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. */
83static int sctp_writeable(struct sock *sk);
84static void sctp_wfree(struct sk_buff *skb);
85static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86				size_t msg_len);
87static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89static int sctp_wait_for_accept(struct sock *sk, long timeo);
90static void sctp_wait_for_close(struct sock *sk, long timeo);
91static void sctp_destruct_sock(struct sock *sk);
92static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93					union sctp_addr *addr, int len);
94static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98static int sctp_send_asconf(struct sctp_association *asoc,
99			    struct sctp_chunk *chunk);
100static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101static int sctp_autobind(struct sock *sk);
102static void sctp_sock_migrate(struct sock *, struct sock *,
103			      struct sctp_association *, sctp_socket_type_t);
104
105static int sctp_memory_pressure;
106static atomic_long_t sctp_memory_allocated;
107struct percpu_counter sctp_sockets_allocated;
108
109static void sctp_enter_memory_pressure(struct sock *sk)
110{
111	sctp_memory_pressure = 1;
112}
113
114
115/* Get the sndbuf space available at the time on the association.  */
116static inline int sctp_wspace(struct sctp_association *asoc)
117{
118	int amt;
119
120	if (asoc->ep->sndbuf_policy)
121		amt = asoc->sndbuf_used;
122	else
123		amt = sk_wmem_alloc_get(asoc->base.sk);
124
125	if (amt >= asoc->base.sk->sk_sndbuf) {
126		if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
127			amt = 0;
128		else {
129			amt = sk_stream_wspace(asoc->base.sk);
130			if (amt < 0)
131				amt = 0;
132		}
133	} else {
134		amt = asoc->base.sk->sk_sndbuf - amt;
135	}
136	return amt;
137}
138
139/* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
142 *
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
146 * tracking.
147 */
148static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
149{
150	struct sctp_association *asoc = chunk->asoc;
151	struct sock *sk = asoc->base.sk;
152
153	/* The sndbuf space is tracked per association.  */
154	sctp_association_hold(asoc);
155
156	skb_set_owner_w(chunk->skb, sk);
157
158	chunk->skb->destructor = sctp_wfree;
159	/* Save the chunk pointer in skb for sctp_wfree to use later.  */
160	skb_shinfo(chunk->skb)->destructor_arg = chunk;
161
162	asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163				sizeof(struct sk_buff) +
164				sizeof(struct sctp_chunk);
165
166	atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167	sk->sk_wmem_queued += chunk->skb->truesize;
168	sk_mem_charge(sk, chunk->skb->truesize);
169}
170
171/* Verify that this is a valid address. */
172static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
173				   int len)
174{
175	struct sctp_af *af;
176
177	/* Verify basic sockaddr. */
178	af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
179	if (!af)
180		return -EINVAL;
181
182	/* Is this a valid SCTP address?  */
183	if (!af->addr_valid(addr, sctp_sk(sk), NULL))
184		return -EINVAL;
185
186	if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
187		return -EINVAL;
188
189	return 0;
190}
191
192/* Look up the association by its id.  If this is not a UDP-style
193 * socket, the ID field is always ignored.
194 */
195struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
196{
197	struct sctp_association *asoc = NULL;
198
199	/* If this is not a UDP-style socket, assoc id should be ignored. */
200	if (!sctp_style(sk, UDP)) {
201		/* Return NULL if the socket state is not ESTABLISHED. It
202		 * could be a TCP-style listening socket or a socket which
203		 * hasn't yet called connect() to establish an association.
204		 */
205		if (!sctp_sstate(sk, ESTABLISHED))
206			return NULL;
207
208		/* Get the first and the only association from the list. */
209		if (!list_empty(&sctp_sk(sk)->ep->asocs))
210			asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211					  struct sctp_association, asocs);
212		return asoc;
213	}
214
215	/* Otherwise this is a UDP-style socket. */
216	if (!id || (id == (sctp_assoc_t)-1))
217		return NULL;
218
219	spin_lock_bh(&sctp_assocs_id_lock);
220	asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221	spin_unlock_bh(&sctp_assocs_id_lock);
222
223	if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
224		return NULL;
225
226	return asoc;
227}
228
229/* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
231 * the same.
232 */
233static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234					      struct sockaddr_storage *addr,
235					      sctp_assoc_t id)
236{
237	struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238	struct sctp_transport *transport;
239	union sctp_addr *laddr = (union sctp_addr *)addr;
240
241	addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
242					       laddr,
243					       &transport);
244
245	if (!addr_asoc)
246		return NULL;
247
248	id_asoc = sctp_id2assoc(sk, id);
249	if (id_asoc && (id_asoc != addr_asoc))
250		return NULL;
251
252	sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
253						(union sctp_addr *)addr);
254
255	return transport;
256}
257
258/* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
260 *
261 *   ret = bind(int sd, struct sockaddr *addr, int addrlen);
262 *
263 *   sd      - the socket descriptor returned by socket().
264 *   addr    - the address structure (struct sockaddr_in or struct
265 *             sockaddr_in6 [RFC 2553]),
266 *   addr_len - the size of the address structure.
267 */
268static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
269{
270	int retval = 0;
271
272	lock_sock(sk);
273
274	pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
275		 addr, addr_len);
276
277	/* Disallow binding twice. */
278	if (!sctp_sk(sk)->ep->base.bind_addr.port)
279		retval = sctp_do_bind(sk, (union sctp_addr *)addr,
280				      addr_len);
281	else
282		retval = -EINVAL;
283
284	release_sock(sk);
285
286	return retval;
287}
288
289static long sctp_get_port_local(struct sock *, union sctp_addr *);
290
291/* Verify this is a valid sockaddr. */
292static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293					union sctp_addr *addr, int len)
294{
295	struct sctp_af *af;
296
297	/* Check minimum size.  */
298	if (len < sizeof (struct sockaddr))
299		return NULL;
300
301	/* V4 mapped address are really of AF_INET family */
302	if (addr->sa.sa_family == AF_INET6 &&
303	    ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304		if (!opt->pf->af_supported(AF_INET, opt))
305			return NULL;
306	} else {
307		/* Does this PF support this AF? */
308		if (!opt->pf->af_supported(addr->sa.sa_family, opt))
309			return NULL;
310	}
311
312	/* If we get this far, af is valid. */
313	af = sctp_get_af_specific(addr->sa.sa_family);
314
315	if (len < af->sockaddr_len)
316		return NULL;
317
318	return af;
319}
320
321/* Bind a local address either to an endpoint or to an association.  */
322static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
323{
324	struct net *net = sock_net(sk);
325	struct sctp_sock *sp = sctp_sk(sk);
326	struct sctp_endpoint *ep = sp->ep;
327	struct sctp_bind_addr *bp = &ep->base.bind_addr;
328	struct sctp_af *af;
329	unsigned short snum;
330	int ret = 0;
331
332	/* Common sockaddr verification. */
333	af = sctp_sockaddr_af(sp, addr, len);
334	if (!af) {
335		pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336			 __func__, sk, addr, len);
337		return -EINVAL;
338	}
339
340	snum = ntohs(addr->v4.sin_port);
341
342	pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343		 __func__, sk, &addr->sa, bp->port, snum, len);
344
345	/* PF specific bind() address verification. */
346	if (!sp->pf->bind_verify(sp, addr))
347		return -EADDRNOTAVAIL;
348
349	/* We must either be unbound, or bind to the same port.
350	 * It's OK to allow 0 ports if we are already bound.
351	 * We'll just inhert an already bound port in this case
352	 */
353	if (bp->port) {
354		if (!snum)
355			snum = bp->port;
356		else if (snum != bp->port) {
357			pr_debug("%s: new port %d doesn't match existing port "
358				 "%d\n", __func__, snum, bp->port);
359			return -EINVAL;
360		}
361	}
362
363	if (snum && snum < PROT_SOCK &&
364	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
365		return -EACCES;
366
367	/* See if the address matches any of the addresses we may have
368	 * already bound before checking against other endpoints.
369	 */
370	if (sctp_bind_addr_match(bp, addr, sp))
371		return -EINVAL;
372
373	/* Make sure we are allowed to bind here.
374	 * The function sctp_get_port_local() does duplicate address
375	 * detection.
376	 */
377	addr->v4.sin_port = htons(snum);
378	if ((ret = sctp_get_port_local(sk, addr))) {
379		return -EADDRINUSE;
380	}
381
382	/* Refresh ephemeral port.  */
383	if (!bp->port)
384		bp->port = inet_sk(sk)->inet_num;
385
386	/* Add the address to the bind address list.
387	 * Use GFP_ATOMIC since BHs will be disabled.
388	 */
389	ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
390
391	/* Copy back into socket for getsockname() use. */
392	if (!ret) {
393		inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
394		sp->pf->to_sk_saddr(addr, sk);
395	}
396
397	return ret;
398}
399
400 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
401 *
402 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
403 * at any one time.  If a sender, after sending an ASCONF chunk, decides
404 * it needs to transfer another ASCONF Chunk, it MUST wait until the
405 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
406 * subsequent ASCONF. Note this restriction binds each side, so at any
407 * time two ASCONF may be in-transit on any given association (one sent
408 * from each endpoint).
409 */
410static int sctp_send_asconf(struct sctp_association *asoc,
411			    struct sctp_chunk *chunk)
412{
413	struct net 	*net = sock_net(asoc->base.sk);
414	int		retval = 0;
415
416	/* If there is an outstanding ASCONF chunk, queue it for later
417	 * transmission.
418	 */
419	if (asoc->addip_last_asconf) {
420		list_add_tail(&chunk->list, &asoc->addip_chunk_list);
421		goto out;
422	}
423
424	/* Hold the chunk until an ASCONF_ACK is received. */
425	sctp_chunk_hold(chunk);
426	retval = sctp_primitive_ASCONF(net, asoc, chunk);
427	if (retval)
428		sctp_chunk_free(chunk);
429	else
430		asoc->addip_last_asconf = chunk;
431
432out:
433	return retval;
434}
435
436/* Add a list of addresses as bind addresses to local endpoint or
437 * association.
438 *
439 * Basically run through each address specified in the addrs/addrcnt
440 * array/length pair, determine if it is IPv6 or IPv4 and call
441 * sctp_do_bind() on it.
442 *
443 * If any of them fails, then the operation will be reversed and the
444 * ones that were added will be removed.
445 *
446 * Only sctp_setsockopt_bindx() is supposed to call this function.
447 */
448static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
449{
450	int cnt;
451	int retval = 0;
452	void *addr_buf;
453	struct sockaddr *sa_addr;
454	struct sctp_af *af;
455
456	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
457		 addrs, addrcnt);
458
459	addr_buf = addrs;
460	for (cnt = 0; cnt < addrcnt; cnt++) {
461		/* The list may contain either IPv4 or IPv6 address;
462		 * determine the address length for walking thru the list.
463		 */
464		sa_addr = addr_buf;
465		af = sctp_get_af_specific(sa_addr->sa_family);
466		if (!af) {
467			retval = -EINVAL;
468			goto err_bindx_add;
469		}
470
471		retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
472				      af->sockaddr_len);
473
474		addr_buf += af->sockaddr_len;
475
476err_bindx_add:
477		if (retval < 0) {
478			/* Failed. Cleanup the ones that have been added */
479			if (cnt > 0)
480				sctp_bindx_rem(sk, addrs, cnt);
481			return retval;
482		}
483	}
484
485	return retval;
486}
487
488/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
489 * associations that are part of the endpoint indicating that a list of local
490 * addresses are added to the endpoint.
491 *
492 * If any of the addresses is already in the bind address list of the
493 * association, we do not send the chunk for that association.  But it will not
494 * affect other associations.
495 *
496 * Only sctp_setsockopt_bindx() is supposed to call this function.
497 */
498static int sctp_send_asconf_add_ip(struct sock		*sk,
499				   struct sockaddr	*addrs,
500				   int 			addrcnt)
501{
502	struct net *net = sock_net(sk);
503	struct sctp_sock		*sp;
504	struct sctp_endpoint		*ep;
505	struct sctp_association		*asoc;
506	struct sctp_bind_addr		*bp;
507	struct sctp_chunk		*chunk;
508	struct sctp_sockaddr_entry	*laddr;
509	union sctp_addr			*addr;
510	union sctp_addr			saveaddr;
511	void				*addr_buf;
512	struct sctp_af			*af;
513	struct list_head		*p;
514	int 				i;
515	int 				retval = 0;
516
517	if (!net->sctp.addip_enable)
518		return retval;
519
520	sp = sctp_sk(sk);
521	ep = sp->ep;
522
523	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
524		 __func__, sk, addrs, addrcnt);
525
526	list_for_each_entry(asoc, &ep->asocs, asocs) {
527		if (!asoc->peer.asconf_capable)
528			continue;
529
530		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
531			continue;
532
533		if (!sctp_state(asoc, ESTABLISHED))
534			continue;
535
536		/* Check if any address in the packed array of addresses is
537		 * in the bind address list of the association. If so,
538		 * do not send the asconf chunk to its peer, but continue with
539		 * other associations.
540		 */
541		addr_buf = addrs;
542		for (i = 0; i < addrcnt; i++) {
543			addr = addr_buf;
544			af = sctp_get_af_specific(addr->v4.sin_family);
545			if (!af) {
546				retval = -EINVAL;
547				goto out;
548			}
549
550			if (sctp_assoc_lookup_laddr(asoc, addr))
551				break;
552
553			addr_buf += af->sockaddr_len;
554		}
555		if (i < addrcnt)
556			continue;
557
558		/* Use the first valid address in bind addr list of
559		 * association as Address Parameter of ASCONF CHUNK.
560		 */
561		bp = &asoc->base.bind_addr;
562		p = bp->address_list.next;
563		laddr = list_entry(p, struct sctp_sockaddr_entry, list);
564		chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
565						   addrcnt, SCTP_PARAM_ADD_IP);
566		if (!chunk) {
567			retval = -ENOMEM;
568			goto out;
569		}
570
571		/* Add the new addresses to the bind address list with
572		 * use_as_src set to 0.
573		 */
574		addr_buf = addrs;
575		for (i = 0; i < addrcnt; i++) {
576			addr = addr_buf;
577			af = sctp_get_af_specific(addr->v4.sin_family);
578			memcpy(&saveaddr, addr, af->sockaddr_len);
579			retval = sctp_add_bind_addr(bp, &saveaddr,
580						    SCTP_ADDR_NEW, GFP_ATOMIC);
581			addr_buf += af->sockaddr_len;
582		}
583		if (asoc->src_out_of_asoc_ok) {
584			struct sctp_transport *trans;
585
586			list_for_each_entry(trans,
587			    &asoc->peer.transport_addr_list, transports) {
588				/* Clear the source and route cache */
589				dst_release(trans->dst);
590				trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
591				    2*asoc->pathmtu, 4380));
592				trans->ssthresh = asoc->peer.i.a_rwnd;
593				trans->rto = asoc->rto_initial;
594				sctp_max_rto(asoc, trans);
595				trans->rtt = trans->srtt = trans->rttvar = 0;
596				sctp_transport_route(trans, NULL,
597				    sctp_sk(asoc->base.sk));
598			}
599		}
600		retval = sctp_send_asconf(asoc, chunk);
601	}
602
603out:
604	return retval;
605}
606
607/* Remove a list of addresses from bind addresses list.  Do not remove the
608 * last address.
609 *
610 * Basically run through each address specified in the addrs/addrcnt
611 * array/length pair, determine if it is IPv6 or IPv4 and call
612 * sctp_del_bind() on it.
613 *
614 * If any of them fails, then the operation will be reversed and the
615 * ones that were removed will be added back.
616 *
617 * At least one address has to be left; if only one address is
618 * available, the operation will return -EBUSY.
619 *
620 * Only sctp_setsockopt_bindx() is supposed to call this function.
621 */
622static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
623{
624	struct sctp_sock *sp = sctp_sk(sk);
625	struct sctp_endpoint *ep = sp->ep;
626	int cnt;
627	struct sctp_bind_addr *bp = &ep->base.bind_addr;
628	int retval = 0;
629	void *addr_buf;
630	union sctp_addr *sa_addr;
631	struct sctp_af *af;
632
633	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
634		 __func__, sk, addrs, addrcnt);
635
636	addr_buf = addrs;
637	for (cnt = 0; cnt < addrcnt; cnt++) {
638		/* If the bind address list is empty or if there is only one
639		 * bind address, there is nothing more to be removed (we need
640		 * at least one address here).
641		 */
642		if (list_empty(&bp->address_list) ||
643		    (sctp_list_single_entry(&bp->address_list))) {
644			retval = -EBUSY;
645			goto err_bindx_rem;
646		}
647
648		sa_addr = addr_buf;
649		af = sctp_get_af_specific(sa_addr->sa.sa_family);
650		if (!af) {
651			retval = -EINVAL;
652			goto err_bindx_rem;
653		}
654
655		if (!af->addr_valid(sa_addr, sp, NULL)) {
656			retval = -EADDRNOTAVAIL;
657			goto err_bindx_rem;
658		}
659
660		if (sa_addr->v4.sin_port &&
661		    sa_addr->v4.sin_port != htons(bp->port)) {
662			retval = -EINVAL;
663			goto err_bindx_rem;
664		}
665
666		if (!sa_addr->v4.sin_port)
667			sa_addr->v4.sin_port = htons(bp->port);
668
669		/* FIXME - There is probably a need to check if sk->sk_saddr and
670		 * sk->sk_rcv_addr are currently set to one of the addresses to
671		 * be removed. This is something which needs to be looked into
672		 * when we are fixing the outstanding issues with multi-homing
673		 * socket routing and failover schemes. Refer to comments in
674		 * sctp_do_bind(). -daisy
675		 */
676		retval = sctp_del_bind_addr(bp, sa_addr);
677
678		addr_buf += af->sockaddr_len;
679err_bindx_rem:
680		if (retval < 0) {
681			/* Failed. Add the ones that has been removed back */
682			if (cnt > 0)
683				sctp_bindx_add(sk, addrs, cnt);
684			return retval;
685		}
686	}
687
688	return retval;
689}
690
691/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
692 * the associations that are part of the endpoint indicating that a list of
693 * local addresses are removed from the endpoint.
694 *
695 * If any of the addresses is already in the bind address list of the
696 * association, we do not send the chunk for that association.  But it will not
697 * affect other associations.
698 *
699 * Only sctp_setsockopt_bindx() is supposed to call this function.
700 */
701static int sctp_send_asconf_del_ip(struct sock		*sk,
702				   struct sockaddr	*addrs,
703				   int			addrcnt)
704{
705	struct net *net = sock_net(sk);
706	struct sctp_sock	*sp;
707	struct sctp_endpoint	*ep;
708	struct sctp_association	*asoc;
709	struct sctp_transport	*transport;
710	struct sctp_bind_addr	*bp;
711	struct sctp_chunk	*chunk;
712	union sctp_addr		*laddr;
713	void			*addr_buf;
714	struct sctp_af		*af;
715	struct sctp_sockaddr_entry *saddr;
716	int 			i;
717	int 			retval = 0;
718	int			stored = 0;
719
720	chunk = NULL;
721	if (!net->sctp.addip_enable)
722		return retval;
723
724	sp = sctp_sk(sk);
725	ep = sp->ep;
726
727	pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
728		 __func__, sk, addrs, addrcnt);
729
730	list_for_each_entry(asoc, &ep->asocs, asocs) {
731
732		if (!asoc->peer.asconf_capable)
733			continue;
734
735		if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
736			continue;
737
738		if (!sctp_state(asoc, ESTABLISHED))
739			continue;
740
741		/* Check if any address in the packed array of addresses is
742		 * not present in the bind address list of the association.
743		 * If so, do not send the asconf chunk to its peer, but
744		 * continue with other associations.
745		 */
746		addr_buf = addrs;
747		for (i = 0; i < addrcnt; i++) {
748			laddr = addr_buf;
749			af = sctp_get_af_specific(laddr->v4.sin_family);
750			if (!af) {
751				retval = -EINVAL;
752				goto out;
753			}
754
755			if (!sctp_assoc_lookup_laddr(asoc, laddr))
756				break;
757
758			addr_buf += af->sockaddr_len;
759		}
760		if (i < addrcnt)
761			continue;
762
763		/* Find one address in the association's bind address list
764		 * that is not in the packed array of addresses. This is to
765		 * make sure that we do not delete all the addresses in the
766		 * association.
767		 */
768		bp = &asoc->base.bind_addr;
769		laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
770					       addrcnt, sp);
771		if ((laddr == NULL) && (addrcnt == 1)) {
772			if (asoc->asconf_addr_del_pending)
773				continue;
774			asoc->asconf_addr_del_pending =
775			    kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
776			if (asoc->asconf_addr_del_pending == NULL) {
777				retval = -ENOMEM;
778				goto out;
779			}
780			asoc->asconf_addr_del_pending->sa.sa_family =
781				    addrs->sa_family;
782			asoc->asconf_addr_del_pending->v4.sin_port =
783				    htons(bp->port);
784			if (addrs->sa_family == AF_INET) {
785				struct sockaddr_in *sin;
786
787				sin = (struct sockaddr_in *)addrs;
788				asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
789			} else if (addrs->sa_family == AF_INET6) {
790				struct sockaddr_in6 *sin6;
791
792				sin6 = (struct sockaddr_in6 *)addrs;
793				asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
794			}
795
796			pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
797				 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
798				 asoc->asconf_addr_del_pending);
799
800			asoc->src_out_of_asoc_ok = 1;
801			stored = 1;
802			goto skip_mkasconf;
803		}
804
805		if (laddr == NULL)
806			return -EINVAL;
807
808		/* We do not need RCU protection throughout this loop
809		 * because this is done under a socket lock from the
810		 * setsockopt call.
811		 */
812		chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
813						   SCTP_PARAM_DEL_IP);
814		if (!chunk) {
815			retval = -ENOMEM;
816			goto out;
817		}
818
819skip_mkasconf:
820		/* Reset use_as_src flag for the addresses in the bind address
821		 * list that are to be deleted.
822		 */
823		addr_buf = addrs;
824		for (i = 0; i < addrcnt; i++) {
825			laddr = addr_buf;
826			af = sctp_get_af_specific(laddr->v4.sin_family);
827			list_for_each_entry(saddr, &bp->address_list, list) {
828				if (sctp_cmp_addr_exact(&saddr->a, laddr))
829					saddr->state = SCTP_ADDR_DEL;
830			}
831			addr_buf += af->sockaddr_len;
832		}
833
834		/* Update the route and saddr entries for all the transports
835		 * as some of the addresses in the bind address list are
836		 * about to be deleted and cannot be used as source addresses.
837		 */
838		list_for_each_entry(transport, &asoc->peer.transport_addr_list,
839					transports) {
840			dst_release(transport->dst);
841			sctp_transport_route(transport, NULL,
842					     sctp_sk(asoc->base.sk));
843		}
844
845		if (stored)
846			/* We don't need to transmit ASCONF */
847			continue;
848		retval = sctp_send_asconf(asoc, chunk);
849	}
850out:
851	return retval;
852}
853
854/* set addr events to assocs in the endpoint.  ep and addr_wq must be locked */
855int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
856{
857	struct sock *sk = sctp_opt2sk(sp);
858	union sctp_addr *addr;
859	struct sctp_af *af;
860
861	/* It is safe to write port space in caller. */
862	addr = &addrw->a;
863	addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
864	af = sctp_get_af_specific(addr->sa.sa_family);
865	if (!af)
866		return -EINVAL;
867	if (sctp_verify_addr(sk, addr, af->sockaddr_len))
868		return -EINVAL;
869
870	if (addrw->state == SCTP_ADDR_NEW)
871		return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
872	else
873		return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
874}
875
876/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
877 *
878 * API 8.1
879 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
880 *                int flags);
881 *
882 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
883 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
884 * or IPv6 addresses.
885 *
886 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
887 * Section 3.1.2 for this usage.
888 *
889 * addrs is a pointer to an array of one or more socket addresses. Each
890 * address is contained in its appropriate structure (i.e. struct
891 * sockaddr_in or struct sockaddr_in6) the family of the address type
892 * must be used to distinguish the address length (note that this
893 * representation is termed a "packed array" of addresses). The caller
894 * specifies the number of addresses in the array with addrcnt.
895 *
896 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
897 * -1, and sets errno to the appropriate error code.
898 *
899 * For SCTP, the port given in each socket address must be the same, or
900 * sctp_bindx() will fail, setting errno to EINVAL.
901 *
902 * The flags parameter is formed from the bitwise OR of zero or more of
903 * the following currently defined flags:
904 *
905 * SCTP_BINDX_ADD_ADDR
906 *
907 * SCTP_BINDX_REM_ADDR
908 *
909 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
910 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
911 * addresses from the association. The two flags are mutually exclusive;
912 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
913 * not remove all addresses from an association; sctp_bindx() will
914 * reject such an attempt with EINVAL.
915 *
916 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
917 * additional addresses with an endpoint after calling bind().  Or use
918 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
919 * socket is associated with so that no new association accepted will be
920 * associated with those addresses. If the endpoint supports dynamic
921 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
922 * endpoint to send the appropriate message to the peer to change the
923 * peers address lists.
924 *
925 * Adding and removing addresses from a connected association is
926 * optional functionality. Implementations that do not support this
927 * functionality should return EOPNOTSUPP.
928 *
929 * Basically do nothing but copying the addresses from user to kernel
930 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
931 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
932 * from userspace.
933 *
934 * We don't use copy_from_user() for optimization: we first do the
935 * sanity checks (buffer size -fast- and access check-healthy
936 * pointer); if all of those succeed, then we can alloc the memory
937 * (expensive operation) needed to copy the data to kernel. Then we do
938 * the copying without checking the user space area
939 * (__copy_from_user()).
940 *
941 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
942 * it.
943 *
944 * sk        The sk of the socket
945 * addrs     The pointer to the addresses in user land
946 * addrssize Size of the addrs buffer
947 * op        Operation to perform (add or remove, see the flags of
948 *           sctp_bindx)
949 *
950 * Returns 0 if ok, <0 errno code on error.
951 */
952static int sctp_setsockopt_bindx(struct sock *sk,
953				 struct sockaddr __user *addrs,
954				 int addrs_size, int op)
955{
956	struct sockaddr *kaddrs;
957	int err;
958	int addrcnt = 0;
959	int walk_size = 0;
960	struct sockaddr *sa_addr;
961	void *addr_buf;
962	struct sctp_af *af;
963
964	pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
965		 __func__, sk, addrs, addrs_size, op);
966
967	if (unlikely(addrs_size <= 0))
968		return -EINVAL;
969
970	/* Check the user passed a healthy pointer.  */
971	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
972		return -EFAULT;
973
974	/* Alloc space for the address array in kernel memory.  */
975	kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
976	if (unlikely(!kaddrs))
977		return -ENOMEM;
978
979	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
980		kfree(kaddrs);
981		return -EFAULT;
982	}
983
984	/* Walk through the addrs buffer and count the number of addresses. */
985	addr_buf = kaddrs;
986	while (walk_size < addrs_size) {
987		if (walk_size + sizeof(sa_family_t) > addrs_size) {
988			kfree(kaddrs);
989			return -EINVAL;
990		}
991
992		sa_addr = addr_buf;
993		af = sctp_get_af_specific(sa_addr->sa_family);
994
995		/* If the address family is not supported or if this address
996		 * causes the address buffer to overflow return EINVAL.
997		 */
998		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
999			kfree(kaddrs);
1000			return -EINVAL;
1001		}
1002		addrcnt++;
1003		addr_buf += af->sockaddr_len;
1004		walk_size += af->sockaddr_len;
1005	}
1006
1007	/* Do the work. */
1008	switch (op) {
1009	case SCTP_BINDX_ADD_ADDR:
1010		err = sctp_bindx_add(sk, kaddrs, addrcnt);
1011		if (err)
1012			goto out;
1013		err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1014		break;
1015
1016	case SCTP_BINDX_REM_ADDR:
1017		err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1018		if (err)
1019			goto out;
1020		err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1021		break;
1022
1023	default:
1024		err = -EINVAL;
1025		break;
1026	}
1027
1028out:
1029	kfree(kaddrs);
1030
1031	return err;
1032}
1033
1034/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1035 *
1036 * Common routine for handling connect() and sctp_connectx().
1037 * Connect will come in with just a single address.
1038 */
1039static int __sctp_connect(struct sock *sk,
1040			  struct sockaddr *kaddrs,
1041			  int addrs_size,
1042			  sctp_assoc_t *assoc_id)
1043{
1044	struct net *net = sock_net(sk);
1045	struct sctp_sock *sp;
1046	struct sctp_endpoint *ep;
1047	struct sctp_association *asoc = NULL;
1048	struct sctp_association *asoc2;
1049	struct sctp_transport *transport;
1050	union sctp_addr to;
1051	sctp_scope_t scope;
1052	long timeo;
1053	int err = 0;
1054	int addrcnt = 0;
1055	int walk_size = 0;
1056	union sctp_addr *sa_addr = NULL;
1057	void *addr_buf;
1058	unsigned short port;
1059	unsigned int f_flags = 0;
1060
1061	sp = sctp_sk(sk);
1062	ep = sp->ep;
1063
1064	/* connect() cannot be done on a socket that is already in ESTABLISHED
1065	 * state - UDP-style peeled off socket or a TCP-style socket that
1066	 * is already connected.
