blob: b811691c35bf4eba764efa3eba3b0daeac6eca64 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* SCTP kernel reference 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 reference 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 * The SCTP reference 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 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60#include <linux/config.h>
61#include <linux/types.h>
62#include <linux/kernel.h>
63#include <linux/wait.h>
64#include <linux/time.h>
65#include <linux/ip.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080066#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/fcntl.h>
68#include <linux/poll.h>
69#include <linux/init.h>
70#include <linux/crypto.h>
71
72#include <net/ip.h>
73#include <net/icmp.h>
74#include <net/route.h>
75#include <net/ipv6.h>
76#include <net/inet_common.h>
77
78#include <linux/socket.h> /* for sa_family_t */
79#include <net/sock.h>
80#include <net/sctp/sctp.h>
81#include <net/sctp/sm.h>
82
83/* WARNING: Please do not remove the SCTP_STATIC attribute to
84 * any of the functions below as they are used to export functions
85 * used by a project regression testsuite.
86 */
87
88/* Forward declarations for internal helper functions. */
89static int sctp_writeable(struct sock *sk);
90static void sctp_wfree(struct sk_buff *skb);
91static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 size_t msg_len);
93static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
94static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
95static int sctp_wait_for_accept(struct sock *sk, long timeo);
96static void sctp_wait_for_close(struct sock *sk, long timeo);
97static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
98 union sctp_addr *addr, int len);
99static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
100static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
101static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
102static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
103static int sctp_send_asconf(struct sctp_association *asoc,
104 struct sctp_chunk *chunk);
105static int sctp_do_bind(struct sock *, union sctp_addr *, int);
106static int sctp_autobind(struct sock *sk);
107static void sctp_sock_migrate(struct sock *, struct sock *,
108 struct sctp_association *, sctp_socket_type_t);
109static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110
111extern kmem_cache_t *sctp_bucket_cachep;
112
113/* Get the sndbuf space available at the time on the association. */
114static inline int sctp_wspace(struct sctp_association *asoc)
115{
116 struct sock *sk = asoc->base.sk;
117 int amt = 0;
118
Neil Horman4eb701d2005-04-28 12:02:04 -0700119 if (asoc->ep->sndbuf_policy) {
120 /* make sure that no association uses more than sk_sndbuf */
121 amt = sk->sk_sndbuf - asoc->sndbuf_used;
122 } else {
123 /* do socket level accounting */
124 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
125 }
126
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 if (amt < 0)
128 amt = 0;
Neil Horman4eb701d2005-04-28 12:02:04 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 return amt;
131}
132
133/* Increment the used sndbuf space count of the corresponding association by
134 * the size of the outgoing data chunk.
135 * Also, set the skb destructor for sndbuf accounting later.
136 *
137 * Since it is always 1-1 between chunk and skb, and also a new skb is always
138 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
139 * destructor in the data chunk skb for the purpose of the sndbuf space
140 * tracking.
141 */
142static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
143{
144 struct sctp_association *asoc = chunk->asoc;
145 struct sock *sk = asoc->base.sk;
146
147 /* The sndbuf space is tracked per association. */
148 sctp_association_hold(asoc);
149
Neil Horman4eb701d2005-04-28 12:02:04 -0700150 skb_set_owner_w(chunk->skb, sk);
151
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 chunk->skb->destructor = sctp_wfree;
153 /* Save the chunk pointer in skb for sctp_wfree to use later. */
154 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
155
Neil Horman4eb701d2005-04-28 12:02:04 -0700156 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
157 sizeof(struct sk_buff) +
158 sizeof(struct sctp_chunk);
159
Neil Horman4eb701d2005-04-28 12:02:04 -0700160 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161}
162
163/* Verify that this is a valid address. */
164static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
165 int len)
166{
167 struct sctp_af *af;
168
169 /* Verify basic sockaddr. */
170 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
171 if (!af)
172 return -EINVAL;
173
174 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700175 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 return -EINVAL;
177
178 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
179 return -EINVAL;
180
181 return 0;
182}
183
184/* Look up the association by its id. If this is not a UDP-style
185 * socket, the ID field is always ignored.
186 */
187struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
188{
189 struct sctp_association *asoc = NULL;
190
191 /* If this is not a UDP-style socket, assoc id should be ignored. */
192 if (!sctp_style(sk, UDP)) {
193 /* Return NULL if the socket state is not ESTABLISHED. It
194 * could be a TCP-style listening socket or a socket which
195 * hasn't yet called connect() to establish an association.
196 */
197 if (!sctp_sstate(sk, ESTABLISHED))
198 return NULL;
199
200 /* Get the first and the only association from the list. */
201 if (!list_empty(&sctp_sk(sk)->ep->asocs))
202 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
203 struct sctp_association, asocs);
204 return asoc;
205 }
206
207 /* Otherwise this is a UDP-style socket. */
208 if (!id || (id == (sctp_assoc_t)-1))
209 return NULL;
210
211 spin_lock_bh(&sctp_assocs_id_lock);
212 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
213 spin_unlock_bh(&sctp_assocs_id_lock);
214
215 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
216 return NULL;
217
218 return asoc;
219}
220
221/* Look up the transport from an address and an assoc id. If both address and
222 * id are specified, the associations matching the address and the id should be
223 * the same.
224 */
225static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
226 struct sockaddr_storage *addr,
227 sctp_assoc_t id)
228{
229 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
230 struct sctp_transport *transport;
231 union sctp_addr *laddr = (union sctp_addr *)addr;
232
233 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
234 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
235 (union sctp_addr *)addr,
236 &transport);
237 laddr->v4.sin_port = htons(laddr->v4.sin_port);
238
239 if (!addr_asoc)
240 return NULL;
241
242 id_asoc = sctp_id2assoc(sk, id);
243 if (id_asoc && (id_asoc != addr_asoc))
244 return NULL;
245
246 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
247 (union sctp_addr *)addr);
248
249 return transport;
250}
251
252/* API 3.1.2 bind() - UDP Style Syntax
253 * The syntax of bind() is,
254 *
255 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
256 *
257 * sd - the socket descriptor returned by socket().
258 * addr - the address structure (struct sockaddr_in or struct
259 * sockaddr_in6 [RFC 2553]),
260 * addr_len - the size of the address structure.
261 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700262SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263{
264 int retval = 0;
265
266 sctp_lock_sock(sk);
267
Frank Filz3f7a87d2005-06-20 13:14:57 -0700268 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
269 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
271 /* Disallow binding twice. */
272 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700273 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 addr_len);
275 else
276 retval = -EINVAL;
277
278 sctp_release_sock(sk);
279
280 return retval;
281}
282
283static long sctp_get_port_local(struct sock *, union sctp_addr *);
284
285/* Verify this is a valid sockaddr. */
286static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
287 union sctp_addr *addr, int len)
288{
289 struct sctp_af *af;
290
291 /* Check minimum size. */
292 if (len < sizeof (struct sockaddr))
293 return NULL;
294
295 /* Does this PF support this AF? */
296 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
297 return NULL;
298
299 /* If we get this far, af is valid. */
300 af = sctp_get_af_specific(addr->sa.sa_family);
301
302 if (len < af->sockaddr_len)
303 return NULL;
304
305 return af;
306}
307
308/* Bind a local address either to an endpoint or to an association. */
309SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
310{
311 struct sctp_sock *sp = sctp_sk(sk);
312 struct sctp_endpoint *ep = sp->ep;
313 struct sctp_bind_addr *bp = &ep->base.bind_addr;
314 struct sctp_af *af;
315 unsigned short snum;
316 int ret = 0;
317
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 /* Common sockaddr verification. */
319 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700320 if (!af) {
321 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
322 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700324 }
325
326 snum = ntohs(addr->v4.sin_port);
327
328 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
329 ", port: %d, new port: %d, len: %d)\n",
330 sk,
331 addr,
332 bp->port, snum,
333 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
335 /* PF specific bind() address verification. */
336 if (!sp->pf->bind_verify(sp, addr))
337 return -EADDRNOTAVAIL;
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 /* We must either be unbound, or bind to the same port. */
340 if (bp->port && (snum != bp->port)) {
341 SCTP_DEBUG_PRINTK("sctp_do_bind:"
342 " New port %d does not match existing port "
343 "%d.\n", snum, bp->port);
344 return -EINVAL;
345 }
346
347 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
348 return -EACCES;
349
350 /* Make sure we are allowed to bind here.
351 * The function sctp_get_port_local() does duplicate address
352 * detection.
353 */
354 if ((ret = sctp_get_port_local(sk, addr))) {
355 if (ret == (long) sk) {
356 /* This endpoint has a conflicting address. */
357 return -EINVAL;
358 } else {
359 return -EADDRINUSE;
360 }
361 }
362
363 /* Refresh ephemeral port. */
364 if (!bp->port)
365 bp->port = inet_sk(sk)->num;
366
367 /* Add the address to the bind address list. */
368 sctp_local_bh_disable();
369 sctp_write_lock(&ep->base.addr_lock);
370
371 /* Use GFP_ATOMIC since BHs are disabled. */
372 addr->v4.sin_port = ntohs(addr->v4.sin_port);
373 ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC);
374 addr->v4.sin_port = htons(addr->v4.sin_port);
375 sctp_write_unlock(&ep->base.addr_lock);
376 sctp_local_bh_enable();
377
378 /* Copy back into socket for getsockname() use. */
379 if (!ret) {
380 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
381 af->to_sk_saddr(addr, sk);
382 }
383
384 return ret;
385}
386
387 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
388 *
389 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
390 * at any one time. If a sender, after sending an ASCONF chunk, decides
391 * it needs to transfer another ASCONF Chunk, it MUST wait until the
392 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
393 * subsequent ASCONF. Note this restriction binds each side, so at any
394 * time two ASCONF may be in-transit on any given association (one sent
395 * from each endpoint).
396 */
397static int sctp_send_asconf(struct sctp_association *asoc,
398 struct sctp_chunk *chunk)
399{
400 int retval = 0;
401
402 /* If there is an outstanding ASCONF chunk, queue it for later
403 * transmission.
404 */
405 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700406 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 goto out;
408 }
409
410 /* Hold the chunk until an ASCONF_ACK is received. */
411 sctp_chunk_hold(chunk);
412 retval = sctp_primitive_ASCONF(asoc, chunk);
413 if (retval)
414 sctp_chunk_free(chunk);
415 else
416 asoc->addip_last_asconf = chunk;
417
418out:
419 return retval;
420}
421
422/* Add a list of addresses as bind addresses to local endpoint or
423 * association.
424 *
425 * Basically run through each address specified in the addrs/addrcnt
426 * array/length pair, determine if it is IPv6 or IPv4 and call
427 * sctp_do_bind() on it.
428 *
429 * If any of them fails, then the operation will be reversed and the
430 * ones that were added will be removed.
431 *
432 * Only sctp_setsockopt_bindx() is supposed to call this function.
433 */
434int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
435{
436 int cnt;
437 int retval = 0;
438 void *addr_buf;
439 struct sockaddr *sa_addr;
440 struct sctp_af *af;
441
442 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
443 sk, addrs, addrcnt);
444
445 addr_buf = addrs;
446 for (cnt = 0; cnt < addrcnt; cnt++) {
447 /* The list may contain either IPv4 or IPv6 address;
448 * determine the address length for walking thru the list.
449 */
450 sa_addr = (struct sockaddr *)addr_buf;
451 af = sctp_get_af_specific(sa_addr->sa_family);
452 if (!af) {
453 retval = -EINVAL;
454 goto err_bindx_add;
455 }
456
457 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
458 af->sockaddr_len);
459
460 addr_buf += af->sockaddr_len;
461
462err_bindx_add:
463 if (retval < 0) {
464 /* Failed. Cleanup the ones that have been added */
465 if (cnt > 0)
466 sctp_bindx_rem(sk, addrs, cnt);
467 return retval;
468 }
469 }
470
471 return retval;
472}
473
474/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
475 * associations that are part of the endpoint indicating that a list of local
476 * addresses are added to the endpoint.
477 *
478 * If any of the addresses is already in the bind address list of the
479 * association, we do not send the chunk for that association. But it will not
480 * affect other associations.
481 *
482 * Only sctp_setsockopt_bindx() is supposed to call this function.
483 */
484static int sctp_send_asconf_add_ip(struct sock *sk,
485 struct sockaddr *addrs,
486 int addrcnt)
487{
488 struct sctp_sock *sp;
489 struct sctp_endpoint *ep;
490 struct sctp_association *asoc;
491 struct sctp_bind_addr *bp;
492 struct sctp_chunk *chunk;
493 struct sctp_sockaddr_entry *laddr;
494 union sctp_addr *addr;
495 void *addr_buf;
496 struct sctp_af *af;
497 struct list_head *pos;
498 struct list_head *p;
499 int i;
500 int retval = 0;
501
502 if (!sctp_addip_enable)
503 return retval;
504
505 sp = sctp_sk(sk);
506 ep = sp->ep;
507
508 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
509 __FUNCTION__, sk, addrs, addrcnt);
510
511 list_for_each(pos, &ep->asocs) {
512 asoc = list_entry(pos, struct sctp_association, asocs);
513
514 if (!asoc->peer.asconf_capable)
515 continue;
516
517 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
518 continue;
519
520 if (!sctp_state(asoc, ESTABLISHED))
521 continue;
522
523 /* Check if any address in the packed array of addresses is
524 * in the bind address list of the association. If so,
525 * do not send the asconf chunk to its peer, but continue with
526 * other associations.
527 */
528 addr_buf = addrs;
529 for (i = 0; i < addrcnt; i++) {
530 addr = (union sctp_addr *)addr_buf;
531 af = sctp_get_af_specific(addr->v4.sin_family);
532 if (!af) {
533 retval = -EINVAL;
534 goto out;
535 }
536
537 if (sctp_assoc_lookup_laddr(asoc, addr))
538 break;
539
540 addr_buf += af->sockaddr_len;
541 }
542 if (i < addrcnt)
543 continue;
544
545 /* Use the first address in bind addr list of association as
546 * Address Parameter of ASCONF CHUNK.
547 */
548 sctp_read_lock(&asoc->base.addr_lock);
549 bp = &asoc->base.bind_addr;
550 p = bp->address_list.next;
551 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
552 sctp_read_unlock(&asoc->base.addr_lock);
553
554 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
555 addrcnt, SCTP_PARAM_ADD_IP);
556 if (!chunk) {
557 retval = -ENOMEM;
558 goto out;
559 }
560
561 retval = sctp_send_asconf(asoc, chunk);
562
563 /* FIXME: After sending the add address ASCONF chunk, we
564 * cannot append the address to the association's binding
565 * address list, because the new address may be used as the
566 * source of a message sent to the peer before the ASCONF
567 * chunk is received by the peer. So we should wait until
568 * ASCONF_ACK is received.
569 */
570 }
571
572out:
573 return retval;
574}
575
576/* Remove a list of addresses from bind addresses list. Do not remove the
577 * last address.
578 *
579 * Basically run through each address specified in the addrs/addrcnt
580 * array/length pair, determine if it is IPv6 or IPv4 and call
581 * sctp_del_bind() on it.
582 *
583 * If any of them fails, then the operation will be reversed and the
584 * ones that were removed will be added back.
585 *
586 * At least one address has to be left; if only one address is
587 * available, the operation will return -EBUSY.
588 *
589 * Only sctp_setsockopt_bindx() is supposed to call this function.
590 */
591int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
592{
593 struct sctp_sock *sp = sctp_sk(sk);
594 struct sctp_endpoint *ep = sp->ep;
595 int cnt;
596 struct sctp_bind_addr *bp = &ep->base.bind_addr;
597 int retval = 0;
598 union sctp_addr saveaddr;
599 void *addr_buf;
600 struct sockaddr *sa_addr;
601 struct sctp_af *af;
602
603 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
604 sk, addrs, addrcnt);
605
606 addr_buf = addrs;
607 for (cnt = 0; cnt < addrcnt; cnt++) {
608 /* If the bind address list is empty or if there is only one
609 * bind address, there is nothing more to be removed (we need
610 * at least one address here).
611 */
612 if (list_empty(&bp->address_list) ||
613 (sctp_list_single_entry(&bp->address_list))) {
614 retval = -EBUSY;
615 goto err_bindx_rem;
616 }
617
618 /* The list may contain either IPv4 or IPv6 address;
619 * determine the address length to copy the address to
620 * saveaddr.
621 */
622 sa_addr = (struct sockaddr *)addr_buf;
623 af = sctp_get_af_specific(sa_addr->sa_family);
624 if (!af) {
625 retval = -EINVAL;
626 goto err_bindx_rem;
627 }
628 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
629 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
630 if (saveaddr.v4.sin_port != bp->port) {
631 retval = -EINVAL;
632 goto err_bindx_rem;
633 }
634
635 /* FIXME - There is probably a need to check if sk->sk_saddr and
636 * sk->sk_rcv_addr are currently set to one of the addresses to
637 * be removed. This is something which needs to be looked into
638 * when we are fixing the outstanding issues with multi-homing
639 * socket routing and failover schemes. Refer to comments in
640 * sctp_do_bind(). -daisy
641 */
642 sctp_local_bh_disable();
643 sctp_write_lock(&ep->base.addr_lock);
644
645 retval = sctp_del_bind_addr(bp, &saveaddr);
646
647 sctp_write_unlock(&ep->base.addr_lock);
648 sctp_local_bh_enable();
649
650 addr_buf += af->sockaddr_len;
651err_bindx_rem:
652 if (retval < 0) {
653 /* Failed. Add the ones that has been removed back */
654 if (cnt > 0)
655 sctp_bindx_add(sk, addrs, cnt);
656 return retval;
657 }
658 }
659
660 return retval;
661}
662
663/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
664 * the associations that are part of the endpoint indicating that a list of
665 * local addresses are removed from the endpoint.
666 *
667 * If any of the addresses is already in the bind address list of the
668 * association, we do not send the chunk for that association. But it will not
669 * affect other associations.
670 *
671 * Only sctp_setsockopt_bindx() is supposed to call this function.
672 */
673static int sctp_send_asconf_del_ip(struct sock *sk,
674 struct sockaddr *addrs,
675 int addrcnt)
676{
677 struct sctp_sock *sp;
678 struct sctp_endpoint *ep;
679 struct sctp_association *asoc;
680 struct sctp_bind_addr *bp;
681 struct sctp_chunk *chunk;
682 union sctp_addr *laddr;
683 void *addr_buf;
684 struct sctp_af *af;
685 struct list_head *pos;
686 int i;
687 int retval = 0;
688
689 if (!sctp_addip_enable)
690 return retval;
691
692 sp = sctp_sk(sk);
693 ep = sp->ep;
694
695 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
696 __FUNCTION__, sk, addrs, addrcnt);
697
698 list_for_each(pos, &ep->asocs) {
699 asoc = list_entry(pos, struct sctp_association, asocs);
700
701 if (!asoc->peer.asconf_capable)
702 continue;
703
704 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
705 continue;
706
707 if (!sctp_state(asoc, ESTABLISHED))
708 continue;
709
710 /* Check if any address in the packed array of addresses is
711 * not present in the bind address list of the association.
