldsk)->inet_num)];
7021 	local_bh_disable();
7022 	spin_lock(&head->lock);
7023 	pp = sctp_sk(oldsk)->bind_hash;
7024 	sk_add_bind_node(newsk, &pp->owner);
7025 	sctp_sk(newsk)->bind_hash = pp;
7026 	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7027 	spin_unlock(&head->lock);
7028 	local_bh_enable();
7029 
7030 	/* Copy the bind_addr list from the original endpoint to the new
7031 	 * endpoint so that we can handle restarts properly
7032 	 */
7033 	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7034 				&oldsp->ep->base.bind_addr, GFP_KERNEL);
7035 
7036 	/* Move any messages in the old socket's receive queue that are for the
7037 	 * peeled off association to the new socket's receive queue.
7038 	 */
7039 	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7040 		event = sctp_skb2event(skb);
7041 		if (event->asoc == assoc) {
7042 			__skb_unlink(skb, &oldsk->sk_receive_queue);
7043 			__skb_queue_tail(&newsk->sk_receive_queue, skb);
7044 			sctp_skb_set_owner_r_frag(skb, newsk);
7045 		}
7046 	}
7047 
7048 	/* Clean up any messages pending delivery due to partial
7049 	 * delivery.   Three cases:
7050 	 * 1) No partial deliver;  no work.
7051 	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7052 	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7053 	 */
7054 	skb_queue_head_init(&newsp->pd_lobby);
7055 	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7056 
7057 	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7058 		struct sk_buff_head *queue;
7059 
7060 		/* Decide which queue to move pd_lobby skbs to. */
7061 		if (assoc->ulpq.pd_mode) {
7062 			queue = &newsp->pd_lobby;
7063 		} else
7064 			queue = &newsk->sk_receive_queue;
7065 
7066 		/* Walk through the pd_lobby, looking for skbs that
7067 		 * need moved to the new socket.
7068 		 */
7069 		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7070 			event = sctp_skb2event(skb);
7071 			if (event->asoc == assoc) {
7072 				__skb_unlink(skb, &oldsp->pd_lobby);
7073 				__skb_queue_tail(queue, skb);
7074 				sctp_skb_set_owner_r_frag(skb, newsk);
7075 			}
7076 		}
7077 
7078 		/* Clear up any skbs waiting for the partial
7079 		 * delivery to finish.
7080 		 */
7081 		if (assoc->ulpq.pd_mode)
7082 			sctp_clear_pd(oldsk, NULL);
7083 
7084 	}
7085 
7086 	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7087 		sctp_skb_set_owner_r_frag(skb, newsk);
7088 
7089 	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7090 		sctp_skb_set_owner_r_frag(skb, newsk);
7091 
7092 	/* Set the type of socket to indicate that it is peeled off from the
7093 	 * original UDP-style socket or created with the accept() call on a
7094 	 * TCP-style socket..
7095 	 */
7096 	newsp->type = type;
7097 
7098 	/* Mark the new socket "in-use" by the user so that any packets
7099 	 * that may arrive on the association after we've moved it are
7100 	 * queued to the backlog.  This prevents a potential race between
7101 	 * backlog processing on the old socket and new-packet processing
7102 	 * on the new socket.
7103 	 *
7104 	 * The caller has just allocated newsk so we can guarantee that other
7105 	 * paths won't try to lock it and then oldsk.
7106 	 */
7107 	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7108 	sctp_assoc_migrate(assoc, newsk);
7109 
7110 	/* If the association on the newsk is already closed before accept()
7111 	 * is called, set RCV_SHUTDOWN flag.
7112 	 */
7113 	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7114 		newsk->sk_shutdown |= RCV_SHUTDOWN;
7115 
7116 	newsk->sk_state = SCTP_SS_ESTABLISHED;
7117 	release_sock(newsk);
7118 }
7119 
7120 
7121 /* This proto struct describes the ULP interface for SCTP.  */
7122 struct proto sctp_prot = {
7123 	.name        =	"SCTP",
7124 	.owner       =	THIS_MODULE,
7125 	.close       =	sctp_close,
7126 	.connect     =	sctp_connect,
7127 	.disconnect  =	sctp_disconnect,
7128 	.accept      =	sctp_accept,
7129 	.ioctl       =	sctp_ioctl,
7130 	.init        =	sctp_init_sock,
7131 	.destroy     =	sctp_destroy_sock,
7132 	.shutdown    =	sctp_shutdown,
7133 	.setsockopt  =	sctp_setsockopt,
7134 	.getsockopt  =	sctp_getsockopt,
7135 	.sendmsg     =	sctp_sendmsg,
7136 	.recvmsg     =	sctp_recvmsg,
7137 	.bind        =	sctp_bind,
7138 	.backlog_rcv =	sctp_backlog_rcv,
7139 	.hash        =	sctp_hash,
7140 	.unhash      =	sctp_unhash,
7141 	.get_port    =	sctp_get_port,
7142 	.obj_size    =  sizeof(struct sctp_sock),
7143 	.sysctl_mem  =  sysctl_sctp_mem,
7144 	.sysctl_rmem =  sysctl_sctp_rmem,
7145 	.sysctl_wmem =  sysctl_sctp_wmem,
7146 	.memory_pressure = &sctp_memory_pressure,
7147 	.enter_memory_pressure = sctp_enter_memory_pressure,
7148 	.memory_allocated = &sctp_memory_allocated,
7149 	.sockets_allocated = &sctp_sockets_allocated,
7150 };
7151 
7152 #if IS_ENABLED(CONFIG_IPV6)
7153 
7154 struct proto sctpv6_prot = {
7155 	.name		= "SCTPv6",
7156 	.owner		= THIS_MODULE,
7157 	.close		= sctp_close,
7158 	.connect	= sctp_connect,
7159 	.disconnect	= sctp_disconnect,
7160 	.accept		= sctp_accept,
7161 	.ioctl		= sctp_ioctl,
7162 	.init		= sctp_init_sock,
7163 	.destroy	= sctp_destroy_sock,
7164 	.shutdown	= sctp_shutdown,
7165 	.setsockopt	= sctp_setsockopt,
7166 	.getsockopt	= sctp_getsockopt,
7167 	.sendmsg	= sctp_sendmsg,
7168 	.recvmsg	= sctp_recvmsg,
7169 	.bind		= sctp_bind,
7170 	.backlog_rcv	= sctp_backlog_rcv,
7171 	.hash		= sctp_hash,
7172 	.unhash		= sctp_unhash,
7173 	.get_port	= sctp_get_port,
7174 	.obj_size	= sizeof(struct sctp6_sock),
7175 	.sysctl_mem	= sysctl_sctp_mem,
7176 	.sysctl_rmem	= sysctl_sctp_rmem,
7177 	.sysctl_wmem	= sysctl_sctp_wmem,
7178 	.memory_pressure = &sctp_memory_pressure,
7179 	.enter_memory_pressure = sctp_enter_memory_pressure,
7180 	.memory_allocated = &sctp_memory_allocated,
7181 	.sockets_allocated = &sctp_sockets_allocated,
7182 };
7183 #endif /* IS_ENABLED(CONFIG_IPV6) */
7184