1067	 * It cannot be done even on a TCP-style listening socket.
1068	 */
1069	if (sctp_sstate(sk, ESTABLISHED) ||
1070	    (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1071		err = -EISCONN;
1072		goto out_free;
1073	}
1074
1075	/* Walk through the addrs buffer and count the number of addresses. */
1076	addr_buf = kaddrs;
1077	while (walk_size < addrs_size) {
1078		struct sctp_af *af;
1079
1080		if (walk_size + sizeof(sa_family_t) > addrs_size) {
1081			err = -EINVAL;
1082			goto out_free;
1083		}
1084
1085		sa_addr = addr_buf;
1086		af = sctp_get_af_specific(sa_addr->sa.sa_family);
1087
1088		/* If the address family is not supported or if this address
1089		 * causes the address buffer to overflow return EINVAL.
1090		 */
1091		if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1092			err = -EINVAL;
1093			goto out_free;
1094		}
1095
1096		port = ntohs(sa_addr->v4.sin_port);
1097
1098		/* Save current address so we can work with it */
1099		memcpy(&to, sa_addr, af->sockaddr_len);
1100
1101		err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1102		if (err)
1103			goto out_free;
1104
1105		/* Make sure the destination port is correctly set
1106		 * in all addresses.
1107		 */
1108		if (asoc && asoc->peer.port && asoc->peer.port != port) {
1109			err = -EINVAL;
1110			goto out_free;
1111		}
1112
1113		/* Check if there already is a matching association on the
1114		 * endpoint (other than the one created here).
1115		 */
1116		asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1117		if (asoc2 && asoc2 != asoc) {
1118			if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1119				err = -EISCONN;
1120			else
1121				err = -EALREADY;
1122			goto out_free;
1123		}
1124
1125		/* If we could not find a matching association on the endpoint,
1126		 * make sure that there is no peeled-off association matching
1127		 * the peer address even on another socket.
1128		 */
1129		if (sctp_endpoint_is_peeled_off(ep, &to)) {
1130			err = -EADDRNOTAVAIL;
1131			goto out_free;
1132		}
1133
1134		if (!asoc) {
1135			/* If a bind() or sctp_bindx() is not called prior to
1136			 * an sctp_connectx() call, the system picks an
1137			 * ephemeral port and will choose an address set
1138			 * equivalent to binding with a wildcard address.
1139			 */
1140			if (!ep->base.bind_addr.port) {
1141				if (sctp_autobind(sk)) {
1142					err = -EAGAIN;
1143					goto out_free;
1144				}
1145			} else {
1146				/*
1147				 * If an unprivileged user inherits a 1-many
1148				 * style socket with open associations on a
1149				 * privileged port, it MAY be permitted to
1150				 * accept new associations, but it SHOULD NOT
1151				 * be permitted to open new associations.
1152				 */
1153				if (ep->base.bind_addr.port < PROT_SOCK &&
1154				    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1155					err = -EACCES;
1156					goto out_free;
1157				}
1158			}
1159
1160			scope = sctp_scope(&to);
1161			asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1162			if (!asoc) {
1163				err = -ENOMEM;
1164				goto out_free;
1165			}
1166
1167			err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1168							      GFP_KERNEL);
1169			if (err < 0) {
1170				goto out_free;
1171			}
1172
1173		}
1174
1175		/* Prime the peer's transport structures.  */
1176		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1177						SCTP_UNKNOWN);
1178		if (!transport) {
1179			err = -ENOMEM;
1180			goto out_free;
1181		}
1182
1183		addrcnt++;
1184		addr_buf += af->sockaddr_len;
1185		walk_size += af->sockaddr_len;
1186	}
1187
1188	/* In case the user of sctp_connectx() wants an association
1189	 * id back, assign one now.
1190	 */
1191	if (assoc_id) {
1192		err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1193		if (err < 0)
1194			goto out_free;
1195	}
1196
1197	err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1198	if (err < 0) {
1199		goto out_free;
1200	}
1201
1202	/* Initialize sk's dport and daddr for getpeername() */
1203	inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1204	sp->pf->to_sk_daddr(sa_addr, sk);
1205	sk->sk_err = 0;
1206
1207	/* in-kernel sockets don't generally have a file allocated to them
1208	 * if all they do is call sock_create_kern().
1209	 */
1210	if (sk->sk_socket->file)
1211		f_flags = sk->sk_socket->file->f_flags;
1212
1213	timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1214
1215	err = sctp_wait_for_connect(asoc, &timeo);
1216	if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1217		*assoc_id = asoc->assoc_id;
1218
1219	/* Don't free association on exit. */
1220	asoc = NULL;
1221
1222out_free:
1223	pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1224		 __func__, asoc, kaddrs, err);
1225
1226	if (asoc) {
1227		/* sctp_primitive_ASSOCIATE may have added this association
1228		 * To the hash table, try to unhash it, just in case, its a noop
1229		 * if it wasn't hashed so we're safe
1230		 */
1231		sctp_unhash_established(asoc);
1232		sctp_association_free(asoc);
1233	}
1234	return err;
1235}
1236
1237/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1238 *
1239 * API 8.9
1240 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1241 * 			sctp_assoc_t *asoc);
1242 *
1243 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1244 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1245 * or IPv6 addresses.
1246 *
1247 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1248 * Section 3.1.2 for this usage.
1249 *
1250 * addrs is a pointer to an array of one or more socket addresses. Each
1251 * address is contained in its appropriate structure (i.e. struct
1252 * sockaddr_in or struct sockaddr_in6) the family of the address type
1253 * must be used to distengish the address length (note that this
1254 * representation is termed a "packed array" of addresses). The caller
1255 * specifies the number of addresses in the array with addrcnt.
1256 *
1257 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1258 * the association id of the new association.  On failure, sctp_connectx()
1259 * returns -1, and sets errno to the appropriate error code.  The assoc_id
1260 * is not touched by the kernel.
1261 *
1262 * For SCTP, the port given in each socket address must be the same, or
1263 * sctp_connectx() will fail, setting errno to EINVAL.
1264 *
1265 * An application can use sctp_connectx to initiate an association with
1266 * an endpoint that is multi-homed.  Much like sctp_bindx() this call
1267 * allows a caller to specify multiple addresses at which a peer can be
1268 * reached.  The way the SCTP stack uses the list of addresses to set up
1269 * the association is implementation dependent.  This function only
1270 * specifies that the stack will try to make use of all the addresses in
1271 * the list when needed.
1272 *
1273 * Note that the list of addresses passed in is only used for setting up
1274 * the association.  It does not necessarily equal the set of addresses
1275 * the peer uses for the resulting association.  If the caller wants to
1276 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1277 * retrieve them after the association has been set up.
1278 *
1279 * Basically do nothing but copying the addresses from user to kernel
1280 * land and invoking either sctp_connectx(). This is used for tunneling
1281 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1282 *
1283 * We don't use copy_from_user() for optimization: we first do the
1284 * sanity checks (buffer size -fast- and access check-healthy
1285 * pointer); if all of those succeed, then we can alloc the memory
1286 * (expensive operation) needed to copy the data to kernel. Then we do
1287 * the copying without checking the user space area
1288 * (__copy_from_user()).
1289 *
1290 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1291 * it.
1292 *
1293 * sk        The sk of the socket
1294 * addrs     The pointer to the addresses in user land
1295 * addrssize Size of the addrs buffer
1296 *
1297 * Returns >=0 if ok, <0 errno code on error.
1298 */
1299static int __sctp_setsockopt_connectx(struct sock *sk,
1300				      struct sockaddr __user *addrs,
1301				      int addrs_size,
1302				      sctp_assoc_t *assoc_id)
1303{
1304	struct sockaddr *kaddrs;
1305	gfp_t gfp = GFP_KERNEL;
1306	int err = 0;
1307
1308	pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1309		 __func__, sk, addrs, addrs_size);
1310
1311	if (unlikely(addrs_size <= 0))
1312		return -EINVAL;
1313
1314	/* Check the user passed a healthy pointer.  */
1315	if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1316		return -EFAULT;
1317
1318	/* Alloc space for the address array in kernel memory.  */
1319	if (sk->sk_socket->file)
1320		gfp = GFP_USER | __GFP_NOWARN;
1321	kaddrs = kmalloc(addrs_size, gfp);
1322	if (unlikely(!kaddrs))
1323		return -ENOMEM;
1324
1325	if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1326		err = -EFAULT;
1327	} else {
1328		err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1329	}
1330
1331	kfree(kaddrs);
1332
1333	return err;
1334}
1335
1336/*
1337 * This is an older interface.  It's kept for backward compatibility
1338 * to the option that doesn't provide association id.
1339 */
1340static int sctp_setsockopt_connectx_old(struct sock *sk,
1341					struct sockaddr __user *addrs,
1342					int addrs_size)
1343{
1344	return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1345}
1346
1347/*
1348 * New interface for the API.  The since the API is done with a socket
1349 * option, to make it simple we feed back the association id is as a return
1350 * indication to the call.  Error is always negative and association id is
1351 * always positive.
1352 */
1353static int sctp_setsockopt_connectx(struct sock *sk,
1354				    struct sockaddr __user *addrs,
1355				    int addrs_size)
1356{
1357	sctp_assoc_t assoc_id = 0;
1358	int err = 0;
1359
1360	err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1361
1362	if (err)
1363		return err;
1364	else
1365		return assoc_id;
1366}
1367
1368/*
1369 * New (hopefully final) interface for the API.
1370 * We use the sctp_getaddrs_old structure so that use-space library
1371 * can avoid any unnecessary allocations. The only different part
1372 * is that we store the actual length of the address buffer into the
1373 * addrs_num structure member. That way we can re-use the existing
1374 * code.
1375 */
1376#ifdef CONFIG_COMPAT
1377struct compat_sctp_getaddrs_old {
1378	sctp_assoc_t	assoc_id;
1379	s32		addr_num;
1380	compat_uptr_t	addrs;		/* struct sockaddr * */
1381};
1382#endif
1383
1384static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1385				     char __user *optval,
1386				     int __user *optlen)
1387{
1388	struct sctp_getaddrs_old param;
1389	sctp_assoc_t assoc_id = 0;
1390	int err = 0;
1391
1392#ifdef CONFIG_COMPAT
1393	if (is_compat_task()) {
1394		struct compat_sctp_getaddrs_old param32;
1395
1396		if (len < sizeof(param32))
1397			return -EINVAL;
1398		if (copy_from_user(&param32, optval, sizeof(param32)))
1399			return -EFAULT;
1400
1401		param.assoc_id = param32.assoc_id;
1402		param.addr_num = param32.addr_num;
1403		param.addrs = compat_ptr(param32.addrs);
1404	} else
1405#endif
1406	{
1407		if (len < sizeof(param))
1408			return -EINVAL;
1409		if (copy_from_user(&param, optval, sizeof(param)))
1410			return -EFAULT;
1411	}
1412
1413	err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1414					 param.addrs, param.addr_num,
1415					 &assoc_id);
1416	if (err == 0 || err == -EINPROGRESS) {
1417		if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1418			return -EFAULT;
1419		if (put_user(sizeof(assoc_id), optlen))
1420			return -EFAULT;
1421	}
1422
1423	return err;
1424}
1425
1426/* API 3.1.4 close() - UDP Style Syntax
1427 * Applications use close() to perform graceful shutdown (as described in
1428 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1429 * by a UDP-style socket.
1430 *
1431 * The syntax is
1432 *
1433 *   ret = close(int sd);
1434 *
1435 *   sd      - the socket descriptor of the associations to be closed.
1436 *
1437 * To gracefully shutdown a specific association represented by the
1438 * UDP-style socket, an application should use the sendmsg() call,
1439 * passing no user data, but including the appropriate flag in the
1440 * ancillary data (see Section xxxx).
1441 *
1442 * If sd in the close() call is a branched-off socket representing only
1443 * one association, the shutdown is performed on that association only.
1444 *
1445 * 4.1.6 close() - TCP Style Syntax
1446 *
1447 * Applications use close() to gracefully close down an association.
1448 *
1449 * The syntax is:
1450 *
1451 *    int close(int sd);
1452 *
1453 *      sd      - the socket descriptor of the association to be closed.
1454 *
1455 * After an application calls close() on a socket descriptor, no further
1456 * socket operations will succeed on that descriptor.
1457 *
1458 * API 7.1.4 SO_LINGER
1459 *
1460 * An application using the TCP-style socket can use this option to
1461 * perform the SCTP ABORT primitive.  The linger option structure is:
1462 *
1463 *  struct  linger {
1464 *     int     l_onoff;                // option on/off
1465 *     int     l_linger;               // linger time
1466 * };
1467 *
1468 * To enable the option, set l_onoff to 1.  If the l_linger value is set
1469 * to 0, calling close() is the same as the ABORT primitive.  If the
1470 * value is set to a negative value, the setsockopt() call will return
1471 * an error.  If the value is set to a positive value linger_time, the
1472 * close() can be blocked for at most linger_time ms.  If the graceful
1473 * shutdown phase does not finish during this period, close() will
1474 * return but the graceful shutdown phase continues in the system.
1475 */
1476static void sctp_close(struct sock *sk, long timeout)
1477{
1478	struct net *net = sock_net(sk);
1479	struct sctp_endpoint *ep;
1480	struct sctp_association *asoc;
1481	struct list_head *pos, *temp;
1482	unsigned int data_was_unread;
1483
1484	pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1485
1486	lock_sock(sk);
1487	sk->sk_shutdown = SHUTDOWN_MASK;
1488	sk->sk_state = SCTP_SS_CLOSING;
1489
1490	ep = sctp_sk(sk)->ep;
1491
1492	/* Clean up any skbs sitting on the receive queue.  */
1493	data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1494	data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1495
1496	/* Walk all associations on an endpoint.  */
1497	list_for_each_safe(pos, temp, &ep->asocs) {
1498		asoc = list_entry(pos, struct sctp_association, asocs);
1499
1500		if (sctp_style(sk, TCP)) {
1501			/* A closed association can still be in the list if
1502			 * it belongs to a TCP-style listening socket that is
1503			 * not yet accepted. If so, free it. If not, send an
1504			 * ABORT or SHUTDOWN based on the linger options.
1505			 */
1506			if (sctp_state(asoc, CLOSED)) {
1507				sctp_unhash_established(asoc);
1508				sctp_association_free(asoc);
1509				continue;
1510			}
1511		}
1512
1513		if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1514		    !skb_queue_empty(&asoc->ulpq.reasm) ||
1515		    (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1516			struct sctp_chunk *chunk;
1517
1518			chunk = sctp_make_abort_user(asoc, NULL, 0);
1519			sctp_primitive_ABORT(net, asoc, chunk);
1520		} else
1521			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1522	}
1523
1524	/* On a TCP-style socket, block for at most linger_time if set. */
1525	if (sctp_style(sk, TCP) && timeout)
1526		sctp_wait_for_close(sk, timeout);
1527
1528	/* This will run the backlog queue.  */
1529	release_sock(sk);
1530
1531	/* Supposedly, no process has access to the socket, but
1532	 * the net layers still may.
1533	 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534	 * held and that should be grabbed before socket lock.
1535	 */
1536	spin_lock_bh(&net->sctp.addr_wq_lock);
1537	bh_lock_sock(sk);
1538
1539	/* Hold the sock, since sk_common_release() will put sock_put()
1540	 * and we have just a little more cleanup.
1541	 */
1542	sock_hold(sk);
1543	sk_common_release(sk);
1544
1545	bh_unlock_sock(sk);
1546	spin_unlock_bh(&net->sctp.addr_wq_lock);
1547
1548	sock_put(sk);
1549
1550	SCTP_DBG_OBJCNT_DEC(sock);
1551}
1552
1553/* Handle EPIPE error. */
1554static int sctp_error(struct sock *sk, int flags, int err)
1555{
1556	if (err == -EPIPE)
1557		err = sock_error(sk) ? : -EPIPE;
1558	if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559		send_sig(SIGPIPE, current, 0);
1560	return err;
1561}
1562
1563/* API 3.1.3 sendmsg() - UDP Style Syntax
1564 *
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1567 *
1568 *  ssize_t sendmsg(int socket, const struct msghdr *message,
1569 *                  int flags);
1570 *
1571 *  socket  - the socket descriptor of the endpoint.
1572 *  message - pointer to the msghdr structure which contains a single
1573 *            user message and possibly some ancillary data.
1574 *
1575 *            See Section 5 for complete description of the data
1576 *            structures.
1577 *
1578 *  flags   - flags sent or received with the user message, see Section
1579 *            5 for complete description of the flags.
1580 *
1581 * Note:  This function could use a rewrite especially when explicit
1582 * connect support comes in.
1583 */
1584/* BUG:  We do not implement the equivalent of sk_stream_wait_memory(). */
1585
1586static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1587
1588static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1589{
1590	struct net *net = sock_net(sk);
1591	struct sctp_sock *sp;
1592	struct sctp_endpoint *ep;
1593	struct sctp_association *new_asoc = NULL, *asoc = NULL;
1594	struct sctp_transport *transport, *chunk_tp;
1595	struct sctp_chunk *chunk;
1596	union sctp_addr to;
1597	struct sockaddr *msg_name = NULL;
1598	struct sctp_sndrcvinfo default_sinfo;
1599	struct sctp_sndrcvinfo *sinfo;
1600	struct sctp_initmsg *sinit;
1601	sctp_assoc_t associd = 0;
1602	sctp_cmsgs_t cmsgs = { NULL };
1603	sctp_scope_t scope;
1604	bool fill_sinfo_ttl = false, wait_connect = false;
1605	struct sctp_datamsg *datamsg;
1606	int msg_flags = msg->msg_flags;
1607	__u16 sinfo_flags = 0;
1608	long timeo;
1609	int err;
1610
1611	err = 0;
1612	sp = sctp_sk(sk);
1613	ep = sp->ep;
1614
1615	pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1616		 msg, msg_len, ep);
1617
1618	/* We cannot send a message over a TCP-style listening socket. */
1619	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1620		err = -EPIPE;
1621		goto out_nounlock;
1622	}
1623
1624	/* Parse out the SCTP CMSGs.  */
1625	err = sctp_msghdr_parse(msg, &cmsgs);
1626	if (err) {
1627		pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1628		goto out_nounlock;
1629	}
1630
1631	/* Fetch the destination address for this packet.  This
1632	 * address only selects the association--it is not necessarily
1633	 * the address we will send to.
1634	 * For a peeled-off socket, msg_name is ignored.
1635	 */
1636	if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637		int msg_namelen = msg->msg_namelen;
1638
1639		err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640				       msg_namelen);
1641		if (err)
1642			return err;
1643
1644		if (msg_namelen > sizeof(to))
1645			msg_namelen = sizeof(to);
1646		memcpy(&to, msg->msg_name, msg_namelen);
1647		msg_name = msg->msg_name;
1648	}
1649
1650	sinit = cmsgs.init;
1651	if (cmsgs.sinfo != NULL) {
1652		memset(&default_sinfo, 0, sizeof(default_sinfo));
1653		default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1654		default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1655		default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1656		default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1657		default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1658
1659		sinfo = &default_sinfo;
1660		fill_sinfo_ttl = true;
1661	} else {
1662		sinfo = cmsgs.srinfo;
1663	}
1664	/* Did the user specify SNDINFO/SNDRCVINFO? */
1665	if (sinfo) {
1666		sinfo_flags = sinfo->sinfo_flags;
1667		associd = sinfo->sinfo_assoc_id;
1668	}
1669
1670	pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1671		 msg_len, sinfo_flags);
1672
1673	/* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674	if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1675		err = -EINVAL;
1676		goto out_nounlock;
1677	}
1678
1679	/* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680	 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681	 * If SCTP_ABORT is set, the message length could be non zero with
1682	 * the msg_iov set to the user abort reason.
1683	 */
1684	if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1685	    (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1686		err = -EINVAL;
1687		goto out_nounlock;
1688	}
1689
1690	/* If SCTP_ADDR_OVER is set, there must be an address
1691	 * specified in msg_name.
1692	 */
1693	if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1694		err = -EINVAL;
1695		goto out_nounlock;
1696	}
1697
1698	transport = NULL;
1699
1700	pr_debug("%s: about to look up association\n", __func__);
1701
1702	lock_sock(sk);
1703
1704	/* If a msg_name has been specified, assume this is to be used.  */
1705	if (msg_name) {
1706		/* Look for a matching association on the endpoint. */
1707		asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1708		if (!asoc) {
1709			/* If we could not find a matching association on the
1710			 * endpoint, make sure that it is not a TCP-style
1711			 * socket that already has an association or there is
1712			 * no peeled-off association on another socket.
1713			 */
1714			if ((sctp_style(sk, TCP) &&
1715			     sctp_sstate(sk, ESTABLISHED)) ||
1716			    sctp_endpoint_is_peeled_off(ep, &to)) {
1717				err = -EADDRNOTAVAIL;
1718				goto out_unlock;
1719			}
1720		}
1721	} else {
1722		asoc = sctp_id2assoc(sk, associd);
1723		if (!asoc) {
1724			err = -EPIPE;
1725			goto out_unlock;
1726		}
1727	}
1728
1729	if (asoc) {
1730		pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1731
1732		/* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1733		 * socket that has an association in CLOSED state. This can
1734		 * happen when an accepted socket has an association that is
1735		 * already CLOSED.
1736		 */
1737		if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1738			err = -EPIPE;
1739			goto out_unlock;
1740		}
1741
1742		if (sinfo_flags & SCTP_EOF) {
1743			pr_debug("%s: shutting down association:%p\n",
1744				 __func__, asoc);
1745
1746			sctp_primitive_SHUTDOWN(net, asoc, NULL);
1747			err = 0;
1748			goto out_unlock;
1749		}
1750		if (sinfo_flags & SCTP_ABORT) {
1751
1752			chunk = sctp_make_abort_user(asoc, msg, msg_len);
1753			if (!chunk) {
1754				err = -ENOMEM;
1755				goto out_unlock;
1756			}
1757
1758			pr_debug("%s: aborting association:%p\n",
1759				 __func__, asoc);
1760
1761			sctp_primitive_ABORT(net, asoc, chunk);
1762			err = 0;
1763			goto out_unlock;
1764		}
1765	}
1766
1767	/* Do we need to create the association?  */
1768	if (!asoc) {
1769		pr_debug("%s: there is no association yet\n", __func__);
1770
1771		if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1772			err = -EINVAL;
1773			goto out_unlock;
1774		}
1775
1776		/* Check for invalid stream against the stream counts,
1777		 * either the default or the user specified stream counts.
1778		 */
1779		if (sinfo) {
1780			if (!sinit || !sinit->sinit_num_ostreams) {
1781				/* Check against the defaults. */
1782				if (sinfo->sinfo_stream >=
1783				    sp->initmsg.sinit_num_ostreams) {
1784					err = -EINVAL;
1785					goto out_unlock;
1786				}
1787			} else {
1788				/* Check against the requested.  */
1789				if (sinfo->sinfo_stream >=
1790				    sinit->sinit_num_ostreams) {
1791					err = -EINVAL;
1792					goto out_unlock;
1793				}
1794			}
1795		}
1796
1797		/*
1798		 * API 3.1.2 bind() - UDP Style Syntax
1799		 * If a bind() or sctp_bindx() is not called prior to a
1800		 * sendmsg() call that initiates a new association, the
1801		 * system picks an ephemeral port and will choose an address
1802		 * set equivalent to binding with a wildcard address.
1803		 */
1804		if (!ep->base.bind_addr.port) {
1805			if (sctp_autobind(sk)) {
1806				err = -EAGAIN;
1807				goto out_unlock;
1808			}
1809		} else {
1810			/*
1811			 * If an unprivileged user inherits a one-to-many
1812			 * style socket with open associations on a privileged
1813			 * port, it MAY be permitted to accept new associations,
1814			 * but it SHOULD NOT be permitted to open new
1815			 * associations.
1816			 */
1817			if (ep->base.bind_addr.port < PROT_SOCK &&
1818			    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1819				err = -EACCES;
1820				goto out_unlock;
1821			}
1822		}
1823
1824		scope = sctp_scope(&to);
1825		new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1826		if (!new_asoc) {
1827			err = -ENOMEM;
1828			goto out_unlock;
1829		}
1830		asoc = new_asoc;
1831		err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1832		if (err < 0) {
1833			err = -ENOMEM;
1834			goto out_free;
1835		}
1836
1837		/* If the SCTP_INIT ancillary data is specified, set all
1838		 * the association init values accordingly.
1839		 */
1840		if (sinit) {
1841			if (sinit->sinit_num_ostreams) {
1842				asoc->c.sinit_num_ostreams =
1843					sinit->sinit_num_ostreams;
1844			}
1845			if (sinit->sinit_max_instreams) {
1846				asoc->c.sinit_max_instreams =
1847					sinit->sinit_max_instreams;
1848			}
1849			if (sinit->sinit_max_attempts) {
1850				asoc->max_init_attempts
1851					= sinit->sinit_max_attempts;
1852			}
1853			if (sinit->sinit_max_init_timeo) {
1854				asoc->max_init_timeo =
1855				 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1856			}
1857		}
1858
1859		/* Prime the peer's transport structures.  */
1860		transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1861		if (!transport) {
1862			err = -ENOMEM;
1863			goto out_free;
1864		}
1865	}
1866
1867	/* ASSERT: we have a valid association at this point.  */
1868	pr_debug("%s: we have a valid association\n", __func__);
1869
1870	if (!sinfo) {
1871		/* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1872		 * one with some defaults.
1873		 */
1874		memset(&default_sinfo, 0, sizeof(default_sinfo));
1875		default_sinfo.sinfo_stream = asoc->default_stream;
1876		default_sinfo.sinfo_flags = asoc->default_flags;
1877		default_sinfo.sinfo_ppid = asoc->default_ppid;
1878		default_sinfo.sinfo_context = asoc->default_context;
1879		default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1880		default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1881
1882		sinfo = &default_sinfo;
1883	} else if (fill_sinfo_ttl) {
1884		/* In case SNDINFO was specified, we still need to fill
1885		 * it with a default ttl from the assoc here.
1886		 */
1887		sinfo->sinfo_timetolive = asoc->default_timetolive;
1888	}
1889
1890	/* API 7.1.7, the sndbuf size per association bounds the
1891	 * maximum size of data that can be sent in a single send call.
1892	 */
1893	if (msg_len > sk->sk_sndbuf) {
1894		err = -EMSGSIZE;
1895		goto out_free;
1896	}
1897
1898	if (asoc->pmtu_pending)
1899		sctp_assoc_pending_pmtu(sk, asoc);
1900
1901	/* If fragmentation is disabled and the message length exceeds the
1902	 * association fragmentation point, return EMSGSIZE.  The I-D
1903	 * does not specify what this error is, but this looks like
1904	 * a great fit.
1905	 */
1906	if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1907		err = -EMSGSIZE;
1908		goto out_free;
1909	}
1910
1911	/* Check for invalid stream. */
1912	if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1913		err = -EINVAL;
1914		goto out_free;
1915	}
1916
1917	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1918	if (!sctp_wspace(asoc)) {
1919		err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1920		if (err)
1921			goto out_free;
1922	}
1923
1924	/* If an address is passed with the sendto/sendmsg call, it is used
1925	 * to override the primary destination address in the TCP model, or
1926	 * when SCTP_ADDR_OVER flag is set in the UDP model.
1927	 */
1928	if ((sctp_style(sk, TCP) && msg_name) ||
1929	    (sinfo_flags & SCTP_ADDR_OVER)) {
1930		chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1931		if (!chunk_tp) {
1932			err = -EINVAL;
1933			goto out_free;
1934		}
1935	} else
1936		chunk_tp = NULL;
1937
1938	/* Auto-connect, if we aren't connected already. */
1939	if (sctp_state(asoc, CLOSED)) {
1940		err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1941		if (err < 0)
1942			goto out_free;
1943
1944		wait_connect = true;
1945		pr_debug("%s: we associated primitively\n", __func__);
1946	}
1947
1948	/* Break the message into multiple chunks of maximum size. */
1949	datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1950	if (IS_ERR(datamsg)) {
1951		err = PTR_ERR(datamsg);
1952		goto out_free;
1953	}
1954
1955	/* Now send the (possibly) fragmented message. */
1956	list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1957		/* Do accounting for the write space.  */
1958		sctp_set_owner_w(chunk);
1959
1960		chunk->transport = chunk_tp;
1961	}
1962
1963	/* Send it to the lower layers.  Note:  all chunks
1964	 * must either fail or succeed.   The lower layer
1965	 * works that way today.  Keep it that way or this
1966	 * breaks.