712 * If so, do not send the asconf chunk to its peer, but
713 * continue with other associations.
714 */
715 addr_buf = addrs;
716 for (i = 0; i < addrcnt; i++) {
717 laddr = (union sctp_addr *)addr_buf;
718 af = sctp_get_af_specific(laddr->v4.sin_family);
719 if (!af) {
720 retval = -EINVAL;
721 goto out;
722 }
723
724 if (!sctp_assoc_lookup_laddr(asoc, laddr))
725 break;
726
727 addr_buf += af->sockaddr_len;
728 }
729 if (i < addrcnt)
730 continue;
731
732 /* Find one address in the association's bind address list
733 * that is not in the packed array of addresses. This is to
734 * make sure that we do not delete all the addresses in the
735 * association.
736 */
737 sctp_read_lock(&asoc->base.addr_lock);
738 bp = &asoc->base.bind_addr;
739 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
740 addrcnt, sp);
741 sctp_read_unlock(&asoc->base.addr_lock);
742 if (!laddr)
743 continue;
744
745 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
746 SCTP_PARAM_DEL_IP);
747 if (!chunk) {
748 retval = -ENOMEM;
749 goto out;
750 }
751
752 retval = sctp_send_asconf(asoc, chunk);
753
754 /* FIXME: After sending the delete address ASCONF chunk, we
755 * cannot remove the addresses from the association's bind
756 * address list, because there maybe some packet send to
757 * the delete addresses, so we should wait until ASCONF_ACK
758 * packet is received.
759 */
760 }
761out:
762 return retval;
763}
764
765/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
766 *
767 * API 8.1
768 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
769 * int flags);
770 *
771 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
772 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
773 * or IPv6 addresses.
774 *
775 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
776 * Section 3.1.2 for this usage.
777 *
778 * addrs is a pointer to an array of one or more socket addresses. Each
779 * address is contained in its appropriate structure (i.e. struct
780 * sockaddr_in or struct sockaddr_in6) the family of the address type
781 * must be used to distengish the address length (note that this
782 * representation is termed a "packed array" of addresses). The caller
783 * specifies the number of addresses in the array with addrcnt.
784 *
785 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
786 * -1, and sets errno to the appropriate error code.
787 *
788 * For SCTP, the port given in each socket address must be the same, or
789 * sctp_bindx() will fail, setting errno to EINVAL.
790 *
791 * The flags parameter is formed from the bitwise OR of zero or more of
792 * the following currently defined flags:
793 *
794 * SCTP_BINDX_ADD_ADDR
795 *
796 * SCTP_BINDX_REM_ADDR
797 *
798 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
799 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
800 * addresses from the association. The two flags are mutually exclusive;
801 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
802 * not remove all addresses from an association; sctp_bindx() will
803 * reject such an attempt with EINVAL.
804 *
805 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
806 * additional addresses with an endpoint after calling bind(). Or use
807 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
808 * socket is associated with so that no new association accepted will be
809 * associated with those addresses. If the endpoint supports dynamic
810 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
811 * endpoint to send the appropriate message to the peer to change the
812 * peers address lists.
813 *
814 * Adding and removing addresses from a connected association is
815 * optional functionality. Implementations that do not support this
816 * functionality should return EOPNOTSUPP.
817 *
818 * Basically do nothing but copying the addresses from user to kernel
819 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700820 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
821 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 *
823 * We don't use copy_from_user() for optimization: we first do the
824 * sanity checks (buffer size -fast- and access check-healthy
825 * pointer); if all of those succeed, then we can alloc the memory
826 * (expensive operation) needed to copy the data to kernel. Then we do
827 * the copying without checking the user space area
828 * (__copy_from_user()).
829 *
830 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
831 * it.
832 *
833 * sk The sk of the socket
834 * addrs The pointer to the addresses in user land
835 * addrssize Size of the addrs buffer
836 * op Operation to perform (add or remove, see the flags of
837 * sctp_bindx)
838 *
839 * Returns 0 if ok, <0 errno code on error.
840 */
841SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
842 struct sockaddr __user *addrs,
843 int addrs_size, int op)
844{
845 struct sockaddr *kaddrs;
846 int err;
847 int addrcnt = 0;
848 int walk_size = 0;
849 struct sockaddr *sa_addr;
850 void *addr_buf;
851 struct sctp_af *af;
852
853 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
854 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
855
856 if (unlikely(addrs_size <= 0))
857 return -EINVAL;
858
859 /* Check the user passed a healthy pointer. */
860 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
861 return -EFAULT;
862
863 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800864 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 if (unlikely(!kaddrs))
866 return -ENOMEM;
867
868 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
869 kfree(kaddrs);
870 return -EFAULT;
871 }
872
873 /* Walk through the addrs buffer and count the number of addresses. */
874 addr_buf = kaddrs;
875 while (walk_size < addrs_size) {
876 sa_addr = (struct sockaddr *)addr_buf;
877 af = sctp_get_af_specific(sa_addr->sa_family);
878
879 /* If the address family is not supported or if this address
880 * causes the address buffer to overflow return EINVAL.
881 */
882 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
883 kfree(kaddrs);
884 return -EINVAL;
885 }
886 addrcnt++;
887 addr_buf += af->sockaddr_len;
888 walk_size += af->sockaddr_len;
889 }
890
891 /* Do the work. */
892 switch (op) {
893 case SCTP_BINDX_ADD_ADDR:
894 err = sctp_bindx_add(sk, kaddrs, addrcnt);
895 if (err)
896 goto out;
897 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
898 break;
899
900 case SCTP_BINDX_REM_ADDR:
901 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
902 if (err)
903 goto out;
904 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
905 break;
906
907 default:
908 err = -EINVAL;
909 break;
910 };
911
912out:
913 kfree(kaddrs);
914
915 return err;
916}
917
Frank Filz3f7a87d2005-06-20 13:14:57 -0700918/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
919 *
920 * Common routine for handling connect() and sctp_connectx().
921 * Connect will come in with just a single address.
922 */
923static int __sctp_connect(struct sock* sk,
924 struct sockaddr *kaddrs,
925 int addrs_size)
926{
927 struct sctp_sock *sp;
928 struct sctp_endpoint *ep;
929 struct sctp_association *asoc = NULL;
930 struct sctp_association *asoc2;
931 struct sctp_transport *transport;
932 union sctp_addr to;
933 struct sctp_af *af;
934 sctp_scope_t scope;
935 long timeo;
936 int err = 0;
937 int addrcnt = 0;
938 int walk_size = 0;
939 struct sockaddr *sa_addr;
940 void *addr_buf;
941
942 sp = sctp_sk(sk);
943 ep = sp->ep;
944
945 /* connect() cannot be done on a socket that is already in ESTABLISHED
946 * state - UDP-style peeled off socket or a TCP-style socket that
947 * is already connected.
948 * It cannot be done even on a TCP-style listening socket.
949 */
950 if (sctp_sstate(sk, ESTABLISHED) ||
951 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
952 err = -EISCONN;
953 goto out_free;
954 }
955
956 /* Walk through the addrs buffer and count the number of addresses. */
957 addr_buf = kaddrs;
958 while (walk_size < addrs_size) {
959 sa_addr = (struct sockaddr *)addr_buf;
960 af = sctp_get_af_specific(sa_addr->sa_family);
961
962 /* If the address family is not supported or if this address
963 * causes the address buffer to overflow return EINVAL.
964 */
965 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
966 err = -EINVAL;
967 goto out_free;
968 }
969
970 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
971 af->sockaddr_len);
972 if (err)
973 goto out_free;
974
975 memcpy(&to, sa_addr, af->sockaddr_len);
976 to.v4.sin_port = ntohs(to.v4.sin_port);
977
978 /* Check if there already is a matching association on the
979 * endpoint (other than the one created here).
980 */
981 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
982 if (asoc2 && asoc2 != asoc) {
983 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
984 err = -EISCONN;
985 else
986 err = -EALREADY;
987 goto out_free;
988 }
989
990 /* If we could not find a matching association on the endpoint,
991 * make sure that there is no peeled-off association matching
992 * the peer address even on another socket.
993 */
994 if (sctp_endpoint_is_peeled_off(ep, &to)) {
995 err = -EADDRNOTAVAIL;
996 goto out_free;
997 }
998
999 if (!asoc) {
1000 /* If a bind() or sctp_bindx() is not called prior to
1001 * an sctp_connectx() call, the system picks an
1002 * ephemeral port and will choose an address set
1003 * equivalent to binding with a wildcard address.
1004 */
1005 if (!ep->base.bind_addr.port) {
1006 if (sctp_autobind(sk)) {
1007 err = -EAGAIN;
1008 goto out_free;
1009 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001010 } else {
1011 /*
1012 * If an unprivileged user inherits a 1-many
1013 * style socket with open associations on a
1014 * privileged port, it MAY be permitted to
1015 * accept new associations, but it SHOULD NOT
1016 * be permitted to open new associations.
1017 */
1018 if (ep->base.bind_addr.port < PROT_SOCK &&
1019 !capable(CAP_NET_BIND_SERVICE)) {
1020 err = -EACCES;
1021 goto out_free;
1022 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001023 }
1024
1025 scope = sctp_scope(&to);
1026 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1027 if (!asoc) {
1028 err = -ENOMEM;
1029 goto out_free;
1030 }
1031 }
1032
1033 /* Prime the peer's transport structures. */
1034 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1035 SCTP_UNKNOWN);
1036 if (!transport) {
1037 err = -ENOMEM;
1038 goto out_free;
1039 }
1040
1041 addrcnt++;
1042 addr_buf += af->sockaddr_len;
1043 walk_size += af->sockaddr_len;
1044 }
1045
1046 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1047 if (err < 0) {
1048 goto out_free;
1049 }
1050
1051 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1052 if (err < 0) {
1053 goto out_free;
1054 }
1055
1056 /* Initialize sk's dport and daddr for getpeername() */
1057 inet_sk(sk)->dport = htons(asoc->peer.port);
1058 af = sctp_get_af_specific(to.sa.sa_family);
1059 af->to_sk_daddr(&to, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001060 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001061
1062 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1063 err = sctp_wait_for_connect(asoc, &timeo);
1064
1065 /* Don't free association on exit. */
1066 asoc = NULL;
1067
1068out_free:
1069
1070 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1071 " kaddrs: %p err: %d\n",
1072 asoc, kaddrs, err);
1073 if (asoc)
1074 sctp_association_free(asoc);
1075 return err;
1076}
1077
1078/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1079 *
1080 * API 8.9
1081 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1082 *
1083 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1084 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1085 * or IPv6 addresses.
1086 *
1087 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1088 * Section 3.1.2 for this usage.
1089 *
1090 * addrs is a pointer to an array of one or more socket addresses. Each
1091 * address is contained in its appropriate structure (i.e. struct
1092 * sockaddr_in or struct sockaddr_in6) the family of the address type
1093 * must be used to distengish the address length (note that this
1094 * representation is termed a "packed array" of addresses). The caller
1095 * specifies the number of addresses in the array with addrcnt.
1096 *
1097 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1098 * -1, and sets errno to the appropriate error code.
1099 *
1100 * For SCTP, the port given in each socket address must be the same, or
1101 * sctp_connectx() will fail, setting errno to EINVAL.
1102 *
1103 * An application can use sctp_connectx to initiate an association with
1104 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1105 * allows a caller to specify multiple addresses at which a peer can be
1106 * reached. The way the SCTP stack uses the list of addresses to set up
1107 * the association is implementation dependant. This function only
1108 * specifies that the stack will try to make use of all the addresses in
1109 * the list when needed.
1110 *
1111 * Note that the list of addresses passed in is only used for setting up
1112 * the association. It does not necessarily equal the set of addresses
1113 * the peer uses for the resulting association. If the caller wants to
1114 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1115 * retrieve them after the association has been set up.
1116 *
1117 * Basically do nothing but copying the addresses from user to kernel
1118 * land and invoking either sctp_connectx(). This is used for tunneling
1119 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1120 *
1121 * We don't use copy_from_user() for optimization: we first do the
1122 * sanity checks (buffer size -fast- and access check-healthy
1123 * pointer); if all of those succeed, then we can alloc the memory
1124 * (expensive operation) needed to copy the data to kernel. Then we do
1125 * the copying without checking the user space area
1126 * (__copy_from_user()).
1127 *
1128 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1129 * it.
1130 *
1131 * sk The sk of the socket
1132 * addrs The pointer to the addresses in user land
1133 * addrssize Size of the addrs buffer
1134 *
1135 * Returns 0 if ok, <0 errno code on error.
1136 */
1137SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1138 struct sockaddr __user *addrs,
1139 int addrs_size)
1140{
1141 int err = 0;
1142 struct sockaddr *kaddrs;
1143
1144 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1145 __FUNCTION__, sk, addrs, addrs_size);
1146
1147 if (unlikely(addrs_size <= 0))
1148 return -EINVAL;
1149
1150 /* Check the user passed a healthy pointer. */
1151 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1152 return -EFAULT;
1153
1154 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001155 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001156 if (unlikely(!kaddrs))
1157 return -ENOMEM;
1158
1159 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1160 err = -EFAULT;
1161 } else {
1162 err = __sctp_connect(sk, kaddrs, addrs_size);
1163 }
1164
1165 kfree(kaddrs);
1166 return err;
1167}
1168
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169/* API 3.1.4 close() - UDP Style Syntax
1170 * Applications use close() to perform graceful shutdown (as described in
1171 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1172 * by a UDP-style socket.
1173 *
1174 * The syntax is
1175 *
1176 * ret = close(int sd);
1177 *
1178 * sd - the socket descriptor of the associations to be closed.
1179 *
1180 * To gracefully shutdown a specific association represented by the
1181 * UDP-style socket, an application should use the sendmsg() call,
1182 * passing no user data, but including the appropriate flag in the
1183 * ancillary data (see Section xxxx).
1184 *
1185 * If sd in the close() call is a branched-off socket representing only
1186 * one association, the shutdown is performed on that association only.
1187 *
1188 * 4.1.6 close() - TCP Style Syntax
1189 *
1190 * Applications use close() to gracefully close down an association.
1191 *
1192 * The syntax is:
1193 *
1194 * int close(int sd);
1195 *
1196 * sd - the socket descriptor of the association to be closed.
1197 *
1198 * After an application calls close() on a socket descriptor, no further
1199 * socket operations will succeed on that descriptor.
1200 *
1201 * API 7.1.4 SO_LINGER
1202 *
1203 * An application using the TCP-style socket can use this option to
1204 * perform the SCTP ABORT primitive. The linger option structure is:
1205 *
1206 * struct linger {
1207 * int l_onoff; // option on/off
1208 * int l_linger; // linger time
1209 * };
1210 *
1211 * To enable the option, set l_onoff to 1. If the l_linger value is set
1212 * to 0, calling close() is the same as the ABORT primitive. If the
1213 * value is set to a negative value, the setsockopt() call will return
1214 * an error. If the value is set to a positive value linger_time, the
1215 * close() can be blocked for at most linger_time ms. If the graceful
1216 * shutdown phase does not finish during this period, close() will
1217 * return but the graceful shutdown phase continues in the system.
1218 */
1219SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1220{
1221 struct sctp_endpoint *ep;
1222 struct sctp_association *asoc;
1223 struct list_head *pos, *temp;
1224
1225 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1226
1227 sctp_lock_sock(sk);
1228 sk->sk_shutdown = SHUTDOWN_MASK;
1229
1230 ep = sctp_sk(sk)->ep;
1231
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001232 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 list_for_each_safe(pos, temp, &ep->asocs) {
1234 asoc = list_entry(pos, struct sctp_association, asocs);
1235
1236 if (sctp_style(sk, TCP)) {
1237 /* A closed association can still be in the list if
1238 * it belongs to a TCP-style listening socket that is
1239 * not yet accepted. If so, free it. If not, send an
1240 * ABORT or SHUTDOWN based on the linger options.
1241 */
1242 if (sctp_state(asoc, CLOSED)) {
1243 sctp_unhash_established(asoc);
1244 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001245 continue;
1246 }
1247 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001249 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)
1250 sctp_primitive_ABORT(asoc, NULL);
1251 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 sctp_primitive_SHUTDOWN(asoc, NULL);
1253 }
1254
1255 /* Clean up any skbs sitting on the receive queue. */
1256 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1257 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1258
1259 /* On a TCP-style socket, block for at most linger_time if set. */
1260 if (sctp_style(sk, TCP) && timeout)
1261 sctp_wait_for_close(sk, timeout);
1262
1263 /* This will run the backlog queue. */
1264 sctp_release_sock(sk);
1265
1266 /* Supposedly, no process has access to the socket, but
1267 * the net layers still may.
1268 */
1269 sctp_local_bh_disable();
1270 sctp_bh_lock_sock(sk);
1271
1272 /* Hold the sock, since sk_common_release() will put sock_put()
1273 * and we have just a little more cleanup.
1274 */
1275 sock_hold(sk);
1276 sk_common_release(sk);
1277
1278 sctp_bh_unlock_sock(sk);
1279 sctp_local_bh_enable();
1280
1281 sock_put(sk);
1282
1283 SCTP_DBG_OBJCNT_DEC(sock);
1284}
1285
1286/* Handle EPIPE error. */
1287static int sctp_error(struct sock *sk, int flags, int err)
1288{
1289 if (err == -EPIPE)
1290 err = sock_error(sk) ? : -EPIPE;
1291 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1292 send_sig(SIGPIPE, current, 0);
1293 return err;
1294}
1295
1296/* API 3.1.3 sendmsg() - UDP Style Syntax
1297 *
1298 * An application uses sendmsg() and recvmsg() calls to transmit data to
1299 * and receive data from its peer.
1300 *
1301 * ssize_t sendmsg(int socket, const struct msghdr *message,
1302 * int flags);
1303 *
1304 * socket - the socket descriptor of the endpoint.
1305 * message - pointer to the msghdr structure which contains a single
1306 * user message and possibly some ancillary data.
1307 *
1308 * See Section 5 for complete description of the data
1309 * structures.
1310 *
1311 * flags - flags sent or received with the user message, see Section
1312 * 5 for complete description of the flags.
1313 *
1314 * Note: This function could use a rewrite especially when explicit
1315 * connect support comes in.
1316 */
1317/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1318
1319SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1320
1321SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1322 struct msghdr *msg, size_t msg_len)
1323{
1324 struct sctp_sock *sp;
1325 struct sctp_endpoint *ep;
1326 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1327 struct sctp_transport *transport, *chunk_tp;
1328 struct sctp_chunk *chunk;
1329 union sctp_addr to;
1330 struct sockaddr *msg_name = NULL;
1331 struct sctp_sndrcvinfo default_sinfo = { 0 };
1332 struct sctp_sndrcvinfo *sinfo;
1333 struct sctp_initmsg *sinit;
1334 sctp_assoc_t associd = 0;
1335 sctp_cmsgs_t cmsgs = { NULL };
1336 int err;
1337 sctp_scope_t scope;
1338 long timeo;
1339 __u16 sinfo_flags = 0;
1340 struct sctp_datamsg *datamsg;
1341 struct list_head *pos;
1342 int msg_flags = msg->msg_flags;
1343
1344 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1345 sk, msg, msg_len);
1346
1347 err = 0;
1348 sp = sctp_sk(sk);
1349 ep = sp->ep;
1350
Frank Filz3f7a87d2005-06-20 13:14:57 -07001351 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
1353 /* We cannot send a message over a TCP-style listening socket. */
1354 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1355 err = -EPIPE;
1356 goto out_nounlock;
1357 }
1358
1359 /* Parse out the SCTP CMSGs. */
1360 err = sctp_msghdr_parse(msg, &cmsgs);
1361
1362 if (err) {
1363 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1364 goto out_nounlock;
1365 }
1366
1367 /* Fetch the destination address for this packet. This
1368 * address only selects the association--it is not necessarily
1369 * the address we will send to.
1370 * For a peeled-off socket, msg_name is ignored.
1371 */
1372 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1373 int msg_namelen = msg->msg_namelen;
1374
1375 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1376 msg_namelen);
1377 if (err)
1378 return err;
1379
1380 if (msg_namelen > sizeof(to))
1381 msg_namelen = sizeof(to);
1382 memcpy(&to, msg->msg_name, msg_namelen);
1383 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1384 "0x%x:%u.\n",
1385 to.v4.sin_addr.s_addr, to.v4.sin_port);
1386
1387 to.v4.sin_port = ntohs(to.v4.sin_port);
1388 msg_name = msg->msg_name;
1389 }
1390
1391 sinfo = cmsgs.info;
1392 sinit = cmsgs.init;
1393
1394 /* Did the user specify SNDRCVINFO? */
1395 if (sinfo) {
1396 sinfo_flags = sinfo->sinfo_flags;
1397 associd = sinfo->sinfo_assoc_id;
1398 }
1399
1400 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1401 msg_len, sinfo_flags);
1402
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001403 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1404 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 err = -EINVAL;
1406 goto out_nounlock;
1407 }
1408
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001409 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1410 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1411 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 * the msg_iov set to the user abort reason.
1413 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001414 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1415 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 err = -EINVAL;
1417 goto out_nounlock;
1418 }
1419
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001420 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 * specified in msg_name.
1422 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001423 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 err = -EINVAL;
1425 goto out_nounlock;
1426 }
1427
1428 transport = NULL;
1429
1430 SCTP_DEBUG_PRINTK("About to look up association.\n");
1431
1432 sctp_lock_sock(sk);
1433
1434 /* If a msg_name has been specified, assume this is to be used. */
1435 if (msg_name) {
1436 /* Look for a matching association on the endpoint. */
1437 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1438 if (!asoc) {
1439 /* If we could not find a matching association on the
1440 * endpoint, make sure that it is not a TCP-style
1441 * socket that already has an association or there is
1442 * no peeled-off association on another socket.
1443 */
1444 if ((sctp_style(sk, TCP) &&
1445 sctp_sstate(sk, ESTABLISHED)) ||
1446 sctp_endpoint_is_peeled_off(ep, &to)) {
1447 err = -EADDRNOTAVAIL;
1448 goto out_unlock;
1449 }
1450 }
1451 } else {
1452 asoc = sctp_id2assoc(sk, associd);
1453 if (!asoc) {
1454 err = -EPIPE;
1455 goto out_unlock;
1456 }
1457 }
1458
1459 if (asoc) {
1460 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1461
1462 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1463 * socket that has an association in CLOSED state. This can
1464 * happen when an accepted socket has an association that is
1465 * already CLOSED.
1466 */
1467 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1468 err = -EPIPE;
1469 goto out_unlock;
1470 }
1471
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001472 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1474 asoc);
1475 sctp_primitive_SHUTDOWN(asoc, NULL);
1476 err = 0;
1477 goto out_unlock;
1478 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001479 if (sinfo_flags & SCTP_ABORT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1481 sctp_primitive_ABORT(asoc, msg);
1482 err = 0;
1483 goto out_unlock;
1484 }
1485 }
1486
1487 /* Do we need to create the association? */
1488 if (!asoc) {
1489 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1490
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001491 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 err = -EINVAL;
1493 goto out_unlock;
1494 }
1495
1496 /* Check for invalid stream against the stream counts,
1497 * either the default or the user specified stream counts.
1498 */
1499 if (sinfo) {
1500 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1501 /* Check against the defaults. */
1502 if (sinfo->sinfo_stream >=
1503 sp->initmsg.sinit_num_ostreams) {
1504 err = -EINVAL;
1505 goto out_unlock;
1506 }
1507 } else {
1508 /* Check against the requested. */
1509 if (sinfo->sinfo_stream >=
1510 sinit->sinit_num_ostreams) {
1511 err = -EINVAL;
1512 goto out_unlock;
1513 }
1514 }
1515 }
1516
1517 /*
1518 * API 3.1.2 bind() - UDP Style Syntax
1519 * If a bind() or sctp_bindx() is not called prior to a
1520 * sendmsg() call that initiates a new association, the
1521 * system picks an ephemeral port and will choose an address
1522 * set equivalent to binding with a wildcard address.
1523 */
1524 if (!ep->base.bind_addr.port) {
1525 if (sctp_autobind(sk)) {
1526 err = -EAGAIN;
1527 goto out_unlock;
1528 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001529 } else {
1530 /*
1531 * If an unprivileged user inherits a one-to-many
1532 * style socket with open associations on a privileged
1533 * port, it MAY be permitted to accept new associations,
1534 * but it SHOULD NOT be permitted to open new
1535 * associations.
1536 */
1537 if (ep->base.bind_addr.port < PROT_SOCK &&
1538 !capable(CAP_NET_BIND_SERVICE)) {
1539 err = -EACCES;
1540 goto out_unlock;
1541 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 }
1543
1544 scope = sctp_scope(&to);
1545 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1546 if (!new_asoc) {
1547 err = -ENOMEM;
1548 goto out_unlock;
1549 }
1550 asoc = new_asoc;
1551
1552 /* If the SCTP_INIT ancillary data is specified, set all
1553 * the association init values accordingly.
1554 */
1555 if (sinit) {
1556 if (sinit->sinit_num_ostreams) {
1557 asoc->c.sinit_num_ostreams =
1558 sinit->sinit_num_ostreams;
1559 }
1560 if (sinit->sinit_max_instreams) {
1561 asoc->c.sinit_max_instreams =
1562 sinit->sinit_max_instreams;
1563 }
1564 if (sinit->sinit_max_attempts) {
1565 asoc->max_init_attempts
1566 = sinit->sinit_max_attempts;
1567 }
1568 if (sinit->sinit_max_init_timeo) {
1569 asoc->max_init_timeo =
1570 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1571 }
1572 }
1573
1574 /* Prime the peer's transport structures. */
Frank Filz3f7a87d2005-06-20 13:14:57 -07001575 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (!transport) {
1577 err = -ENOMEM;
1578 goto out_free;
1579 }
1580 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1581 if (err < 0) {
1582 err = -ENOMEM;
1583 goto out_free;
1584 }
1585 }
1586
1587 /* ASSERT: we have a valid association at this point. */
1588 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1589
1590 if (!sinfo) {
1591 /* If the user didn't specify SNDRCVINFO, make up one with
1592 * some defaults.
1593 */
1594 default_sinfo.sinfo_stream = asoc->default_stream;
1595 default_sinfo.sinfo_flags = asoc->default_flags;
1596 default_sinfo.sinfo_ppid = asoc->default_ppid;
1597 default_sinfo.sinfo_context = asoc->default_context;
1598 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1599 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1600 sinfo = &default_sinfo;
1601 }
1602
1603 /* API 7.1.7, the sndbuf size per association bounds the
1604 * maximum size of data that can be sent in a single send call.
1605 */
1606 if (msg_len > sk->sk_sndbuf) {
1607 err = -EMSGSIZE;
1608 goto out_free;
1609 }
1610
1611 /* If fragmentation is disabled and the message length exceeds the
1612 * association fragmentation point, return EMSGSIZE. The I-D
1613 * does not specify what this error is, but this looks like
1614 * a great fit.
1615 */
1616 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1617 err = -EMSGSIZE;
1618 goto out_free;
1619 }
1620
1621 if (sinfo) {
1622 /* Check for invalid stream. */
1623 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1624 err = -EINVAL;
1625 goto out_free;
1626 }
1627 }
1628
1629 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1630 if (!sctp_wspace(asoc)) {
1631 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1632 if (err)
1633 goto out_free;
1634 }
1635
1636 /* If an address is passed with the sendto/sendmsg call, it is used
1637 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001638 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 */
1640 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001641 (sinfo_flags & SCTP_ADDR_OVER)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1643 if (!chunk_tp) {
1644 err = -EINVAL;
1645 goto out_free;
1646 }
1647 } else
1648 chunk_tp = NULL;
1649
1650 /* Auto-connect, if we aren't connected already. */
1651 if (sctp_state(asoc, CLOSED)) {
1652 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1653 if (err < 0)
1654 goto out_free;
1655 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1656 }
1657
1658 /* Break the message into multiple chunks of maximum size. */
1659 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1660 if (!datamsg) {
1661 err = -ENOMEM;
1662 goto out_free;
1663 }
1664
1665 /* Now send the (possibly) fragmented message. */
1666 list_for_each(pos, &datamsg->chunks) {
1667 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1668 sctp_datamsg_track(chunk);
1669
1670 /* Do accounting for the write space. */
1671 sctp_set_owner_w(chunk);
1672
1673 chunk->transport = chunk_tp;
1674
1675 /* Send it to the lower layers. Note: all chunks
1676 * must either fail or succeed. The lower layer
1677 * works that way today. Keep it that way or this
1678 * breaks.
1679 */
1680 err = sctp_primitive_SEND(asoc, chunk);
1681 /* Did the lower layer accept the chunk? */
1682 if (err)
1683 sctp_chunk_free(chunk);
1684 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1685 }
1686
1687 sctp_datamsg_free(datamsg);
1688 if (err)
1689 goto out_free;
1690 else
1691 err = msg_len;
1692
1693 /* If we are already past ASSOCIATE, the lower
1694 * layers are responsible for association cleanup.
1695 */
1696 goto out_unlock;
1697
1698out_free:
1699 if (new_asoc)
1700 sctp_association_free(asoc);
1701out_unlock:
1702 sctp_release_sock(sk);
1703
1704out_nounlock:
1705 return sctp_error(sk, msg_flags, err);
1706
1707#if 0
1708do_sock_err:
1709 if (msg_len)
1710 err = msg_len;
1711 else
1712 err = sock_error(sk);
1713 goto out;
1714
1715do_interrupted:
1716 if (msg_len)
1717 err = msg_len;
1718 goto out;
1719#endif /* 0 */
1720}
1721
1722/* This is an extended version of skb_pull() that removes the data from the
1723 * start of a skb even when data is spread across the list of skb's in the
1724 * frag_list. len specifies the total amount of data that needs to be removed.
1725 * when 'len' bytes could be removed from the skb, it returns 0.
1726 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1727 * could not be removed.
1728 */
1729static int sctp_skb_pull(struct sk_buff *skb, int len)
1730{
1731 struct sk_buff *list;
1732 int skb_len = skb_headlen(skb);
1733 int rlen;
1734
1735 if (len <= skb_len) {
1736 __skb_pull(skb, len);
1737 return 0;
1738 }
1739 len -= skb_len;
1740 __skb_pull(skb, skb_len);
1741
1742 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1743 rlen = sctp_skb_pull(list, len);
1744 skb->len -= (len-rlen);
1745 skb->data_len -= (len-rlen);
1746
1747 if (!rlen)
1748 return 0;
1749
1750 len = rlen;
1751 }
1752
1753 return len;
1754}
1755
1756/* API 3.1.3 recvmsg() - UDP Style Syntax
1757 *
1758 * ssize_t recvmsg(int socket, struct msghdr *message,
1759 * int flags);
1760 *
1761 * socket - the socket descriptor of the endpoint.
1762 * message - pointer to the msghdr structure which contains a single
1763 * user message and possibly some ancillary data.
1764 *
1765 * See Section 5 for complete description of the data
1766 * structures.
1767 *
1768 * flags - flags sent or received with the user message, see Section
1769 * 5 for complete description of the flags.
1770 */
1771static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1772
1773SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1774 struct msghdr *msg, size_t len, int noblock,
1775 int flags, int *addr_len)
1776{
1777 struct sctp_ulpevent *event = NULL;
1778 struct sctp_sock *sp = sctp_sk(sk);
1779 struct sk_buff *skb;
1780 int copied;
1781 int err = 0;
1782 int skb_len;
1783
1784 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1785 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1786 "len", len, "knoblauch", noblock,
1787 "flags", flags, "addr_len", addr_len);
1788
1789 sctp_lock_sock(sk);
1790
1791 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1792 err = -ENOTCONN;
1793 goto out;
1794 }
1795
1796 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1797 if (!skb)
1798 goto out;
1799
1800 /* Get the total length of the skb including any skb's in the
1801 * frag_list.
1802 */
1803 skb_len = skb->len;
1804
1805 copied = skb_len;
1806 if (copied > len)
1807 copied = len;
1808
1809 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1810
1811 event = sctp_skb2event(skb);
1812
1813 if (err)
1814 goto out_free;
1815
1816 sock_recv_timestamp(msg, sk, skb);
1817 if (sctp_ulpevent_is_notification(event)) {
1818 msg->msg_flags |= MSG_NOTIFICATION;
1819 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1820 } else {
1821 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1822 }
1823
1824 /* Check if we allow SCTP_SNDRCVINFO. */
1825 if (sp->subscribe.sctp_data_io_event)
1826 sctp_ulpevent_read_sndrcvinfo(event, msg);
1827#if 0
1828 /* FIXME: we should be calling IP/IPv6 layers. */
1829 if (sk->sk_protinfo.af_inet.cmsg_flags)
1830 ip_cmsg_recv(msg, skb);
1831#endif
1832
1833 err = copied;
1834
1835 /* If skb's length exceeds the user's buffer, update the skb and
1836 * push it back to the receive_queue so that the next call to
1837 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1838 */
1839 if (skb_len > copied) {
1840 msg->msg_flags &= ~MSG_EOR;
1841 if (flags & MSG_PEEK)
1842 goto out_free;
1843 sctp_skb_pull(skb, copied);
1844 skb_queue_head(&sk->sk_receive_queue, skb);
1845
1846 /* When only partial message is copied to the user, increase
1847 * rwnd by that amount. If all the data in the skb is read,
1848 * rwnd is updated when the event is freed.
1849 */
1850 sctp_assoc_rwnd_increase(event->asoc, copied);
1851 goto out;
1852 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1853 (event->msg_flags & MSG_EOR))
1854 msg->msg_flags |= MSG_EOR;
1855 else
1856 msg->msg_flags &= ~MSG_EOR;
1857
1858out_free:
1859 if (flags & MSG_PEEK) {
1860 /* Release the skb reference acquired after peeking the skb in
1861 * sctp_skb_recv_datagram().
1862 */
1863 kfree_skb(skb);
1864 } else {
1865 /* Free the event which includes releasing the reference to
1866 * the owner of the skb, freeing the skb and updating the
1867 * rwnd.
1868 */
1869 sctp_ulpevent_free(event);
1870 }
1871out:
1872 sctp_release_sock(sk);
1873 return err;
1874}
1875
1876/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1877 *
1878 * This option is a on/off flag. If enabled no SCTP message
1879 * fragmentation will be performed. Instead if a message being sent
1880 * exceeds the current PMTU size, the message will NOT be sent and
1881 * instead a error will be indicated to the user.
1882 */
1883static int sctp_setsockopt_disable_fragments(struct sock *sk,
1884 char __user *optval, int optlen)
1885{
1886 int val;
1887
1888 if (optlen < sizeof(int))
1889 return -EINVAL;
1890
1891 if (get_user(val, (int __user *)optval))
1892 return -EFAULT;
1893
1894 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1895
1896 return 0;
1897}
1898
1899static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1900 int optlen)
1901{
1902 if (optlen != sizeof(struct sctp_event_subscribe))
1903 return -EINVAL;
1904 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1905 return -EFAULT;
1906 return 0;
1907}
1908
1909/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1910 *
1911 * This socket option is applicable to the UDP-style socket only. When
1912 * set it will cause associations that are idle for more than the
1913 * specified number of seconds to automatically close. An association
1914 * being idle is defined an association that has NOT sent or received
1915 * user data. The special value of '0' indicates that no automatic
1916 * close of any associations should be performed. The option expects an
1917 * integer defining the number of seconds of idle time before an
1918 * association is closed.
1919 */
1920static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1921 int optlen)
1922{
1923 struct sctp_sock *sp = sctp_sk(sk);
1924
1925 /* Applicable to UDP-style socket only */
1926 if (sctp_style(sk, TCP))
1927 return -EOPNOTSUPP;
1928 if (optlen != sizeof(int))
1929 return -EINVAL;
1930 if (copy_from_user(&sp->autoclose, optval, optlen))
1931 return -EFAULT;
1932
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 return 0;
1934}
1935
1936/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1937 *
1938 * Applications can enable or disable heartbeats for any peer address of
1939 * an association, modify an address's heartbeat interval, force a
1940 * heartbeat to be sent immediately, and adjust the address's maximum
1941 * number of retransmissions sent before an address is considered
1942 * unreachable. The following structure is used to access and modify an
1943 * address's parameters:
1944 *
1945 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08001946 * sctp_assoc_t spp_assoc_id;
1947 * struct sockaddr_storage spp_address;
1948 * uint32_t spp_hbinterval;
1949 * uint16_t spp_pathmaxrxt;
1950 * uint32_t spp_pathmtu;
1951 * uint32_t spp_sackdelay;
1952 * uint32_t spp_flags;
1953 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08001955 * spp_assoc_id - (one-to-many style socket) This is filled in the
1956 * application, and identifies the association for
1957 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 * spp_address - This specifies which address is of interest.
1959 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08001960 * in milliseconds. If a value of zero
1961 * is present in this field then no changes are to
1962 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 * spp_pathmaxrxt - This contains the maximum number of
1964 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08001965 * considered unreachable. If a value of zero
1966 * is present in this field then no changes are to
1967 * be made to this parameter.
1968 * spp_pathmtu - When Path MTU discovery is disabled the value
1969 * specified here will be the "fixed" path mtu.
1970 * Note that if the spp_address field is empty
1971 * then all associations on this address will
1972 * have this fixed path mtu set upon them.
1973 *
1974 * spp_sackdelay - When delayed sack is enabled, this value specifies
1975 * the number of milliseconds that sacks will be delayed
1976 * for. This value will apply to all addresses of an
1977 * association if the spp_address field is empty. Note
1978 * also, that if delayed sack is enabled and this
1979 * value is set to 0, no change is made to the last
1980 * recorded delayed sack timer value.
1981 *
1982 * spp_flags - These flags are used to control various features
1983 * on an association. The flag field may contain
1984 * zero or more of the following options.
1985 *
1986 * SPP_HB_ENABLE - Enable heartbeats on the
1987 * specified address. Note that if the address
1988 * field is empty all addresses for the association
1989 * have heartbeats enabled upon them.