1967	 */
1968	err = sctp_primitive_SEND(net, asoc, datamsg);
1969	sctp_datamsg_put(datamsg);
1970	/* Did the lower layer accept the chunk? */
1971	if (err)
1972		goto out_free;
1973
1974	pr_debug("%s: we sent primitively\n", __func__);
1975
1976	err = msg_len;
1977
1978	if (unlikely(wait_connect)) {
1979		timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1980		sctp_wait_for_connect(asoc, &timeo);
1981	}
1982
1983	/* If we are already past ASSOCIATE, the lower
1984	 * layers are responsible for association cleanup.
1985	 */
1986	goto out_unlock;
1987
1988out_free:
1989	if (new_asoc) {
1990		sctp_unhash_established(asoc);
1991		sctp_association_free(asoc);
1992	}
1993out_unlock:
1994	release_sock(sk);
1995
1996out_nounlock:
1997	return sctp_error(sk, msg_flags, err);
1998
1999#if 0
2000do_sock_err:
2001	if (msg_len)
2002		err = msg_len;
2003	else
2004		err = sock_error(sk);
2005	goto out;
2006
2007do_interrupted:
2008	if (msg_len)
2009		err = msg_len;
2010	goto out;
2011#endif /* 0 */
2012}
2013
2014/* This is an extended version of skb_pull() that removes the data from the
2015 * start of a skb even when data is spread across the list of skb's in the
2016 * frag_list. len specifies the total amount of data that needs to be removed.
2017 * when 'len' bytes could be removed from the skb, it returns 0.
2018 * If 'len' exceeds the total skb length,  it returns the no. of bytes that
2019 * could not be removed.
2020 */
2021static int sctp_skb_pull(struct sk_buff *skb, int len)
2022{
2023	struct sk_buff *list;
2024	int skb_len = skb_headlen(skb);
2025	int rlen;
2026
2027	if (len <= skb_len) {
2028		__skb_pull(skb, len);
2029		return 0;
2030	}
2031	len -= skb_len;
2032	__skb_pull(skb, skb_len);
2033
2034	skb_walk_frags(skb, list) {
2035		rlen = sctp_skb_pull(list, len);
2036		skb->len -= (len-rlen);
2037		skb->data_len -= (len-rlen);
2038
2039		if (!rlen)
2040			return 0;
2041
2042		len = rlen;
2043	}
2044
2045	return len;
2046}
2047
2048/* API 3.1.3  recvmsg() - UDP Style Syntax
2049 *
2050 *  ssize_t recvmsg(int socket, struct msghdr *message,
2051 *                    int flags);
2052 *
2053 *  socket  - the socket descriptor of the endpoint.
2054 *  message - pointer to the msghdr structure which contains a single
2055 *            user message and possibly some ancillary data.
2056 *
2057 *            See Section 5 for complete description of the data
2058 *            structures.
2059 *
2060 *  flags   - flags sent or received with the user message, see Section
2061 *            5 for complete description of the flags.
2062 */
2063static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2064			int noblock, int flags, int *addr_len)
2065{
2066	struct sctp_ulpevent *event = NULL;
2067	struct sctp_sock *sp = sctp_sk(sk);
2068	struct sk_buff *skb;
2069	int copied;
2070	int err = 0;
2071	int skb_len;
2072
2073	pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2074		 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2075		 addr_len);
2076
2077	lock_sock(sk);
2078
2079	if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2080		err = -ENOTCONN;
2081		goto out;
2082	}
2083
2084	skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2085	if (!skb)
2086		goto out;
2087
2088	/* Get the total length of the skb including any skb's in the
2089	 * frag_list.
2090	 */
2091	skb_len = skb->len;
2092
2093	copied = skb_len;
2094	if (copied > len)
2095		copied = len;
2096
2097	err = skb_copy_datagram_msg(skb, 0, msg, copied);
2098
2099	event = sctp_skb2event(skb);
2100
2101	if (err)
2102		goto out_free;
2103
2104	sock_recv_ts_and_drops(msg, sk, skb);
2105	if (sctp_ulpevent_is_notification(event)) {
2106		msg->msg_flags |= MSG_NOTIFICATION;
2107		sp->pf->event_msgname(event, msg->msg_name, addr_len);
2108	} else {
2109		sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2110	}
2111
2112	/* Check if we allow SCTP_NXTINFO. */
2113	if (sp->recvnxtinfo)
2114		sctp_ulpevent_read_nxtinfo(event, msg, sk);
2115	/* Check if we allow SCTP_RCVINFO. */
2116	if (sp->recvrcvinfo)
2117		sctp_ulpevent_read_rcvinfo(event, msg);
2118	/* Check if we allow SCTP_SNDRCVINFO. */
2119	if (sp->subscribe.sctp_data_io_event)
2120		sctp_ulpevent_read_sndrcvinfo(event, msg);
2121
2122	err = copied;
2123
2124	/* If skb's length exceeds the user's buffer, update the skb and
2125	 * push it back to the receive_queue so that the next call to
2126	 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2127	 */
2128	if (skb_len > copied) {
2129		msg->msg_flags &= ~MSG_EOR;
2130		if (flags & MSG_PEEK)
2131			goto out_free;
2132		sctp_skb_pull(skb, copied);
2133		skb_queue_head(&sk->sk_receive_queue, skb);
2134
2135		/* When only partial message is copied to the user, increase
2136		 * rwnd by that amount. If all the data in the skb is read,
2137		 * rwnd is updated when the event is freed.
2138		 */
2139		if (!sctp_ulpevent_is_notification(event))
2140			sctp_assoc_rwnd_increase(event->asoc, copied);
2141		goto out;
2142	} else if ((event->msg_flags & MSG_NOTIFICATION) ||
2143		   (event->msg_flags & MSG_EOR))
2144		msg->msg_flags |= MSG_EOR;
2145	else
2146		msg->msg_flags &= ~MSG_EOR;
2147
2148out_free:
2149	if (flags & MSG_PEEK) {
2150		/* Release the skb reference acquired after peeking the skb in
2151		 * sctp_skb_recv_datagram().
2152		 */
2153		kfree_skb(skb);
2154	} else {
2155		/* Free the event which includes releasing the reference to
2156		 * the owner of the skb, freeing the skb and updating the
2157		 * rwnd.
2158		 */
2159		sctp_ulpevent_free(event);
2160	}
2161out:
2162	release_sock(sk);
2163	return err;
2164}
2165
2166/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2167 *
2168 * This option is a on/off flag.  If enabled no SCTP message
2169 * fragmentation will be performed.  Instead if a message being sent
2170 * exceeds the current PMTU size, the message will NOT be sent and
2171 * instead a error will be indicated to the user.
2172 */
2173static int sctp_setsockopt_disable_fragments(struct sock *sk,
2174					     char __user *optval,
2175					     unsigned int optlen)
2176{
2177	int val;
2178
2179	if (optlen < sizeof(int))
2180		return -EINVAL;
2181
2182	if (get_user(val, (int __user *)optval))
2183		return -EFAULT;
2184
2185	sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2186
2187	return 0;
2188}
2189
2190static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2191				  unsigned int optlen)
2192{
2193	struct sctp_association *asoc;
2194	struct sctp_ulpevent *event;
2195
2196	if (optlen > sizeof(struct sctp_event_subscribe))
2197		return -EINVAL;
2198	if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2199		return -EFAULT;
2200
2201	/* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2202	 * if there is no data to be sent or retransmit, the stack will
2203	 * immediately send up this notification.
2204	 */
2205	if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2206				       &sctp_sk(sk)->subscribe)) {
2207		asoc = sctp_id2assoc(sk, 0);
2208
2209		if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2210			event = sctp_ulpevent_make_sender_dry_event(asoc,
2211					GFP_ATOMIC);
2212			if (!event)
2213				return -ENOMEM;
2214
2215			sctp_ulpq_tail_event(&asoc->ulpq, event);
2216		}
2217	}
2218
2219	return 0;
2220}
2221
2222/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2223 *
2224 * This socket option is applicable to the UDP-style socket only.  When
2225 * set it will cause associations that are idle for more than the
2226 * specified number of seconds to automatically close.  An association
2227 * being idle is defined an association that has NOT sent or received
2228 * user data.  The special value of '0' indicates that no automatic
2229 * close of any associations should be performed.  The option expects an
2230 * integer defining the number of seconds of idle time before an
2231 * association is closed.
2232 */
2233static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2234				     unsigned int optlen)
2235{
2236	struct sctp_sock *sp = sctp_sk(sk);
2237	struct net *net = sock_net(sk);
2238
2239	/* Applicable to UDP-style socket only */
2240	if (sctp_style(sk, TCP))
2241		return -EOPNOTSUPP;
2242	if (optlen != sizeof(int))
2243		return -EINVAL;
2244	if (copy_from_user(&sp->autoclose, optval, optlen))
2245		return -EFAULT;
2246
2247	if (sp->autoclose > net->sctp.max_autoclose)
2248		sp->autoclose = net->sctp.max_autoclose;
2249
2250	return 0;
2251}
2252
2253/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2254 *
2255 * Applications can enable or disable heartbeats for any peer address of
2256 * an association, modify an address's heartbeat interval, force a
2257 * heartbeat to be sent immediately, and adjust the address's maximum
2258 * number of retransmissions sent before an address is considered
2259 * unreachable.  The following structure is used to access and modify an
2260 * address's parameters:
2261 *
2262 *  struct sctp_paddrparams {
2263 *     sctp_assoc_t            spp_assoc_id;
2264 *     struct sockaddr_storage spp_address;
2265 *     uint32_t                spp_hbinterval;
2266 *     uint16_t                spp_pathmaxrxt;
2267 *     uint32_t                spp_pathmtu;
2268 *     uint32_t                spp_sackdelay;
2269 *     uint32_t                spp_flags;
2270 * };
2271 *
2272 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
2273 *                     application, and identifies the association for
2274 *                     this query.
2275 *   spp_address     - This specifies which address is of interest.
2276 *   spp_hbinterval  - This contains the value of the heartbeat interval,
2277 *                     in milliseconds.  If a  value of zero
2278 *                     is present in this field then no changes are to
2279 *                     be made to this parameter.
2280 *   spp_pathmaxrxt  - This contains the maximum number of
2281 *                     retransmissions before this address shall be
2282 *                     considered unreachable. If a  value of zero
2283 *                     is present in this field then no changes are to
2284 *                     be made to this parameter.
2285 *   spp_pathmtu     - When Path MTU discovery is disabled the value
2286 *                     specified here will be the "fixed" path mtu.
2287 *                     Note that if the spp_address field is empty
2288 *                     then all associations on this address will
2289 *                     have this fixed path mtu set upon them.
2290 *
2291 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
2292 *                     the number of milliseconds that sacks will be delayed
2293 *                     for. This value will apply to all addresses of an
2294 *                     association if the spp_address field is empty. Note
2295 *                     also, that if delayed sack is enabled and this
2296 *                     value is set to 0, no change is made to the last
2297 *                     recorded delayed sack timer value.
2298 *
2299 *   spp_flags       - These flags are used to control various features
2300 *                     on an association. The flag field may contain
2301 *                     zero or more of the following options.
2302 *
2303 *                     SPP_HB_ENABLE  - Enable heartbeats on the
2304 *                     specified address. Note that if the address
2305 *                     field is empty all addresses for the association
2306 *                     have heartbeats enabled upon them.
2307 *
2308 *                     SPP_HB_DISABLE - Disable heartbeats on the
2309 *                     speicifed address. Note that if the address
2310 *                     field is empty all addresses for the association
2311 *                     will have their heartbeats disabled. Note also
2312 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
2313 *                     mutually exclusive, only one of these two should
2314 *                     be specified. Enabling both fields will have
2315 *                     undetermined results.
2316 *
2317 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
2318 *                     to be made immediately.
2319 *
2320 *                     SPP_HB_TIME_IS_ZERO - Specify's that the time for
2321 *                     heartbeat delayis to be set to the value of 0
2322 *                     milliseconds.
2323 *
2324 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
2325 *                     discovery upon the specified address. Note that
2326 *                     if the address feild is empty then all addresses
2327 *                     on the association are effected.
2328 *
2329 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
2330 *                     discovery upon the specified address. Note that
2331 *                     if the address feild is empty then all addresses
2332 *                     on the association are effected. Not also that
2333 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2334 *                     exclusive. Enabling both will have undetermined
2335 *                     results.
2336 *
2337 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
2338 *                     on delayed sack. The time specified in spp_sackdelay
2339 *                     is used to specify the sack delay for this address. Note
2340 *                     that if spp_address is empty then all addresses will
2341 *                     enable delayed sack and take on the sack delay
2342 *                     value specified in spp_sackdelay.
2343 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
2344 *                     off delayed sack. If the spp_address field is blank then
2345 *                     delayed sack is disabled for the entire association. Note
2346 *                     also that this field is mutually exclusive to
2347 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
2348 *                     results.
2349 */
2350static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2351				       struct sctp_transport   *trans,
2352				       struct sctp_association *asoc,
2353				       struct sctp_sock        *sp,
2354				       int                      hb_change,
2355				       int                      pmtud_change,
2356				       int                      sackdelay_change)
2357{
2358	int error;
2359
2360	if (params->spp_flags & SPP_HB_DEMAND && trans) {
2361		struct net *net = sock_net(trans->asoc->base.sk);
2362
2363		error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2364		if (error)
2365			return error;
2366	}
2367
2368	/* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2369	 * this field is ignored.  Note also that a value of zero indicates
2370	 * the current setting should be left unchanged.
2371	 */
2372	if (params->spp_flags & SPP_HB_ENABLE) {
2373
2374		/* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2375		 * set.  This lets us use 0 value when this flag
2376		 * is set.
2377		 */
2378		if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2379			params->spp_hbinterval = 0;
2380
2381		if (params->spp_hbinterval ||
2382		    (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2383			if (trans) {
2384				trans->hbinterval =
2385				    msecs_to_jiffies(params->spp_hbinterval);
2386			} else if (asoc) {
2387				asoc->hbinterval =
2388				    msecs_to_jiffies(params->spp_hbinterval);
2389			} else {
2390				sp->hbinterval = params->spp_hbinterval;
2391			}
2392		}
2393	}
2394
2395	if (hb_change) {
2396		if (trans) {
2397			trans->param_flags =
2398				(trans->param_flags & ~SPP_HB) | hb_change;
2399		} else if (asoc) {
2400			asoc->param_flags =
2401				(asoc->param_flags & ~SPP_HB) | hb_change;
2402		} else {
2403			sp->param_flags =
2404				(sp->param_flags & ~SPP_HB) | hb_change;
2405		}
2406	}
2407
2408	/* When Path MTU discovery is disabled the value specified here will
2409	 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2410	 * include the flag SPP_PMTUD_DISABLE for this field to have any
2411	 * effect).
2412	 */
2413	if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2414		if (trans) {
2415			trans->pathmtu = params->spp_pathmtu;
2416			sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2417		} else if (asoc) {
2418			asoc->pathmtu = params->spp_pathmtu;
2419			sctp_frag_point(asoc, params->spp_pathmtu);
2420		} else {
2421			sp->pathmtu = params->spp_pathmtu;
2422		}
2423	}
2424
2425	if (pmtud_change) {
2426		if (trans) {
2427			int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2428				(params->spp_flags & SPP_PMTUD_ENABLE);
2429			trans->param_flags =
2430				(trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2431			if (update) {
2432				sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2433				sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2434			}
2435		} else if (asoc) {
2436			asoc->param_flags =
2437				(asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2438		} else {
2439			sp->param_flags =
2440				(sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2441		}
2442	}
2443
2444	/* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2445	 * value of this field is ignored.  Note also that a value of zero
2446	 * indicates the current setting should be left unchanged.
2447	 */
2448	if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2449		if (trans) {
2450			trans->sackdelay =
2451				msecs_to_jiffies(params->spp_sackdelay);
2452		} else if (asoc) {
2453			asoc->sackdelay =
2454				msecs_to_jiffies(params->spp_sackdelay);
2455		} else {
2456			sp->sackdelay = params->spp_sackdelay;
2457		}
2458	}
2459
2460	if (sackdelay_change) {
2461		if (trans) {
2462			trans->param_flags =
2463				(trans->param_flags & ~SPP_SACKDELAY) |
2464				sackdelay_change;
2465		} else if (asoc) {
2466			asoc->param_flags =
2467				(asoc->param_flags & ~SPP_SACKDELAY) |
2468				sackdelay_change;
2469		} else {
2470			sp->param_flags =
2471				(sp->param_flags & ~SPP_SACKDELAY) |
2472				sackdelay_change;
2473		}
2474	}
2475
2476	/* Note that a value of zero indicates the current setting should be
2477	   left unchanged.
2478	 */
2479	if (params->spp_pathmaxrxt) {
2480		if (trans) {
2481			trans->pathmaxrxt = params->spp_pathmaxrxt;
2482		} else if (asoc) {
2483			asoc->pathmaxrxt = params->spp_pathmaxrxt;
2484		} else {
2485			sp->pathmaxrxt = params->spp_pathmaxrxt;
2486		}
2487	}
2488
2489	return 0;
2490}
2491
2492static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2493					    char __user *optval,
2494					    unsigned int optlen)
2495{
2496	struct sctp_paddrparams  params;
2497	struct sctp_transport   *trans = NULL;
2498	struct sctp_association *asoc = NULL;
2499	struct sctp_sock        *sp = sctp_sk(sk);
2500	int error;
2501	int hb_change, pmtud_change, sackdelay_change;
2502
2503	if (optlen != sizeof(struct sctp_paddrparams))
2504		return -EINVAL;
2505
2506	if (copy_from_user(&params, optval, optlen))
2507		return -EFAULT;
2508
2509	/* Validate flags and value parameters. */
2510	hb_change        = params.spp_flags & SPP_HB;
2511	pmtud_change     = params.spp_flags & SPP_PMTUD;
2512	sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2513
2514	if (hb_change        == SPP_HB ||
2515	    pmtud_change     == SPP_PMTUD ||
2516	    sackdelay_change == SPP_SACKDELAY ||
2517	    params.spp_sackdelay > 500 ||
2518	    (params.spp_pathmtu &&
2519	     params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2520		return -EINVAL;
2521
2522	/* If an address other than INADDR_ANY is specified, and
2523	 * no transport is found, then the request is invalid.
2524	 */
2525	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2526		trans = sctp_addr_id2transport(sk, &params.spp_address,
2527					       params.spp_assoc_id);
2528		if (!trans)
2529			return -EINVAL;
2530	}
2531
2532	/* Get association, if assoc_id != 0 and the socket is a one
2533	 * to many style socket, and an association was not found, then
2534	 * the id was invalid.
2535	 */
2536	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2537	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2538		return -EINVAL;
2539
2540	/* Heartbeat demand can only be sent on a transport or
2541	 * association, but not a socket.
2542	 */
2543	if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2544		return -EINVAL;
2545
2546	/* Process parameters. */
2547	error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2548					    hb_change, pmtud_change,
2549					    sackdelay_change);
2550
2551	if (error)
2552		return error;
2553
2554	/* If changes are for association, also apply parameters to each
2555	 * transport.
2556	 */
2557	if (!trans && asoc) {
2558		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2559				transports) {
2560			sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2561						    hb_change, pmtud_change,
2562						    sackdelay_change);
2563		}
2564	}
2565
2566	return 0;
2567}
2568
2569static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2570{
2571	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2572}
2573
2574static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2575{
2576	return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2577}
2578
2579/*
2580 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
2581 *
2582 * This option will effect the way delayed acks are performed.  This
2583 * option allows you to get or set the delayed ack time, in
2584 * milliseconds.  It also allows changing the delayed ack frequency.
2585 * Changing the frequency to 1 disables the delayed sack algorithm.  If
2586 * the assoc_id is 0, then this sets or gets the endpoints default
2587 * values.  If the assoc_id field is non-zero, then the set or get
2588 * effects the specified association for the one to many model (the
2589 * assoc_id field is ignored by the one to one model).  Note that if
2590 * sack_delay or sack_freq are 0 when setting this option, then the
2591 * current values will remain unchanged.
2592 *
2593 * struct sctp_sack_info {
2594 *     sctp_assoc_t            sack_assoc_id;
2595 *     uint32_t                sack_delay;
2596 *     uint32_t                sack_freq;
2597 * };
2598 *
2599 * sack_assoc_id -  This parameter, indicates which association the user
2600 *    is performing an action upon.  Note that if this field's value is
2601 *    zero then the endpoints default value is changed (effecting future
2602 *    associations only).
2603 *
2604 * sack_delay -  This parameter contains the number of milliseconds that
2605 *    the user is requesting the delayed ACK timer be set to.  Note that
2606 *    this value is defined in the standard to be between 200 and 500
2607 *    milliseconds.
2608 *
2609 * sack_freq -  This parameter contains the number of packets that must
2610 *    be received before a sack is sent without waiting for the delay
2611 *    timer to expire.  The default value for this is 2, setting this
2612 *    value to 1 will disable the delayed sack algorithm.
2613 */
2614
2615static int sctp_setsockopt_delayed_ack(struct sock *sk,
2616				       char __user *optval, unsigned int optlen)
2617{
2618	struct sctp_sack_info    params;
2619	struct sctp_transport   *trans = NULL;
2620	struct sctp_association *asoc = NULL;
2621	struct sctp_sock        *sp = sctp_sk(sk);
2622
2623	if (optlen == sizeof(struct sctp_sack_info)) {
2624		if (copy_from_user(&params, optval, optlen))
2625			return -EFAULT;
2626
2627		if (params.sack_delay == 0 && params.sack_freq == 0)
2628			return 0;
2629	} else if (optlen == sizeof(struct sctp_assoc_value)) {
2630		pr_warn_ratelimited(DEPRECATED
2631				    "%s (pid %d) "
2632				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2633				    "Use struct sctp_sack_info instead\n",
2634				    current->comm, task_pid_nr(current));
2635		if (copy_from_user(&params, optval, optlen))
2636			return -EFAULT;
2637
2638		if (params.sack_delay == 0)
2639			params.sack_freq = 1;
2640		else
2641			params.sack_freq = 0;
2642	} else
2643		return -EINVAL;
2644
2645	/* Validate value parameter. */
2646	if (params.sack_delay > 500)
2647		return -EINVAL;
2648
2649	/* Get association, if sack_assoc_id != 0 and the socket is a one
2650	 * to many style socket, and an association was not found, then
2651	 * the id was invalid.
2652	 */
2653	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2654	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2655		return -EINVAL;
2656
2657	if (params.sack_delay) {
2658		if (asoc) {
2659			asoc->sackdelay =
2660				msecs_to_jiffies(params.sack_delay);
2661			asoc->param_flags =
2662				sctp_spp_sackdelay_enable(asoc->param_flags);
2663		} else {
2664			sp->sackdelay = params.sack_delay;
2665			sp->param_flags =
2666				sctp_spp_sackdelay_enable(sp->param_flags);
2667		}
2668	}
2669
2670	if (params.sack_freq == 1) {
2671		if (asoc) {
2672			asoc->param_flags =
2673				sctp_spp_sackdelay_disable(asoc->param_flags);
2674		} else {
2675			sp->param_flags =
2676				sctp_spp_sackdelay_disable(sp->param_flags);
2677		}
2678	} else if (params.sack_freq > 1) {
2679		if (asoc) {
2680			asoc->sackfreq = params.sack_freq;
2681			asoc->param_flags =
2682				sctp_spp_sackdelay_enable(asoc->param_flags);
2683		} else {
2684			sp->sackfreq = params.sack_freq;
2685			sp->param_flags =
2686				sctp_spp_sackdelay_enable(sp->param_flags);
2687		}
2688	}
2689
2690	/* If change is for association, also apply to each transport. */
2691	if (asoc) {
2692		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2693				transports) {
2694			if (params.sack_delay) {
2695				trans->sackdelay =
2696					msecs_to_jiffies(params.sack_delay);
2697				trans->param_flags =
2698					sctp_spp_sackdelay_enable(trans->param_flags);
2699			}
2700			if (params.sack_freq == 1) {
2701				trans->param_flags =
2702					sctp_spp_sackdelay_disable(trans->param_flags);
2703			} else if (params.sack_freq > 1) {
2704				trans->sackfreq = params.sack_freq;
2705				trans->param_flags =
2706					sctp_spp_sackdelay_enable(trans->param_flags);
2707			}
2708		}
2709	}
2710
2711	return 0;
2712}
2713
2714/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2715 *
2716 * Applications can specify protocol parameters for the default association
2717 * initialization.  The option name argument to setsockopt() and getsockopt()
2718 * is SCTP_INITMSG.
2719 *
2720 * Setting initialization parameters is effective only on an unconnected
2721 * socket (for UDP-style sockets only future associations are effected
2722 * by the change).  With TCP-style sockets, this option is inherited by
2723 * sockets derived from a listener socket.
2724 */
2725static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2726{
2727	struct sctp_initmsg sinit;
2728	struct sctp_sock *sp = sctp_sk(sk);
2729
2730	if (optlen != sizeof(struct sctp_initmsg))
2731		return -EINVAL;
2732	if (copy_from_user(&sinit, optval, optlen))
2733		return -EFAULT;
2734
2735	if (sinit.sinit_num_ostreams)
2736		sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2737	if (sinit.sinit_max_instreams)
2738		sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2739	if (sinit.sinit_max_attempts)
2740		sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2741	if (sinit.sinit_max_init_timeo)
2742		sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2743
2744	return 0;
2745}
2746
2747/*
2748 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2749 *
2750 *   Applications that wish to use the sendto() system call may wish to
2751 *   specify a default set of parameters that would normally be supplied
2752 *   through the inclusion of ancillary data.  This socket option allows
2753 *   such an application to set the default sctp_sndrcvinfo structure.
2754 *   The application that wishes to use this socket option simply passes
2755 *   in to this call the sctp_sndrcvinfo structure defined in Section
2756 *   5.2.2) The input parameters accepted by this call include
2757 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2758 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
2759 *   to this call if the caller is using the UDP model.
2760 */
2761static int sctp_setsockopt_default_send_param(struct sock *sk,
2762					      char __user *optval,
2763					      unsigned int optlen)
2764{
2765	struct sctp_sock *sp = sctp_sk(sk);
2766	struct sctp_association *asoc;
2767	struct sctp_sndrcvinfo info;
2768
2769	if (optlen != sizeof(info))
2770		return -EINVAL;
2771	if (copy_from_user(&info, optval, optlen))
2772		return -EFAULT;
2773	if (info.sinfo_flags &
2774	    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2775	      SCTP_ABORT | SCTP_EOF))
2776		return -EINVAL;
2777
2778	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2779	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2780		return -EINVAL;
2781	if (asoc) {
2782		asoc->default_stream = info.sinfo_stream;
2783		asoc->default_flags = info.sinfo_flags;
2784		asoc->default_ppid = info.sinfo_ppid;
2785		asoc->default_context = info.sinfo_context;
2786		asoc->default_timetolive = info.sinfo_timetolive;
2787	} else {
2788		sp->default_stream = info.sinfo_stream;
2789		sp->default_flags = info.sinfo_flags;
2790		sp->default_ppid = info.sinfo_ppid;
2791		sp->default_context = info.sinfo_context;
2792		sp->default_timetolive = info.sinfo_timetolive;
2793	}
2794
2795	return 0;
2796}
2797
2798/* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2799 * (SCTP_DEFAULT_SNDINFO)
2800 */
2801static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2802					   char __user *optval,
2803					   unsigned int optlen)
2804{
2805	struct sctp_sock *sp = sctp_sk(sk);
2806	struct sctp_association *asoc;
2807	struct sctp_sndinfo info;
2808
2809	if (optlen != sizeof(info))
2810		return -EINVAL;
2811	if (copy_from_user(&info, optval, optlen))
2812		return -EFAULT;
2813	if (info.snd_flags &
2814	    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2815	      SCTP_ABORT | SCTP_EOF))
2816		return -EINVAL;
2817
2818	asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2819	if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2820		return -EINVAL;
2821	if (asoc) {
2822		asoc->default_stream = info.snd_sid;
2823		asoc->default_flags = info.snd_flags;
2824		asoc->default_ppid = info.snd_ppid;
2825		asoc->default_context = info.snd_context;
2826	} else {
2827		sp->default_stream = info.snd_sid;
2828		sp->default_flags = info.snd_flags;
2829		sp->default_ppid = info.snd_ppid;
2830		sp->default_context = info.snd_context;
2831	}
2832
2833	return 0;
2834}
2835
2836/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2837 *
2838 * Requests that the local SCTP stack use the enclosed peer address as
2839 * the association primary.  The enclosed address must be one of the
2840 * association peer's addresses.
2841 */
2842static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2843					unsigned int optlen)
2844{
2845	struct sctp_prim prim;
2846	struct sctp_transport *trans;
2847
2848	if (optlen != sizeof(struct sctp_prim))
2849		return -EINVAL;
2850
2851	if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2852		return -EFAULT;
2853
2854	trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2855	if (!trans)
2856		return -EINVAL;
2857
2858	sctp_assoc_set_primary(trans->asoc, trans);
2859
2860	return 0;
2861}
2862
2863/*
2864 * 7.1.5 SCTP_NODELAY
2865 *
2866 * Turn on/off any Nagle-like algorithm.  This means that packets are
2867 * generally sent as soon as possible and no unnecessary delays are
2868 * introduced, at the cost of more packets in the network.  Expects an
2869 *  integer boolean flag.