1990 *
1991 * SPP_HB_DISABLE - Disable heartbeats on the
1992 * speicifed address. Note that if the address
1993 * field is empty all addresses for the association
1994 * will have their heartbeats disabled. Note also
1995 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
1996 * mutually exclusive, only one of these two should
1997 * be specified. Enabling both fields will have
1998 * undetermined results.
1999 *
2000 * SPP_HB_DEMAND - Request a user initiated heartbeat
2001 * to be made immediately.
2002 *
2003 * SPP_PMTUD_ENABLE - This field will enable PMTU
2004 * discovery upon the specified address. Note that
2005 * if the address feild is empty then all addresses
2006 * on the association are effected.
2007 *
2008 * SPP_PMTUD_DISABLE - This field will disable PMTU
2009 * discovery upon the specified address. Note that
2010 * if the address feild is empty then all addresses
2011 * on the association are effected. Not also that
2012 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2013 * exclusive. Enabling both will have undetermined
2014 * results.
2015 *
2016 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2017 * on delayed sack. The time specified in spp_sackdelay
2018 * is used to specify the sack delay for this address. Note
2019 * that if spp_address is empty then all addresses will
2020 * enable delayed sack and take on the sack delay
2021 * value specified in spp_sackdelay.
2022 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2023 * off delayed sack. If the spp_address field is blank then
2024 * delayed sack is disabled for the entire association. Note
2025 * also that this field is mutually exclusive to
2026 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2027 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08002029int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2030 struct sctp_transport *trans,
2031 struct sctp_association *asoc,
2032 struct sctp_sock *sp,
2033 int hb_change,
2034 int pmtud_change,
2035 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 int error;
2038
Frank Filz52ccb8e2005-12-22 11:36:46 -08002039 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2041 if (error)
2042 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 }
2044
Frank Filz52ccb8e2005-12-22 11:36:46 -08002045 if (params->spp_hbinterval) {
2046 if (trans) {
2047 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2048 } else if (asoc) {
2049 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2050 } else {
2051 sp->hbinterval = params->spp_hbinterval;
2052 }
2053 }
2054
2055 if (hb_change) {
2056 if (trans) {
2057 trans->param_flags =
2058 (trans->param_flags & ~SPP_HB) | hb_change;
2059 } else if (asoc) {
2060 asoc->param_flags =
2061 (asoc->param_flags & ~SPP_HB) | hb_change;
2062 } else {
2063 sp->param_flags =
2064 (sp->param_flags & ~SPP_HB) | hb_change;
2065 }
2066 }
2067
2068 if (params->spp_pathmtu) {
2069 if (trans) {
2070 trans->pathmtu = params->spp_pathmtu;
2071 sctp_assoc_sync_pmtu(asoc);
2072 } else if (asoc) {
2073 asoc->pathmtu = params->spp_pathmtu;
2074 sctp_frag_point(sp, params->spp_pathmtu);
2075 } else {
2076 sp->pathmtu = params->spp_pathmtu;
2077 }
2078 }
2079
2080 if (pmtud_change) {
2081 if (trans) {
2082 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2083 (params->spp_flags & SPP_PMTUD_ENABLE);
2084 trans->param_flags =
2085 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2086 if (update) {
2087 sctp_transport_pmtu(trans);
2088 sctp_assoc_sync_pmtu(asoc);
2089 }
2090 } else if (asoc) {
2091 asoc->param_flags =
2092 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2093 } else {
2094 sp->param_flags =
2095 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2096 }
2097 }
2098
2099 if (params->spp_sackdelay) {
2100 if (trans) {
2101 trans->sackdelay =
2102 msecs_to_jiffies(params->spp_sackdelay);
2103 } else if (asoc) {
2104 asoc->sackdelay =
2105 msecs_to_jiffies(params->spp_sackdelay);
2106 } else {
2107 sp->sackdelay = params->spp_sackdelay;
2108 }
2109 }
2110
2111 if (sackdelay_change) {
2112 if (trans) {
2113 trans->param_flags =
2114 (trans->param_flags & ~SPP_SACKDELAY) |
2115 sackdelay_change;
2116 } else if (asoc) {
2117 asoc->param_flags =
2118 (asoc->param_flags & ~SPP_SACKDELAY) |
2119 sackdelay_change;
2120 } else {
2121 sp->param_flags =
2122 (sp->param_flags & ~SPP_SACKDELAY) |
2123 sackdelay_change;
2124 }
2125 }
2126
2127 if (params->spp_pathmaxrxt) {
2128 if (trans) {
2129 trans->pathmaxrxt = params->spp_pathmaxrxt;
2130 } else if (asoc) {
2131 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2132 } else {
2133 sp->pathmaxrxt = params->spp_pathmaxrxt;
2134 }
2135 }
2136
2137 return 0;
2138}
2139
2140static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2141 char __user *optval, int optlen)
2142{
2143 struct sctp_paddrparams params;
2144 struct sctp_transport *trans = NULL;
2145 struct sctp_association *asoc = NULL;
2146 struct sctp_sock *sp = sctp_sk(sk);
2147 int error;
2148 int hb_change, pmtud_change, sackdelay_change;
2149
2150 if (optlen != sizeof(struct sctp_paddrparams))
2151 return - EINVAL;
2152
2153 if (copy_from_user(&params, optval, optlen))
2154 return -EFAULT;
2155
2156 /* Validate flags and value parameters. */
2157 hb_change = params.spp_flags & SPP_HB;
2158 pmtud_change = params.spp_flags & SPP_PMTUD;
2159 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2160
2161 if (hb_change == SPP_HB ||
2162 pmtud_change == SPP_PMTUD ||
2163 sackdelay_change == SPP_SACKDELAY ||
2164 params.spp_sackdelay > 500 ||
2165 (params.spp_pathmtu
2166 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2167 return -EINVAL;
2168
2169 /* If an address other than INADDR_ANY is specified, and
2170 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08002172 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2173 trans = sctp_addr_id2transport(sk, &params.spp_address,
2174 params.spp_assoc_id);
2175 if (!trans)
2176 return -EINVAL;
2177 }
2178
2179 /* Get association, if assoc_id != 0 and the socket is a one
2180 * to many style socket, and an association was not found, then
2181 * the id was invalid.
2182 */
2183 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2184 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2185 return -EINVAL;
2186
2187 /* Heartbeat demand can only be sent on a transport or
2188 * association, but not a socket.
2189 */
2190 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2191 return -EINVAL;
2192
2193 /* Process parameters. */
2194 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2195 hb_change, pmtud_change,
2196 sackdelay_change);
2197
2198 if (error)
2199 return error;
2200
2201 /* If changes are for association, also apply parameters to each
2202 * transport.
2203 */
2204 if (!trans && asoc) {
2205 struct list_head *pos;
2206
2207 list_for_each(pos, &asoc->peer.transport_addr_list) {
2208 trans = list_entry(pos, struct sctp_transport,
2209 transports);
2210 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2211 hb_change, pmtud_change,
2212 sackdelay_change);
2213 }
2214 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215
2216 return 0;
2217}
2218
Frank Filz77086102005-12-22 11:37:30 -08002219/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2220 *
2221 * This options will get or set the delayed ack timer. The time is set
2222 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2223 * endpoints default delayed ack timer value. If the assoc_id field is
2224 * non-zero, then the set or get effects the specified association.
2225 *
2226 * struct sctp_assoc_value {
2227 * sctp_assoc_t assoc_id;
2228 * uint32_t assoc_value;
2229 * };
2230 *
2231 * assoc_id - This parameter, indicates which association the
2232 * user is preforming an action upon. Note that if
2233 * this field's value is zero then the endpoints
2234 * default value is changed (effecting future
2235 * associations only).
2236 *
2237 * assoc_value - This parameter contains the number of milliseconds
2238 * that the user is requesting the delayed ACK timer
2239 * be set to. Note that this value is defined in
2240 * the standard to be between 200 and 500 milliseconds.
2241 *
2242 * Note: a value of zero will leave the value alone,
2243 * but disable SACK delay. A non-zero value will also
2244 * enable SACK delay.
2245 */
2246
2247static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2248 char __user *optval, int optlen)
2249{
2250 struct sctp_assoc_value params;
2251 struct sctp_transport *trans = NULL;
2252 struct sctp_association *asoc = NULL;
2253 struct sctp_sock *sp = sctp_sk(sk);
2254
2255 if (optlen != sizeof(struct sctp_assoc_value))
2256 return - EINVAL;
2257
2258 if (copy_from_user(&params, optval, optlen))
2259 return -EFAULT;
2260
2261 /* Validate value parameter. */
2262 if (params.assoc_value > 500)
2263 return -EINVAL;
2264
2265 /* Get association, if assoc_id != 0 and the socket is a one
2266 * to many style socket, and an association was not found, then
2267 * the id was invalid.
2268 */
2269 asoc = sctp_id2assoc(sk, params.assoc_id);
2270 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2271 return -EINVAL;
2272
2273 if (params.assoc_value) {
2274 if (asoc) {
2275 asoc->sackdelay =
2276 msecs_to_jiffies(params.assoc_value);
2277 asoc->param_flags =
2278 (asoc->param_flags & ~SPP_SACKDELAY) |
2279 SPP_SACKDELAY_ENABLE;
2280 } else {
2281 sp->sackdelay = params.assoc_value;
2282 sp->param_flags =
2283 (sp->param_flags & ~SPP_SACKDELAY) |
2284 SPP_SACKDELAY_ENABLE;
2285 }
2286 } else {
2287 if (asoc) {
2288 asoc->param_flags =
2289 (asoc->param_flags & ~SPP_SACKDELAY) |
2290 SPP_SACKDELAY_DISABLE;
2291 } else {
2292 sp->param_flags =
2293 (sp->param_flags & ~SPP_SACKDELAY) |
2294 SPP_SACKDELAY_DISABLE;
2295 }
2296 }
2297
2298 /* If change is for association, also apply to each transport. */
2299 if (asoc) {
2300 struct list_head *pos;
2301
2302 list_for_each(pos, &asoc->peer.transport_addr_list) {
2303 trans = list_entry(pos, struct sctp_transport,
2304 transports);
2305 if (params.assoc_value) {
2306 trans->sackdelay =
2307 msecs_to_jiffies(params.assoc_value);
2308 trans->param_flags =
2309 (trans->param_flags & ~SPP_SACKDELAY) |
2310 SPP_SACKDELAY_ENABLE;
2311 } else {
2312 trans->param_flags =
2313 (trans->param_flags & ~SPP_SACKDELAY) |
2314 SPP_SACKDELAY_DISABLE;
2315 }
2316 }
2317 }
2318
2319 return 0;
2320}
2321
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2323 *
2324 * Applications can specify protocol parameters for the default association
2325 * initialization. The option name argument to setsockopt() and getsockopt()
2326 * is SCTP_INITMSG.
2327 *
2328 * Setting initialization parameters is effective only on an unconnected
2329 * socket (for UDP-style sockets only future associations are effected
2330 * by the change). With TCP-style sockets, this option is inherited by
2331 * sockets derived from a listener socket.
2332 */
2333static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2334{
2335 struct sctp_initmsg sinit;
2336 struct sctp_sock *sp = sctp_sk(sk);
2337
2338 if (optlen != sizeof(struct sctp_initmsg))
2339 return -EINVAL;
2340 if (copy_from_user(&sinit, optval, optlen))
2341 return -EFAULT;
2342
2343 if (sinit.sinit_num_ostreams)
2344 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2345 if (sinit.sinit_max_instreams)
2346 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2347 if (sinit.sinit_max_attempts)
2348 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2349 if (sinit.sinit_max_init_timeo)
2350 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2351
2352 return 0;
2353}
2354
2355/*
2356 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2357 *
2358 * Applications that wish to use the sendto() system call may wish to
2359 * specify a default set of parameters that would normally be supplied
2360 * through the inclusion of ancillary data. This socket option allows
2361 * such an application to set the default sctp_sndrcvinfo structure.
2362 * The application that wishes to use this socket option simply passes
2363 * in to this call the sctp_sndrcvinfo structure defined in Section
2364 * 5.2.2) The input parameters accepted by this call include
2365 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2366 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2367 * to this call if the caller is using the UDP model.
2368 */
2369static int sctp_setsockopt_default_send_param(struct sock *sk,
2370 char __user *optval, int optlen)
2371{
2372 struct sctp_sndrcvinfo info;
2373 struct sctp_association *asoc;
2374 struct sctp_sock *sp = sctp_sk(sk);
2375
2376 if (optlen != sizeof(struct sctp_sndrcvinfo))
2377 return -EINVAL;
2378 if (copy_from_user(&info, optval, optlen))
2379 return -EFAULT;
2380
2381 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2382 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2383 return -EINVAL;
2384
2385 if (asoc) {
2386 asoc->default_stream = info.sinfo_stream;
2387 asoc->default_flags = info.sinfo_flags;
2388 asoc->default_ppid = info.sinfo_ppid;
2389 asoc->default_context = info.sinfo_context;
2390 asoc->default_timetolive = info.sinfo_timetolive;
2391 } else {
2392 sp->default_stream = info.sinfo_stream;
2393 sp->default_flags = info.sinfo_flags;
2394 sp->default_ppid = info.sinfo_ppid;
2395 sp->default_context = info.sinfo_context;
2396 sp->default_timetolive = info.sinfo_timetolive;
2397 }
2398
2399 return 0;
2400}
2401
2402/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2403 *
2404 * Requests that the local SCTP stack use the enclosed peer address as
2405 * the association primary. The enclosed address must be one of the
2406 * association peer's addresses.
2407 */
2408static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2409 int optlen)
2410{
2411 struct sctp_prim prim;
2412 struct sctp_transport *trans;
2413
2414 if (optlen != sizeof(struct sctp_prim))
2415 return -EINVAL;
2416
2417 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2418 return -EFAULT;
2419
2420 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2421 if (!trans)
2422 return -EINVAL;
2423
2424 sctp_assoc_set_primary(trans->asoc, trans);
2425
2426 return 0;
2427}
2428
2429/*
2430 * 7.1.5 SCTP_NODELAY
2431 *
2432 * Turn on/off any Nagle-like algorithm. This means that packets are
2433 * generally sent as soon as possible and no unnecessary delays are
2434 * introduced, at the cost of more packets in the network. Expects an
2435 * integer boolean flag.
2436 */
2437static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2438 int optlen)
2439{
2440 int val;
2441
2442 if (optlen < sizeof(int))
2443 return -EINVAL;
2444 if (get_user(val, (int __user *)optval))
2445 return -EFAULT;
2446
2447 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2448 return 0;
2449}
2450
2451/*
2452 *
2453 * 7.1.1 SCTP_RTOINFO
2454 *
2455 * The protocol parameters used to initialize and bound retransmission
2456 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2457 * and modify these parameters.
2458 * All parameters are time values, in milliseconds. A value of 0, when
2459 * modifying the parameters, indicates that the current value should not
2460 * be changed.
2461 *
2462 */
2463static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2464 struct sctp_rtoinfo rtoinfo;
2465 struct sctp_association *asoc;
2466
2467 if (optlen != sizeof (struct sctp_rtoinfo))
2468 return -EINVAL;
2469
2470 if (copy_from_user(&rtoinfo, optval, optlen))
2471 return -EFAULT;
2472
2473 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2474
2475 /* Set the values to the specific association */
2476 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2477 return -EINVAL;
2478
2479 if (asoc) {
2480 if (rtoinfo.srto_initial != 0)
2481 asoc->rto_initial =
2482 msecs_to_jiffies(rtoinfo.srto_initial);
2483 if (rtoinfo.srto_max != 0)
2484 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2485 if (rtoinfo.srto_min != 0)
2486 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2487 } else {
2488 /* If there is no association or the association-id = 0
2489 * set the values to the endpoint.
2490 */
2491 struct sctp_sock *sp = sctp_sk(sk);
2492
2493 if (rtoinfo.srto_initial != 0)
2494 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2495 if (rtoinfo.srto_max != 0)
2496 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2497 if (rtoinfo.srto_min != 0)
2498 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2499 }
2500
2501 return 0;
2502}
2503
2504/*
2505 *
2506 * 7.1.2 SCTP_ASSOCINFO
2507 *
2508 * This option is used to tune the the maximum retransmission attempts
2509 * of the association.
2510 * Returns an error if the new association retransmission value is
2511 * greater than the sum of the retransmission value of the peer.
2512 * See [SCTP] for more information.
2513 *
2514 */
2515static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2516{
2517
2518 struct sctp_assocparams assocparams;
2519 struct sctp_association *asoc;
2520
2521 if (optlen != sizeof(struct sctp_assocparams))
2522 return -EINVAL;
2523 if (copy_from_user(&assocparams, optval, optlen))
2524 return -EFAULT;
2525
2526 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2527
2528 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2529 return -EINVAL;
2530
2531 /* Set the values to the specific association */
2532 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002533 if (assocparams.sasoc_asocmaxrxt != 0) {
2534 __u32 path_sum = 0;
2535 int paths = 0;
2536 struct list_head *pos;
2537 struct sctp_transport *peer_addr;
2538
2539 list_for_each(pos, &asoc->peer.transport_addr_list) {
2540 peer_addr = list_entry(pos,
2541 struct sctp_transport,
2542 transports);
2543 path_sum += peer_addr->pathmaxrxt;
2544 paths++;
2545 }
2546
2547 /* Only validate asocmaxrxt if we have more then
2548 * one path/transport. We do this because path
2549 * retransmissions are only counted when we have more
2550 * then one path.
2551 */
2552 if (paths > 1 &&
2553 assocparams.sasoc_asocmaxrxt > path_sum)
2554 return -EINVAL;
2555
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002557 }
2558
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 if (assocparams.sasoc_cookie_life != 0) {
2560 asoc->cookie_life.tv_sec =
2561 assocparams.sasoc_cookie_life / 1000;
2562 asoc->cookie_life.tv_usec =
2563 (assocparams.sasoc_cookie_life % 1000)
2564 * 1000;
2565 }
2566 } else {
2567 /* Set the values to the endpoint */
2568 struct sctp_sock *sp = sctp_sk(sk);
2569
2570 if (assocparams.sasoc_asocmaxrxt != 0)
2571 sp->assocparams.sasoc_asocmaxrxt =
2572 assocparams.sasoc_asocmaxrxt;
2573 if (assocparams.sasoc_cookie_life != 0)
2574 sp->assocparams.sasoc_cookie_life =
2575 assocparams.sasoc_cookie_life;
2576 }
2577 return 0;
2578}
2579
2580/*
2581 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2582 *
2583 * This socket option is a boolean flag which turns on or off mapped V4
2584 * addresses. If this option is turned on and the socket is type
2585 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2586 * If this option is turned off, then no mapping will be done of V4
2587 * addresses and a user will receive both PF_INET6 and PF_INET type
2588 * addresses on the socket.
2589 */
2590static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2591{
2592 int val;
2593 struct sctp_sock *sp = sctp_sk(sk);
2594
2595 if (optlen < sizeof(int))
2596 return -EINVAL;
2597 if (get_user(val, (int __user *)optval))
2598 return -EFAULT;
2599 if (val)
2600 sp->v4mapped = 1;
2601 else
2602 sp->v4mapped = 0;
2603
2604 return 0;
2605}
2606
2607/*
2608 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2609 *
2610 * This socket option specifies the maximum size to put in any outgoing
2611 * SCTP chunk. If a message is larger than this size it will be
2612 * fragmented by SCTP into the specified size. Note that the underlying
2613 * SCTP implementation may fragment into smaller sized chunks when the
2614 * PMTU of the underlying association is smaller than the value set by
2615 * the user.