2870 */
2871static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2872				   unsigned int optlen)
2873{
2874	int val;
2875
2876	if (optlen < sizeof(int))
2877		return -EINVAL;
2878	if (get_user(val, (int __user *)optval))
2879		return -EFAULT;
2880
2881	sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2882	return 0;
2883}
2884
2885/*
2886 *
2887 * 7.1.1 SCTP_RTOINFO
2888 *
2889 * The protocol parameters used to initialize and bound retransmission
2890 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2891 * and modify these parameters.
2892 * All parameters are time values, in milliseconds.  A value of 0, when
2893 * modifying the parameters, indicates that the current value should not
2894 * be changed.
2895 *
2896 */
2897static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2898{
2899	struct sctp_rtoinfo rtoinfo;
2900	struct sctp_association *asoc;
2901	unsigned long rto_min, rto_max;
2902	struct sctp_sock *sp = sctp_sk(sk);
2903
2904	if (optlen != sizeof (struct sctp_rtoinfo))
2905		return -EINVAL;
2906
2907	if (copy_from_user(&rtoinfo, optval, optlen))
2908		return -EFAULT;
2909
2910	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2911
2912	/* Set the values to the specific association */
2913	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2914		return -EINVAL;
2915
2916	rto_max = rtoinfo.srto_max;
2917	rto_min = rtoinfo.srto_min;
2918
2919	if (rto_max)
2920		rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2921	else
2922		rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2923
2924	if (rto_min)
2925		rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2926	else
2927		rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2928
2929	if (rto_min > rto_max)
2930		return -EINVAL;
2931
2932	if (asoc) {
2933		if (rtoinfo.srto_initial != 0)
2934			asoc->rto_initial =
2935				msecs_to_jiffies(rtoinfo.srto_initial);
2936		asoc->rto_max = rto_max;
2937		asoc->rto_min = rto_min;
2938	} else {
2939		/* If there is no association or the association-id = 0
2940		 * set the values to the endpoint.
2941		 */
2942		if (rtoinfo.srto_initial != 0)
2943			sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2944		sp->rtoinfo.srto_max = rto_max;
2945		sp->rtoinfo.srto_min = rto_min;
2946	}
2947
2948	return 0;
2949}
2950
2951/*
2952 *
2953 * 7.1.2 SCTP_ASSOCINFO
2954 *
2955 * This option is used to tune the maximum retransmission attempts
2956 * of the association.
2957 * Returns an error if the new association retransmission value is
2958 * greater than the sum of the retransmission value  of the peer.
2959 * See [SCTP] for more information.
2960 *
2961 */
2962static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2963{
2964
2965	struct sctp_assocparams assocparams;
2966	struct sctp_association *asoc;
2967
2968	if (optlen != sizeof(struct sctp_assocparams))
2969		return -EINVAL;
2970	if (copy_from_user(&assocparams, optval, optlen))
2971		return -EFAULT;
2972
2973	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2974
2975	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2976		return -EINVAL;
2977
2978	/* Set the values to the specific association */
2979	if (asoc) {
2980		if (assocparams.sasoc_asocmaxrxt != 0) {
2981			__u32 path_sum = 0;
2982			int   paths = 0;
2983			struct sctp_transport *peer_addr;
2984
2985			list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2986					transports) {
2987				path_sum += peer_addr->pathmaxrxt;
2988				paths++;
2989			}
2990
2991			/* Only validate asocmaxrxt if we have more than
2992			 * one path/transport.  We do this because path
2993			 * retransmissions are only counted when we have more
2994			 * then one path.
2995			 */
2996			if (paths > 1 &&
2997			    assocparams.sasoc_asocmaxrxt > path_sum)
2998				return -EINVAL;
2999
3000			asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3001		}
3002
3003		if (assocparams.sasoc_cookie_life != 0)
3004			asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3005	} else {
3006		/* Set the values to the endpoint */
3007		struct sctp_sock *sp = sctp_sk(sk);
3008
3009		if (assocparams.sasoc_asocmaxrxt != 0)
3010			sp->assocparams.sasoc_asocmaxrxt =
3011						assocparams.sasoc_asocmaxrxt;
3012		if (assocparams.sasoc_cookie_life != 0)
3013			sp->assocparams.sasoc_cookie_life =
3014						assocparams.sasoc_cookie_life;
3015	}
3016	return 0;
3017}
3018
3019/*
3020 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3021 *
3022 * This socket option is a boolean flag which turns on or off mapped V4
3023 * addresses.  If this option is turned on and the socket is type
3024 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3025 * If this option is turned off, then no mapping will be done of V4
3026 * addresses and a user will receive both PF_INET6 and PF_INET type
3027 * addresses on the socket.
3028 */
3029static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3030{
3031	int val;
3032	struct sctp_sock *sp = sctp_sk(sk);
3033
3034	if (optlen < sizeof(int))
3035		return -EINVAL;
3036	if (get_user(val, (int __user *)optval))
3037		return -EFAULT;
3038	if (val)
3039		sp->v4mapped = 1;
3040	else
3041		sp->v4mapped = 0;
3042
3043	return 0;
3044}
3045
3046/*
3047 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3048 * This option will get or set the maximum size to put in any outgoing
3049 * SCTP DATA chunk.  If a message is larger than this size it will be
3050 * fragmented by SCTP into the specified size.  Note that the underlying
3051 * SCTP implementation may fragment into smaller sized chunks when the
3052 * PMTU of the underlying association is smaller than the value set by
3053 * the user.  The default value for this option is '0' which indicates
3054 * the user is NOT limiting fragmentation and only the PMTU will effect
3055 * SCTP's choice of DATA chunk size.  Note also that values set larger
3056 * than the maximum size of an IP datagram will effectively let SCTP
3057 * control fragmentation (i.e. the same as setting this option to 0).
3058 *
3059 * The following structure is used to access and modify this parameter:
3060 *
3061 * struct sctp_assoc_value {
3062 *   sctp_assoc_t assoc_id;
3063 *   uint32_t assoc_value;
3064 * };
3065 *
3066 * assoc_id:  This parameter is ignored for one-to-one style sockets.
3067 *    For one-to-many style sockets this parameter indicates which
3068 *    association the user is performing an action upon.  Note that if
3069 *    this field's value is zero then the endpoints default value is
3070 *    changed (effecting future associations only).
3071 * assoc_value:  This parameter specifies the maximum size in bytes.
3072 */
3073static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3074{
3075	struct sctp_assoc_value params;
3076	struct sctp_association *asoc;
3077	struct sctp_sock *sp = sctp_sk(sk);
3078	int val;
3079
3080	if (optlen == sizeof(int)) {
3081		pr_warn_ratelimited(DEPRECATED
3082				    "%s (pid %d) "
3083				    "Use of int in maxseg socket option.\n"
3084				    "Use struct sctp_assoc_value instead\n",
3085				    current->comm, task_pid_nr(current));
3086		if (copy_from_user(&val, optval, optlen))
3087			return -EFAULT;
3088		params.assoc_id = 0;
3089	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3090		if (copy_from_user(&params, optval, optlen))
3091			return -EFAULT;
3092		val = params.assoc_value;
3093	} else
3094		return -EINVAL;
3095
3096	if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3097		return -EINVAL;
3098
3099	asoc = sctp_id2assoc(sk, params.assoc_id);
3100	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3101		return -EINVAL;
3102
3103	if (asoc) {
3104		if (val == 0) {
3105			val = asoc->pathmtu;
3106			val -= sp->pf->af->net_header_len;
3107			val -= sizeof(struct sctphdr) +
3108					sizeof(struct sctp_data_chunk);
3109		}
3110		asoc->user_frag = val;
3111		asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3112	} else {
3113		sp->user_frag = val;
3114	}
3115
3116	return 0;
3117}
3118
3119
3120/*
3121 *  7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3122 *
3123 *   Requests that the peer mark the enclosed address as the association
3124 *   primary. The enclosed address must be one of the association's
3125 *   locally bound addresses. The following structure is used to make a
3126 *   set primary request:
3127 */
3128static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3129					     unsigned int optlen)
3130{
3131	struct net *net = sock_net(sk);
3132	struct sctp_sock	*sp;
3133	struct sctp_association	*asoc = NULL;
3134	struct sctp_setpeerprim	prim;
3135	struct sctp_chunk	*chunk;
3136	struct sctp_af		*af;
3137	int 			err;
3138
3139	sp = sctp_sk(sk);
3140
3141	if (!net->sctp.addip_enable)
3142		return -EPERM;
3143
3144	if (optlen != sizeof(struct sctp_setpeerprim))
3145		return -EINVAL;
3146
3147	if (copy_from_user(&prim, optval, optlen))
3148		return -EFAULT;
3149
3150	asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3151	if (!asoc)
3152		return -EINVAL;
3153
3154	if (!asoc->peer.asconf_capable)
3155		return -EPERM;
3156
3157	if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3158		return -EPERM;
3159
3160	if (!sctp_state(asoc, ESTABLISHED))
3161		return -ENOTCONN;
3162
3163	af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3164	if (!af)
3165		return -EINVAL;
3166
3167	if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3168		return -EADDRNOTAVAIL;
3169
3170	if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3171		return -EADDRNOTAVAIL;
3172
3173	/* Create an ASCONF chunk with SET_PRIMARY parameter	*/
3174	chunk = sctp_make_asconf_set_prim(asoc,
3175					  (union sctp_addr *)&prim.sspp_addr);
3176	if (!chunk)
3177		return -ENOMEM;
3178
3179	err = sctp_send_asconf(asoc, chunk);
3180
3181	pr_debug("%s: we set peer primary addr primitively\n", __func__);
3182
3183	return err;
3184}
3185
3186static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3187					    unsigned int optlen)
3188{
3189	struct sctp_setadaptation adaptation;
3190
3191	if (optlen != sizeof(struct sctp_setadaptation))
3192		return -EINVAL;
3193	if (copy_from_user(&adaptation, optval, optlen))
3194		return -EFAULT;
3195
3196	sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3197
3198	return 0;
3199}
3200
3201/*
3202 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
3203 *
3204 * The context field in the sctp_sndrcvinfo structure is normally only
3205 * used when a failed message is retrieved holding the value that was
3206 * sent down on the actual send call.  This option allows the setting of
3207 * a default context on an association basis that will be received on
3208 * reading messages from the peer.  This is especially helpful in the
3209 * one-2-many model for an application to keep some reference to an
3210 * internal state machine that is processing messages on the
3211 * association.  Note that the setting of this value only effects
3212 * received messages from the peer and does not effect the value that is
3213 * saved with outbound messages.
3214 */
3215static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3216				   unsigned int optlen)
3217{
3218	struct sctp_assoc_value params;
3219	struct sctp_sock *sp;
3220	struct sctp_association *asoc;
3221
3222	if (optlen != sizeof(struct sctp_assoc_value))
3223		return -EINVAL;
3224	if (copy_from_user(&params, optval, optlen))
3225		return -EFAULT;
3226
3227	sp = sctp_sk(sk);
3228
3229	if (params.assoc_id != 0) {
3230		asoc = sctp_id2assoc(sk, params.assoc_id);
3231		if (!asoc)
3232			return -EINVAL;
3233		asoc->default_rcv_context = params.assoc_value;
3234	} else {
3235		sp->default_rcv_context = params.assoc_value;
3236	}
3237
3238	return 0;
3239}
3240
3241/*
3242 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3243 *
3244 * This options will at a minimum specify if the implementation is doing
3245 * fragmented interleave.  Fragmented interleave, for a one to many
3246 * socket, is when subsequent calls to receive a message may return
3247 * parts of messages from different associations.  Some implementations
3248 * may allow you to turn this value on or off.  If so, when turned off,
3249 * no fragment interleave will occur (which will cause a head of line
3250 * blocking amongst multiple associations sharing the same one to many
3251 * socket).  When this option is turned on, then each receive call may
3252 * come from a different association (thus the user must receive data
3253 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3254 * association each receive belongs to.
3255 *
3256 * This option takes a boolean value.  A non-zero value indicates that
3257 * fragmented interleave is on.  A value of zero indicates that
3258 * fragmented interleave is off.
3259 *
3260 * Note that it is important that an implementation that allows this
3261 * option to be turned on, have it off by default.  Otherwise an unaware
3262 * application using the one to many model may become confused and act
3263 * incorrectly.
3264 */
3265static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3266					       char __user *optval,
3267					       unsigned int optlen)
3268{
3269	int val;
3270
3271	if (optlen != sizeof(int))
3272		return -EINVAL;
3273	if (get_user(val, (int __user *)optval))
3274		return -EFAULT;
3275
3276	sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3277
3278	return 0;
3279}
3280
3281/*
3282 * 8.1.21.  Set or Get the SCTP Partial Delivery Point
3283 *       (SCTP_PARTIAL_DELIVERY_POINT)
3284 *
3285 * This option will set or get the SCTP partial delivery point.  This
3286 * point is the size of a message where the partial delivery API will be
3287 * invoked to help free up rwnd space for the peer.  Setting this to a
3288 * lower value will cause partial deliveries to happen more often.  The
3289 * calls argument is an integer that sets or gets the partial delivery
3290 * point.  Note also that the call will fail if the user attempts to set
3291 * this value larger than the socket receive buffer size.
3292 *
3293 * Note that any single message having a length smaller than or equal to
3294 * the SCTP partial delivery point will be delivered in one single read
3295 * call as long as the user provided buffer is large enough to hold the
3296 * message.
3297 */
3298static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3299						  char __user *optval,
3300						  unsigned int optlen)
3301{
3302	u32 val;
3303
3304	if (optlen != sizeof(u32))
3305		return -EINVAL;
3306	if (get_user(val, (int __user *)optval))
3307		return -EFAULT;
3308
3309	/* Note: We double the receive buffer from what the user sets
3310	 * it to be, also initial rwnd is based on rcvbuf/2.
3311	 */
3312	if (val > (sk->sk_rcvbuf >> 1))
3313		return -EINVAL;
3314
3315	sctp_sk(sk)->pd_point = val;
3316
3317	return 0; /* is this the right error code? */
3318}
3319
3320/*
3321 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
3322 *
3323 * This option will allow a user to change the maximum burst of packets
3324 * that can be emitted by this association.  Note that the default value
3325 * is 4, and some implementations may restrict this setting so that it
3326 * can only be lowered.
3327 *
3328 * NOTE: This text doesn't seem right.  Do this on a socket basis with
3329 * future associations inheriting the socket value.
3330 */
3331static int sctp_setsockopt_maxburst(struct sock *sk,
3332				    char __user *optval,
3333				    unsigned int optlen)
3334{
3335	struct sctp_assoc_value params;
3336	struct sctp_sock *sp;
3337	struct sctp_association *asoc;
3338	int val;
3339	int assoc_id = 0;
3340
3341	if (optlen == sizeof(int)) {
3342		pr_warn_ratelimited(DEPRECATED
3343				    "%s (pid %d) "
3344				    "Use of int in max_burst socket option deprecated.\n"
3345				    "Use struct sctp_assoc_value instead\n",
3346				    current->comm, task_pid_nr(current));
3347		if (copy_from_user(&val, optval, optlen))
3348			return -EFAULT;
3349	} else if (optlen == sizeof(struct sctp_assoc_value)) {
3350		if (copy_from_user(&params, optval, optlen))
3351			return -EFAULT;
3352		val = params.assoc_value;
3353		assoc_id = params.assoc_id;
3354	} else
3355		return -EINVAL;
3356
3357	sp = sctp_sk(sk);
3358
3359	if (assoc_id != 0) {
3360		asoc = sctp_id2assoc(sk, assoc_id);
3361		if (!asoc)
3362			return -EINVAL;
3363		asoc->max_burst = val;
3364	} else
3365		sp->max_burst = val;
3366
3367	return 0;
3368}
3369
3370/*
3371 * 7.1.18.  Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3372 *
3373 * This set option adds a chunk type that the user is requesting to be
3374 * received only in an authenticated way.  Changes to the list of chunks
3375 * will only effect future associations on the socket.
3376 */
3377static int sctp_setsockopt_auth_chunk(struct sock *sk,
3378				      char __user *optval,
3379				      unsigned int optlen)
3380{
3381	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3382	struct sctp_authchunk val;
3383
3384	if (!ep->auth_enable)
3385		return -EACCES;
3386
3387	if (optlen != sizeof(struct sctp_authchunk))
3388		return -EINVAL;
3389	if (copy_from_user(&val, optval, optlen))
3390		return -EFAULT;
3391
3392	switch (val.sauth_chunk) {
3393	case SCTP_CID_INIT:
3394	case SCTP_CID_INIT_ACK:
3395	case SCTP_CID_SHUTDOWN_COMPLETE:
3396	case SCTP_CID_AUTH:
3397		return -EINVAL;
3398	}
3399
3400	/* add this chunk id to the endpoint */
3401	return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3402}
3403
3404/*
3405 * 7.1.19.  Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3406 *
3407 * This option gets or sets the list of HMAC algorithms that the local
3408 * endpoint requires the peer to use.
3409 */
3410static int sctp_setsockopt_hmac_ident(struct sock *sk,
3411				      char __user *optval,
3412				      unsigned int optlen)
3413{
3414	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3415	struct sctp_hmacalgo *hmacs;
3416	u32 idents;
3417	int err;
3418
3419	if (!ep->auth_enable)
3420		return -EACCES;
3421
3422	if (optlen < sizeof(struct sctp_hmacalgo))
3423		return -EINVAL;
3424
3425	hmacs = memdup_user(optval, optlen);
3426	if (IS_ERR(hmacs))
3427		return PTR_ERR(hmacs);
3428
3429	idents = hmacs->shmac_num_idents;
3430	if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3431	    (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3432		err = -EINVAL;
3433		goto out;
3434	}
3435
3436	err = sctp_auth_ep_set_hmacs(ep, hmacs);
3437out:
3438	kfree(hmacs);
3439	return err;
3440}
3441
3442/*
3443 * 7.1.20.  Set a shared key (SCTP_AUTH_KEY)
3444 *
3445 * This option will set a shared secret key which is used to build an
3446 * association shared key.
3447 */
3448static int sctp_setsockopt_auth_key(struct sock *sk,
3449				    char __user *optval,
3450				    unsigned int optlen)
3451{
3452	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3453	struct sctp_authkey *authkey;
3454	struct sctp_association *asoc;
3455	int ret;
3456
3457	if (!ep->auth_enable)
3458		return -EACCES;
3459
3460	if (optlen <= sizeof(struct sctp_authkey))
3461		return -EINVAL;
3462
3463	authkey = memdup_user(optval, optlen);
3464	if (IS_ERR(authkey))
3465		return PTR_ERR(authkey);
3466
3467	if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3468		ret = -EINVAL;
3469		goto out;
3470	}
3471
3472	asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3473	if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3474		ret = -EINVAL;
3475		goto out;
3476	}
3477
3478	ret = sctp_auth_set_key(ep, asoc, authkey);
3479out:
3480	kzfree(authkey);
3481	return ret;
3482}
3483
3484/*
3485 * 7.1.21.  Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3486 *
3487 * This option will get or set the active shared key to be used to build
3488 * the association shared key.
3489 */
3490static int sctp_setsockopt_active_key(struct sock *sk,
3491				      char __user *optval,
3492				      unsigned int optlen)
3493{
3494	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3495	struct sctp_authkeyid val;
3496	struct sctp_association *asoc;
3497
3498	if (!ep->auth_enable)
3499		return -EACCES;
3500
3501	if (optlen != sizeof(struct sctp_authkeyid))
3502		return -EINVAL;
3503	if (copy_from_user(&val, optval, optlen))
3504		return -EFAULT;
3505
3506	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3507	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3508		return -EINVAL;
3509
3510	return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3511}
3512
3513/*
3514 * 7.1.22.  Delete a shared key (SCTP_AUTH_DELETE_KEY)
3515 *
3516 * This set option will delete a shared secret key from use.
3517 */
3518static int sctp_setsockopt_del_key(struct sock *sk,
3519				   char __user *optval,
3520				   unsigned int optlen)
3521{
3522	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3523	struct sctp_authkeyid val;
3524	struct sctp_association *asoc;
3525
3526	if (!ep->auth_enable)
3527		return -EACCES;
3528
3529	if (optlen != sizeof(struct sctp_authkeyid))
3530		return -EINVAL;
3531	if (copy_from_user(&val, optval, optlen))
3532		return -EFAULT;
3533
3534	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3535	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3536		return -EINVAL;
3537
3538	return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3539
3540}
3541
3542/*
3543 * 8.1.23 SCTP_AUTO_ASCONF
3544 *
3545 * This option will enable or disable the use of the automatic generation of
3546 * ASCONF chunks to add and delete addresses to an existing association.  Note
3547 * that this option has two caveats namely: a) it only affects sockets that
3548 * are bound to all addresses available to the SCTP stack, and b) the system
3549 * administrator may have an overriding control that turns the ASCONF feature
3550 * off no matter what setting the socket option may have.
3551 * This option expects an integer boolean flag, where a non-zero value turns on
3552 * the option, and a zero value turns off the option.
3553 * Note. In this implementation, socket operation overrides default parameter
3554 * being set by sysctl as well as FreeBSD implementation
3555 */
3556static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3557					unsigned int optlen)
3558{
3559	int val;
3560	struct sctp_sock *sp = sctp_sk(sk);
3561
3562	if (optlen < sizeof(int))
3563		return -EINVAL;
3564	if (get_user(val, (int __user *)optval))
3565		return -EFAULT;
3566	if (!sctp_is_ep_boundall(sk) && val)
3567		return -EINVAL;
3568	if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3569		return 0;
3570
3571	spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3572	if (val == 0 && sp->do_auto_asconf) {
3573		list_del(&sp->auto_asconf_list);
3574		sp->do_auto_asconf = 0;
3575	} else if (val && !sp->do_auto_asconf) {
3576		list_add_tail(&sp->auto_asconf_list,
3577		    &sock_net(sk)->sctp.auto_asconf_splist);
3578		sp->do_auto_asconf = 1;
3579	}
3580	spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3581	return 0;
3582}
3583
3584/*
3585 * SCTP_PEER_ADDR_THLDS
3586 *
3587 * This option allows us to alter the partially failed threshold for one or all
3588 * transports in an association.  See Section 6.1 of:
3589 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3590 */
3591static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3592					    char __user *optval,
3593					    unsigned int optlen)
3594{
3595	struct sctp_paddrthlds val;
3596	struct sctp_transport *trans;
3597	struct sctp_association *asoc;
3598
3599	if (optlen < sizeof(struct sctp_paddrthlds))
3600		return -EINVAL;
3601	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3602			   sizeof(struct sctp_paddrthlds)))
3603		return -EFAULT;
3604
3605
3606	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3607		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3608		if (!asoc)
3609			return -ENOENT;
3610		list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3611				    transports) {
3612			if (val.spt_pathmaxrxt)
3613				trans->pathmaxrxt = val.spt_pathmaxrxt;
3614			trans->pf_retrans = val.spt_pathpfthld;
3615		}
3616
3617		if (val.spt_pathmaxrxt)
3618			asoc->pathmaxrxt = val.spt_pathmaxrxt;
3619		asoc->pf_retrans = val.spt_pathpfthld;
3620	} else {
3621		trans = sctp_addr_id2transport(sk, &val.spt_address,
3622					       val.spt_assoc_id);
3623		if (!trans)
3624			return -ENOENT;
3625
3626		if (val.spt_pathmaxrxt)
3627			trans->pathmaxrxt = val.spt_pathmaxrxt;
3628		trans->pf_retrans = val.spt_pathpfthld;
3629	}
3630
3631	return 0;
3632}
3633
3634static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3635				       char __user *optval,
3636				       unsigned int optlen)
3637{
3638	int val;
3639
3640	if (optlen < sizeof(int))
3641		return -EINVAL;
3642	if (get_user(val, (int __user *) optval))
3643		return -EFAULT;
3644
3645	sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3646
3647	return 0;
3648}
3649
3650static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3651				       char __user *optval,
3652				       unsigned int optlen)
3653{
3654	int val;
3655
3656	if (optlen < sizeof(int))
3657		return -EINVAL;
3658	if (get_user(val, (int __user *) optval))
3659		return -EFAULT;
3660
3661	sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3662
3663	return 0;
3664}
3665
3666/* API 6.2 setsockopt(), getsockopt()
3667 *
3668 * Applications use setsockopt() and getsockopt() to set or retrieve
3669 * socket options.  Socket options are used to change the default
3670 * behavior of sockets calls.  They are described in Section 7.
3671 *
3672 * The syntax is:
3673 *
3674 *   ret = getsockopt(int sd, int level, int optname, void __user *optval,
3675 *                    int __user *optlen);
3676 *   ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3677 *                    int optlen);
3678 *
3679 *   sd      - the socket descript.
3680 *   level   - set to IPPROTO_SCTP for all SCTP options.
3681 *   optname - the option name.
3682 *   optval  - the buffer to store the value of the option.
3683 *   optlen  - the size of the buffer.
3684 */
3685static int sctp_setsockopt(struct sock *sk, int level, int optname,
3686			   char __user *optval, unsigned int optlen)
3687{
3688	int retval = 0;
3689
3690	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3691
3692	/* I can hardly begin to describe how wrong this is.  This is
3693	 * so broken as to be worse than useless.  The API draft
3694	 * REALLY is NOT helpful here...  I am not convinced that the
3695	 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3696	 * are at all well-founded.
3697	 */
3698	if (level != SOL_SCTP) {
3699		struct sctp_af *af = sctp_sk(sk)->pf->af;
3700		retval = af->setsockopt(sk, level, optname, optval, optlen);
3701		goto out_nounlock;
3702	}
3703
3704	lock_sock(sk);
3705
3706	switch (optname) {
3707	case SCTP_SOCKOPT_BINDX_ADD:
3708		/* 'optlen' is the size of the addresses buffer. */
3709		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3710					       optlen, SCTP_BINDX_ADD_ADDR);
3711		break;
3712
3713	case SCTP_SOCKOPT_BINDX_REM:
3714		/* 'optlen' is the size of the addresses buffer. */
3715		retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3716					       optlen, SCTP_BINDX_REM_ADDR);
3717		break;
3718
3719	case SCTP_SOCKOPT_CONNECTX_OLD:
3720		/* 'optlen' is the size of the addresses buffer. */
3721		retval = sctp_setsockopt_connectx_old(sk,
3722					    (struct sockaddr __user *)optval,
3723					    optlen);
3724		break;
3725
3726	case SCTP_SOCKOPT_CONNECTX:
3727		/* 'optlen' is the size of the addresses buffer. */
3728		retval = sctp_setsockopt_connectx(sk,
3729					    (struct sockaddr __user *)optval,
3730					    optlen);
3731		break;
3732
3733	case SCTP_DISABLE_FRAGMENTS:
3734		retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3735		break;
3736
3737	case SCTP_EVENTS:
3738		retval = sctp_setsockopt_events(sk, optval, optlen);
3739		break;
3740
3741	case SCTP_AUTOCLOSE:
3742		retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3743		break;
3744
3745	case SCTP_PEER_ADDR_PARAMS:
3746		retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3747		break;
3748
3749	case SCTP_DELAYED_SACK:
3750		retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3751		break;
3752	case SCTP_PARTIAL_DELIVERY_POINT:
3753		retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3754		break;
3755
3756	case SCTP_INITMSG:
3757		retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3758		break;
3759	case SCTP_DEFAULT_SEND_PARAM:
3760		retval = sctp_setsockopt_default_send_param(sk, optval,
3761							    optlen);
3762		break;
3763	case SCTP_DEFAULT_SNDINFO:
3764		retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3765		break;
3766	case SCTP_PRIMARY_ADDR:
3767		retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3768		break;
3769	case SCTP_SET_PEER_PRIMARY_ADDR:
3770		retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3771		break;
3772	case SCTP_NODELAY:
3773		retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3774		break;
3775	case SCTP_RTOINFO:
3776		retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3777		break;
3778	case SCTP_ASSOCINFO:
3779		retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3780		break;
3781	case SCTP_I_WANT_MAPPED_V4_ADDR:
3782		retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3783		break;
3784	case SCTP_MAXSEG:
3785		retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3786		break;
3787	case SCTP_ADAPTATION_LAYER:
3788		retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3789		break;
3790	case SCTP_CONTEXT:
3791		retval = sctp_setsockopt_context(sk, optval, optlen);
3792		break;
3793	case SCTP_FRAGMENT_INTERLEAVE:
3794		retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3795		break;
3796	case SCTP_MAX_BURST:
3797		retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3798		break;
3799	case SCTP_AUTH_CHUNK:
3800		retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3801		break;
3802	case SCTP_HMAC_IDENT:
3803		retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3804		break;
3805	case SCTP_AUTH_KEY:
3806		retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3807		break;
3808	case SCTP_AUTH_ACTIVE_KEY:
3809		retval = sctp_setsockopt_active_key(sk, optval, optlen);
3810		break;
3811	case SCTP_AUTH_DELETE_KEY:
3812		retval = sctp_setsockopt_del_key(sk, optval, optlen);
3813		break;
3814	case SCTP_AUTO_ASCONF:
3815		retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3816		break;
3817	case SCTP_PEER_ADDR_THLDS:
3818		retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3819		break;
3820	case SCTP_RECVRCVINFO:
3821		retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3822		break;
3823	case SCTP_RECVNXTINFO:
3824		retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3825		break;
3826	default:
3827		retval = -ENOPROTOOPT;
3828		break;
3829	}
3830
3831	release_sock(sk);
3832
3833out_nounlock:
3834	return retval;
3835}
3836
3837/* API 3.1.6 connect() - UDP Style Syntax
3838 *
3839 * An application may use the connect() call in the UDP model to initiate an
3840 * association without sending data.