2616 */
2617static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2618{
2619 struct sctp_association *asoc;
2620 struct list_head *pos;
2621 struct sctp_sock *sp = sctp_sk(sk);
2622 int val;
2623
2624 if (optlen < sizeof(int))
2625 return -EINVAL;
2626 if (get_user(val, (int __user *)optval))
2627 return -EFAULT;
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002628 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 return -EINVAL;
2630 sp->user_frag = val;
2631
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002632 /* Update the frag_point of the existing associations. */
2633 list_for_each(pos, &(sp->ep->asocs)) {
2634 asoc = list_entry(pos, struct sctp_association, asocs);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002635 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002636 }
2637
2638 return 0;
2639}
2640
2641
2642/*
2643 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2644 *
2645 * Requests that the peer mark the enclosed address as the association
2646 * primary. The enclosed address must be one of the association's
2647 * locally bound addresses. The following structure is used to make a
2648 * set primary request:
2649 */
2650static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2651 int optlen)
2652{
2653 struct sctp_sock *sp;
2654 struct sctp_endpoint *ep;
2655 struct sctp_association *asoc = NULL;
2656 struct sctp_setpeerprim prim;
2657 struct sctp_chunk *chunk;
2658 int err;
2659
2660 sp = sctp_sk(sk);
2661 ep = sp->ep;
2662
2663 if (!sctp_addip_enable)
2664 return -EPERM;
2665
2666 if (optlen != sizeof(struct sctp_setpeerprim))
2667 return -EINVAL;
2668
2669 if (copy_from_user(&prim, optval, optlen))
2670 return -EFAULT;
2671
2672 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2673 if (!asoc)
2674 return -EINVAL;
2675
2676 if (!asoc->peer.asconf_capable)
2677 return -EPERM;
2678
2679 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2680 return -EPERM;
2681
2682 if (!sctp_state(asoc, ESTABLISHED))
2683 return -ENOTCONN;
2684
2685 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2686 return -EADDRNOTAVAIL;
2687
2688 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2689 chunk = sctp_make_asconf_set_prim(asoc,
2690 (union sctp_addr *)&prim.sspp_addr);
2691 if (!chunk)
2692 return -ENOMEM;
2693
2694 err = sctp_send_asconf(asoc, chunk);
2695
2696 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2697
2698 return err;
2699}
2700
2701static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2702 int optlen)
2703{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002704 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002706 if (optlen != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002707 return -EINVAL;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002708 if (copy_from_user(&adaption, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002709 return -EFAULT;
2710
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002711 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712
2713 return 0;
2714}
2715
2716/* API 6.2 setsockopt(), getsockopt()
2717 *
2718 * Applications use setsockopt() and getsockopt() to set or retrieve
2719 * socket options. Socket options are used to change the default
2720 * behavior of sockets calls. They are described in Section 7.
2721 *
2722 * The syntax is:
2723 *
2724 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2725 * int __user *optlen);
2726 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2727 * int optlen);
2728 *
2729 * sd - the socket descript.
2730 * level - set to IPPROTO_SCTP for all SCTP options.
2731 * optname - the option name.
2732 * optval - the buffer to store the value of the option.
2733 * optlen - the size of the buffer.
2734 */
2735SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2736 char __user *optval, int optlen)
2737{
2738 int retval = 0;
2739
2740 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2741 sk, optname);
2742
2743 /* I can hardly begin to describe how wrong this is. This is
2744 * so broken as to be worse than useless. The API draft
2745 * REALLY is NOT helpful here... I am not convinced that the
2746 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2747 * are at all well-founded.
2748 */
2749 if (level != SOL_SCTP) {
2750 struct sctp_af *af = sctp_sk(sk)->pf->af;
2751 retval = af->setsockopt(sk, level, optname, optval, optlen);
2752 goto out_nounlock;
2753 }
2754
2755 sctp_lock_sock(sk);
2756
2757 switch (optname) {
2758 case SCTP_SOCKOPT_BINDX_ADD:
2759 /* 'optlen' is the size of the addresses buffer. */
2760 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2761 optlen, SCTP_BINDX_ADD_ADDR);
2762 break;
2763
2764 case SCTP_SOCKOPT_BINDX_REM:
2765 /* 'optlen' is the size of the addresses buffer. */
2766 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2767 optlen, SCTP_BINDX_REM_ADDR);
2768 break;
2769
Frank Filz3f7a87d2005-06-20 13:14:57 -07002770 case SCTP_SOCKOPT_CONNECTX:
2771 /* 'optlen' is the size of the addresses buffer. */
2772 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2773 optlen);
2774 break;
2775
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776 case SCTP_DISABLE_FRAGMENTS:
2777 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2778 break;
2779
2780 case SCTP_EVENTS:
2781 retval = sctp_setsockopt_events(sk, optval, optlen);
2782 break;
2783
2784 case SCTP_AUTOCLOSE:
2785 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2786 break;
2787
2788 case SCTP_PEER_ADDR_PARAMS:
2789 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2790 break;
2791
Frank Filz77086102005-12-22 11:37:30 -08002792 case SCTP_DELAYED_ACK_TIME:
2793 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2794 break;
2795
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796 case SCTP_INITMSG:
2797 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2798 break;
2799 case SCTP_DEFAULT_SEND_PARAM:
2800 retval = sctp_setsockopt_default_send_param(sk, optval,
2801 optlen);
2802 break;
2803 case SCTP_PRIMARY_ADDR:
2804 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2805 break;
2806 case SCTP_SET_PEER_PRIMARY_ADDR:
2807 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2808 break;
2809 case SCTP_NODELAY:
2810 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2811 break;
2812 case SCTP_RTOINFO:
2813 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2814 break;
2815 case SCTP_ASSOCINFO:
2816 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2817 break;
2818 case SCTP_I_WANT_MAPPED_V4_ADDR:
2819 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2820 break;
2821 case SCTP_MAXSEG:
2822 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2823 break;
2824 case SCTP_ADAPTION_LAYER:
2825 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2826 break;
2827
2828 default:
2829 retval = -ENOPROTOOPT;
2830 break;
2831 };
2832
2833 sctp_release_sock(sk);
2834
2835out_nounlock:
2836 return retval;
2837}
2838
2839/* API 3.1.6 connect() - UDP Style Syntax
2840 *
2841 * An application may use the connect() call in the UDP model to initiate an
2842 * association without sending data.
2843 *
2844 * The syntax is:
2845 *
2846 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2847 *
2848 * sd: the socket descriptor to have a new association added to.
2849 *
2850 * nam: the address structure (either struct sockaddr_in or struct
2851 * sockaddr_in6 defined in RFC2553 [7]).
2852 *
2853 * len: the size of the address.
2854 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07002855SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856 int addr_len)
2857{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07002859 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002860
2861 sctp_lock_sock(sk);
2862
Frank Filz3f7a87d2005-06-20 13:14:57 -07002863 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2864 __FUNCTION__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865
Frank Filz3f7a87d2005-06-20 13:14:57 -07002866 /* Validate addr_len before calling common connect/connectx routine. */
2867 af = sctp_get_af_specific(addr->sa_family);
2868 if (!af || addr_len < af->sockaddr_len) {
2869 err = -EINVAL;
2870 } else {
2871 /* Pass correct addr len to common routine (so it knows there
2872 * is only one address being passed.
2873 */
2874 err = __sctp_connect(sk, addr, af->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875 }
2876
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878 return err;
2879}
2880
2881/* FIXME: Write comments. */
2882SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2883{
2884 return -EOPNOTSUPP; /* STUB */
2885}
2886
2887/* 4.1.4 accept() - TCP Style Syntax
2888 *
2889 * Applications use accept() call to remove an established SCTP
2890 * association from the accept queue of the endpoint. A new socket
2891 * descriptor will be returned from accept() to represent the newly
2892 * formed association.
2893 */
2894SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2895{
2896 struct sctp_sock *sp;
2897 struct sctp_endpoint *ep;
2898 struct sock *newsk = NULL;
2899 struct sctp_association *asoc;
2900 long timeo;
2901 int error = 0;
2902
2903 sctp_lock_sock(sk);
2904
2905 sp = sctp_sk(sk);
2906 ep = sp->ep;
2907
2908 if (!sctp_style(sk, TCP)) {
2909 error = -EOPNOTSUPP;
2910 goto out;
2911 }
2912
2913 if (!sctp_sstate(sk, LISTENING)) {
2914 error = -EINVAL;
2915 goto out;
2916 }
2917
2918 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2919
2920 error = sctp_wait_for_accept(sk, timeo);
2921 if (error)
2922 goto out;
2923
2924 /* We treat the list of associations on the endpoint as the accept
2925 * queue and pick the first association on the list.
2926 */
2927 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2928
2929 newsk = sp->pf->create_accept_sk(sk, asoc);
2930 if (!newsk) {
2931 error = -ENOMEM;
2932 goto out;
2933 }
2934
2935 /* Populate the fields of the newsk from the oldsk and migrate the
2936 * asoc to the newsk.
2937 */
2938 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2939
2940out:
2941 sctp_release_sock(sk);
2942 *err = error;
2943 return newsk;
2944}
2945
2946/* The SCTP ioctl handler. */
2947SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2948{
2949 return -ENOIOCTLCMD;
2950}
2951
2952/* This is the function which gets called during socket creation to
2953 * initialized the SCTP-specific portion of the sock.
2954 * The sock structure should already be zero-filled memory.
2955 */
2956SCTP_STATIC int sctp_init_sock(struct sock *sk)
2957{
2958 struct sctp_endpoint *ep;
2959 struct sctp_sock *sp;
2960
2961 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
2962
2963 sp = sctp_sk(sk);
2964
2965 /* Initialize the SCTP per socket area. */
2966 switch (sk->sk_type) {
2967 case SOCK_SEQPACKET:
2968 sp->type = SCTP_SOCKET_UDP;
2969 break;
2970 case SOCK_STREAM:
2971 sp->type = SCTP_SOCKET_TCP;
2972 break;
2973 default:
2974 return -ESOCKTNOSUPPORT;
2975 }
2976
2977 /* Initialize default send parameters. These parameters can be
2978 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2979 */
2980 sp->default_stream = 0;
2981 sp->default_ppid = 0;
2982 sp->default_flags = 0;
2983 sp->default_context = 0;
2984 sp->default_timetolive = 0;
2985
2986 /* Initialize default setup parameters. These parameters
2987 * can be modified with the SCTP_INITMSG socket option or
2988 * overridden by the SCTP_INIT CMSG.
2989 */
2990 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
2991 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
2992 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
2993 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
2994
2995 /* Initialize default RTO related parameters. These parameters can
2996 * be modified for with the SCTP_RTOINFO socket option.
2997 */
2998 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
2999 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
3000 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
3001
3002 /* Initialize default association related parameters. These parameters
3003 * can be modified with the SCTP_ASSOCINFO socket option.
3004 */
3005 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3006 sp->assocparams.sasoc_number_peer_destinations = 0;
3007 sp->assocparams.sasoc_peer_rwnd = 0;
3008 sp->assocparams.sasoc_local_rwnd = 0;
3009 sp->assocparams.sasoc_cookie_life =
3010 jiffies_to_msecs(sctp_valid_cookie_life);
3011
3012 /* Initialize default event subscriptions. By default, all the
3013 * options are off.
3014 */
3015 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3016
3017 /* Default Peer Address Parameters. These defaults can
3018 * be modified via SCTP_PEER_ADDR_PARAMS
3019 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08003020 sp->hbinterval = jiffies_to_msecs(sctp_hb_interval);
3021 sp->pathmaxrxt = sctp_max_retrans_path;
3022 sp->pathmtu = 0; // allow default discovery
Vlad Yasevich8116ffa2006-01-17 11:55:17 -08003023 sp->sackdelay = jiffies_to_msecs(sctp_sack_timeout);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003024 sp->param_flags = SPP_HB_ENABLE |
3025 SPP_PMTUD_ENABLE |
3026 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003027
3028 /* If enabled no SCTP message fragmentation will be performed.
3029 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3030 */
3031 sp->disable_fragments = 0;
3032
3033 /* Turn on/off any Nagle-like algorithm. */
3034 sp->nodelay = 1;
3035
3036 /* Enable by default. */
3037 sp->v4mapped = 1;
3038
3039 /* Auto-close idle associations after the configured
3040 * number of seconds. A value of 0 disables this
3041 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3042 * for UDP-style sockets only.
3043 */
3044 sp->autoclose = 0;
3045
3046 /* User specified fragmentation limit. */
3047 sp->user_frag = 0;
3048
3049 sp->adaption_ind = 0;
3050
3051 sp->pf = sctp_get_pf_specific(sk->sk_family);
3052
3053 /* Control variables for partial data delivery. */
3054 sp->pd_mode = 0;
3055 skb_queue_head_init(&sp->pd_lobby);
3056
3057 /* Create a per socket endpoint structure. Even if we
3058 * change the data structure relationships, this may still
3059 * be useful for storing pre-connect address information.
3060 */
3061 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3062 if (!ep)
3063 return -ENOMEM;
3064
3065 sp->ep = ep;
3066 sp->hmac = NULL;
3067
3068 SCTP_DBG_OBJCNT_INC(sock);
3069 return 0;
3070}
3071
3072/* Cleanup any SCTP per socket resources. */
3073SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3074{
3075 struct sctp_endpoint *ep;
3076
3077 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3078
3079 /* Release our hold on the endpoint. */
3080 ep = sctp_sk(sk)->ep;
3081 sctp_endpoint_free(ep);
3082
3083 return 0;
3084}
3085
3086/* API 4.1.7 shutdown() - TCP Style Syntax
3087 * int shutdown(int socket, int how);
3088 *
3089 * sd - the socket descriptor of the association to be closed.
3090 * how - Specifies the type of shutdown. The values are
3091 * as follows:
3092 * SHUT_RD
3093 * Disables further receive operations. No SCTP
3094 * protocol action is taken.
3095 * SHUT_WR
3096 * Disables further send operations, and initiates
3097 * the SCTP shutdown sequence.
3098 * SHUT_RDWR
3099 * Disables further send and receive operations
3100 * and initiates the SCTP shutdown sequence.
3101 */
3102SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3103{
3104 struct sctp_endpoint *ep;
3105 struct sctp_association *asoc;
3106
3107 if (!sctp_style(sk, TCP))
3108 return;
3109
3110 if (how & SEND_SHUTDOWN) {
3111 ep = sctp_sk(sk)->ep;
3112 if (!list_empty(&ep->asocs)) {
3113 asoc = list_entry(ep->asocs.next,
3114 struct sctp_association, asocs);
3115 sctp_primitive_SHUTDOWN(asoc, NULL);
3116 }
3117 }
3118}
3119
3120/* 7.2.1 Association Status (SCTP_STATUS)
3121
3122 * Applications can retrieve current status information about an
3123 * association, including association state, peer receiver window size,
3124 * number of unacked data chunks, and number of data chunks pending
3125 * receipt. This information is read-only.
3126 */
3127static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3128 char __user *optval,
3129 int __user *optlen)
3130{
3131 struct sctp_status status;
3132 struct sctp_association *asoc = NULL;
3133 struct sctp_transport *transport;
3134 sctp_assoc_t associd;
3135 int retval = 0;
3136
3137 if (len != sizeof(status)) {
3138 retval = -EINVAL;
3139 goto out;
3140 }
3141
3142 if (copy_from_user(&status, optval, sizeof(status))) {
3143 retval = -EFAULT;
3144 goto out;
3145 }
3146
3147 associd = status.sstat_assoc_id;
3148 asoc = sctp_id2assoc(sk, associd);
3149 if (!asoc) {
3150 retval = -EINVAL;
3151 goto out;
3152 }
3153
3154 transport = asoc->peer.primary_path;
3155
3156 status.sstat_assoc_id = sctp_assoc2id(asoc);
3157 status.sstat_state = asoc->state;
3158 status.sstat_rwnd = asoc->peer.rwnd;
3159 status.sstat_unackdata = asoc->unack_data;
3160
3161 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3162 status.sstat_instrms = asoc->c.sinit_max_instreams;
3163 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3164 status.sstat_fragmentation_point = asoc->frag_point;
3165 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3166 memcpy(&status.sstat_primary.spinfo_address,
3167 &(transport->ipaddr), sizeof(union sctp_addr));
3168 /* Map ipv4 address into v4-mapped-on-v6 address. */
3169 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3170 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003171 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003172 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3173 status.sstat_primary.spinfo_srtt = transport->srtt;
3174 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003175 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003176
Frank Filz3f7a87d2005-06-20 13:14:57 -07003177 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3178 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3179
Linus Torvalds1da177e2005-04-16 15:20:36 -07003180 if (put_user(len, optlen)) {
3181 retval = -EFAULT;
3182 goto out;
3183 }
3184
3185 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3186 len, status.sstat_state, status.sstat_rwnd,
3187 status.sstat_assoc_id);
3188
3189 if (copy_to_user(optval, &status, len)) {
3190 retval = -EFAULT;
3191 goto out;
3192 }
3193
3194out:
3195 return (retval);
3196}
3197
3198
3199/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3200 *
3201 * Applications can retrieve information about a specific peer address
3202 * of an association, including its reachability state, congestion
3203 * window, and retransmission timer values. This information is
3204 * read-only.
3205 */
3206static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3207 char __user *optval,
3208 int __user *optlen)
3209{
3210 struct sctp_paddrinfo pinfo;
3211 struct sctp_transport *transport;
3212 int retval = 0;
3213
3214 if (len != sizeof(pinfo)) {
3215 retval = -EINVAL;
3216 goto out;
3217 }
3218
3219 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3220 retval = -EFAULT;
3221 goto out;
3222 }
3223
3224 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3225 pinfo.spinfo_assoc_id);
3226 if (!transport)
3227 return -EINVAL;
3228
3229 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003230 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003231 pinfo.spinfo_cwnd = transport->cwnd;
3232 pinfo.spinfo_srtt = transport->srtt;
3233 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003234 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003235
Frank Filz3f7a87d2005-06-20 13:14:57 -07003236 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3237 pinfo.spinfo_state = SCTP_ACTIVE;
3238
Linus Torvalds1da177e2005-04-16 15:20:36 -07003239 if (put_user(len, optlen)) {
3240 retval = -EFAULT;
3241 goto out;
3242 }
3243
3244 if (copy_to_user(optval, &pinfo, len)) {
3245 retval = -EFAULT;
3246 goto out;
3247 }
3248
3249out:
3250 return (retval);
3251}
3252
3253/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3254 *
3255 * This option is a on/off flag. If enabled no SCTP message
3256 * fragmentation will be performed. Instead if a message being sent
3257 * exceeds the current PMTU size, the message will NOT be sent and
3258 * instead a error will be indicated to the user.