3841 *
3842 * The syntax is:
3843 *
3844 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3845 *
3846 * sd: the socket descriptor to have a new association added to.
3847 *
3848 * nam: the address structure (either struct sockaddr_in or struct
3849 *    sockaddr_in6 defined in RFC2553 [7]).
3850 *
3851 * len: the size of the address.
3852 */
3853static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3854			int addr_len)
3855{
3856	int err = 0;
3857	struct sctp_af *af;
3858
3859	lock_sock(sk);
3860
3861	pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3862		 addr, addr_len);
3863
3864	/* Validate addr_len before calling common connect/connectx routine. */
3865	af = sctp_get_af_specific(addr->sa_family);
3866	if (!af || addr_len < af->sockaddr_len) {
3867		err = -EINVAL;
3868	} else {
3869		/* Pass correct addr len to common routine (so it knows there
3870		 * is only one address being passed.
3871		 */
3872		err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3873	}
3874
3875	release_sock(sk);
3876	return err;
3877}
3878
3879/* FIXME: Write comments. */
3880static int sctp_disconnect(struct sock *sk, int flags)
3881{
3882	return -EOPNOTSUPP; /* STUB */
3883}
3884
3885/* 4.1.4 accept() - TCP Style Syntax
3886 *
3887 * Applications use accept() call to remove an established SCTP
3888 * association from the accept queue of the endpoint.  A new socket
3889 * descriptor will be returned from accept() to represent the newly
3890 * formed association.
3891 */
3892static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3893{
3894	struct sctp_sock *sp;
3895	struct sctp_endpoint *ep;
3896	struct sock *newsk = NULL;
3897	struct sctp_association *asoc;
3898	long timeo;
3899	int error = 0;
3900
3901	lock_sock(sk);
3902
3903	sp = sctp_sk(sk);
3904	ep = sp->ep;
3905
3906	if (!sctp_style(sk, TCP)) {
3907		error = -EOPNOTSUPP;
3908		goto out;
3909	}
3910
3911	if (!sctp_sstate(sk, LISTENING)) {
3912		error = -EINVAL;
3913		goto out;
3914	}
3915
3916	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3917
3918	error = sctp_wait_for_accept(sk, timeo);
3919	if (error)
3920		goto out;
3921
3922	/* We treat the list of associations on the endpoint as the accept
3923	 * queue and pick the first association on the list.
3924	 */
3925	asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3926
3927	newsk = sp->pf->create_accept_sk(sk, asoc);
3928	if (!newsk) {
3929		error = -ENOMEM;
3930		goto out;
3931	}
3932
3933	/* Populate the fields of the newsk from the oldsk and migrate the
3934	 * asoc to the newsk.
3935	 */
3936	sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3937
3938out:
3939	release_sock(sk);
3940	*err = error;
3941	return newsk;
3942}
3943
3944/* The SCTP ioctl handler. */
3945static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3946{
3947	int rc = -ENOTCONN;
3948
3949	lock_sock(sk);
3950
3951	/*
3952	 * SEQPACKET-style sockets in LISTENING state are valid, for
3953	 * SCTP, so only discard TCP-style sockets in LISTENING state.
3954	 */
3955	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3956		goto out;
3957
3958	switch (cmd) {
3959	case SIOCINQ: {
3960		struct sk_buff *skb;
3961		unsigned int amount = 0;
3962
3963		skb = skb_peek(&sk->sk_receive_queue);
3964		if (skb != NULL) {
3965			/*
3966			 * We will only return the amount of this packet since
3967			 * that is all that will be read.
3968			 */
3969			amount = skb->len;
3970		}
3971		rc = put_user(amount, (int __user *)arg);
3972		break;
3973	}
3974	default:
3975		rc = -ENOIOCTLCMD;
3976		break;
3977	}
3978out:
3979	release_sock(sk);
3980	return rc;
3981}
3982
3983/* This is the function which gets called during socket creation to
3984 * initialized the SCTP-specific portion of the sock.
3985 * The sock structure should already be zero-filled memory.
3986 */
3987static int sctp_init_sock(struct sock *sk)
3988{
3989	struct net *net = sock_net(sk);
3990	struct sctp_sock *sp;
3991
3992	pr_debug("%s: sk:%p\n", __func__, sk);
3993
3994	sp = sctp_sk(sk);
3995
3996	/* Initialize the SCTP per socket area.  */
3997	switch (sk->sk_type) {
3998	case SOCK_SEQPACKET:
3999		sp->type = SCTP_SOCKET_UDP;
4000		break;
4001	case SOCK_STREAM:
4002		sp->type = SCTP_SOCKET_TCP;
4003		break;
4004	default:
4005		return -ESOCKTNOSUPPORT;
4006	}
4007
4008	/* Initialize default send parameters. These parameters can be
4009	 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4010	 */
4011	sp->default_stream = 0;
4012	sp->default_ppid = 0;
4013	sp->default_flags = 0;
4014	sp->default_context = 0;
4015	sp->default_timetolive = 0;
4016
4017	sp->default_rcv_context = 0;
4018	sp->max_burst = net->sctp.max_burst;
4019
4020	sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4021
4022	/* Initialize default setup parameters. These parameters
4023	 * can be modified with the SCTP_INITMSG socket option or
4024	 * overridden by the SCTP_INIT CMSG.
4025	 */
4026	sp->initmsg.sinit_num_ostreams   = sctp_max_outstreams;
4027	sp->initmsg.sinit_max_instreams  = sctp_max_instreams;
4028	sp->initmsg.sinit_max_attempts   = net->sctp.max_retrans_init;
4029	sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4030
4031	/* Initialize default RTO related parameters.  These parameters can
4032	 * be modified for with the SCTP_RTOINFO socket option.
4033	 */
4034	sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4035	sp->rtoinfo.srto_max     = net->sctp.rto_max;
4036	sp->rtoinfo.srto_min     = net->sctp.rto_min;
4037
4038	/* Initialize default association related parameters. These parameters
4039	 * can be modified with the SCTP_ASSOCINFO socket option.
4040	 */
4041	sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4042	sp->assocparams.sasoc_number_peer_destinations = 0;
4043	sp->assocparams.sasoc_peer_rwnd = 0;
4044	sp->assocparams.sasoc_local_rwnd = 0;
4045	sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4046
4047	/* Initialize default event subscriptions. By default, all the
4048	 * options are off.
4049	 */
4050	memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4051
4052	/* Default Peer Address Parameters.  These defaults can
4053	 * be modified via SCTP_PEER_ADDR_PARAMS
4054	 */
4055	sp->hbinterval  = net->sctp.hb_interval;
4056	sp->pathmaxrxt  = net->sctp.max_retrans_path;
4057	sp->pathmtu     = 0; /* allow default discovery */
4058	sp->sackdelay   = net->sctp.sack_timeout;
4059	sp->sackfreq	= 2;
4060	sp->param_flags = SPP_HB_ENABLE |
4061			  SPP_PMTUD_ENABLE |
4062			  SPP_SACKDELAY_ENABLE;
4063
4064	/* If enabled no SCTP message fragmentation will be performed.
4065	 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4066	 */
4067	sp->disable_fragments = 0;
4068
4069	/* Enable Nagle algorithm by default.  */
4070	sp->nodelay           = 0;
4071
4072	sp->recvrcvinfo = 0;
4073	sp->recvnxtinfo = 0;
4074
4075	/* Enable by default. */
4076	sp->v4mapped          = 1;
4077
4078	/* Auto-close idle associations after the configured
4079	 * number of seconds.  A value of 0 disables this
4080	 * feature.  Configure through the SCTP_AUTOCLOSE socket option,
4081	 * for UDP-style sockets only.
4082	 */
4083	sp->autoclose         = 0;
4084
4085	/* User specified fragmentation limit. */
4086	sp->user_frag         = 0;
4087
4088	sp->adaptation_ind = 0;
4089
4090	sp->pf = sctp_get_pf_specific(sk->sk_family);
4091
4092	/* Control variables for partial data delivery. */
4093	atomic_set(&sp->pd_mode, 0);
4094	skb_queue_head_init(&sp->pd_lobby);
4095	sp->frag_interleave = 0;
4096
4097	/* Create a per socket endpoint structure.  Even if we
4098	 * change the data structure relationships, this may still
4099	 * be useful for storing pre-connect address information.
4100	 */
4101	sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4102	if (!sp->ep)
4103		return -ENOMEM;
4104
4105	sp->hmac = NULL;
4106
4107	sk->sk_destruct = sctp_destruct_sock;
4108
4109	SCTP_DBG_OBJCNT_INC(sock);
4110
4111	local_bh_disable();
4112	percpu_counter_inc(&sctp_sockets_allocated);
4113	sock_prot_inuse_add(net, sk->sk_prot, 1);
4114
4115	/* Nothing can fail after this block, otherwise
4116	 * sctp_destroy_sock() will be called without addr_wq_lock held
4117	 */
4118	if (net->sctp.default_auto_asconf) {
4119		spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4120		list_add_tail(&sp->auto_asconf_list,
4121		    &net->sctp.auto_asconf_splist);
4122		sp->do_auto_asconf = 1;
4123		spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4124	} else {
4125		sp->do_auto_asconf = 0;
4126	}
4127
4128	local_bh_enable();
4129
4130	return 0;
4131}
4132
4133/* Cleanup any SCTP per socket resources. Must be called with
4134 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4135 */
4136static void sctp_destroy_sock(struct sock *sk)
4137{
4138	struct sctp_sock *sp;
4139
4140	pr_debug("%s: sk:%p\n", __func__, sk);
4141
4142	/* Release our hold on the endpoint. */
4143	sp = sctp_sk(sk);
4144	/* This could happen during socket init, thus we bail out
4145	 * early, since the rest of the below is not setup either.
4146	 */
4147	if (sp->ep == NULL)
4148		return;
4149
4150	if (sp->do_auto_asconf) {
4151		sp->do_auto_asconf = 0;
4152		list_del(&sp->auto_asconf_list);
4153	}
4154	sctp_endpoint_free(sp->ep);
4155	local_bh_disable();
4156	percpu_counter_dec(&sctp_sockets_allocated);
4157	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4158	local_bh_enable();
4159}
4160
4161/* Triggered when there are no references on the socket anymore */
4162static void sctp_destruct_sock(struct sock *sk)
4163{
4164	struct sctp_sock *sp = sctp_sk(sk);
4165
4166	/* Free up the HMAC transform. */
4167	crypto_free_hash(sp->hmac);
4168
4169	inet_sock_destruct(sk);
4170}
4171
4172/* API 4.1.7 shutdown() - TCP Style Syntax
4173 *     int shutdown(int socket, int how);
4174 *
4175 *     sd      - the socket descriptor of the association to be closed.
4176 *     how     - Specifies the type of shutdown.  The  values  are
4177 *               as follows:
4178 *               SHUT_RD
4179 *                     Disables further receive operations. No SCTP
4180 *                     protocol action is taken.
4181 *               SHUT_WR
4182 *                     Disables further send operations, and initiates
4183 *                     the SCTP shutdown sequence.
4184 *               SHUT_RDWR
4185 *                     Disables further send  and  receive  operations
4186 *                     and initiates the SCTP shutdown sequence.
4187 */
4188static void sctp_shutdown(struct sock *sk, int how)
4189{
4190	struct net *net = sock_net(sk);
4191	struct sctp_endpoint *ep;
4192	struct sctp_association *asoc;
4193
4194	if (!sctp_style(sk, TCP))
4195		return;
4196
4197	if (how & SEND_SHUTDOWN) {
4198		ep = sctp_sk(sk)->ep;
4199		if (!list_empty(&ep->asocs)) {
4200			asoc = list_entry(ep->asocs.next,
4201					  struct sctp_association, asocs);
4202			sctp_primitive_SHUTDOWN(net, asoc, NULL);
4203		}
4204	}
4205}
4206
4207/* 7.2.1 Association Status (SCTP_STATUS)
4208
4209 * Applications can retrieve current status information about an
4210 * association, including association state, peer receiver window size,
4211 * number of unacked data chunks, and number of data chunks pending
4212 * receipt.  This information is read-only.
4213 */
4214static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4215				       char __user *optval,
4216				       int __user *optlen)
4217{
4218	struct sctp_status status;
4219	struct sctp_association *asoc = NULL;
4220	struct sctp_transport *transport;
4221	sctp_assoc_t associd;
4222	int retval = 0;
4223
4224	if (len < sizeof(status)) {
4225		retval = -EINVAL;
4226		goto out;
4227	}
4228
4229	len = sizeof(status);
4230	if (copy_from_user(&status, optval, len)) {
4231		retval = -EFAULT;
4232		goto out;
4233	}
4234
4235	associd = status.sstat_assoc_id;
4236	asoc = sctp_id2assoc(sk, associd);
4237	if (!asoc) {
4238		retval = -EINVAL;
4239		goto out;
4240	}
4241
4242	transport = asoc->peer.primary_path;
4243
4244	status.sstat_assoc_id = sctp_assoc2id(asoc);
4245	status.sstat_state = sctp_assoc_to_state(asoc);
4246	status.sstat_rwnd =  asoc->peer.rwnd;
4247	status.sstat_unackdata = asoc->unack_data;
4248
4249	status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4250	status.sstat_instrms = asoc->c.sinit_max_instreams;
4251	status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4252	status.sstat_fragmentation_point = asoc->frag_point;
4253	status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4254	memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4255			transport->af_specific->sockaddr_len);
4256	/* Map ipv4 address into v4-mapped-on-v6 address.  */
4257	sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4258		(union sctp_addr *)&status.sstat_primary.spinfo_address);
4259	status.sstat_primary.spinfo_state = transport->state;
4260	status.sstat_primary.spinfo_cwnd = transport->cwnd;
4261	status.sstat_primary.spinfo_srtt = transport->srtt;
4262	status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4263	status.sstat_primary.spinfo_mtu = transport->pathmtu;
4264
4265	if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4266		status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4267
4268	if (put_user(len, optlen)) {
4269		retval = -EFAULT;
4270		goto out;
4271	}
4272
4273	pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4274		 __func__, len, status.sstat_state, status.sstat_rwnd,
4275		 status.sstat_assoc_id);
4276
4277	if (copy_to_user(optval, &status, len)) {
4278		retval = -EFAULT;
4279		goto out;
4280	}
4281
4282out:
4283	return retval;
4284}
4285
4286
4287/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4288 *
4289 * Applications can retrieve information about a specific peer address
4290 * of an association, including its reachability state, congestion
4291 * window, and retransmission timer values.  This information is
4292 * read-only.
4293 */
4294static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4295					  char __user *optval,
4296					  int __user *optlen)
4297{
4298	struct sctp_paddrinfo pinfo;
4299	struct sctp_transport *transport;
4300	int retval = 0;
4301
4302	if (len < sizeof(pinfo)) {
4303		retval = -EINVAL;
4304		goto out;
4305	}
4306
4307	len = sizeof(pinfo);
4308	if (copy_from_user(&pinfo, optval, len)) {
4309		retval = -EFAULT;
4310		goto out;
4311	}
4312
4313	transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4314					   pinfo.spinfo_assoc_id);
4315	if (!transport)
4316		return -EINVAL;
4317
4318	pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4319	pinfo.spinfo_state = transport->state;
4320	pinfo.spinfo_cwnd = transport->cwnd;
4321	pinfo.spinfo_srtt = transport->srtt;
4322	pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4323	pinfo.spinfo_mtu = transport->pathmtu;
4324
4325	if (pinfo.spinfo_state == SCTP_UNKNOWN)
4326		pinfo.spinfo_state = SCTP_ACTIVE;
4327
4328	if (put_user(len, optlen)) {
4329		retval = -EFAULT;
4330		goto out;
4331	}
4332
4333	if (copy_to_user(optval, &pinfo, len)) {
4334		retval = -EFAULT;
4335		goto out;
4336	}
4337
4338out:
4339	return retval;
4340}
4341
4342/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4343 *
4344 * This option is a on/off flag.  If enabled no SCTP message
4345 * fragmentation will be performed.  Instead if a message being sent
4346 * exceeds the current PMTU size, the message will NOT be sent and
4347 * instead a error will be indicated to the user.
4348 */
4349static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4350					char __user *optval, int __user *optlen)
4351{
4352	int val;
4353
4354	if (len < sizeof(int))
4355		return -EINVAL;
4356
4357	len = sizeof(int);
4358	val = (sctp_sk(sk)->disable_fragments == 1);
4359	if (put_user(len, optlen))
4360		return -EFAULT;
4361	if (copy_to_user(optval, &val, len))
4362		return -EFAULT;
4363	return 0;
4364}
4365
4366/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4367 *
4368 * This socket option is used to specify various notifications and
4369 * ancillary data the user wishes to receive.
4370 */
4371static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4372				  int __user *optlen)
4373{
4374	if (len <= 0)
4375		return -EINVAL;
4376	if (len > sizeof(struct sctp_event_subscribe))
4377		len = sizeof(struct sctp_event_subscribe);
4378	if (put_user(len, optlen))
4379		return -EFAULT;
4380	if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4381		return -EFAULT;
4382	return 0;
4383}
4384
4385/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4386 *
4387 * This socket option is applicable to the UDP-style socket only.  When
4388 * set it will cause associations that are idle for more than the
4389 * specified number of seconds to automatically close.  An association
4390 * being idle is defined an association that has NOT sent or received
4391 * user data.  The special value of '0' indicates that no automatic
4392 * close of any associations should be performed.  The option expects an
4393 * integer defining the number of seconds of idle time before an
4394 * association is closed.
4395 */
4396static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4397{
4398	/* Applicable to UDP-style socket only */
4399	if (sctp_style(sk, TCP))
4400		return -EOPNOTSUPP;
4401	if (len < sizeof(int))
4402		return -EINVAL;
4403	len = sizeof(int);
4404	if (put_user(len, optlen))
4405		return -EFAULT;
4406	if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4407		return -EFAULT;
4408	return 0;
4409}
4410
4411/* Helper routine to branch off an association to a new socket.  */
4412int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4413{
4414	struct sctp_association *asoc = sctp_id2assoc(sk, id);
4415	struct sctp_sock *sp = sctp_sk(sk);
4416	struct socket *sock;
4417	int err = 0;
4418
4419	if (!asoc)
4420		return -EINVAL;
4421
4422	/* An association cannot be branched off from an already peeled-off
4423	 * socket, nor is this supported for tcp style sockets.
4424	 */
4425	if (!sctp_style(sk, UDP))
4426		return -EINVAL;
4427
4428	/* Create a new socket.  */
4429	err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4430	if (err < 0)
4431		return err;
4432
4433	sctp_copy_sock(sock->sk, sk, asoc);
4434
4435	/* Make peeled-off sockets more like 1-1 accepted sockets.
4436	 * Set the daddr and initialize id to something more random
4437	 */
4438	sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4439
4440	/* Populate the fields of the newsk from the oldsk and migrate the
4441	 * asoc to the newsk.
4442	 */
4443	sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4444
4445	*sockp = sock;
4446
4447	return err;
4448}
4449EXPORT_SYMBOL(sctp_do_peeloff);
4450
4451static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4452{
4453	sctp_peeloff_arg_t peeloff;
4454	struct socket *newsock;
4455	struct file *newfile;
4456	int retval = 0;
4457
4458	if (len < sizeof(sctp_peeloff_arg_t))
4459		return -EINVAL;
4460	len = sizeof(sctp_peeloff_arg_t);
4461	if (copy_from_user(&peeloff, optval, len))
4462		return -EFAULT;
4463
4464	retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4465	if (retval < 0)
4466		goto out;
4467
4468	/* Map the socket to an unused fd that can be returned to the user.  */
4469	retval = get_unused_fd_flags(0);
4470	if (retval < 0) {
4471		sock_release(newsock);
4472		goto out;
4473	}
4474
4475	newfile = sock_alloc_file(newsock, 0, NULL);
4476	if (IS_ERR(newfile)) {
4477		put_unused_fd(retval);
4478		sock_release(newsock);
4479		return PTR_ERR(newfile);
4480	}
4481
4482	pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4483		 retval);
4484
4485	/* Return the fd mapped to the new socket.  */
4486	if (put_user(len, optlen)) {
4487		fput(newfile);
4488		put_unused_fd(retval);
4489		return -EFAULT;
4490	}
4491	peeloff.sd = retval;
4492	if (copy_to_user(optval, &peeloff, len)) {
4493		fput(newfile);
4494		put_unused_fd(retval);
4495		return -EFAULT;
4496	}
4497	fd_install(retval, newfile);
4498out:
4499	return retval;
4500}
4501
4502/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4503 *
4504 * Applications can enable or disable heartbeats for any peer address of
4505 * an association, modify an address's heartbeat interval, force a
4506 * heartbeat to be sent immediately, and adjust the address's maximum
4507 * number of retransmissions sent before an address is considered
4508 * unreachable.  The following structure is used to access and modify an
4509 * address's parameters:
4510 *
4511 *  struct sctp_paddrparams {
4512 *     sctp_assoc_t            spp_assoc_id;
4513 *     struct sockaddr_storage spp_address;
4514 *     uint32_t                spp_hbinterval;
4515 *     uint16_t                spp_pathmaxrxt;
4516 *     uint32_t                spp_pathmtu;
4517 *     uint32_t                spp_sackdelay;
4518 *     uint32_t                spp_flags;
4519 * };
4520 *
4521 *   spp_assoc_id    - (one-to-many style socket) This is filled in the
4522 *                     application, and identifies the association for
4523 *                     this query.
4524 *   spp_address     - This specifies which address is of interest.
4525 *   spp_hbinterval  - This contains the value of the heartbeat interval,
4526 *                     in milliseconds.  If a  value of zero
4527 *                     is present in this field then no changes are to
4528 *                     be made to this parameter.
4529 *   spp_pathmaxrxt  - This contains the maximum number of
4530 *                     retransmissions before this address shall be
4531 *                     considered unreachable. If a  value of zero
4532 *                     is present in this field then no changes are to
4533 *                     be made to this parameter.
4534 *   spp_pathmtu     - When Path MTU discovery is disabled the value
4535 *                     specified here will be the "fixed" path mtu.
4536 *                     Note that if the spp_address field is empty
4537 *                     then all associations on this address will
4538 *                     have this fixed path mtu set upon them.
4539 *
4540 *   spp_sackdelay   - When delayed sack is enabled, this value specifies
4541 *                     the number of milliseconds that sacks will be delayed
4542 *                     for. This value will apply to all addresses of an
4543 *                     association if the spp_address field is empty. Note
4544 *                     also, that if delayed sack is enabled and this
4545 *                     value is set to 0, no change is made to the last
4546 *                     recorded delayed sack timer value.
4547 *
4548 *   spp_flags       - These flags are used to control various features
4549 *                     on an association. The flag field may contain
4550 *                     zero or more of the following options.
4551 *
4552 *                     SPP_HB_ENABLE  - Enable heartbeats on the
4553 *                     specified address. Note that if the address
4554 *                     field is empty all addresses for the association
4555 *                     have heartbeats enabled upon them.
4556 *
4557 *                     SPP_HB_DISABLE - Disable heartbeats on the
4558 *                     speicifed address. Note that if the address
4559 *                     field is empty all addresses for the association
4560 *                     will have their heartbeats disabled. Note also
4561 *                     that SPP_HB_ENABLE and SPP_HB_DISABLE are
4562 *                     mutually exclusive, only one of these two should
4563 *                     be specified. Enabling both fields will have
4564 *                     undetermined results.
4565 *
4566 *                     SPP_HB_DEMAND - Request a user initiated heartbeat
4567 *                     to be made immediately.
4568 *
4569 *                     SPP_PMTUD_ENABLE - This field will enable PMTU
4570 *                     discovery upon the specified address. Note that
4571 *                     if the address feild is empty then all addresses
4572 *                     on the association are effected.
4573 *
4574 *                     SPP_PMTUD_DISABLE - This field will disable PMTU
4575 *                     discovery upon the specified address. Note that
4576 *                     if the address feild is empty then all addresses
4577 *                     on the association are effected. Not also that
4578 *                     SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4579 *                     exclusive. Enabling both will have undetermined
4580 *                     results.
4581 *
4582 *                     SPP_SACKDELAY_ENABLE - Setting this flag turns
4583 *                     on delayed sack. The time specified in spp_sackdelay
4584 *                     is used to specify the sack delay for this address. Note
4585 *                     that if spp_address is empty then all addresses will
4586 *                     enable delayed sack and take on the sack delay
4587 *                     value specified in spp_sackdelay.
4588 *                     SPP_SACKDELAY_DISABLE - Setting this flag turns
4589 *                     off delayed sack. If the spp_address field is blank then
4590 *                     delayed sack is disabled for the entire association. Note
4591 *                     also that this field is mutually exclusive to
4592 *                     SPP_SACKDELAY_ENABLE, setting both will have undefined
4593 *                     results.
4594 */
4595static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4596					    char __user *optval, int __user *optlen)
4597{
4598	struct sctp_paddrparams  params;
4599	struct sctp_transport   *trans = NULL;
4600	struct sctp_association *asoc = NULL;
4601	struct sctp_sock        *sp = sctp_sk(sk);
4602
4603	if (len < sizeof(struct sctp_paddrparams))
4604		return -EINVAL;
4605	len = sizeof(struct sctp_paddrparams);
4606	if (copy_from_user(&params, optval, len))
4607		return -EFAULT;
4608
4609	/* If an address other than INADDR_ANY is specified, and
4610	 * no transport is found, then the request is invalid.
4611	 */
4612	if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4613		trans = sctp_addr_id2transport(sk, &params.spp_address,
4614					       params.spp_assoc_id);
4615		if (!trans) {
4616			pr_debug("%s: failed no transport\n", __func__);
4617			return -EINVAL;
4618		}
4619	}
4620
4621	/* Get association, if assoc_id != 0 and the socket is a one
4622	 * to many style socket, and an association was not found, then
4623	 * the id was invalid.
4624	 */
4625	asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4626	if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4627		pr_debug("%s: failed no association\n", __func__);
4628		return -EINVAL;
4629	}
4630
4631	if (trans) {
4632		/* Fetch transport values. */
4633		params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4634		params.spp_pathmtu    = trans->pathmtu;
4635		params.spp_pathmaxrxt = trans->pathmaxrxt;
4636		params.spp_sackdelay  = jiffies_to_msecs(trans->sackdelay);
4637
4638		/*draft-11 doesn't say what to return in spp_flags*/
4639		params.spp_flags      = trans->param_flags;
4640	} else if (asoc) {
4641		/* Fetch association values. */
4642		params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4643		params.spp_pathmtu    = asoc->pathmtu;
4644		params.spp_pathmaxrxt = asoc->pathmaxrxt;
4645		params.spp_sackdelay  = jiffies_to_msecs(asoc->sackdelay);
4646
4647		/*draft-11 doesn't say what to return in spp_flags*/
4648		params.spp_flags      = asoc->param_flags;
4649	} else {
4650		/* Fetch socket values. */
4651		params.spp_hbinterval = sp->hbinterval;
4652		params.spp_pathmtu    = sp->pathmtu;
4653		params.spp_sackdelay  = sp->sackdelay;
4654		params.spp_pathmaxrxt = sp->pathmaxrxt;
4655
4656		/*draft-11 doesn't say what to return in spp_flags*/
4657		params.spp_flags      = sp->param_flags;
4658	}
4659
4660	if (copy_to_user(optval, &params, len))
4661		return -EFAULT;
4662
4663	if (put_user(len, optlen))
4664		return -EFAULT;
4665
4666	return 0;
4667}
4668
4669/*
4670 * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
4671 *
4672 * This option will effect the way delayed acks are performed.  This
4673 * option allows you to get or set the delayed ack time, in
4674 * milliseconds.  It also allows changing the delayed ack frequency.
4675 * Changing the frequency to 1 disables the delayed sack algorithm.  If
4676 * the assoc_id is 0, then this sets or gets the endpoints default
4677 * values.  If the assoc_id field is non-zero, then the set or get
4678 * effects the specified association for the one to many model (the
4679 * assoc_id field is ignored by the one to one model).  Note that if
4680 * sack_delay or sack_freq are 0 when setting this option, then the
4681 * current values will remain unchanged.