3259 */
3260static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3261 char __user *optval, int __user *optlen)
3262{
3263 int val;
3264
3265 if (len < sizeof(int))
3266 return -EINVAL;
3267
3268 len = sizeof(int);
3269 val = (sctp_sk(sk)->disable_fragments == 1);
3270 if (put_user(len, optlen))
3271 return -EFAULT;
3272 if (copy_to_user(optval, &val, len))
3273 return -EFAULT;
3274 return 0;
3275}
3276
3277/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3278 *
3279 * This socket option is used to specify various notifications and
3280 * ancillary data the user wishes to receive.
3281 */
3282static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3283 int __user *optlen)
3284{
3285 if (len != sizeof(struct sctp_event_subscribe))
3286 return -EINVAL;
3287 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3288 return -EFAULT;
3289 return 0;
3290}
3291
3292/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3293 *
3294 * This socket option is applicable to the UDP-style socket only. When
3295 * set it will cause associations that are idle for more than the
3296 * specified number of seconds to automatically close. An association
3297 * being idle is defined an association that has NOT sent or received
3298 * user data. The special value of '0' indicates that no automatic
3299 * close of any associations should be performed. The option expects an
3300 * integer defining the number of seconds of idle time before an
3301 * association is closed.
3302 */
3303static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3304{
3305 /* Applicable to UDP-style socket only */
3306 if (sctp_style(sk, TCP))
3307 return -EOPNOTSUPP;
3308 if (len != sizeof(int))
3309 return -EINVAL;
3310 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3311 return -EFAULT;
3312 return 0;
3313}
3314
3315/* Helper routine to branch off an association to a new socket. */
3316SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3317 struct socket **sockp)
3318{
3319 struct sock *sk = asoc->base.sk;
3320 struct socket *sock;
3321 int err = 0;
3322
3323 /* An association cannot be branched off from an already peeled-off
3324 * socket, nor is this supported for tcp style sockets.
3325 */
3326 if (!sctp_style(sk, UDP))
3327 return -EINVAL;
3328
3329 /* Create a new socket. */
3330 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3331 if (err < 0)
3332 return err;
3333
3334 /* Populate the fields of the newsk from the oldsk and migrate the
3335 * asoc to the newsk.
3336 */
3337 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3338 *sockp = sock;
3339
3340 return err;
3341}
3342
3343static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3344{
3345 sctp_peeloff_arg_t peeloff;
3346 struct socket *newsock;
3347 int retval = 0;
3348 struct sctp_association *asoc;
3349
3350 if (len != sizeof(sctp_peeloff_arg_t))
3351 return -EINVAL;
3352 if (copy_from_user(&peeloff, optval, len))
3353 return -EFAULT;
3354
3355 asoc = sctp_id2assoc(sk, peeloff.associd);
3356 if (!asoc) {
3357 retval = -EINVAL;
3358 goto out;
3359 }
3360
3361 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3362
3363 retval = sctp_do_peeloff(asoc, &newsock);
3364 if (retval < 0)
3365 goto out;
3366
3367 /* Map the socket to an unused fd that can be returned to the user. */
3368 retval = sock_map_fd(newsock);
3369 if (retval < 0) {
3370 sock_release(newsock);
3371 goto out;
3372 }
3373
3374 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3375 __FUNCTION__, sk, asoc, newsock->sk, retval);
3376
3377 /* Return the fd mapped to the new socket. */
3378 peeloff.sd = retval;
3379 if (copy_to_user(optval, &peeloff, len))
3380 retval = -EFAULT;
3381
3382out:
3383 return retval;
3384}
3385
3386/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3387 *
3388 * Applications can enable or disable heartbeats for any peer address of
3389 * an association, modify an address's heartbeat interval, force a
3390 * heartbeat to be sent immediately, and adjust the address's maximum
3391 * number of retransmissions sent before an address is considered
3392 * unreachable. The following structure is used to access and modify an
3393 * address's parameters:
3394 *
3395 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08003396 * sctp_assoc_t spp_assoc_id;
3397 * struct sockaddr_storage spp_address;
3398 * uint32_t spp_hbinterval;
3399 * uint16_t spp_pathmaxrxt;
3400 * uint32_t spp_pathmtu;
3401 * uint32_t spp_sackdelay;
3402 * uint32_t spp_flags;
3403 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07003404 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08003405 * spp_assoc_id - (one-to-many style socket) This is filled in the
3406 * application, and identifies the association for
3407 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003408 * spp_address - This specifies which address is of interest.
3409 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003410 * in milliseconds. If a value of zero
3411 * is present in this field then no changes are to
3412 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003413 * spp_pathmaxrxt - This contains the maximum number of
3414 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08003415 * considered unreachable. If a value of zero
3416 * is present in this field then no changes are to
3417 * be made to this parameter.
3418 * spp_pathmtu - When Path MTU discovery is disabled the value
3419 * specified here will be the "fixed" path mtu.
3420 * Note that if the spp_address field is empty
3421 * then all associations on this address will
3422 * have this fixed path mtu set upon them.
3423 *
3424 * spp_sackdelay - When delayed sack is enabled, this value specifies
3425 * the number of milliseconds that sacks will be delayed
3426 * for. This value will apply to all addresses of an
3427 * association if the spp_address field is empty. Note
3428 * also, that if delayed sack is enabled and this
3429 * value is set to 0, no change is made to the last
3430 * recorded delayed sack timer value.
3431 *
3432 * spp_flags - These flags are used to control various features
3433 * on an association. The flag field may contain
3434 * zero or more of the following options.
3435 *
3436 * SPP_HB_ENABLE - Enable heartbeats on the
3437 * specified address. Note that if the address
3438 * field is empty all addresses for the association
3439 * have heartbeats enabled upon them.
3440 *
3441 * SPP_HB_DISABLE - Disable heartbeats on the
3442 * speicifed address. Note that if the address
3443 * field is empty all addresses for the association
3444 * will have their heartbeats disabled. Note also
3445 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3446 * mutually exclusive, only one of these two should
3447 * be specified. Enabling both fields will have
3448 * undetermined results.
3449 *
3450 * SPP_HB_DEMAND - Request a user initiated heartbeat
3451 * to be made immediately.
3452 *
3453 * SPP_PMTUD_ENABLE - This field will enable PMTU
3454 * discovery upon the specified address. Note that
3455 * if the address feild is empty then all addresses
3456 * on the association are effected.
3457 *
3458 * SPP_PMTUD_DISABLE - This field will disable PMTU
3459 * discovery upon the specified address. Note that
3460 * if the address feild is empty then all addresses
3461 * on the association are effected. Not also that
3462 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3463 * exclusive. Enabling both will have undetermined
3464 * results.
3465 *
3466 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3467 * on delayed sack. The time specified in spp_sackdelay
3468 * is used to specify the sack delay for this address. Note
3469 * that if spp_address is empty then all addresses will
3470 * enable delayed sack and take on the sack delay
3471 * value specified in spp_sackdelay.
3472 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3473 * off delayed sack. If the spp_address field is blank then
3474 * delayed sack is disabled for the entire association. Note
3475 * also that this field is mutually exclusive to
3476 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3477 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003478 */
3479static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003480 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003481{
Frank Filz52ccb8e2005-12-22 11:36:46 -08003482 struct sctp_paddrparams params;
3483 struct sctp_transport *trans = NULL;
3484 struct sctp_association *asoc = NULL;
3485 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003486
3487 if (len != sizeof(struct sctp_paddrparams))
3488 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003489
Linus Torvalds1da177e2005-04-16 15:20:36 -07003490 if (copy_from_user(&params, optval, len))
3491 return -EFAULT;
3492
Frank Filz52ccb8e2005-12-22 11:36:46 -08003493 /* If an address other than INADDR_ANY is specified, and
3494 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003495 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08003496 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3497 trans = sctp_addr_id2transport(sk, &params.spp_address,
3498 params.spp_assoc_id);
3499 if (!trans) {
3500 SCTP_DEBUG_PRINTK("Failed no transport\n");
3501 return -EINVAL;
3502 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003503 }
3504
Frank Filz52ccb8e2005-12-22 11:36:46 -08003505 /* Get association, if assoc_id != 0 and the socket is a one
3506 * to many style socket, and an association was not found, then
3507 * the id was invalid.
3508 */
3509 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3510 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3511 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003512 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003513 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003514
Frank Filz52ccb8e2005-12-22 11:36:46 -08003515 if (trans) {
3516 /* Fetch transport values. */
3517 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3518 params.spp_pathmtu = trans->pathmtu;
3519 params.spp_pathmaxrxt = trans->pathmaxrxt;
3520 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003521
Frank Filz52ccb8e2005-12-22 11:36:46 -08003522 /*draft-11 doesn't say what to return in spp_flags*/
3523 params.spp_flags = trans->param_flags;
3524 } else if (asoc) {
3525 /* Fetch association values. */
3526 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3527 params.spp_pathmtu = asoc->pathmtu;
3528 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3529 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003530
Frank Filz52ccb8e2005-12-22 11:36:46 -08003531 /*draft-11 doesn't say what to return in spp_flags*/
3532 params.spp_flags = asoc->param_flags;
3533 } else {
3534 /* Fetch socket values. */
3535 params.spp_hbinterval = sp->hbinterval;
3536 params.spp_pathmtu = sp->pathmtu;
3537 params.spp_sackdelay = sp->sackdelay;
3538 params.spp_pathmaxrxt = sp->pathmaxrxt;
3539
3540 /*draft-11 doesn't say what to return in spp_flags*/
3541 params.spp_flags = sp->param_flags;
3542 }
3543
Linus Torvalds1da177e2005-04-16 15:20:36 -07003544 if (copy_to_user(optval, &params, len))
3545 return -EFAULT;
3546
3547 if (put_user(len, optlen))
3548 return -EFAULT;
3549
3550 return 0;
3551}
3552
Frank Filz77086102005-12-22 11:37:30 -08003553/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3554 *
3555 * This options will get or set the delayed ack timer. The time is set
3556 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3557 * endpoints default delayed ack timer value. If the assoc_id field is
3558 * non-zero, then the set or get effects the specified association.
3559 *
3560 * struct sctp_assoc_value {
3561 * sctp_assoc_t assoc_id;
3562 * uint32_t assoc_value;
3563 * };
3564 *
3565 * assoc_id - This parameter, indicates which association the
3566 * user is preforming an action upon. Note that if
3567 * this field's value is zero then the endpoints
3568 * default value is changed (effecting future
3569 * associations only).
3570 *
3571 * assoc_value - This parameter contains the number of milliseconds
3572 * that the user is requesting the delayed ACK timer
3573 * be set to. Note that this value is defined in
3574 * the standard to be between 200 and 500 milliseconds.
3575 *
3576 * Note: a value of zero will leave the value alone,
3577 * but disable SACK delay. A non-zero value will also
3578 * enable SACK delay.
3579 */
3580static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3581 char __user *optval,
3582 int __user *optlen)
3583{
3584 struct sctp_assoc_value params;
3585 struct sctp_association *asoc = NULL;
3586 struct sctp_sock *sp = sctp_sk(sk);
3587
3588 if (len != sizeof(struct sctp_assoc_value))
3589 return - EINVAL;
3590
3591 if (copy_from_user(&params, optval, len))
3592 return -EFAULT;
3593
3594 /* Get association, if assoc_id != 0 and the socket is a one
3595 * to many style socket, and an association was not found, then
3596 * the id was invalid.
3597 */
3598 asoc = sctp_id2assoc(sk, params.assoc_id);
3599 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3600 return -EINVAL;
3601
3602 if (asoc) {
3603 /* Fetch association values. */
3604 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3605 params.assoc_value = jiffies_to_msecs(
3606 asoc->sackdelay);
3607 else
3608 params.assoc_value = 0;
3609 } else {
3610 /* Fetch socket values. */
3611 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3612 params.assoc_value = sp->sackdelay;
3613 else
3614 params.assoc_value = 0;
3615 }
3616
3617 if (copy_to_user(optval, &params, len))
3618 return -EFAULT;
3619
3620 if (put_user(len, optlen))
3621 return -EFAULT;
3622
3623 return 0;
3624}
3625
Linus Torvalds1da177e2005-04-16 15:20:36 -07003626/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3627 *
3628 * Applications can specify protocol parameters for the default association
3629 * initialization. The option name argument to setsockopt() and getsockopt()
3630 * is SCTP_INITMSG.
3631 *
3632 * Setting initialization parameters is effective only on an unconnected
3633 * socket (for UDP-style sockets only future associations are effected
3634 * by the change). With TCP-style sockets, this option is inherited by
3635 * sockets derived from a listener socket.
3636 */
3637static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3638{
3639 if (len != sizeof(struct sctp_initmsg))
3640 return -EINVAL;
3641 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3642 return -EFAULT;
3643 return 0;
3644}
3645
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003646static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3647 char __user *optval,
3648 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003649{
3650 sctp_assoc_t id;
3651 struct sctp_association *asoc;
3652 struct list_head *pos;
3653 int cnt = 0;
3654
3655 if (len != sizeof(sctp_assoc_t))
3656 return -EINVAL;
3657
3658 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3659 return -EFAULT;
3660
3661 /* For UDP-style sockets, id specifies the association to query. */
3662 asoc = sctp_id2assoc(sk, id);
3663 if (!asoc)
3664 return -EINVAL;
3665
3666 list_for_each(pos, &asoc->peer.transport_addr_list) {
3667 cnt ++;
3668 }
3669
3670 return cnt;
3671}
3672
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003673/*
3674 * Old API for getting list of peer addresses. Does not work for 32-bit
3675 * programs running on a 64-bit kernel
3676 */
3677static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3678 char __user *optval,
3679 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003680{
3681 struct sctp_association *asoc;
3682 struct list_head *pos;
3683 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003684 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003685 struct sctp_transport *from;
3686 void __user *to;
3687 union sctp_addr temp;
3688 struct sctp_sock *sp = sctp_sk(sk);
3689 int addrlen;
3690
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003691 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003692 return -EINVAL;
3693
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003694 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003695 return -EFAULT;
3696
3697 if (getaddrs.addr_num <= 0) return -EINVAL;
3698
3699 /* For UDP-style sockets, id specifies the association to query. */
3700 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3701 if (!asoc)
3702 return -EINVAL;
3703
3704 to = (void __user *)getaddrs.addrs;
3705 list_for_each(pos, &asoc->peer.transport_addr_list) {
3706 from = list_entry(pos, struct sctp_transport, transports);
3707 memcpy(&temp, &from->ipaddr, sizeof(temp));
3708 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3709 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3710 temp.v4.sin_port = htons(temp.v4.sin_port);
3711 if (copy_to_user(to, &temp, addrlen))
3712 return -EFAULT;
3713 to += addrlen ;
3714 cnt ++;
3715 if (cnt >= getaddrs.addr_num) break;
3716 }
3717 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003718 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003719 return -EFAULT;
3720
3721 return 0;
3722}
3723
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003724static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3725 char __user *optval, int __user *optlen)
3726{
3727 struct sctp_association *asoc;
3728 struct list_head *pos;
3729 int cnt = 0;
3730 struct sctp_getaddrs getaddrs;
3731 struct sctp_transport *from;
3732 void __user *to;
3733 union sctp_addr temp;
3734 struct sctp_sock *sp = sctp_sk(sk);
3735 int addrlen;
3736 size_t space_left;
3737 int bytes_copied;
3738
3739 if (len < sizeof(struct sctp_getaddrs))
3740 return -EINVAL;
3741
3742 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3743 return -EFAULT;
3744
3745 /* For UDP-style sockets, id specifies the association to query. */
3746 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3747 if (!asoc)
3748 return -EINVAL;
3749
3750 to = optval + offsetof(struct sctp_getaddrs,addrs);
3751 space_left = len - sizeof(struct sctp_getaddrs) -
3752 offsetof(struct sctp_getaddrs,addrs);
3753
3754 list_for_each(pos, &asoc->peer.transport_addr_list) {
3755 from = list_entry(pos, struct sctp_transport, transports);
3756 memcpy(&temp, &from->ipaddr, sizeof(temp));
3757 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3758 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3759 if(space_left < addrlen)
3760 return -ENOMEM;
3761 temp.v4.sin_port = htons(temp.v4.sin_port);
3762 if (copy_to_user(to, &temp, addrlen))
3763 return -EFAULT;
3764 to += addrlen;
3765 cnt++;
3766 space_left -= addrlen;
3767 }
3768
3769 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3770 return -EFAULT;
3771 bytes_copied = ((char __user *)to) - optval;
3772 if (put_user(bytes_copied, optlen))
3773 return -EFAULT;
3774
3775 return 0;
3776}
3777
3778static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3779 char __user *optval,
3780 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003781{
3782 sctp_assoc_t id;
3783 struct sctp_bind_addr *bp;
3784 struct sctp_association *asoc;
3785 struct list_head *pos;
3786 struct sctp_sockaddr_entry *addr;
3787 rwlock_t *addr_lock;
3788 unsigned long flags;
3789 int cnt = 0;
3790
3791 if (len != sizeof(sctp_assoc_t))
3792 return -EINVAL;
3793
3794 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3795 return -EFAULT;
3796
3797 /*
3798 * For UDP-style sockets, id specifies the association to query.
3799 * If the id field is set to the value '0' then the locally bound
3800 * addresses are returned without regard to any particular
3801 * association.
3802 */
3803 if (0 == id) {
3804 bp = &sctp_sk(sk)->ep->base.bind_addr;
3805 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3806 } else {
3807 asoc = sctp_id2assoc(sk, id);
3808 if (!asoc)
3809 return -EINVAL;
3810 bp = &asoc->base.bind_addr;
3811 addr_lock = &asoc->base.addr_lock;
3812 }
3813
3814 sctp_read_lock(addr_lock);
3815
3816 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3817 * addresses from the global local address list.
3818 */
3819 if (sctp_list_single_entry(&bp->address_list)) {
3820 addr = list_entry(bp->address_list.next,
3821 struct sctp_sockaddr_entry, list);
3822 if (sctp_is_any(&addr->a)) {
3823 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3824 list_for_each(pos, &sctp_local_addr_list) {
3825 addr = list_entry(pos,
3826 struct sctp_sockaddr_entry,
3827 list);
3828 if ((PF_INET == sk->sk_family) &&
3829 (AF_INET6 == addr->a.sa.sa_family))
3830 continue;
3831 cnt++;
3832 }
3833 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3834 flags);
3835 } else {
3836 cnt = 1;
3837 }
3838 goto done;
3839 }
3840
3841 list_for_each(pos, &bp->address_list) {
3842 cnt ++;
3843 }
3844
3845done:
3846 sctp_read_unlock(addr_lock);
3847 return cnt;
3848}
3849
3850/* Helper function that copies local addresses to user and returns the number
3851 * of addresses copied.
3852 */
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003853static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3854 void __user *to)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003855{
3856 struct list_head *pos;
3857 struct sctp_sockaddr_entry *addr;
3858 unsigned long flags;
3859 union sctp_addr temp;
3860 int cnt = 0;
3861 int addrlen;
3862
3863 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3864 list_for_each(pos, &sctp_local_addr_list) {
3865 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3866 if ((PF_INET == sk->sk_family) &&
3867 (AF_INET6 == addr->a.sa.sa_family))
3868 continue;
3869 memcpy(&temp, &addr->a, sizeof(temp));
3870 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3871 &temp);
3872 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3873 temp.v4.sin_port = htons(port);
3874 if (copy_to_user(to, &temp, addrlen)) {
3875 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3876 flags);
3877 return -EFAULT;
3878 }
3879 to += addrlen;
3880 cnt ++;
3881 if (cnt >= max_addrs) break;
3882 }
3883 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3884
3885 return cnt;
3886}
3887
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003888static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
Al Virod3a880e2005-12-15 09:18:30 +00003889 void __user **to, size_t space_left)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003890{
3891 struct list_head *pos;
3892 struct sctp_sockaddr_entry *addr;
3893 unsigned long flags;
3894 union sctp_addr temp;
3895 int cnt = 0;
3896 int addrlen;
3897
3898 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3899 list_for_each(pos, &sctp_local_addr_list) {
3900 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3901 if ((PF_INET == sk->sk_family) &&
3902 (AF_INET6 == addr->a.sa.sa_family))
3903 continue;
3904 memcpy(&temp, &addr->a, sizeof(temp));
3905 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3906 &temp);
3907 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3908 if(space_left<addrlen)
3909 return -ENOMEM;
3910 temp.v4.sin_port = htons(port);
3911 if (copy_to_user(*to, &temp, addrlen)) {
3912 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3913 flags);
3914 return -EFAULT;
3915 }
3916 *to += addrlen;
3917 cnt ++;
3918 space_left -= addrlen;
3919 }
3920 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3921
3922 return cnt;
3923}
3924
3925/* Old API for getting list of local addresses. Does not work for 32-bit
3926 * programs running on a 64-bit kernel
3927 */
3928static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3929 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003930{
3931 struct sctp_bind_addr *bp;
3932 struct sctp_association *asoc;
3933 struct list_head *pos;
3934 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003935 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003936 struct sctp_sockaddr_entry *addr;
3937 void __user *to;
3938 union sctp_addr temp;
3939 struct sctp_sock *sp = sctp_sk(sk);
3940 int addrlen;
3941 rwlock_t *addr_lock;
3942 int err = 0;
3943
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003944 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003945 return -EINVAL;
3946
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003947 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003948 return -EFAULT;
3949
3950 if (getaddrs.addr_num <= 0) return -EINVAL;
3951 /*
3952 * For UDP-style sockets, id specifies the association to query.
3953 * If the id field is set to the value '0' then the locally bound
3954 * addresses are returned without regard to any particular
3955 * association.
3956 */
3957 if (0 == getaddrs.assoc_id) {
3958 bp = &sctp_sk(sk)->ep->base.bind_addr;
3959 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3960 } else {
3961 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3962 if (!asoc)
3963 return -EINVAL;
3964 bp = &asoc->base.bind_addr;
3965 addr_lock = &asoc->base.addr_lock;
3966 }
3967
3968 to = getaddrs.addrs;
3969
3970 sctp_read_lock(addr_lock);
3971
3972 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3973 * addresses from the global local address list.
3974 */
3975 if (sctp_list_single_entry(&bp->address_list)) {
3976 addr = list_entry(bp->address_list.next,
3977 struct sctp_sockaddr_entry, list);
3978 if (sctp_is_any(&addr->a)) {
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003979 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
3980 getaddrs.addr_num,
3981 to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003982 if (cnt < 0) {
3983 err = cnt;
3984 goto unlock;
3985 }
3986 goto copy_getaddrs;
3987 }
3988 }
3989
3990 list_for_each(pos, &bp->address_list) {
3991 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3992 memcpy(&temp, &addr->a, sizeof(temp));
3993 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3994 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3995 temp.v4.sin_port = htons(temp.v4.sin_port);
3996 if (copy_to_user(to, &temp, addrlen)) {
3997 err = -EFAULT;
3998 goto unlock;
3999 }
4000 to += addrlen;
4001 cnt ++;
4002 if (cnt >= getaddrs.addr_num) break;
4003 }
4004
4005copy_getaddrs:
4006 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004007 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004008 err = -EFAULT;
4009
4010unlock:
4011 sctp_read_unlock(addr_lock);
4012 return err;
4013}
4014
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004015static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4016 char __user *optval, int __user *optlen)
4017{
4018 struct sctp_bind_addr *bp;
4019 struct sctp_association *asoc;
4020 struct list_head *pos;
4021 int cnt = 0;
4022 struct sctp_getaddrs getaddrs;
4023 struct sctp_sockaddr_entry *addr;
4024 void __user *to;
4025 union sctp_addr temp;
4026 struct sctp_sock *sp = sctp_sk(sk);
4027 int addrlen;
4028 rwlock_t *addr_lock;
4029 int err = 0;
4030 size_t space_left;
4031 int bytes_copied;
4032
4033 if (len <= sizeof(struct sctp_getaddrs))
4034 return -EINVAL;
4035
4036 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4037 return -EFAULT;
4038
4039 /*
4040 * For UDP-style sockets, id specifies the association to query.
4041 * If the id field is set to the value '0' then the locally bound
4042 * addresses are returned without regard to any particular
4043 * association.
4044 */
4045 if (0 == getaddrs.assoc_id) {
4046 bp = &sctp_sk(sk)->ep->base.bind_addr;
4047 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4048 } else {
4049 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4050 if (!asoc)
4051 return -EINVAL;
4052 bp = &asoc->base.bind_addr;
4053 addr_lock = &asoc->base.addr_lock;
4054 }
4055
4056 to = optval + offsetof(struct sctp_getaddrs,addrs);
4057 space_left = len - sizeof(struct sctp_getaddrs) -
4058 offsetof(struct sctp_getaddrs,addrs);
4059
4060 sctp_read_lock(addr_lock);
4061
4062 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4063 * addresses from the global local address list.
4064 */
4065 if (sctp_list_single_entry(&bp->address_list)) {
4066 addr = list_entry(bp->address_list.next,
4067 struct sctp_sockaddr_entry, list);
4068 if (sctp_is_any(&addr->a)) {
4069 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4070 &to, space_left);
4071 if (cnt < 0) {
4072 err = cnt;
4073 goto unlock;
4074 }
4075 goto copy_getaddrs;
4076 }
4077 }
4078
4079 list_for_each(pos, &bp->address_list) {
4080 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4081 memcpy(&temp, &addr->a, sizeof(temp));
4082 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4083 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4084 if(space_left < addrlen)
4085 return -ENOMEM; /*fixme: right error?*/
4086 temp.v4.sin_port = htons(temp.v4.sin_port);
4087 if (copy_to_user(to, &temp, addrlen)) {
4088 err = -EFAULT;
4089 goto unlock;
4090 }
4091 to += addrlen;
4092 cnt ++;
4093 space_left -= addrlen;
4094 }
4095
4096copy_getaddrs:
4097 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4098 return -EFAULT;
4099 bytes_copied = ((char __user *)to) - optval;
4100 if (put_user(bytes_copied, optlen))
4101 return -EFAULT;
4102
4103unlock:
4104 sctp_read_unlock(addr_lock);
4105 return err;
4106}
4107
Linus Torvalds1da177e2005-04-16 15:20:36 -07004108/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4109 *
4110 * Requests that the local SCTP stack use the enclosed peer address as
4111 * the association primary. The enclosed address must be one of the
4112 * association peer's addresses.
4113 */
4114static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4115 char __user *optval, int __user *optlen)
4116{
4117 struct sctp_prim prim;
4118 struct sctp_association *asoc;
4119 struct sctp_sock *sp = sctp_sk(sk);
4120
4121 if (len != sizeof(struct sctp_prim))
4122 return -EINVAL;
4123
4124 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4125 return -EFAULT;
4126
4127 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4128 if (!asoc)
4129 return -EINVAL;
4130
4131 if (!asoc->peer.primary_path)
4132 return -ENOTCONN;
4133
4134 asoc->peer.primary_path->ipaddr.v4.sin_port =
4135 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
4136 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4137 sizeof(union sctp_addr));
4138 asoc->peer.primary_path->ipaddr.v4.sin_port =
4139 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
4140
4141 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4142 (union sctp_addr *)&prim.ssp_addr);
4143
4144 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4145 return -EFAULT;
4146
4147 return 0;
4148}
4149
4150/*
4151 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4152 *
4153 * Requests that the local endpoint set the specified Adaption Layer
4154 * Indication parameter for all future INIT and INIT-ACK exchanges.
4155 */
4156static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4157 char __user *optval, int __user *optlen)
4158{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004159 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004160
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004161 if (len != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004162 return -EINVAL;
4163
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004164 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4165 if (copy_to_user(optval, &adaption, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004166 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004167
Linus Torvalds1da177e2005-04-16 15:20:36 -07004168 return 0;
4169}
4170
4171/*
4172 *
4173 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4174 *
4175 * Applications that wish to use the sendto() system call may wish to
4176 * specify a default set of parameters that would normally be supplied
4177 * through the inclusion of ancillary data. This socket option allows
4178 * such an application to set the default sctp_sndrcvinfo structure.
4179
4180
4181 * The application that wishes to use this socket option simply passes
4182 * in to this call the sctp_sndrcvinfo structure defined in Section
4183 * 5.2.2) The input parameters accepted by this call include
4184 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4185 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4186 * to this call if the caller is using the UDP model.
4187 *
4188 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4189 */
4190static int sctp_getsockopt_default_send_param(struct sock *sk,
4191 int len, char __user *optval,
4192 int __user *optlen)
4193{
4194 struct sctp_sndrcvinfo info;
4195 struct sctp_association *asoc;
4196 struct sctp_sock *sp = sctp_sk(sk);
4197
4198 if (len != sizeof(struct sctp_sndrcvinfo))
4199 return -EINVAL;
4200 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4201 return -EFAULT;
4202
4203 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4204 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4205 return -EINVAL;
4206
4207 if (asoc) {
4208 info.sinfo_stream = asoc->default_stream;
4209 info.sinfo_flags = asoc->default_flags;
4210 info.sinfo_ppid = asoc->default_ppid;
4211 info.sinfo_context = asoc->default_context;
4212 info.sinfo_timetolive = asoc->default_timetolive;
4213 } else {
4214 info.sinfo_stream = sp->default_stream;
4215 info.sinfo_flags = sp->default_flags;
4216 info.sinfo_ppid = sp->default_ppid;
4217 info.sinfo_context = sp->default_context;
4218 info.sinfo_timetolive = sp->default_timetolive;
4219 }
4220
4221 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4222 return -EFAULT;
4223
4224 return 0;
4225}
4226
4227/*
4228 *
4229 * 7.1.5 SCTP_NODELAY
4230 *
4231 * Turn on/off any Nagle-like algorithm. This means that packets are
4232 * generally sent as soon as possible and no unnecessary delays are
4233 * introduced, at the cost of more packets in the network. Expects an
4234 * integer boolean flag.
4235 */
4236
4237static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4238 char __user *optval, int __user *optlen)
4239{
4240 int val;
4241
4242 if (len < sizeof(int))
4243 return -EINVAL;
4244
4245 len = sizeof(int);
4246 val = (sctp_sk(sk)->nodelay == 1);
4247 if (put_user(len, optlen))
4248 return -EFAULT;
4249 if (copy_to_user(optval, &val, len))
4250 return -EFAULT;
4251 return 0;
4252}
4253
4254/*
4255 *
4256 * 7.1.1 SCTP_RTOINFO
4257 *
4258 * The protocol parameters used to initialize and bound retransmission
4259 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4260 * and modify these parameters.
4261 * All parameters are time values, in milliseconds. A value of 0, when
4262 * modifying the parameters, indicates that the current value should not
4263 * be changed.
4264 *
4265 */
4266static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4267 char __user *optval,
4268 int __user *optlen) {
4269 struct sctp_rtoinfo rtoinfo;
4270 struct sctp_association *asoc;
4271
4272 if (len != sizeof (struct sctp_rtoinfo))
4273 return -EINVAL;
4274
4275 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4276 return -EFAULT;
4277
4278 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4279
4280 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4281 return -EINVAL;
4282
4283 /* Values corresponding to the specific association. */
4284 if (asoc) {
4285 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4286 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4287 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4288 } else {
4289 /* Values corresponding to the endpoint. */
4290 struct sctp_sock *sp = sctp_sk(sk);
4291
4292 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4293 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4294 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4295 }
4296
4297 if (put_user(len, optlen))
4298 return -EFAULT;
4299
4300 if (copy_to_user(optval, &rtoinfo, len))
4301 return -EFAULT;
4302
4303 return 0;
4304}
4305
4306/*
4307 *
4308 * 7.1.2 SCTP_ASSOCINFO
4309 *
4310 * This option is used to tune the the maximum retransmission attempts
4311 * of the association.
4312 * Returns an error if the new association retransmission value is
4313 * greater than the sum of the retransmission value of the peer.
4314 * See [SCTP] for more information.
4315 *
4316 */
4317static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4318 char __user *optval,
4319 int __user *optlen)
4320{
4321
4322 struct sctp_assocparams assocparams;
4323 struct sctp_association *asoc;
4324 struct list_head *pos;
4325 int cnt = 0;
4326
4327 if (len != sizeof (struct sctp_assocparams))
4328 return -EINVAL;
4329
4330 if (copy_from_user(&assocparams, optval,
4331 sizeof (struct sctp_assocparams)))
4332 return -EFAULT;
4333
4334 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4335
4336 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4337 return -EINVAL;
4338
4339 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07004340 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004341 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4342 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4343 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4344 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4345 * 1000) +
4346 (asoc->cookie_life.tv_usec
4347 / 1000);
4348
4349 list_for_each(pos, &asoc->peer.transport_addr_list) {
4350 cnt ++;
4351 }
4352
4353 assocparams.sasoc_number_peer_destinations = cnt;
4354 } else {
4355 /* Values corresponding to the endpoint */
4356 struct sctp_sock *sp = sctp_sk(sk);
4357
4358 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4359 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4360 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4361 assocparams.sasoc_cookie_life =
4362 sp->assocparams.sasoc_cookie_life;
4363 assocparams.sasoc_number_peer_destinations =
4364 sp->assocparams.
4365 sasoc_number_peer_destinations;
4366 }
4367
4368 if (put_user(len, optlen))
4369 return -EFAULT;
4370
4371 if (copy_to_user(optval, &assocparams, len))
4372 return -EFAULT;
4373
4374 return 0;
4375}
4376
4377/*
4378 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4379 *
4380 * This socket option is a boolean flag which turns on or off mapped V4
4381 * addresses. If this option is turned on and the socket is type
4382 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4383 * If this option is turned off, then no mapping will be done of V4
4384 * addresses and a user will receive both PF_INET6 and PF_INET type
4385 * addresses on the socket.
4386 */
4387static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4388 char __user *optval, int __user *optlen)
4389{
4390 int val;
4391 struct sctp_sock *sp = sctp_sk(sk);
4392
4393 if (len < sizeof(int))
4394 return -EINVAL;
4395
4396 len = sizeof(int);
4397 val = sp->v4mapped;
4398 if (put_user(len, optlen))
4399 return -EFAULT;
4400 if (copy_to_user(optval, &val, len))
4401 return -EFAULT;
4402
4403 return 0;
4404}
4405
4406/*
4407 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4408 *
4409 * This socket option specifies the maximum size to put in any outgoing
4410 * SCTP chunk. If a message is larger than this size it will be
4411 * fragmented by SCTP into the specified size. Note that the underlying
4412 * SCTP implementation may fragment into smaller sized chunks when the
4413 * PMTU of the underlying association is smaller than the value set by
4414 * the user.
4415 */
4416static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4417 char __user *optval, int __user *optlen)
4418{
4419 int val;
4420
4421 if (len < sizeof(int))
4422 return -EINVAL;
4423
4424 len = sizeof(int);
4425
4426 val = sctp_sk(sk)->user_frag;
4427 if (put_user(len, optlen))
4428 return -EFAULT;
4429 if (copy_to_user(optval, &val, len))
4430 return -EFAULT;
4431
4432 return 0;
4433}
4434
4435SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4436 char __user *optval, int __user *optlen)
4437{
4438 int retval = 0;
4439 int len;
4440
Frank Filz3f7a87d2005-06-20 13:14:57 -07004441 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4442 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004443
4444 /* I can hardly begin to describe how wrong this is. This is
4445 * so broken as to be worse than useless. The API draft
4446 * REALLY is NOT helpful here... I am not convinced that the
4447 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4448 * are at all well-founded.
4449 */
4450 if (level != SOL_SCTP) {
4451 struct sctp_af *af = sctp_sk(sk)->pf->af;
4452
4453 retval = af->getsockopt(sk, level, optname, optval, optlen);
4454 return retval;
4455 }
4456
4457 if (get_user(len, optlen))
4458 return -EFAULT;
4459
4460 sctp_lock_sock(sk);
4461
4462 switch (optname) {
4463 case SCTP_STATUS:
4464 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4465 break;
4466 case SCTP_DISABLE_FRAGMENTS:
4467 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4468 optlen);
4469 break;
4470 case SCTP_EVENTS:
4471 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4472 break;
4473 case SCTP_AUTOCLOSE:
4474 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4475 break;
4476 case SCTP_SOCKOPT_PEELOFF:
4477 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4478 break;
4479 case SCTP_PEER_ADDR_PARAMS:
4480 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4481 optlen);
4482 break;
Frank Filz77086102005-12-22 11:37:30 -08004483 case SCTP_DELAYED_ACK_TIME:
4484 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4485 optlen);
4486 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004487 case SCTP_INITMSG:
4488 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4489 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004490 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4491 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4492 optlen);
4493 break;
4494 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4495 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4496 optlen);
4497 break;
4498 case SCTP_GET_PEER_ADDRS_OLD:
4499 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004500 optlen);
4501 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004502 case SCTP_GET_LOCAL_ADDRS_OLD:
4503 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004504 optlen);
4505 break;
4506 case SCTP_GET_PEER_ADDRS:
4507 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4508 optlen);
4509 break;
4510 case SCTP_GET_LOCAL_ADDRS:
4511 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4512 optlen);
4513 break;
4514 case SCTP_DEFAULT_SEND_PARAM:
4515 retval = sctp_getsockopt_default_send_param(sk, len,
4516 optval, optlen);
4517 break;
4518 case SCTP_PRIMARY_ADDR:
4519 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4520 break;
4521 case SCTP_NODELAY:
4522 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4523 break;
4524 case SCTP_RTOINFO:
4525 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4526 break;
4527 case SCTP_ASSOCINFO:
4528 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4529 break;
4530 case SCTP_I_WANT_MAPPED_V4_ADDR:
4531 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4532 break;
4533 case SCTP_MAXSEG:
4534 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4535 break;
4536 case SCTP_GET_PEER_ADDR_INFO:
4537 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4538 optlen);
4539 break;
4540 case SCTP_ADAPTION_LAYER:
4541 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4542 optlen);
4543 break;
4544 default:
4545 retval = -ENOPROTOOPT;
4546 break;
4547 };
4548
4549 sctp_release_sock(sk);
4550 return retval;
4551}
4552
4553static void sctp_hash(struct sock *sk)
4554{
4555 /* STUB */
4556}
4557
4558static void sctp_unhash(struct sock *sk)
4559{
4560 /* STUB */
4561}
4562
4563/* Check if port is acceptable. Possibly find first available port.