4682 *
4683 * struct sctp_sack_info {
4684 *     sctp_assoc_t            sack_assoc_id;
4685 *     uint32_t                sack_delay;
4686 *     uint32_t                sack_freq;
4687 * };
4688 *
4689 * sack_assoc_id -  This parameter, indicates which association the user
4690 *    is performing an action upon.  Note that if this field's value is
4691 *    zero then the endpoints default value is changed (effecting future
4692 *    associations only).
4693 *
4694 * sack_delay -  This parameter contains the number of milliseconds that
4695 *    the user is requesting the delayed ACK timer be set to.  Note that
4696 *    this value is defined in the standard to be between 200 and 500
4697 *    milliseconds.
4698 *
4699 * sack_freq -  This parameter contains the number of packets that must
4700 *    be received before a sack is sent without waiting for the delay
4701 *    timer to expire.  The default value for this is 2, setting this
4702 *    value to 1 will disable the delayed sack algorithm.
4703 */
4704static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4705					    char __user *optval,
4706					    int __user *optlen)
4707{
4708	struct sctp_sack_info    params;
4709	struct sctp_association *asoc = NULL;
4710	struct sctp_sock        *sp = sctp_sk(sk);
4711
4712	if (len >= sizeof(struct sctp_sack_info)) {
4713		len = sizeof(struct sctp_sack_info);
4714
4715		if (copy_from_user(&params, optval, len))
4716			return -EFAULT;
4717	} else if (len == sizeof(struct sctp_assoc_value)) {
4718		pr_warn_ratelimited(DEPRECATED
4719				    "%s (pid %d) "
4720				    "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4721				    "Use struct sctp_sack_info instead\n",
4722				    current->comm, task_pid_nr(current));
4723		if (copy_from_user(&params, optval, len))
4724			return -EFAULT;
4725	} else
4726		return -EINVAL;
4727
4728	/* Get association, if sack_assoc_id != 0 and the socket is a one
4729	 * to many style socket, and an association was not found, then
4730	 * the id was invalid.
4731	 */
4732	asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4733	if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4734		return -EINVAL;
4735
4736	if (asoc) {
4737		/* Fetch association values. */
4738		if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4739			params.sack_delay = jiffies_to_msecs(
4740				asoc->sackdelay);
4741			params.sack_freq = asoc->sackfreq;
4742
4743		} else {
4744			params.sack_delay = 0;
4745			params.sack_freq = 1;
4746		}
4747	} else {
4748		/* Fetch socket values. */
4749		if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4750			params.sack_delay  = sp->sackdelay;
4751			params.sack_freq = sp->sackfreq;
4752		} else {
4753			params.sack_delay  = 0;
4754			params.sack_freq = 1;
4755		}
4756	}
4757
4758	if (copy_to_user(optval, &params, len))
4759		return -EFAULT;
4760
4761	if (put_user(len, optlen))
4762		return -EFAULT;
4763
4764	return 0;
4765}
4766
4767/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4768 *
4769 * Applications can specify protocol parameters for the default association
4770 * initialization.  The option name argument to setsockopt() and getsockopt()
4771 * is SCTP_INITMSG.
4772 *
4773 * Setting initialization parameters is effective only on an unconnected
4774 * socket (for UDP-style sockets only future associations are effected
4775 * by the change).  With TCP-style sockets, this option is inherited by
4776 * sockets derived from a listener socket.
4777 */
4778static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4779{
4780	if (len < sizeof(struct sctp_initmsg))
4781		return -EINVAL;
4782	len = sizeof(struct sctp_initmsg);
4783	if (put_user(len, optlen))
4784		return -EFAULT;
4785	if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4786		return -EFAULT;
4787	return 0;
4788}
4789
4790
4791static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4792				      char __user *optval, int __user *optlen)
4793{
4794	struct sctp_association *asoc;
4795	int cnt = 0;
4796	struct sctp_getaddrs getaddrs;
4797	struct sctp_transport *from;
4798	void __user *to;
4799	union sctp_addr temp;
4800	struct sctp_sock *sp = sctp_sk(sk);
4801	int addrlen;
4802	size_t space_left;
4803	int bytes_copied;
4804
4805	if (len < sizeof(struct sctp_getaddrs))
4806		return -EINVAL;
4807
4808	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4809		return -EFAULT;
4810
4811	/* For UDP-style sockets, id specifies the association to query.  */
4812	asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4813	if (!asoc)
4814		return -EINVAL;
4815
4816	to = optval + offsetof(struct sctp_getaddrs, addrs);
4817	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4818
4819	list_for_each_entry(from, &asoc->peer.transport_addr_list,
4820				transports) {
4821		memcpy(&temp, &from->ipaddr, sizeof(temp));
4822		addrlen = sctp_get_pf_specific(sk->sk_family)
4823			      ->addr_to_user(sp, &temp);
4824		if (space_left < addrlen)
4825			return -ENOMEM;
4826		if (copy_to_user(to, &temp, addrlen))
4827			return -EFAULT;
4828		to += addrlen;
4829		cnt++;
4830		space_left -= addrlen;
4831	}
4832
4833	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4834		return -EFAULT;
4835	bytes_copied = ((char __user *)to) - optval;
4836	if (put_user(bytes_copied, optlen))
4837		return -EFAULT;
4838
4839	return 0;
4840}
4841
4842static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4843			    size_t space_left, int *bytes_copied)
4844{
4845	struct sctp_sockaddr_entry *addr;
4846	union sctp_addr temp;
4847	int cnt = 0;
4848	int addrlen;
4849	struct net *net = sock_net(sk);
4850
4851	rcu_read_lock();
4852	list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4853		if (!addr->valid)
4854			continue;
4855
4856		if ((PF_INET == sk->sk_family) &&
4857		    (AF_INET6 == addr->a.sa.sa_family))
4858			continue;
4859		if ((PF_INET6 == sk->sk_family) &&
4860		    inet_v6_ipv6only(sk) &&
4861		    (AF_INET == addr->a.sa.sa_family))
4862			continue;
4863		memcpy(&temp, &addr->a, sizeof(temp));
4864		if (!temp.v4.sin_port)
4865			temp.v4.sin_port = htons(port);
4866
4867		addrlen = sctp_get_pf_specific(sk->sk_family)
4868			      ->addr_to_user(sctp_sk(sk), &temp);
4869
4870		if (space_left < addrlen) {
4871			cnt =  -ENOMEM;
4872			break;
4873		}
4874		memcpy(to, &temp, addrlen);
4875
4876		to += addrlen;
4877		cnt++;
4878		space_left -= addrlen;
4879		*bytes_copied += addrlen;
4880	}
4881	rcu_read_unlock();
4882
4883	return cnt;
4884}
4885
4886
4887static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4888				       char __user *optval, int __user *optlen)
4889{
4890	struct sctp_bind_addr *bp;
4891	struct sctp_association *asoc;
4892	int cnt = 0;
4893	struct sctp_getaddrs getaddrs;
4894	struct sctp_sockaddr_entry *addr;
4895	void __user *to;
4896	union sctp_addr temp;
4897	struct sctp_sock *sp = sctp_sk(sk);
4898	int addrlen;
4899	int err = 0;
4900	size_t space_left;
4901	int bytes_copied = 0;
4902	void *addrs;
4903	void *buf;
4904
4905	if (len < sizeof(struct sctp_getaddrs))
4906		return -EINVAL;
4907
4908	if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4909		return -EFAULT;
4910
4911	/*
4912	 *  For UDP-style sockets, id specifies the association to query.
4913	 *  If the id field is set to the value '0' then the locally bound
4914	 *  addresses are returned without regard to any particular
4915	 *  association.
4916	 */
4917	if (0 == getaddrs.assoc_id) {
4918		bp = &sctp_sk(sk)->ep->base.bind_addr;
4919	} else {
4920		asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4921		if (!asoc)
4922			return -EINVAL;
4923		bp = &asoc->base.bind_addr;
4924	}
4925
4926	to = optval + offsetof(struct sctp_getaddrs, addrs);
4927	space_left = len - offsetof(struct sctp_getaddrs, addrs);
4928
4929	addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
4930	if (!addrs)
4931		return -ENOMEM;
4932
4933	/* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4934	 * addresses from the global local address list.
4935	 */
4936	if (sctp_list_single_entry(&bp->address_list)) {
4937		addr = list_entry(bp->address_list.next,
4938				  struct sctp_sockaddr_entry, list);
4939		if (sctp_is_any(sk, &addr->a)) {
4940			cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4941						space_left, &bytes_copied);
4942			if (cnt < 0) {
4943				err = cnt;
4944				goto out;
4945			}
4946			goto copy_getaddrs;
4947		}
4948	}
4949
4950	buf = addrs;
4951	/* Protection on the bound address list is not needed since
4952	 * in the socket option context we hold a socket lock and
4953	 * thus the bound address list can't change.
4954	 */
4955	list_for_each_entry(addr, &bp->address_list, list) {
4956		memcpy(&temp, &addr->a, sizeof(temp));
4957		addrlen = sctp_get_pf_specific(sk->sk_family)
4958			      ->addr_to_user(sp, &temp);
4959		if (space_left < addrlen) {
4960			err =  -ENOMEM; /*fixme: right error?*/
4961			goto out;
4962		}
4963		memcpy(buf, &temp, addrlen);
4964		buf += addrlen;
4965		bytes_copied += addrlen;
4966		cnt++;
4967		space_left -= addrlen;
4968	}
4969
4970copy_getaddrs:
4971	if (copy_to_user(to, addrs, bytes_copied)) {
4972		err = -EFAULT;
4973		goto out;
4974	}
4975	if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4976		err = -EFAULT;
4977		goto out;
4978	}
4979	if (put_user(bytes_copied, optlen))
4980		err = -EFAULT;
4981out:
4982	kfree(addrs);
4983	return err;
4984}
4985
4986/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4987 *
4988 * Requests that the local SCTP stack use the enclosed peer address as
4989 * the association primary.  The enclosed address must be one of the
4990 * association peer's addresses.
4991 */
4992static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4993					char __user *optval, int __user *optlen)
4994{
4995	struct sctp_prim prim;
4996	struct sctp_association *asoc;
4997	struct sctp_sock *sp = sctp_sk(sk);
4998
4999	if (len < sizeof(struct sctp_prim))
5000		return -EINVAL;
5001
5002	len = sizeof(struct sctp_prim);
5003
5004	if (copy_from_user(&prim, optval, len))
5005		return -EFAULT;
5006
5007	asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5008	if (!asoc)
5009		return -EINVAL;
5010
5011	if (!asoc->peer.primary_path)
5012		return -ENOTCONN;
5013
5014	memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5015		asoc->peer.primary_path->af_specific->sockaddr_len);
5016
5017	sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5018			(union sctp_addr *)&prim.ssp_addr);
5019
5020	if (put_user(len, optlen))
5021		return -EFAULT;
5022	if (copy_to_user(optval, &prim, len))
5023		return -EFAULT;
5024
5025	return 0;
5026}
5027
5028/*
5029 * 7.1.11  Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5030 *
5031 * Requests that the local endpoint set the specified Adaptation Layer
5032 * Indication parameter for all future INIT and INIT-ACK exchanges.
5033 */
5034static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5035				  char __user *optval, int __user *optlen)
5036{
5037	struct sctp_setadaptation adaptation;
5038
5039	if (len < sizeof(struct sctp_setadaptation))
5040		return -EINVAL;
5041
5042	len = sizeof(struct sctp_setadaptation);
5043
5044	adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5045
5046	if (put_user(len, optlen))
5047		return -EFAULT;
5048	if (copy_to_user(optval, &adaptation, len))
5049		return -EFAULT;
5050
5051	return 0;
5052}
5053
5054/*
5055 *
5056 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5057 *
5058 *   Applications that wish to use the sendto() system call may wish to
5059 *   specify a default set of parameters that would normally be supplied
5060 *   through the inclusion of ancillary data.  This socket option allows
5061 *   such an application to set the default sctp_sndrcvinfo structure.
5062
5063
5064 *   The application that wishes to use this socket option simply passes
5065 *   in to this call the sctp_sndrcvinfo structure defined in Section
5066 *   5.2.2) The input parameters accepted by this call include
5067 *   sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5068 *   sinfo_timetolive.  The user must provide the sinfo_assoc_id field in
5069 *   to this call if the caller is using the UDP model.
5070 *
5071 *   For getsockopt, it get the default sctp_sndrcvinfo structure.
5072 */
5073static int sctp_getsockopt_default_send_param(struct sock *sk,
5074					int len, char __user *optval,
5075					int __user *optlen)
5076{
5077	struct sctp_sock *sp = sctp_sk(sk);
5078	struct sctp_association *asoc;
5079	struct sctp_sndrcvinfo info;
5080
5081	if (len < sizeof(info))
5082		return -EINVAL;
5083
5084	len = sizeof(info);
5085
5086	if (copy_from_user(&info, optval, len))
5087		return -EFAULT;
5088
5089	asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5090	if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5091		return -EINVAL;
5092	if (asoc) {
5093		info.sinfo_stream = asoc->default_stream;
5094		info.sinfo_flags = asoc->default_flags;
5095		info.sinfo_ppid = asoc->default_ppid;
5096		info.sinfo_context = asoc->default_context;
5097		info.sinfo_timetolive = asoc->default_timetolive;
5098	} else {
5099		info.sinfo_stream = sp->default_stream;
5100		info.sinfo_flags = sp->default_flags;
5101		info.sinfo_ppid = sp->default_ppid;
5102		info.sinfo_context = sp->default_context;
5103		info.sinfo_timetolive = sp->default_timetolive;
5104	}
5105
5106	if (put_user(len, optlen))
5107		return -EFAULT;
5108	if (copy_to_user(optval, &info, len))
5109		return -EFAULT;
5110
5111	return 0;
5112}
5113
5114/* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5115 * (SCTP_DEFAULT_SNDINFO)
5116 */
5117static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5118					   char __user *optval,
5119					   int __user *optlen)
5120{
5121	struct sctp_sock *sp = sctp_sk(sk);
5122	struct sctp_association *asoc;
5123	struct sctp_sndinfo info;
5124
5125	if (len < sizeof(info))
5126		return -EINVAL;
5127
5128	len = sizeof(info);
5129
5130	if (copy_from_user(&info, optval, len))
5131		return -EFAULT;
5132
5133	asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5134	if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5135		return -EINVAL;
5136	if (asoc) {
5137		info.snd_sid = asoc->default_stream;
5138		info.snd_flags = asoc->default_flags;
5139		info.snd_ppid = asoc->default_ppid;
5140		info.snd_context = asoc->default_context;
5141	} else {
5142		info.snd_sid = sp->default_stream;
5143		info.snd_flags = sp->default_flags;
5144		info.snd_ppid = sp->default_ppid;
5145		info.snd_context = sp->default_context;
5146	}
5147
5148	if (put_user(len, optlen))
5149		return -EFAULT;
5150	if (copy_to_user(optval, &info, len))
5151		return -EFAULT;
5152
5153	return 0;
5154}
5155
5156/*
5157 *
5158 * 7.1.5 SCTP_NODELAY
5159 *
5160 * Turn on/off any Nagle-like algorithm.  This means that packets are
5161 * generally sent as soon as possible and no unnecessary delays are
5162 * introduced, at the cost of more packets in the network.  Expects an
5163 * integer boolean flag.
5164 */
5165
5166static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5167				   char __user *optval, int __user *optlen)
5168{
5169	int val;
5170
5171	if (len < sizeof(int))
5172		return -EINVAL;
5173
5174	len = sizeof(int);
5175	val = (sctp_sk(sk)->nodelay == 1);
5176	if (put_user(len, optlen))
5177		return -EFAULT;
5178	if (copy_to_user(optval, &val, len))
5179		return -EFAULT;
5180	return 0;
5181}
5182
5183/*
5184 *
5185 * 7.1.1 SCTP_RTOINFO
5186 *
5187 * The protocol parameters used to initialize and bound retransmission
5188 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5189 * and modify these parameters.
5190 * All parameters are time values, in milliseconds.  A value of 0, when
5191 * modifying the parameters, indicates that the current value should not
5192 * be changed.
5193 *
5194 */
5195static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5196				char __user *optval,
5197				int __user *optlen) {
5198	struct sctp_rtoinfo rtoinfo;
5199	struct sctp_association *asoc;
5200
5201	if (len < sizeof (struct sctp_rtoinfo))
5202		return -EINVAL;
5203
5204	len = sizeof(struct sctp_rtoinfo);
5205
5206	if (copy_from_user(&rtoinfo, optval, len))
5207		return -EFAULT;
5208
5209	asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5210
5211	if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5212		return -EINVAL;
5213
5214	/* Values corresponding to the specific association. */
5215	if (asoc) {
5216		rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5217		rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5218		rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5219	} else {
5220		/* Values corresponding to the endpoint. */
5221		struct sctp_sock *sp = sctp_sk(sk);
5222
5223		rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5224		rtoinfo.srto_max = sp->rtoinfo.srto_max;
5225		rtoinfo.srto_min = sp->rtoinfo.srto_min;
5226	}
5227
5228	if (put_user(len, optlen))
5229		return -EFAULT;
5230
5231	if (copy_to_user(optval, &rtoinfo, len))
5232		return -EFAULT;
5233
5234	return 0;
5235}
5236
5237/*
5238 *
5239 * 7.1.2 SCTP_ASSOCINFO
5240 *
5241 * This option is used to tune the maximum retransmission attempts
5242 * of the association.
5243 * Returns an error if the new association retransmission value is
5244 * greater than the sum of the retransmission value  of the peer.
5245 * See [SCTP] for more information.
5246 *
5247 */
5248static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5249				     char __user *optval,
5250				     int __user *optlen)
5251{
5252
5253	struct sctp_assocparams assocparams;
5254	struct sctp_association *asoc;
5255	struct list_head *pos;
5256	int cnt = 0;
5257
5258	if (len < sizeof (struct sctp_assocparams))
5259		return -EINVAL;
5260
5261	len = sizeof(struct sctp_assocparams);
5262
5263	if (copy_from_user(&assocparams, optval, len))
5264		return -EFAULT;
5265
5266	asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5267
5268	if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5269		return -EINVAL;
5270
5271	/* Values correspoinding to the specific association */
5272	if (asoc) {
5273		assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5274		assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5275		assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5276		assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5277
5278		list_for_each(pos, &asoc->peer.transport_addr_list) {
5279			cnt++;
5280		}
5281
5282		assocparams.sasoc_number_peer_destinations = cnt;
5283	} else {
5284		/* Values corresponding to the endpoint */
5285		struct sctp_sock *sp = sctp_sk(sk);
5286
5287		assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5288		assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5289		assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5290		assocparams.sasoc_cookie_life =
5291					sp->assocparams.sasoc_cookie_life;
5292		assocparams.sasoc_number_peer_destinations =
5293					sp->assocparams.
5294					sasoc_number_peer_destinations;
5295	}
5296
5297	if (put_user(len, optlen))
5298		return -EFAULT;
5299
5300	if (copy_to_user(optval, &assocparams, len))
5301		return -EFAULT;
5302
5303	return 0;
5304}
5305
5306/*
5307 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5308 *
5309 * This socket option is a boolean flag which turns on or off mapped V4
5310 * addresses.  If this option is turned on and the socket is type
5311 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5312 * If this option is turned off, then no mapping will be done of V4
5313 * addresses and a user will receive both PF_INET6 and PF_INET type
5314 * addresses on the socket.
5315 */
5316static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5317				    char __user *optval, int __user *optlen)
5318{
5319	int val;
5320	struct sctp_sock *sp = sctp_sk(sk);
5321
5322	if (len < sizeof(int))
5323		return -EINVAL;
5324
5325	len = sizeof(int);
5326	val = sp->v4mapped;
5327	if (put_user(len, optlen))
5328		return -EFAULT;
5329	if (copy_to_user(optval, &val, len))
5330		return -EFAULT;
5331
5332	return 0;
5333}
5334
5335/*
5336 * 7.1.29.  Set or Get the default context (SCTP_CONTEXT)
5337 * (chapter and verse is quoted at sctp_setsockopt_context())
5338 */
5339static int sctp_getsockopt_context(struct sock *sk, int len,
5340				   char __user *optval, int __user *optlen)
5341{
5342	struct sctp_assoc_value params;
5343	struct sctp_sock *sp;
5344	struct sctp_association *asoc;
5345
5346	if (len < sizeof(struct sctp_assoc_value))
5347		return -EINVAL;
5348
5349	len = sizeof(struct sctp_assoc_value);
5350
5351	if (copy_from_user(&params, optval, len))
5352		return -EFAULT;
5353
5354	sp = sctp_sk(sk);
5355
5356	if (params.assoc_id != 0) {
5357		asoc = sctp_id2assoc(sk, params.assoc_id);
5358		if (!asoc)
5359			return -EINVAL;
5360		params.assoc_value = asoc->default_rcv_context;
5361	} else {
5362		params.assoc_value = sp->default_rcv_context;
5363	}
5364
5365	if (put_user(len, optlen))
5366		return -EFAULT;
5367	if (copy_to_user(optval, &params, len))
5368		return -EFAULT;
5369
5370	return 0;
5371}
5372
5373/*
5374 * 8.1.16.  Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5375 * This option will get or set the maximum size to put in any outgoing
5376 * SCTP DATA chunk.  If a message is larger than this size it will be
5377 * fragmented by SCTP into the specified size.  Note that the underlying
5378 * SCTP implementation may fragment into smaller sized chunks when the
5379 * PMTU of the underlying association is smaller than the value set by
5380 * the user.  The default value for this option is '0' which indicates
5381 * the user is NOT limiting fragmentation and only the PMTU will effect
5382 * SCTP's choice of DATA chunk size.  Note also that values set larger
5383 * than the maximum size of an IP datagram will effectively let SCTP
5384 * control fragmentation (i.e. the same as setting this option to 0).
5385 *
5386 * The following structure is used to access and modify this parameter:
5387 *
5388 * struct sctp_assoc_value {
5389 *   sctp_assoc_t assoc_id;
5390 *   uint32_t assoc_value;
5391 * };
5392 *
5393 * assoc_id:  This parameter is ignored for one-to-one style sockets.
5394 *    For one-to-many style sockets this parameter indicates which
5395 *    association the user is performing an action upon.  Note that if
5396 *    this field's value is zero then the endpoints default value is
5397 *    changed (effecting future associations only).
5398 * assoc_value:  This parameter specifies the maximum size in bytes.
5399 */
5400static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5401				  char __user *optval, int __user *optlen)
5402{
5403	struct sctp_assoc_value params;
5404	struct sctp_association *asoc;
5405
5406	if (len == sizeof(int)) {
5407		pr_warn_ratelimited(DEPRECATED
5408				    "%s (pid %d) "
5409				    "Use of int in maxseg socket option.\n"
5410				    "Use struct sctp_assoc_value instead\n",
5411				    current->comm, task_pid_nr(current));
5412		params.assoc_id = 0;
5413	} else if (len >= sizeof(struct sctp_assoc_value)) {
5414		len = sizeof(struct sctp_assoc_value);
5415		if (copy_from_user(&params, optval, sizeof(params)))
5416			return -EFAULT;
5417	} else
5418		return -EINVAL;
5419
5420	asoc = sctp_id2assoc(sk, params.assoc_id);
5421	if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5422		return -EINVAL;
5423
5424	if (asoc)
5425		params.assoc_value = asoc->frag_point;
5426	else
5427		params.assoc_value = sctp_sk(sk)->user_frag;
5428
5429	if (put_user(len, optlen))
5430		return -EFAULT;
5431	if (len == sizeof(int)) {
5432		if (copy_to_user(optval, &params.assoc_value, len))
5433			return -EFAULT;
5434	} else {
5435		if (copy_to_user(optval, &params, len))
5436			return -EFAULT;
5437	}
5438
5439	return 0;
5440}
5441
5442/*
5443 * 7.1.24.  Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5444 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5445 */
5446static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5447					       char __user *optval, int __user *optlen)
5448{
5449	int val;
5450
5451	if (len < sizeof(int))
5452		return -EINVAL;
5453
5454	len = sizeof(int);
5455
5456	val = sctp_sk(sk)->frag_interleave;
5457	if (put_user(len, optlen))
5458		return -EFAULT;
5459	if (copy_to_user(optval, &val, len))
5460		return -EFAULT;
5461
5462	return 0;
5463}
5464
5465/*
5466 * 7.1.25.  Set or Get the sctp partial delivery point
5467 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5468 */
5469static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5470						  char __user *optval,
5471						  int __user *optlen)
5472{
5473	u32 val;
5474
5475	if (len < sizeof(u32))
5476		return -EINVAL;
5477
5478	len = sizeof(u32);
5479
5480	val = sctp_sk(sk)->pd_point;
5481	if (put_user(len, optlen))
5482		return -EFAULT;
5483	if (copy_to_user(optval, &val, len))
5484		return -EFAULT;
5485
5486	return 0;
5487}
5488
5489/*
5490 * 7.1.28.  Set or Get the maximum burst (SCTP_MAX_BURST)
5491 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5492 */
5493static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5494				    char __user *optval,
5495				    int __user *optlen)
5496{
5497	struct sctp_assoc_value params;
5498	struct sctp_sock *sp;
5499	struct sctp_association *asoc;
5500
5501	if (len == sizeof(int)) {
5502		pr_warn_ratelimited(DEPRECATED
5503				    "%s (pid %d) "
5504				    "Use of int in max_burst socket option.\n"
5505				    "Use struct sctp_assoc_value instead\n",
5506				    current->comm, task_pid_nr(current));
5507		params.assoc_id = 0;
5508	} else if (len >= sizeof(struct sctp_assoc_value)) {
5509		len = sizeof(struct sctp_assoc_value);
5510		if (copy_from_user(&params, optval, len))
5511			return -EFAULT;
5512	} else
5513		return -EINVAL;
5514
5515	sp = sctp_sk(sk);
5516
5517	if (params.assoc_id != 0) {
5518		asoc = sctp_id2assoc(sk, params.assoc_id);
5519		if (!asoc)
5520			return -EINVAL;
5521		params.assoc_value = asoc->max_burst;
5522	} else
5523		params.assoc_value = sp->max_burst;
5524
5525	if (len == sizeof(int)) {
5526		if (copy_to_user(optval, &params.assoc_value, len))
5527			return -EFAULT;
5528	} else {
5529		if (copy_to_user(optval, &params, len))
5530			return -EFAULT;
5531	}
5532
5533	return 0;
5534
5535}
5536
5537static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5538				    char __user *optval, int __user *optlen)
5539{
5540	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5541	struct sctp_hmacalgo  __user *p = (void __user *)optval;
5542	struct sctp_hmac_algo_param *hmacs;
5543	__u16 data_len = 0;
5544	u32 num_idents;
5545	int i;
5546
5547	if (!ep->auth_enable)
5548		return -EACCES;
5549
5550	hmacs = ep->auth_hmacs_list;
5551	data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5552
5553	if (len < sizeof(struct sctp_hmacalgo) + data_len)
5554		return -EINVAL;
5555
5556	len = sizeof(struct sctp_hmacalgo) + data_len;
5557	num_idents = data_len / sizeof(u16);
5558
5559	if (put_user(len, optlen))
5560		return -EFAULT;
5561	if (put_user(num_idents, &p->shmac_num_idents))
5562		return -EFAULT;
5563	for (i = 0; i < num_idents; i++) {
5564		__u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5565
5566		if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5567			return -EFAULT;
5568	}
5569	return 0;
5570}
5571
5572static int sctp_getsockopt_active_key(struct sock *sk, int len,
5573				    char __user *optval, int __user *optlen)
5574{
5575	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5576	struct sctp_authkeyid val;
5577	struct sctp_association *asoc;
5578
5579	if (!ep->auth_enable)
5580		return -EACCES;
5581
5582	if (len < sizeof(struct sctp_authkeyid))
5583		return -EINVAL;
5584	if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5585		return -EFAULT;
5586
5587	asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5588	if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5589		return -EINVAL;
5590
5591	if (asoc)
5592		val.scact_keynumber = asoc->active_key_id;
5593	else
5594		val.scact_keynumber = ep->active_key_id;
5595
5596	len = sizeof(struct sctp_authkeyid);
5597	if (put_user(len, optlen))
5598		return -EFAULT;
5599	if (copy_to_user(optval, &val, len))
5600		return -EFAULT;
5601
5602	return 0;
5603}
5604
5605static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5606				    char __user *optval, int __user *optlen)
5607{
5608	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5609	struct sctp_authchunks __user *p = (void __user *)optval;
5610	struct sctp_authchunks val;
5611	struct sctp_association *asoc;
5612	struct sctp_chunks_param *ch;
5613	u32    num_chunks = 0;
5614	char __user *to;
5615
5616	if (!ep->auth_enable)
5617		return -EACCES;
5618
5619	if (len < sizeof(struct sctp_authchunks))
5620		return -EINVAL;
5621
5622	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5623		return -EFAULT;
5624
5625	to = p->gauth_chunks;
5626	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5627	if (!asoc)
5628		return -EINVAL;
5629
5630	ch = asoc->peer.peer_chunks;
5631	if (!ch)
5632		goto num;
5633
5634	/* See if the user provided enough room for all the data */
5635	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5636	if (len < num_chunks)
5637		return -EINVAL;
5638
5639	if (copy_to_user(to, ch->chunks, num_chunks))
5640		return -EFAULT;
5641num:
5642	len = sizeof(struct sctp_authchunks) + num_chunks;
5643	if (put_user(len, optlen))
5644		return -EFAULT;
5645	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5646		return -EFAULT;
5647	return 0;
5648}
5649
5650static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5651				    char __user *optval, int __user *optlen)
5652{
5653	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5654	struct sctp_authchunks __user *p = (void __user *)optval;
5655	struct sctp_authchunks val;
5656	struct sctp_association *asoc;
5657	struct sctp_chunks_param *ch;
5658	u32    num_chunks = 0;
5659	char __user *to;
5660
5661	if (!ep->auth_enable)
5662		return -EACCES;
5663
5664	if (len < sizeof(struct sctp_authchunks))
5665		return -EINVAL;
5666
5667	if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5668		return -EFAULT;
5669
5670	to = p->gauth_chunks;
5671	asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5672	if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5673		return -EINVAL;
5674
5675	if (asoc)
5676		ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5677	else
5678		ch = ep->auth_chunk_list;
5679
5680	if (!ch)
5681		goto num;
5682
5683	num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5684	if (len < sizeof(struct sctp_authchunks) + num_chunks)
5685		return -EINVAL;
5686
5687	if (copy_to_user(to, ch->chunks, num_chunks))
5688		return -EFAULT;
5689num:
5690	len = sizeof(struct sctp_authchunks) + num_chunks;
5691	if (put_user(len, optlen))
5692		return -EFAULT;
5693	if (put_user(num_chunks, &p->gauth_number_of_chunks))
5694		return -EFAULT;
5695
5696	return 0;
5697}
5698
5699/*
5700 * 8.2.5.  Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5701 * This option gets the current number of associations that are attached
5702 * to a one-to-many style socket.  The option value is an uint32_t.
5703 */
5704static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5705				    char __user *optval, int __user *optlen)
5706{
5707	struct sctp_sock *sp = sctp_sk(sk);
5708	struct sctp_association *asoc;
5709	u32 val = 0;
5710
5711	if (sctp_style(sk, TCP))
5712		return -EOPNOTSUPP;
5713
5714	if (len < sizeof(u32))
5715		return -EINVAL;
5716
5717	len = sizeof(u32);
5718
5719	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5720		val++;
5721	}
5722
5723	if (put_user(len, optlen))
5724		return -EFAULT;
5725	if (copy_to_user(optval, &val, len))
5726		return -EFAULT;
5727
5728	return 0;
5729}
5730
5731/*
5732 * 8.1.23 SCTP_AUTO_ASCONF
5733 * See the corresponding setsockopt entry as description
5734 */
5735static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5736				   char __user *optval, int __user *optlen)
5737{
5738	int val = 0;
5739
5740	if (len < sizeof(int))
5741		return -EINVAL;
5742
5743	len = sizeof(int);
5744	if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5745		val = 1;
5746	if (put_user(len, optlen))
5747		return -EFAULT;
5748	if (copy_to_user(optval, &val, len))
5749		return -EFAULT;
5750	return 0;
5751}
5752
5753/*
5754 * 8.2.6. Get the Current Identifiers of Associations
5755 *        (SCTP_GET_ASSOC_ID_LIST)
5756 *
5757 * This option gets the current list of SCTP association identifiers of
5758 * the SCTP associations handled by a one-to-many style socket.
5759 */
5760static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5761				    char __user *optval, int __user *optlen)
5762{
5763	struct sctp_sock *sp = sctp_sk(sk);
5764	struct sctp_association *asoc;
5765	struct sctp_assoc_ids *ids;
5766	u32 num = 0;
5767
5768	if (sctp_style(sk, TCP))
5769		return -EOPNOTSUPP;
5770
5771	if (len < sizeof(struct sctp_assoc_ids))
5772		return -EINVAL;
5773
5774	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5775		num++;
5776	}
5777
5778	if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5779		return -EINVAL;
5780
5781	len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5782
5783	ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
5784	if (unlikely(!ids))
5785		return -ENOMEM;
5786
5787	ids->gaids_number_of_ids = num;
5788	num = 0;
5789	list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5790		ids->gaids_assoc_id[num++] = asoc->assoc_id;
5791	}
5792
5793	if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5794		kfree(ids);
5795		return -EFAULT;
5796	}
5797
5798	kfree(ids);
5799	return 0;
5800}
5801
5802/*
5803 * SCTP_PEER_ADDR_THLDS
5804 *
5805 * This option allows us to fetch the partially failed threshold for one or all
5806 * transports in an association.  See Section 6.1 of:
5807 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5808 */
5809static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5810					    char __user *optval,
5811					    int len,
5812					    int __user *optlen)
5813{
5814	struct sctp_paddrthlds val;
5815	struct sctp_transport *trans;
5816	struct sctp_association *asoc;
5817
5818	if (len < sizeof(struct sctp_paddrthlds))
5819		return -EINVAL;
5820	len = sizeof(struct sctp_paddrthlds);
5821	if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5822		return -EFAULT;
5823
5824	if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5825		asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5826		if (!asoc)
5827			return -ENOENT;
5828
5829		val.spt_pathpfthld = asoc->pf_retrans;
5830		val.spt_pathmaxrxt = asoc->pathmaxrxt;
5831	} else {
5832		trans = sctp_addr_id2transport(sk, &val.spt_address,
5833					       val.spt_assoc_id);
5834		if (!trans)
5835			return -ENOENT;
5836
5837		val.spt_pathmaxrxt = trans->pathmaxrxt;
5838		val.spt_pathpfthld = trans->pf_retrans;
5839	}
5840
5841	if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5842		return -EFAULT;
5843
5844	return 0;
5845}
5846
5847/*
5848 * SCTP_GET_ASSOC_STATS
5849 *
5850 * This option retrieves local per endpoint statistics. It is modeled
5851 * after OpenSolaris' implementation
5852 */
5853static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5854				       char __user *optval,
5855				       int __user *optlen)
5856{
5857	struct sctp_assoc_stats sas;
5858	struct sctp_association *asoc = NULL;
5859
5860	/* User must provide at least the assoc id */
5861	if (len < sizeof(sctp_assoc_t))
5862		return -EINVAL;
5863
5864	/* Allow the struct to grow and fill in as much as possible */
5865	len = min_t(size_t, len, sizeof(sas));
5866
5867	if (copy_from_user(&sas, optval, len))
5868		return -EFAULT;
5869
5870	asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5871	if (!asoc)
5872		return -EINVAL;
5873
5874	sas.sas_rtxchunks = asoc->stats.rtxchunks;
5875	sas.sas_gapcnt = asoc->stats.gapcnt;
5876	sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5877	sas.sas_osacks = asoc->stats.osacks;
5878	sas.sas_isacks = asoc->stats.isacks;
5879	sas.sas_octrlchunks = asoc->stats.octrlchunks;
5880	sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5881	sas.sas_oodchunks = asoc->stats.oodchunks;
5882	sas.sas_iodchunks = asoc->stats.iodchunks;
5883	sas.sas_ouodchunks = asoc->stats.ouodchunks;
5884	sas.sas_iuodchunks = asoc->stats.iuodchunks;
5885	sas.sas_idupchunks = asoc->stats.idupchunks;
5886	sas.sas_opackets = asoc->stats.opackets;
5887	sas.sas_ipackets = asoc->stats.ipackets;
5888
5889	/* New high max rto observed, will return 0 if not a single
5890	 * RTO update took place. obs_rto_ipaddr will be bogus
5891	 * in such a case
5892	 */
5893	sas.sas_maxrto = asoc->stats.max_obs_rto;
5894	memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5895		sizeof(struct sockaddr_storage));
5896
5897	/* Mark beginning of a new observation period */
5898	asoc->stats.max_obs_rto = asoc->rto_min;
5899
5900	if (put_user(len, optlen))
5901		return -EFAULT;
5902
5903	pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5904
5905	if (copy_to_user(optval, &sas, len))
5906		return -EFAULT;
5907
5908	return 0;
5909}
5910
5911static int sctp_getsockopt_recvrcvinfo(struct sock *sk,	int len,
5912				       char __user *optval,
5913				       int __user *optlen)
5914{
5915	int val = 0;
5916
5917	if (len < sizeof(int))
5918		return -EINVAL;
5919
5920	len = sizeof(int);
5921	if (sctp_sk(sk)->recvrcvinfo)
5922		val = 1;
5923	if (put_user(len, optlen))
5924		return -EFAULT;
5925	if (copy_to_user(optval, &val, len))
5926		return -EFAULT;
5927
5928	return 0;
5929}
5930
5931static int sctp_getsockopt_recvnxtinfo(struct sock *sk,	int len,
5932				       char __user *optval,
5933				       int __user *optlen)
5934{
5935	int val = 0;
5936
5937	if (len < sizeof(int))
5938		return -EINVAL;
5939
5940	len = sizeof(int);
5941	if (sctp_sk(sk)->recvnxtinfo)
5942		val = 1;
5943	if (put_user(len, optlen))
5944		return -EFAULT;
5945	if (copy_to_user(optval, &val, len))
5946		return -EFAULT;
5947
5948	return 0;
5949}
5950
5951static int sctp_getsockopt(struct sock *sk, int level, int optname,
5952			   char __user *optval, int __user *optlen)
5953{
5954	int retval = 0;
5955	int len;
5956
5957	pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5958
5959	/* I can hardly begin to describe how wrong this is.  This is
5960	 * so broken as to be worse than useless.  The API draft
5961	 * REALLY is NOT helpful here...  I am not convinced that the
5962	 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5963	 * are at all well-founded.
5964	 */
5965	if (level != SOL_SCTP) {
5966		struct sctp_af *af = sctp_sk(sk)->pf->af;
5967
5968		retval = af->getsockopt(sk, level, optname, optval, optlen);
5969		return retval;
5970	}
5971
5972	if (get_user(len, optlen))
5973		return -EFAULT;
5974
5975	lock_sock(sk);
5976
5977	switch (optname) {
5978	case SCTP_STATUS:
5979		retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5980		break;
5981	case SCTP_DISABLE_FRAGMENTS:
5982		retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5983							   optlen);
5984		break;
5985	case SCTP_EVENTS:
5986		retval = sctp_getsockopt_events(sk, len, optval, optlen);
5987		break;
5988	case SCTP_AUTOCLOSE:
5989		retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5990		break;
5991	case SCTP_SOCKOPT_PEELOFF:
5992		retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5993		break;
5994	case SCTP_PEER_ADDR_PARAMS:
5995		retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5996							  optlen);
5997		break;
5998	case SCTP_DELAYED_SACK:
5999		retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6000							  optlen);
6001		break;
6002	case SCTP_INITMSG:
6003		retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6004		break;
6005	case SCTP_GET_PEER_ADDRS:
6006		retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6007						    optlen);
6008		break;
6009	case SCTP_GET_LOCAL_ADDRS:
6010		retval = sctp_getsockopt_local_addrs(sk, len, optval,
6011						     optlen);
6012		break;
6013	case SCTP_SOCKOPT_CONNECTX3:
6014		retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6015		break;
6016	case SCTP_DEFAULT_SEND_PARAM:
6017		retval = sctp_getsockopt_default_send_param(sk, len,
6018							    optval, optlen);
6019		break;
6020	case SCTP_DEFAULT_SNDINFO:
6021		retval = sctp_getsockopt_default_sndinfo(sk, len,
6022							 optval, optlen);
6023		break;
6024	case SCTP_PRIMARY_ADDR:
6025		retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6026		break;
6027	case SCTP_NODELAY:
6028		retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6029		break;
6030	case SCTP_RTOINFO:
6031		retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6032		break;
6033	case SCTP_ASSOCINFO:
6034		retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6035		break;
6036	case SCTP_I_WANT_MAPPED_V4_ADDR:
6037		retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6038		break;
6039	case SCTP_MAXSEG:
6040		retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6041		break;
6042	case SCTP_GET_PEER_ADDR_INFO:
6043		retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6044							optlen);
6045		break;
6046	case SCTP_ADAPTATION_LAYER:
6047		retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6048							optlen);
6049		break;
6050	case SCTP_CONTEXT:
6051		retval = sctp_getsockopt_context(sk, len, optval, optlen);
6052		break;
6053	case SCTP_FRAGMENT_INTERLEAVE:
6054		retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6055							     optlen);
6056		break;
6057	case SCTP_PARTIAL_DELIVERY_POINT:
6058		retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6059								optlen);
6060		break;
6061	case SCTP_MAX_BURST:
6062		retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6063		break;
6064	case SCTP_AUTH_KEY:
6065	case SCTP_AUTH_CHUNK:
6066	case SCTP_AUTH_DELETE_KEY:
6067		retval = -EOPNOTSUPP;
6068		break;
6069	case SCTP_HMAC_IDENT:
6070		retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6071		break;
6072	case SCTP_AUTH_ACTIVE_KEY:
6073		retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6074		break;
6075	case SCTP_PEER_AUTH_CHUNKS:
6076		retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6077							optlen);
6078		break;
6079	case SCTP_LOCAL_AUTH_CHUNKS:
6080		retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6081							optlen);
6082		break;
6083	case SCTP_GET_ASSOC_NUMBER:
6084		retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6085		break;
6086	case SCTP_GET_ASSOC_ID_LIST:
6087		retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6088		break;
6089	case SCTP_AUTO_ASCONF:
6090		retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6091		break;
6092	case SCTP_PEER_ADDR_THLDS:
6093		retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6094		break;
6095	case SCTP_GET_ASSOC_STATS:
6096		retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6097		break;
6098	case SCTP_RECVRCVINFO:
6099		retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6100		break;
6101	case SCTP_RECVNXTINFO:
6102		retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6103		break;
6104	default:
6105		retval = -ENOPROTOOPT;
6106		break;
6107	}
6108
6109	release_sock(sk);
6110	return retval;
6111}
6112
6113static void sctp_hash(struct sock *sk)
6114{
6115	/* STUB */
6116}
6117
6118static void sctp_unhash(struct sock *sk)
6119{
6120	/* STUB */
6121}
6122
6123/* Check if port is acceptable.  Possibly find first available port.
6124 *
6125 * The port hash table (contained in the 'global' SCTP protocol storage
6126 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6127 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6128 * list (the list number is the port number hashed out, so as you
6129 * would expect from a hash function, all the ports in a given list have
6130 * such a number that hashes out to the same list number; you were
6131 * expecting that, right?); so each list has a set of ports, with a
6132 * link to the socket (struct sock) that uses it, the port number and
6133 * a fastreuse flag (FIXME: NPI ipg).
6134 */
6135static struct sctp_bind_bucket *sctp_bucket_create(
6136	struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6137
6138static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6139{
6140	struct sctp_bind_hashbucket *head; /* hash list */
6141	struct sctp_bind_bucket *pp;
6142	unsigned short snum;
6143	int ret;
6144
6145	snum = ntohs(addr->v4.sin_port);
6146
6147	pr_debug("%s: begins, snum:%d\n", __func__, snum);
6148
6149	local_bh_disable();
6150
6151	if (snum == 0) {
6152		/* Search for an available port. */
6153		int low, high, remaining, index;
6154		unsigned int rover;
6155		struct net *net = sock_net(sk);
6156
6157		inet_get_local_port_range(net, &low, &high);
6158		remaining = (high - low) + 1;
6159		rover = prandom_u32() % remaining + low;
6160
6161		do {
6162			rover++;
6163			if ((rover < low) || (rover > high))
6164				rover = low;
6165			if (inet_is_local_reserved_port(net, rover))
6166				continue;
6167			index = sctp_phashfn(sock_net(sk), rover);
6168			head = &sctp_port_hashtable[index];
6169			spin_lock(&head->lock);
6170			sctp_for_each_hentry(pp, &head->chain)
6171				if ((pp->port == rover) &&
6172				    net_eq(sock_net(sk), pp->net))
6173					goto next;
6174			break;
6175		next:
6176			spin_unlock(&head->lock);
6177		} while (--remaining > 0);
6178
6179		/* Exhausted local port range during search? */
6180		ret = 1;
6181		if (remaining <= 0)
6182			goto fail;
6183
6184		/* OK, here is the one we will use.  HEAD (the port
6185		 * hash table list entry) is non-NULL and we hold it's
6186		 * mutex.
6187		 */
6188		snum = rover;
6189	} else {
6190		/* We are given an specific port number; we verify
6191		 * that it is not being used. If it is used, we will
6192		 * exahust the search in the hash list corresponding
6193		 * to the port number (snum) - we detect that with the
6194		 * port iterator, pp being NULL.
6195		 */
6196		head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6197		spin_lock(&head->lock);
6198		sctp_for_each_hentry(pp, &head->chain) {
6199			if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6200				goto pp_found;
6201		}
6202	}
6203	pp = NULL;
6204	goto pp_not_found;
6205pp_found:
6206	if (!hlist_empty(&pp->owner)) {
6207		/* We had a port hash table hit - there is an
6208		 * available port (pp != NULL) and it is being
6209		 * used by other socket (pp->owner not empty); that other
6210		 * socket is going to be sk2.
6211		 */
6212		int reuse = sk->sk_reuse;
6213		struct sock *sk2;
6214
6215		pr_debug("%s: found a possible match\n", __func__);
6216
6217		if (pp->fastreuse && sk->sk_reuse &&
6218			sk->sk_state != SCTP_SS_LISTENING)
6219			goto success;
6220
6221		/* Run through the list of sockets bound to the port
6222		 * (pp->port) [via the pointers bind_next and
6223		 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6224		 * we get the endpoint they describe and run through
6225		 * the endpoint's list of IP (v4 or v6) addresses,
6226		 * comparing each of the addresses with the address of
6227		 * the socket sk. If we find a match, then that means
6228		 * that this port/socket (sk) combination are already
6229		 * in an endpoint.
6230		 */
6231		sk_for_each_bound(sk2, &pp->owner) {
6232			struct sctp_endpoint *ep2;
6233			ep2 = sctp_sk(sk2)->ep;
6234
6235			if (sk == sk2 ||
6236			    (reuse && sk2->sk_reuse &&
6237			     sk2->sk_state != SCTP_SS_LISTENING))
6238				continue;
6239
6240			if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6241						 sctp_sk(sk2), sctp_sk(sk))) {
6242				ret = (long)sk2;
6243				goto fail_unlock;
6244			}
6245		}
6246
6247		pr_debug("%s: found a match\n", __func__);
6248	}
6249pp_not_found:
6250	/* If there was a hash table miss, create a new port.  */
6251	ret = 1;
6252	if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6253		goto fail_unlock;
6254
6255	/* In either case (hit or miss), make sure fastreuse is 1 only
6256	 * if sk->sk_reuse is too (that is, if the caller requested
6257	 * SO_REUSEADDR on this socket -sk-).
6258	 */
6259	if (hlist_empty(&pp->owner)) {
6260		if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6261			pp->fastreuse = 1;
6262		else
6263			pp->fastreuse = 0;
6264	} else if (pp->fastreuse &&
6265		(!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6266		pp->fastreuse = 0;
6267
6268	/* We are set, so fill up all the data in the hash table
6269	 * entry, tie the socket list information with the rest of the
6270	 * sockets FIXME: Blurry, NPI (ipg).
6271	 */
6272success:
6273	if (!sctp_sk(sk)->bind_hash) {
6274		inet_sk(sk)->inet_num = snum;
6275		sk_add_bind_node(sk, &pp->owner);
6276		sctp_sk(sk)->bind_hash = pp;
6277	}
6278	ret = 0;
6279
6280fail_unlock:
6281	spin_unlock(&head->lock);
6282
6283fail:
6284	local_bh_enable();
6285	return ret;
6286}
6287
6288/* Assign a 'snum' port to the socket.  If snum == 0, an ephemeral
6289 * port is requested.
6290 */
6291static int sctp_get_port(struct sock *sk, unsigned short snum)
6292{
6293	union sctp_addr addr;
6294	struct sctp_af *af = sctp_sk(sk)->pf->af;
6295
6296	/* Set up a dummy address struct from the sk. */
6297	af->from_sk(&addr, sk);
6298	addr.v4.sin_port = htons(snum);
6299
6300	/* Note: sk->sk_num gets filled in if ephemeral port request. */
6301	return !!sctp_get_port_local(sk, &addr);
6302}
6303
6304/*
6305 *  Move a socket to LISTENING state.
6306 */
6307static int sctp_listen_start(struct sock *sk, int backlog)
6308{
6309	struct sctp_sock *sp = sctp_sk(sk);
6310	struct sctp_endpoint *ep = sp->ep;
6311	struct crypto_hash *tfm = NULL;
6312	char alg[32];
6313
6314	/* Allocate HMAC for generating cookie. */
6315	if (!sp->hmac && sp->sctp_hmac_alg) {
6316		sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6317		tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6318		if (IS_ERR(tfm)) {
6319			net_info_ratelimited("failed to load transform for %s: %ld\n",
6320					     sp->sctp_hmac_alg, PTR_ERR(tfm));
6321			return -ENOSYS;
6322		}
6323		sctp_sk(sk)->hmac = tfm;
6324	}
6325
6326	/*
6327	 * If a bind() or sctp_bindx() is not called prior to a listen()
6328	 * call that allows new associations to be accepted, the system
6329	 * picks an ephemeral port and will choose an address set equivalent
6330	 * to binding with a wildcard address.
6331	 *
6332	 * This is not currently spelled out in the SCTP sockets
6333	 * extensions draft, but follows the practice as seen in TCP
6334	 * sockets.
6335	 *
6336	 */
6337	sk->sk_state = SCTP_SS_LISTENING;
6338	if (!ep->base.bind_addr.port) {
6339		if (sctp_autobind(sk))
6340			return -EAGAIN;
6341	} else {
6342		if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6343			sk->sk_state = SCTP_SS_CLOSED;
6344			return -EADDRINUSE;
6345		}
6346	}
6347
6348	sk->sk_max_ack_backlog = backlog;
6349	sctp_hash_endpoint(ep);
6350	return 0;
6351}
6352
6353/*
6354 * 4.1.3 / 5.1.3 listen()
6355 *
6356 *   By default, new associations are not accepted for UDP style sockets.
6357 *   An application uses listen() to mark a socket as being able to
6358 *   accept new associations.
6359 *
6360 *   On TCP style sockets, applications use listen() to ready the SCTP
6361 *   endpoint for accepting inbound associations.
6362 *
6363 *   On both types of endpoints a backlog of '0' disables listening.
6364 *
6365 *  Move a socket to LISTENING state.
6366 */
6367int sctp_inet_listen(struct socket *sock, int backlog)
6368{
6369	struct sock *sk = sock->sk;
6370	struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6371	int err = -EINVAL;
6372
6373	if (unlikely(backlog < 0))
6374		return err;
6375
6376	lock_sock(sk);
6377
6378	/* Peeled-off sockets are not allowed to listen().  */
6379	if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6380		goto out;
6381
6382	if (sock->state != SS_UNCONNECTED)
6383		goto out;
6384
6385	/* If backlog is zero, disable listening. */
6386	if (!backlog) {
6387		if (sctp_sstate(sk, CLOSED))
6388			goto out;
6389
6390		err = 0;
6391		sctp_unhash_endpoint(ep);
6392		sk->sk_state = SCTP_SS_CLOSED;
6393		if (sk->sk_reuse)
6394			sctp_sk(sk)->bind_hash->fastreuse = 1;
6395		goto out;
6396	}
6397
6398	/* If we are already listening, just update the backlog */
6399	if (sctp_sstate(sk, LISTENING))
6400		sk->sk_max_ack_backlog = backlog;
6401	else {
6402		err = sctp_listen_start(sk, backlog);
6403		if (err)
6404			goto out;
6405	}
6406
6407	err = 0;
6408out:
6409	release_sock(sk);
6410	return err;
6411}
6412
6413/*
6414 * This function is done by modeling the current datagram_poll() and the
6415 * tcp_poll().  Note that, based on these implementations, we don't
6416 * lock the socket in this function, even though it seems that,
6417 * ideally, locking or some other mechanisms can be used to ensure
6418 * the integrity of the counters (sndbuf and wmem_alloc) used
6419 * in this place.  We assume that we don't need locks either until proven
6420 * otherwise.
6421 *
6422 * Another thing to note is that we include the Async I/O support
6423 * here, again, by modeling the current TCP/UDP code.  We don't have
6424 * a good way to test with it yet.
6425 */
6426unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6427{
6428	struct sock *sk = sock->sk;
6429	struct sctp_sock *sp = sctp_sk(sk);
6430	unsigned int mask;
6431
6432	poll_wait(file, sk_sleep(sk), wait);
6433
6434	/* A TCP-style listening socket becomes readable when the accept queue
6435	 * is not empty.
6436	 */
6437	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6438		return (!list_empty(&sp->ep->asocs)) ?
6439			(POLLIN | POLLRDNORM) : 0;
6440
6441	mask = 0;
6442
6443	/* Is there any exceptional events?  */
6444	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6445		mask |= POLLERR |
6446			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6447	if (sk->sk_shutdown & RCV_SHUTDOWN)
6448		mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6449	if (sk->sk_shutdown == SHUTDOWN_MASK)
6450		mask |= POLLHUP;
6451
6452	/* Is it readable?  Reconsider this code with TCP-style support.  */
6453	if (!skb_queue_empty(&sk->sk_receive_queue))
6454		mask |= POLLIN | POLLRDNORM;
6455
6456	/* The association is either gone or not ready.  */
6457	if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6458		return mask;
6459
6460	/* Is it writable?  */
6461	if (sctp_writeable(sk)) {
6462		mask |= POLLOUT | POLLWRNORM;
6463	} else {
6464		sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6465		/*
6466		 * Since the socket is not locked, the buffer
6467		 * might be made available after the writeable check and
6468		 * before the bit is set.  This could cause a lost I/O
6469		 * signal.  tcp_poll() has a race breaker for this race
6470		 * condition.  Based on their implementation, we put
6471		 * in the following code to cover it as well.
6472		 */
6473		if (sctp_writeable(sk))
6474			mask |= POLLOUT | POLLWRNORM;
6475	}
6476	return mask;
6477}
6478
6479/********************************************************************
6480 * 2nd Level Abstractions
6481 ********************************************************************/
6482
6483static struct sctp_bind_bucket *sctp_bucket_create(
6484	struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6485{
6486	struct sctp_bind_bucket *pp;
6487
6488	pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6489	if (pp) {
6490		SCTP_DBG_OBJCNT_INC(bind_bucket);
6491		pp->port = snum;
6492		pp->fastreuse = 0;
6493		INIT_HLIST_HEAD(&pp->owner);
6494		pp->net = net;
6495		hlist_add_head(&pp->node, &head->chain);
6496	}
6497	return pp;
6498}
6499
6500/* Caller must hold hashbucket lock for this tb with local BH disabled */
6501static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6502{
6503	if (pp && hlist_empty(&pp->owner)) {
6504		__hlist_del(&pp->node);
6505		kmem_cache_free(sctp_bucket_cachep, pp);
6506		SCTP_DBG_OBJCNT_DEC(bind_bucket);
6507	}
6508}
6509
6510/* Release this socket's reference to a local port.  */
6511static inline void __sctp_put_port(struct sock *sk)
6512{
6513	struct sctp_bind_hashbucket *head =
6514		&sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6515						  inet_sk(sk)->inet_num)];
6516	struct sctp_bind_bucket *pp;
6517
6518	spin_lock(&head->lock);
6519	pp = sctp_sk(sk)->bind_hash;
6520	__sk_del_bind_node(sk);
6521	sctp_sk(sk)->bind_hash = NULL;
6522	inet_sk(sk)->inet_num = 0;
6523	sctp_bucket_destroy(pp);
6524	spin_unlock(&head->lock);
6525}
6526
6527void sctp_put_port(struct sock *sk)
6528{
6529	local_bh_disable();
6530	__sctp_put_port(sk);
6531	local_bh_enable();
6532}
6533
6534/*
6535 * The system picks an ephemeral port and choose an address set equivalent
6536 * to binding with a wildcard address.
6537 * One of those addresses will be the primary address for the association.
6538 * This automatically enables the multihoming capability of SCTP.
6539 */
6540static int sctp_autobind(struct sock *sk)
6541{
6542	union sctp_addr autoaddr;
6543	struct sctp_af *af;
6544	__be16 port;
6545
6546	/* Initialize a local sockaddr structure to INADDR_ANY. */
6547	af = sctp_sk(sk)->pf->af;
6548
6549	port = htons(inet_sk(sk)->inet_num);
6550	af->inaddr_any(&autoaddr, port);
6551
6552	return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6553}
6554
6555/* Parse out IPPROTO_SCTP CMSG headers.  Perform only minimal validation.
6556 *
6557 * From RFC 2292
6558 * 4.2 The cmsghdr Structure *
6559 *
6560 * When ancillary data is sent or received, any number of ancillary data
6561 * objects can be specified by the msg_control and msg_controllen members of
6562 * the msghdr structure, because each object is preceded by
6563 * a cmsghdr structure defining the object's length (the cmsg_len member).
6564 * Historically Berkeley-derived implementations have passed only one object
6565 * at a time, but this API allows multiple objects to be
6566 * passed in a single call to sendmsg() or recvmsg(). The following example
6567 * shows two ancillary data objects in a control buffer.
6568 *
6569 *   |<--------------------------- msg_controllen -------------------------->|
6570 *   |                                                                       |
6571 *
6572 *   |<----- ancillary data object ----->|<----- ancillary data object ----->|
6573 *
6574 *   |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6575 *   |                                   |                                   |
6576 *
6577 *   |<---------- cmsg_len ---------->|  |<--------- cmsg_len ----------->|  |
6578 *
6579 *   |<--------- CMSG_LEN() --------->|  |<-------- CMSG_LEN() ---------->|  |
6580 *   |                                |  |                                |  |
6581 *
6582 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6583 *   |cmsg_|cmsg_|cmsg_|XX|           |XX|cmsg_|cmsg_|cmsg_|XX|           |XX|
6584 *
6585 *   |len  |level|type |XX|cmsg_data[]|XX|len  |level|type |XX|cmsg_data[]|XX|
6586 *
6587 *   +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6588 *    ^
6589 *    |
6590 *
6591 * msg_control
6592 * points here
6593 */
6594static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6595{
6596	struct cmsghdr *cmsg;
6597	struct msghdr *my_msg = (struct msghdr *)msg;
6598
6599	for_each_cmsghdr(cmsg, my_msg) {
6600		if (!CMSG_OK(my_msg, cmsg))
6601			return -EINVAL;
6602
6603		/* Should we parse this header or ignore?  */
6604		if (cmsg->cmsg_level != IPPROTO_SCTP)
6605			continue;
6606
6607		/* Strictly check lengths following example in SCM code.  */
6608		switch (cmsg->cmsg_type) {
6609		case SCTP_INIT:
6610			/* SCTP Socket API Extension
6611			 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6612			 *
6613			 * This cmsghdr structure provides information for
6614			 * initializing new SCTP associations with sendmsg().
6615			 * The SCTP_INITMSG socket option uses this same data
6616			 * structure.  This structure is not used for
6617			 * recvmsg().
6618			 *
6619			 * cmsg_level    cmsg_type      cmsg_data[]
6620			 * ------------  ------------   ----------------------
6621			 * IPPROTO_SCTP  SCTP_INIT      struct sctp_initmsg
6622			 */
6623			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6624				return -EINVAL;
6625
6626			cmsgs->init = CMSG_DATA(cmsg);
6627			break;
6628
6629		case SCTP_SNDRCV:
6630			/* SCTP Socket API Extension
6631			 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6632			 *
6633			 * This cmsghdr structure specifies SCTP options for
6634			 * sendmsg() and describes SCTP header information
6635			 * about a received message through recvmsg().
6636			 *
6637			 * cmsg_level    cmsg_type      cmsg_data[]
6638			 * ------------  ------------   ----------------------
6639			 * IPPROTO_SCTP  SCTP_SNDRCV    struct sctp_sndrcvinfo
6640			 */
6641			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6642				return -EINVAL;
6643
6644			cmsgs->srinfo = CMSG_DATA(cmsg);
6645
6646			if (cmsgs->srinfo->sinfo_flags &
6647			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6648			      SCTP_SACK_IMMEDIATELY |
6649			      SCTP_ABORT | SCTP_EOF))
6650				return -EINVAL;
6651			break;
6652
6653		case SCTP_SNDINFO:
6654			/* SCTP Socket API Extension
6655			 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6656			 *
6657			 * This cmsghdr structure specifies SCTP options for
6658			 * sendmsg(). This structure and SCTP_RCVINFO replaces
6659			 * SCTP_SNDRCV which has been deprecated.
6660			 *
6661			 * cmsg_level    cmsg_type      cmsg_data[]
6662			 * ------------  ------------   ---------------------
6663			 * IPPROTO_SCTP  SCTP_SNDINFO    struct sctp_sndinfo
6664			 */
6665			if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6666				return -EINVAL;
6667
6668			cmsgs->sinfo = CMSG_DATA(cmsg);
6669
6670			if (cmsgs->sinfo->snd_flags &
6671			    ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6672			      SCTP_SACK_IMMEDIATELY |
6673			      SCTP_ABORT | SCTP_EOF))
6674				return -EINVAL;
6675			break;
6676		default:
6677			return -EINVAL;
6678		}
6679	}
6680
6681	return 0;
6682}
6683
6684/*
6685 * Wait for a packet..
6686 * Note: This function is the same function as in core/datagram.c
6687 * with a few modifications to make lksctp work.
6688 */
6689static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6690{
6691	int error;
6692	DEFINE_WAIT(wait);
6693
6694	prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6695
6696	/* Socket errors? */
6697	error = sock_error(sk);
6698	if (error)
6699		goto out;
6700
6701	if (!skb_queue_empty(&sk->sk_receive_queue))
6702		goto ready;
6703
6704	/* Socket shut down?  */
6705	if (sk->sk_shutdown & RCV_SHUTDOWN)
6706		goto out;
6707
6708	/* Sequenced packets can come disconnected.  If so we report the
6709	 * problem.
6710	 */
6711	error = -ENOTCONN;
6712
6713	/* Is there a good reason to think that we may receive some data?  */
6714	if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6715		goto out;
6716
6717	/* Handle signals.  */
6718	if (signal_pending(current))
6719		goto interrupted;
6720
6721	/* Let another process have a go.  Since we are going to sleep
6722	 * anyway.  Note: This may cause odd behaviors if the message
6723	 * does not fit in the user's buffer, but this seems to be the
6724	 * only way to honor MSG_DONTWAIT realistically.
6725	 */
6726	release_sock(sk);
6727	*timeo_p = schedule_timeout(*timeo_p);
6728	lock_sock(sk);
6729
6730ready:
6731	finish_wait(sk_sleep(sk), &wait);
6732	return 0;
6733
6734interrupted:
6735	error = sock_intr_errno(*timeo_p);
6736
6737out:
6738	finish_wait(sk_sleep(sk), &wait);
6739	*err = error;
6740	return error;
6741}
6742
6743/* Receive a datagram.
6744 * Note: This is pretty much the same routine as in core/datagram.c
6745 * with a few changes to make lksctp work.
6746 */
6747struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6748				       int noblock, int *err)
6749{
6750	int error;
6751	struct sk_buff *skb;
6752	long timeo;
6753
6754	timeo = sock_rcvtimeo(sk, noblock);
6755
6756	pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6757		 MAX_SCHEDULE_TIMEOUT);
6758
6759	do {
6760		/* Again only user level code calls this function,
6761		 * so nothing interrupt level
6762		 * will suddenly eat the receive_queue.
6763		 *
6764		 *  Look at current nfs client by the way...
6765		 *  However, this function was correct in any case. 8)
6766		 */
6767		if (flags & MSG_PEEK) {
6768			spin_lock_bh(&sk->sk_receive_queue.lock);
6769			skb = skb_peek(&sk->sk_receive_queue);
6770			if (skb)
6771				atomic_inc(&skb->users);
6772			spin_unlock_bh(&sk->sk_receive_queue.lock);
6773		} else {
6774			skb = skb_dequeue(&sk->sk_receive_queue);
6775		}
6776
6777		if (skb)
6778			return skb;
6779
6780		/* Caller is allowed not to check sk->sk_err before calling. */
6781		error = sock_error(sk);
6782		if (error)
6783			goto no_packet;
6784
6785		if (sk->sk_shutdown & RCV_SHUTDOWN)
6786			break;
6787
6788		if (sk_can_busy_loop(sk) &&
6789		    sk_busy_loop(sk, noblock))
6790			continue;
6791
6792		/* User doesn't want to wait.  */
6793		error = -EAGAIN;
6794		if (!timeo)
6795			goto no_packet;
6796	} while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6797
6798	return NULL;
6799
6800no_packet:
6801	*err = error;
6802	return NULL;
6803}
6804
6805/* If sndbuf has changed, wake up per association sndbuf waiters.  */
6806static void __sctp_write_space(struct sctp_association *asoc)
6807{
6808	struct sock *sk = asoc->base.sk;
6809
6810	if (sctp_wspace(asoc) <= 0)
6811		return;
6812
6813	if (waitqueue_active(&asoc->wait))
6814		wake_up_interruptible(&asoc->wait);
6815
6816	if (sctp_writeable(sk)) {
6817		struct socket_wq *wq;
6818
6819		rcu_read_lock();
6820		wq = rcu_dereference(sk->sk_wq);
6821		if (wq) {
6822			if (waitqueue_active(&wq->wait))
6823				wake_up_interruptible(&wq->wait);
6824
6825			/* Note that we try to include the Async I/O support
6826			 * here by modeling from the current TCP/UDP code.
6827			 * We have not tested with it yet.
6828			 */
6829			if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6830				sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
6831		}
6832		rcu_read_unlock();
6833	}
6834}
6835
6836static void sctp_wake_up_waiters(struct sock *sk,
6837				 struct sctp_association *asoc)
6838{
6839	struct sctp_association *tmp = asoc;
6840
6841	/* We do accounting for the sndbuf space per association,
6842	 * so we only need to wake our own association.
6843	 */
6844	if (asoc->ep->sndbuf_policy)
6845		return __sctp_write_space(asoc);
6846
6847	/* If association goes down and is just flushing its
6848	 * outq, then just normally notify others.
6849	 */
6850	if (asoc->base.dead)
6851		return sctp_write_space(sk);
6852
6853	/* Accounting for the sndbuf space is per socket, so we
6854	 * need to wake up others, try to be fair and in case of
6855	 * other associations, let them have a go first instead
6856	 * of just doing a sctp_write_space() call.
6857	 *
6858	 * Note that we reach sctp_wake_up_waiters() only when
6859	 * associations free up queued chunks, thus we are under
6860	 * lock and the list of associations on a socket is
6861	 * guaranteed not to change.
6862	 */
6863	for (tmp = list_next_entry(tmp, asocs); 1;
6864	     tmp = list_next_entry(tmp, asocs)) {
6865		/* Manually skip the head element. */
6866		if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6867			continue;
6868		/* Wake up association. */
6869		__sctp_write_space(tmp);
6870		/* We've reached the end. */
6871		if (tmp == asoc)
6872			break;
6873	}
6874}
6875
6876/* Do accounting for the sndbuf space.
6877 * Decrement the used sndbuf space of the corresponding association by the
6878 * data size which was just transmitted(freed).
6879 */
6880static void sctp_wfree(struct sk_buff *skb)
6881{
6882	struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
6883	struct sctp_association *asoc = chunk->asoc;
6884	struct sock *sk = asoc->base.sk;
6885
6886	asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6887				sizeof(struct sk_buff) +
6888				sizeof(struct sctp_chunk);
6889
6890	atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6891
6892	/*
6893	 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6894	 */
6895	sk->sk_wmem_queued   -= skb->truesize;
6896	sk_mem_uncharge(sk, skb->truesize);
6897
6898	sock_wfree(skb);
6899	sctp_wake_up_waiters(sk, asoc);
6900
6901	sctp_association_put(asoc);
6902}
6903
6904/* Do accounting for the receive space on the socket.
6905 * Accounting for the association is done in ulpevent.c
6906 * We set this as a destructor for the cloned data skbs so that
6907 * accounting is done at the correct time.
6908 */
6909void sctp_sock_rfree(struct sk_buff *skb)
6910{
6911	struct sock *sk = skb->sk;
6912	struct sctp_ulpevent *event = sctp_skb2event(skb);
6913
6914	atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6915
6916	/*
6917	 * Mimic the behavior of sock_rfree
6918	 */
6919	sk_mem_uncharge(sk, event->rmem_len);
6920}
6921
6922
6923/* Helper function to wait for space in the sndbuf.  */
6924static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6925				size_t msg_len)
6926{
6927	struct sock *sk = asoc->base.sk;
6928	int err = 0;
6929	long current_timeo = *timeo_p;
6930	DEFINE_WAIT(wait);
6931
6932	pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6933		 *timeo_p, msg_len);
6934
6935	/* Increment the association's refcnt.  */
6936	sctp_association_hold(asoc);
6937
6938	/* Wait on the association specific sndbuf space. */
6939	for (;;) {
6940		prepare_to_wait_exclusive(&asoc->wait, &wait,
6941					  TASK_INTERRUPTIBLE);
6942		if (!*timeo_p)
6943			goto do_nonblock;
6944		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6945		    asoc->base.dead)
6946			goto do_error;
6947		if (signal_pending(current))
6948			goto do_interrupted;
6949		if (msg_len <= sctp_wspace(asoc))
6950			break;
6951
6952		/* Let another process have a go.  Since we are going
6953		 * to sleep anyway.
6954		 */
6955		release_sock(sk);
6956		current_timeo = schedule_timeout(current_timeo);
6957		BUG_ON(sk != asoc->base.sk);
6958		lock_sock(sk);
6959
6960		*timeo_p = current_timeo;
6961	}
6962
6963out:
6964	finish_wait(&asoc->wait, &wait);
6965
6966	/* Release the association's refcnt.  */
6967	sctp_association_put(asoc);
6968
6969	return err;
6970
6971do_error:
6972	err = -EPIPE;
6973	goto out;
6974
6975do_interrupted:
6976	err = sock_intr_errno(*timeo_p);
6977	goto out;
6978
6979do_nonblock:
6980	err = -EAGAIN;
6981	goto out;
6982}
6983
6984void sctp_data_ready(struct sock *sk)
6985{
6986	struct socket_wq *wq;
6987
6988	rcu_read_lock();
6989	wq = rcu_dereference(sk->sk_wq);
6990	if (wq_has_sleeper(wq))
6991		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6992						POLLRDNORM | POLLRDBAND);
6993	sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6994	rcu_read_unlock();
6995}
6996
6997/* If socket sndbuf has changed, wake up all per association waiters.  */
6998void sctp_write_space(struct sock *sk)
6999{
7000	struct sctp_association *asoc;
7001
7002	/* Wake up the tasks in each wait queue.  */
7003	list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7004		__sctp_write_space(asoc);
7005	}
7006}
7007
7008/* Is there any sndbuf space available on the socket?
7009 *
7010 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7011 * associations on the same socket.  For a UDP-style socket with
7012 * multiple associations, it is possible for it to be "unwriteable"
7013 * prematurely.  I assume that this is acceptable because
7014 * a premature "unwriteable" is better than an accidental "writeable" which
7015 * would cause an unwanted block under certain circumstances.  For the 1-1
7016 * UDP-style sockets or TCP-style sockets, this code should work.
7017 *  - Daisy
7018 */
7019static int sctp_writeable(struct sock *sk)
7020{
7021	int amt = 0;
7022
7023	amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7024	if (amt < 0)
7025		amt = 0;
7026	return amt;
7027}
7028
7029/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7030 * returns immediately with EINPROGRESS.
7031 */
7032static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7033{
7034	struct sock *sk = asoc->base.sk;
7035	int err = 0;
7036	long current_timeo = *timeo_p;
7037	DEFINE_WAIT(wait);
7038
7039	pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7040
7041	/* Increment the association's refcnt.  */
7042	sctp_association_hold(asoc);
7043
7044	for (;;) {
7045		prepare_to_wait_exclusive(&asoc->wait, &wait,
7046					  TASK_INTERRUPTIBLE);
7047		if (!*timeo_p)
7048			goto do_nonblock;
7049		if (sk->sk_shutdown & RCV_SHUTDOWN)
7050			break;
7051		if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7052		    asoc->base.dead)
7053			goto do_error;
7054		if (signal_pending(current))
7055			goto do_interrupted;
7056
7057		if (sctp_state(asoc, ESTABLISHED))
7058			break;
7059
7060		/* Let another process have a go.  Since we are going
7061		 * to sleep anyway.
7062		 */
7063		release_sock(sk);
7064		current_timeo = schedule_timeout(current_timeo);
7065		lock_sock(sk);
7066
7067		*timeo_p = current_timeo;
7068	}
7069
7070out:
7071	finish_wait(&asoc->wait, &wait);
7072
7073	/* Release the association's refcnt.  */
7074	sctp_association_put(asoc);
7075
7076	return err;
7077
7078do_error:
7079	if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7080		err = -ETIMEDOUT;
7081	else
7082		err = -ECONNREFUSED;
7083	goto out;
7084
7085do_interrupted:
7086	err = sock_intr_errno(*timeo_p);
7087	goto out;
7088
7089do_nonblock:
7090	err = -EINPROGRESS;
7091	goto out;
7092}
7093
7094static int sctp_wait_for_accept(struct sock *sk, long timeo)
7095{
7096	struct sctp_endpoint *ep;
7097	int err = 0;
7098	DEFINE_WAIT(wait);
7099
7100	ep = sctp_sk(sk)->ep;
7101
7102
7103	for (;;) {
7104		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7105					  TASK_INTERRUPTIBLE);
7106
7107		if (list_empty(&ep->asocs)) {
7108			release_sock(sk);
7109			timeo = schedule_timeout(timeo);
7110			lock_sock(sk);
7111		}
7112
7113		err = -EINVAL;
7114		if (!sctp_sstate(sk, LISTENING))
7115			break;
7116
7117		err = 0;
7118		if (!list_empty(&ep->asocs))
7119			break;
7120
7121		err = sock_intr_errno(timeo);
7122		if (signal_pending(current))
7123			break;
7124
7125		err = -EAGAIN;
7126		if (!timeo)
7127			break;
7128	}
7129
7130	finish_wait(sk_sleep(sk), &wait);
7131
7132	return err;
7133}
7134
7135static void sctp_wait_for_close(struct sock *sk, long timeout)
7136{
7137	DEFINE_WAIT(wait);
7138
7139	do {
7140		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7141		if (list_empty(&sctp_sk(sk)->ep->asocs))
7142			break;
7143		release_sock(sk);
7144		timeout = schedule_timeout(timeout);
7145		lock_sock(sk);
7146	} while (!signal_pending(current) && timeout);
7147
7148	finish_wait(sk_sleep(sk), &wait);
7149}
7150
7151static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7152{
7153	struct sk_buff *frag;
7154
7155	if (!skb->data_len)
7156		goto done;
7157
7158	/* Don't forget the fragments. */
7159	skb_walk_frags(skb, frag)
7160		sctp_skb_set_owner_r_frag(frag, sk);
7161
7162done:
7163	sctp_skb_set_owner_r(skb, sk);
7164}
7165
7166void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7167		    struct sctp_association *asoc)
7168{
7169	struct inet_sock *inet = inet_sk(sk);
7170	struct inet_sock *newinet;
7171
7172	newsk->sk_type = sk->sk_type;
7173	newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7174	newsk->sk_flags = sk->sk_flags;
7175	newsk->sk_tsflags = sk->sk_tsflags;
7176	newsk->sk_no_check_tx = sk->sk_no_check_tx;
7177	newsk->sk_no_check_rx = sk->sk_no_check_rx;
7178	newsk->sk_reuse = sk->sk_reuse;
7179
7180	newsk->sk_shutdown = sk->sk_shutdown;
7181	newsk->sk_destruct = sctp_destruct_sock;
7182	newsk->sk_family = sk->sk_family;
7183	newsk->sk_protocol = IPPROTO_SCTP;
7184	newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7185	newsk->sk_sndbuf = sk->sk_sndbuf;
7186	newsk->sk_rcvbuf = sk->sk_rcvbuf;
7187	newsk->sk_lingertime = sk->sk_lingertime;
7188	newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7189	newsk->sk_sndtimeo = sk->sk_sndtimeo;
7190
7191	newinet = inet_sk(newsk);
7192
7193	/* Initialize sk's sport, dport, rcv_saddr and daddr for
7194	 * getsockname() and getpeername()
7195	 */
7196	newinet->inet_sport = inet->inet_sport;
7197	newinet->inet_saddr = inet->inet_saddr;
7198	newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7199	newinet->inet_dport = htons(asoc->peer.port);
7200	newinet->pmtudisc = inet->pmtudisc;
7201	newinet->inet_id = asoc->next_tsn ^ jiffies;
7202
7203	newinet->uc_ttl = inet->uc_ttl;
7204	newinet->mc_loop = 1;
7205	newinet->mc_ttl = 1;
7206	newinet->mc_index = 0;
7207	newinet->mc_list = NULL;
7208
7209	if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7210		net_enable_timestamp();
7211
7212	security_sk_clone(sk, newsk);
7213}
7214
7215static inline void sctp_copy_descendant(struct sock *sk_to,
7216					const struct sock *sk_from)
7217{
7218	int ancestor_size = sizeof(struct inet_sock) +
7219			    sizeof(struct sctp_sock) -
7220			    offsetof(struct sctp_sock, auto_asconf_list);
7221
7222	if (sk_from->sk_family == PF_INET6)
7223		ancestor_size += sizeof(struct ipv6_pinfo);
7224
7225	__inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7226}
7227
7228/* Populate the fields of the newsk from the oldsk and migrate the assoc
7229 * and its messages to the newsk.
7230 */
7231static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7232			      struct sctp_association *assoc,
7233			      sctp_socket_type_t type)
7234{
7235	struct sctp_sock *oldsp = sctp_sk(oldsk);
7236	struct sctp_sock *newsp = sctp_sk(newsk);
7237	struct sctp_bind_bucket *pp; /* hash list port iterator */
7238	struct sctp_endpoint *newep = newsp->ep;
7239	struct sk_buff *skb, *tmp;
7240	struct sctp_ulpevent *event;
7241	struct sctp_bind_hashbucket *head;
7242
7243	/* Migrate socket buffer sizes and all the socket level options to the
7244	 * new socket.
7245	 */
7246	newsk->sk_sndbuf = oldsk->sk_sndbuf;
7247	newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7248	/* Brute force copy old sctp opt. */
7249	sctp_copy_descendant(newsk, oldsk);
7250
7251	/* Restore the ep value that was overwritten with the above structure
7252	 * copy.
7253	 */
7254	newsp->ep = newep;
7255	newsp->hmac = NULL;
7256
7257	/* Hook this new socket in to the bind_hash list. */
7258	head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7259						 inet_sk(oldsk)->inet_num)];
7260	local_bh_disable();
7261	spin_lock(&head->lock);
7262	pp = sctp_sk(oldsk)->bind_hash;
7263	sk_add_bind_node(newsk, &pp->owner);
7264	sctp_sk(newsk)->bind_hash = pp;
7265	inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7266	spin_unlock(&head->lock);
7267	local_bh_enable();
7268
7269	/* Copy the bind_addr list from the original endpoint to the new
7270	 * endpoint so that we can handle restarts properly
7271	 */
7272	sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7273				&oldsp->ep->base.bind_addr, GFP_KERNEL);
7274
7275	/* Move any messages in the old socket's receive queue that are for the
7276	 * peeled off association to the new socket's receive queue.
7277	 */
7278	sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7279		event = sctp_skb2event(skb);
7280		if (event->asoc == assoc) {
7281			__skb_unlink(skb, &oldsk->sk_receive_queue);
7282			__skb_queue_tail(&newsk->sk_receive_queue, skb);
7283			sctp_skb_set_owner_r_frag(skb, newsk);
7284		}
7285	}
7286
7287	/* Clean up any messages pending delivery due to partial
7288	 * delivery.   Three cases:
7289	 * 1) No partial deliver;  no work.
7290	 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7291	 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7292	 */
7293	skb_queue_head_init(&newsp->pd_lobby);
7294	atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7295
7296	if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7297		struct sk_buff_head *queue;
7298
7299		/* Decide which queue to move pd_lobby skbs to. */
7300		if (assoc->ulpq.pd_mode) {
7301			queue = &newsp->pd_lobby;
7302		} else
7303			queue = &newsk->sk_receive_queue;
7304
7305		/* Walk through the pd_lobby, looking for skbs that
7306		 * need moved to the new socket.
7307		 */
7308		sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7309			event = sctp_skb2event(skb);
7310			if (event->asoc == assoc) {
7311				__skb_unlink(skb, &oldsp->pd_lobby);
7312				__skb_queue_tail(queue, skb);
7313				sctp_skb_set_owner_r_frag(skb, newsk);
7314			}
7315		}
7316
7317		/* Clear up any skbs waiting for the partial
7318		 * delivery to finish.
7319		 */
7320		if (assoc->ulpq.pd_mode)
7321			sctp_clear_pd(oldsk, NULL);
7322
7323	}
7324
7325	sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7326		sctp_skb_set_owner_r_frag(skb, newsk);
7327
7328	sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7329		sctp_skb_set_owner_r_frag(skb, newsk);
7330
7331	/* Set the type of socket to indicate that it is peeled off from the
7332	 * original UDP-style socket or created with the accept() call on a
7333	 * TCP-style socket..
7334	 */
7335	newsp->type = type;
7336
7337	/* Mark the new socket "in-use" by the user so that any packets
7338	 * that may arrive on the association after we've moved it are
7339	 * queued to the backlog.  This prevents a potential race between
7340	 * backlog processing on the old socket and new-packet processing
7341	 * on the new socket.
7342	 *
7343	 * The caller has just allocated newsk so we can guarantee that other
7344	 * paths won't try to lock it and then oldsk.
7345	 */
7346	lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7347	sctp_assoc_migrate(assoc, newsk);
7348
7349	/* If the association on the newsk is already closed before accept()
7350	 * is called, set RCV_SHUTDOWN flag.
7351	 */
7352	if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7353		newsk->sk_shutdown |= RCV_SHUTDOWN;
7354
7355	newsk->sk_state = SCTP_SS_ESTABLISHED;
7356	release_sock(newsk);
7357}
7358
7359
7360/* This proto struct describes the ULP interface for SCTP.  */
7361struct proto sctp_prot = {
7362	.name        =	"SCTP",
7363	.owner       =	THIS_MODULE,
7364	.close       =	sctp_close,
7365	.connect     =	sctp_connect,
7366	.disconnect  =	sctp_disconnect,
7367	.accept      =	sctp_accept,
7368	.ioctl       =	sctp_ioctl,
7369	.init        =	sctp_init_sock,
7370	.destroy     =	sctp_destroy_sock,
7371	.shutdown    =	sctp_shutdown,
7372	.setsockopt  =	sctp_setsockopt,
7373	.getsockopt  =	sctp_getsockopt,
7374	.sendmsg     =	sctp_sendmsg,
7375	.recvmsg     =	sctp_recvmsg,
7376	.bind        =	sctp_bind,
7377	.backlog_rcv =	sctp_backlog_rcv,
7378	.hash        =	sctp_hash,
7379	.unhash      =	sctp_unhash,
7380	.get_port    =	sctp_get_port,
7381	.obj_size    =  sizeof(struct sctp_sock),
7382	.sysctl_mem  =  sysctl_sctp_mem,
7383	.sysctl_rmem =  sysctl_sctp_rmem,
7384	.sysctl_wmem =  sysctl_sctp_wmem,
7385	.memory_pressure = &sctp_memory_pressure,
7386	.enter_memory_pressure = sctp_enter_memory_pressure,
7387	.memory_allocated = &sctp_memory_allocated,
7388	.sockets_allocated = &sctp_sockets_allocated,
7389};
7390
7391#if IS_ENABLED(CONFIG_IPV6)
7392
7393#include <net/transp_v6.h>
7394static void sctp_v6_destroy_sock(struct sock *sk)
7395{
7396	sctp_destroy_sock(sk);
7397	inet6_destroy_sock(sk);
7398}
7399
7400struct proto sctpv6_prot = {
7401	.name		= "SCTPv6",
7402	.owner		= THIS_MODULE,
7403	.close		= sctp_close,
7404	.connect	= sctp_connect,
7405	.disconnect	= sctp_disconnect,
7406	.accept		= sctp_accept,
7407	.ioctl		= sctp_ioctl,
7408	.init		= sctp_init_sock,
7409	.destroy	= sctp_v6_destroy_sock,
7410	.shutdown	= sctp_shutdown,
7411	.setsockopt	= sctp_setsockopt,
7412	.getsockopt	= sctp_getsockopt,
7413	.sendmsg	= sctp_sendmsg,
7414	.recvmsg	= sctp_recvmsg,
7415	.bind		= sctp_bind,
7416	.backlog_rcv	= sctp_backlog_rcv,
7417	.hash		= sctp_hash,
7418	.unhash		= sctp_unhash,
7419	.get_port	= sctp_get_port,
7420	.obj_size	= sizeof(struct sctp6_sock),
7421	.sysctl_mem	= sysctl_sctp_mem,
7422	.sysctl_rmem	= sysctl_sctp_rmem,
7423	.sysctl_wmem	= sysctl_sctp_wmem,
7424	.memory_pressure = &sctp_memory_pressure,
7425	.enter_memory_pressure = sctp_enter_memory_pressure,
7426	.memory_allocated = &sctp_memory_allocated,
7427	.sockets_allocated = &sctp_sockets_allocated,
7428};
7429#endif /* IS_ENABLED(CONFIG_IPV6) */
7430