4564 *
4565 * The port hash table (contained in the 'global' SCTP protocol storage
4566 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4567 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4568 * list (the list number is the port number hashed out, so as you
4569 * would expect from a hash function, all the ports in a given list have
4570 * such a number that hashes out to the same list number; you were
4571 * expecting that, right?); so each list has a set of ports, with a
4572 * link to the socket (struct sock) that uses it, the port number and
4573 * a fastreuse flag (FIXME: NPI ipg).
4574 */
4575static struct sctp_bind_bucket *sctp_bucket_create(
4576 struct sctp_bind_hashbucket *head, unsigned short snum);
4577
4578static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4579{
4580 struct sctp_bind_hashbucket *head; /* hash list */
4581 struct sctp_bind_bucket *pp; /* hash list port iterator */
4582 unsigned short snum;
4583 int ret;
4584
4585 /* NOTE: Remember to put this back to net order. */
4586 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4587 snum = addr->v4.sin_port;
4588
4589 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4590 sctp_local_bh_disable();
4591
4592 if (snum == 0) {
4593 /* Search for an available port.
4594 *
4595 * 'sctp_port_rover' was the last port assigned, so
4596 * we start to search from 'sctp_port_rover +
4597 * 1'. What we do is first check if port 'rover' is
4598 * already in the hash table; if not, we use that; if
4599 * it is, we try next.
4600 */
4601 int low = sysctl_local_port_range[0];
4602 int high = sysctl_local_port_range[1];
4603 int remaining = (high - low) + 1;
4604 int rover;
4605 int index;
4606
4607 sctp_spin_lock(&sctp_port_alloc_lock);
4608 rover = sctp_port_rover;
4609 do {
4610 rover++;
4611 if ((rover < low) || (rover > high))
4612 rover = low;
4613 index = sctp_phashfn(rover);
4614 head = &sctp_port_hashtable[index];
4615 sctp_spin_lock(&head->lock);
4616 for (pp = head->chain; pp; pp = pp->next)
4617 if (pp->port == rover)
4618 goto next;
4619 break;
4620 next:
4621 sctp_spin_unlock(&head->lock);
4622 } while (--remaining > 0);
4623 sctp_port_rover = rover;
4624 sctp_spin_unlock(&sctp_port_alloc_lock);
4625
4626 /* Exhausted local port range during search? */
4627 ret = 1;
4628 if (remaining <= 0)
4629 goto fail;
4630
4631 /* OK, here is the one we will use. HEAD (the port
4632 * hash table list entry) is non-NULL and we hold it's
4633 * mutex.
4634 */
4635 snum = rover;
4636 } else {
4637 /* We are given an specific port number; we verify
4638 * that it is not being used. If it is used, we will
4639 * exahust the search in the hash list corresponding
4640 * to the port number (snum) - we detect that with the
4641 * port iterator, pp being NULL.
4642 */
4643 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4644 sctp_spin_lock(&head->lock);
4645 for (pp = head->chain; pp; pp = pp->next) {
4646 if (pp->port == snum)
4647 goto pp_found;
4648 }
4649 }
4650 pp = NULL;
4651 goto pp_not_found;
4652pp_found:
4653 if (!hlist_empty(&pp->owner)) {
4654 /* We had a port hash table hit - there is an
4655 * available port (pp != NULL) and it is being
4656 * used by other socket (pp->owner not empty); that other
4657 * socket is going to be sk2.
4658 */
4659 int reuse = sk->sk_reuse;
4660 struct sock *sk2;
4661 struct hlist_node *node;
4662
4663 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4664 if (pp->fastreuse && sk->sk_reuse)
4665 goto success;
4666
4667 /* Run through the list of sockets bound to the port
4668 * (pp->port) [via the pointers bind_next and
4669 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4670 * we get the endpoint they describe and run through
4671 * the endpoint's list of IP (v4 or v6) addresses,
4672 * comparing each of the addresses with the address of
4673 * the socket sk. If we find a match, then that means
4674 * that this port/socket (sk) combination are already
4675 * in an endpoint.
4676 */
4677 sk_for_each_bound(sk2, node, &pp->owner) {
4678 struct sctp_endpoint *ep2;
4679 ep2 = sctp_sk(sk2)->ep;
4680
4681 if (reuse && sk2->sk_reuse)
4682 continue;
4683
4684 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4685 sctp_sk(sk))) {
4686 ret = (long)sk2;
4687 goto fail_unlock;
4688 }
4689 }
4690 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4691 }
4692pp_not_found:
4693 /* If there was a hash table miss, create a new port. */
4694 ret = 1;
4695 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4696 goto fail_unlock;
4697
4698 /* In either case (hit or miss), make sure fastreuse is 1 only
4699 * if sk->sk_reuse is too (that is, if the caller requested
4700 * SO_REUSEADDR on this socket -sk-).
4701 */
4702 if (hlist_empty(&pp->owner))
4703 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4704 else if (pp->fastreuse && !sk->sk_reuse)
4705 pp->fastreuse = 0;
4706
4707 /* We are set, so fill up all the data in the hash table
4708 * entry, tie the socket list information with the rest of the
4709 * sockets FIXME: Blurry, NPI (ipg).
4710 */
4711success:
4712 inet_sk(sk)->num = snum;
4713 if (!sctp_sk(sk)->bind_hash) {
4714 sk_add_bind_node(sk, &pp->owner);
4715 sctp_sk(sk)->bind_hash = pp;
4716 }
4717 ret = 0;
4718
4719fail_unlock:
4720 sctp_spin_unlock(&head->lock);
4721
4722fail:
4723 sctp_local_bh_enable();
4724 addr->v4.sin_port = htons(addr->v4.sin_port);
4725 return ret;
4726}
4727
4728/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4729 * port is requested.
4730 */
4731static int sctp_get_port(struct sock *sk, unsigned short snum)
4732{
4733 long ret;
4734 union sctp_addr addr;
4735 struct sctp_af *af = sctp_sk(sk)->pf->af;
4736
4737 /* Set up a dummy address struct from the sk. */
4738 af->from_sk(&addr, sk);
4739 addr.v4.sin_port = htons(snum);
4740
4741 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4742 ret = sctp_get_port_local(sk, &addr);
4743
4744 return (ret ? 1 : 0);
4745}
4746
4747/*
4748 * 3.1.3 listen() - UDP Style Syntax
4749 *
4750 * By default, new associations are not accepted for UDP style sockets.
4751 * An application uses listen() to mark a socket as being able to
4752 * accept new associations.
4753 */
4754SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4755{
4756 struct sctp_sock *sp = sctp_sk(sk);
4757 struct sctp_endpoint *ep = sp->ep;
4758
4759 /* Only UDP style sockets that are not peeled off are allowed to
4760 * listen().
4761 */
4762 if (!sctp_style(sk, UDP))
4763 return -EINVAL;
4764
4765 /* If backlog is zero, disable listening. */
4766 if (!backlog) {
4767 if (sctp_sstate(sk, CLOSED))
4768 return 0;
4769
4770 sctp_unhash_endpoint(ep);
4771 sk->sk_state = SCTP_SS_CLOSED;
4772 }
4773
4774 /* Return if we are already listening. */
4775 if (sctp_sstate(sk, LISTENING))
4776 return 0;
4777
4778 /*
4779 * If a bind() or sctp_bindx() is not called prior to a listen()
4780 * call that allows new associations to be accepted, the system
4781 * picks an ephemeral port and will choose an address set equivalent
4782 * to binding with a wildcard address.
4783 *
4784 * This is not currently spelled out in the SCTP sockets
4785 * extensions draft, but follows the practice as seen in TCP
4786 * sockets.
4787 */
4788 if (!ep->base.bind_addr.port) {
4789 if (sctp_autobind(sk))
4790 return -EAGAIN;
4791 }
4792 sk->sk_state = SCTP_SS_LISTENING;
4793 sctp_hash_endpoint(ep);
4794 return 0;
4795}
4796
4797/*
4798 * 4.1.3 listen() - TCP Style Syntax
4799 *
4800 * Applications uses listen() to ready the SCTP endpoint for accepting
4801 * inbound associations.
4802 */
4803SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4804{
4805 struct sctp_sock *sp = sctp_sk(sk);
4806 struct sctp_endpoint *ep = sp->ep;
4807
4808 /* If backlog is zero, disable listening. */
4809 if (!backlog) {
4810 if (sctp_sstate(sk, CLOSED))
4811 return 0;
4812
4813 sctp_unhash_endpoint(ep);
4814 sk->sk_state = SCTP_SS_CLOSED;
4815 }
4816
4817 if (sctp_sstate(sk, LISTENING))
4818 return 0;
4819
4820 /*
4821 * If a bind() or sctp_bindx() is not called prior to a listen()
4822 * call that allows new associations to be accepted, the system
4823 * picks an ephemeral port and will choose an address set equivalent
4824 * to binding with a wildcard address.
4825 *
4826 * This is not currently spelled out in the SCTP sockets
4827 * extensions draft, but follows the practice as seen in TCP
4828 * sockets.
4829 */
4830 if (!ep->base.bind_addr.port) {
4831 if (sctp_autobind(sk))
4832 return -EAGAIN;
4833 }
4834 sk->sk_state = SCTP_SS_LISTENING;
4835 sk->sk_max_ack_backlog = backlog;
4836 sctp_hash_endpoint(ep);
4837 return 0;
4838}
4839
4840/*
4841 * Move a socket to LISTENING state.
4842 */
4843int sctp_inet_listen(struct socket *sock, int backlog)
4844{
4845 struct sock *sk = sock->sk;
4846 struct crypto_tfm *tfm=NULL;
4847 int err = -EINVAL;
4848
4849 if (unlikely(backlog < 0))
4850 goto out;
4851
4852 sctp_lock_sock(sk);
4853
4854 if (sock->state != SS_UNCONNECTED)
4855 goto out;
4856
4857 /* Allocate HMAC for generating cookie. */
4858 if (sctp_hmac_alg) {
4859 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4860 if (!tfm) {
4861 err = -ENOSYS;
4862 goto out;
4863 }
4864 }
4865
4866 switch (sock->type) {
4867 case SOCK_SEQPACKET:
4868 err = sctp_seqpacket_listen(sk, backlog);
4869 break;
4870 case SOCK_STREAM:
4871 err = sctp_stream_listen(sk, backlog);
4872 break;
4873 default:
4874 break;
4875 };
4876 if (err)
4877 goto cleanup;
4878
4879 /* Store away the transform reference. */
4880 sctp_sk(sk)->hmac = tfm;
4881out:
4882 sctp_release_sock(sk);
4883 return err;
4884cleanup:
Jesper Juhl573dbd92005-09-01 17:44:29 -07004885 sctp_crypto_free_tfm(tfm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004886 goto out;
4887}
4888
4889/*
4890 * This function is done by modeling the current datagram_poll() and the
4891 * tcp_poll(). Note that, based on these implementations, we don't
4892 * lock the socket in this function, even though it seems that,
4893 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08004894 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07004895 * in this place. We assume that we don't need locks either until proven
4896 * otherwise.
4897 *
4898 * Another thing to note is that we include the Async I/O support
4899 * here, again, by modeling the current TCP/UDP code. We don't have
4900 * a good way to test with it yet.
4901 */
4902unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4903{
4904 struct sock *sk = sock->sk;
4905 struct sctp_sock *sp = sctp_sk(sk);
4906 unsigned int mask;
4907
4908 poll_wait(file, sk->sk_sleep, wait);
4909
4910 /* A TCP-style listening socket becomes readable when the accept queue
4911 * is not empty.
4912 */
4913 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4914 return (!list_empty(&sp->ep->asocs)) ?
4915 (POLLIN | POLLRDNORM) : 0;
4916
4917 mask = 0;
4918
4919 /* Is there any exceptional events? */
4920 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4921 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08004922 if (sk->sk_shutdown & RCV_SHUTDOWN)
4923 mask |= POLLRDHUP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004924 if (sk->sk_shutdown == SHUTDOWN_MASK)
4925 mask |= POLLHUP;
4926
4927 /* Is it readable? Reconsider this code with TCP-style support. */
4928 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4929 (sk->sk_shutdown & RCV_SHUTDOWN))
4930 mask |= POLLIN | POLLRDNORM;
4931
4932 /* The association is either gone or not ready. */
4933 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4934 return mask;
4935
4936 /* Is it writable? */
4937 if (sctp_writeable(sk)) {
4938 mask |= POLLOUT | POLLWRNORM;
4939 } else {
4940 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4941 /*
4942 * Since the socket is not locked, the buffer
4943 * might be made available after the writeable check and
4944 * before the bit is set. This could cause a lost I/O
4945 * signal. tcp_poll() has a race breaker for this race
4946 * condition. Based on their implementation, we put
4947 * in the following code to cover it as well.
4948 */
4949 if (sctp_writeable(sk))
4950 mask |= POLLOUT | POLLWRNORM;
4951 }
4952 return mask;
4953}
4954
4955/********************************************************************
4956 * 2nd Level Abstractions
4957 ********************************************************************/
4958
4959static struct sctp_bind_bucket *sctp_bucket_create(
4960 struct sctp_bind_hashbucket *head, unsigned short snum)
4961{
4962 struct sctp_bind_bucket *pp;
4963
4964 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
4965 SCTP_DBG_OBJCNT_INC(bind_bucket);
4966 if (pp) {
4967 pp->port = snum;
4968 pp->fastreuse = 0;
4969 INIT_HLIST_HEAD(&pp->owner);
4970 if ((pp->next = head->chain) != NULL)
4971 pp->next->pprev = &pp->next;
4972 head->chain = pp;
4973 pp->pprev = &head->chain;
4974 }
4975 return pp;
4976}
4977
4978/* Caller must hold hashbucket lock for this tb with local BH disabled */
4979static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
4980{
4981 if (hlist_empty(&pp->owner)) {
4982 if (pp->next)
4983 pp->next->pprev = pp->pprev;
4984 *(pp->pprev) = pp->next;
4985 kmem_cache_free(sctp_bucket_cachep, pp);
4986 SCTP_DBG_OBJCNT_DEC(bind_bucket);
4987 }
4988}
4989
4990/* Release this socket's reference to a local port. */
4991static inline void __sctp_put_port(struct sock *sk)
4992{
4993 struct sctp_bind_hashbucket *head =
4994 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
4995 struct sctp_bind_bucket *pp;
4996
4997 sctp_spin_lock(&head->lock);
4998 pp = sctp_sk(sk)->bind_hash;
4999 __sk_del_bind_node(sk);
5000 sctp_sk(sk)->bind_hash = NULL;
5001 inet_sk(sk)->num = 0;
5002 sctp_bucket_destroy(pp);
5003 sctp_spin_unlock(&head->lock);
5004}
5005
5006void sctp_put_port(struct sock *sk)
5007{
5008 sctp_local_bh_disable();
5009 __sctp_put_port(sk);
5010 sctp_local_bh_enable();
5011}
5012
5013/*
5014 * The system picks an ephemeral port and choose an address set equivalent
5015 * to binding with a wildcard address.
5016 * One of those addresses will be the primary address for the association.
5017 * This automatically enables the multihoming capability of SCTP.
5018 */
5019static int sctp_autobind(struct sock *sk)
5020{
5021 union sctp_addr autoaddr;
5022 struct sctp_af *af;
5023 unsigned short port;
5024
5025 /* Initialize a local sockaddr structure to INADDR_ANY. */
5026 af = sctp_sk(sk)->pf->af;
5027
5028 port = htons(inet_sk(sk)->num);
5029 af->inaddr_any(&autoaddr, port);
5030
5031 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5032}
5033
5034/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5035 *
5036 * From RFC 2292
5037 * 4.2 The cmsghdr Structure *
5038 *
5039 * When ancillary data is sent or received, any number of ancillary data
5040 * objects can be specified by the msg_control and msg_controllen members of
5041 * the msghdr structure, because each object is preceded by
5042 * a cmsghdr structure defining the object's length (the cmsg_len member).
5043 * Historically Berkeley-derived implementations have passed only one object
5044 * at a time, but this API allows multiple objects to be
5045 * passed in a single call to sendmsg() or recvmsg(). The following example
5046 * shows two ancillary data objects in a control buffer.
5047 *
5048 * |<--------------------------- msg_controllen -------------------------->|
5049 * | |
5050 *
5051 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5052 *
5053 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5054 * | | |
5055 *
5056 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5057 *
5058 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5059 * | | | | |
5060 *
5061 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5062 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5063 *
5064 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5065 *
5066 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5067 * ^
5068 * |
5069 *
5070 * msg_control
5071 * points here
5072 */
5073SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5074 sctp_cmsgs_t *cmsgs)
5075{
5076 struct cmsghdr *cmsg;
5077
5078 for (cmsg = CMSG_FIRSTHDR(msg);
5079 cmsg != NULL;
5080 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5081 if (!CMSG_OK(msg, cmsg))
5082 return -EINVAL;
5083
5084 /* Should we parse this header or ignore? */
5085 if (cmsg->cmsg_level != IPPROTO_SCTP)
5086 continue;
5087
5088 /* Strictly check lengths following example in SCM code. */
5089 switch (cmsg->cmsg_type) {
5090 case SCTP_INIT:
5091 /* SCTP Socket API Extension
5092 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5093 *
5094 * This cmsghdr structure provides information for
5095 * initializing new SCTP associations with sendmsg().
5096 * The SCTP_INITMSG socket option uses this same data
5097 * structure. This structure is not used for
5098 * recvmsg().
5099 *
5100 * cmsg_level cmsg_type cmsg_data[]
5101 * ------------ ------------ ----------------------
5102 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5103 */
5104 if (cmsg->cmsg_len !=
5105 CMSG_LEN(sizeof(struct sctp_initmsg)))
5106 return -EINVAL;
5107 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5108 break;
5109
5110 case SCTP_SNDRCV:
5111 /* SCTP Socket API Extension
5112 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5113 *
5114 * This cmsghdr structure specifies SCTP options for
5115 * sendmsg() and describes SCTP header information
5116 * about a received message through recvmsg().
5117 *
5118 * cmsg_level cmsg_type cmsg_data[]
5119 * ------------ ------------ ----------------------
5120 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5121 */
5122 if (cmsg->cmsg_len !=
5123 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5124 return -EINVAL;
5125
5126 cmsgs->info =
5127 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5128
5129 /* Minimally, validate the sinfo_flags. */
5130 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07005131 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5132 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005133 return -EINVAL;
5134 break;
5135
5136 default:
5137 return -EINVAL;
5138 };
5139 }
5140 return 0;
5141}
5142
5143/*
5144 * Wait for a packet..
5145 * Note: This function is the same function as in core/datagram.c
5146 * with a few modifications to make lksctp work.
5147 */
5148static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5149{
5150 int error;
5151 DEFINE_WAIT(wait);
5152
5153 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5154
5155 /* Socket errors? */
5156 error = sock_error(sk);
5157 if (error)
5158 goto out;
5159
5160 if (!skb_queue_empty(&sk->sk_receive_queue))
5161 goto ready;
5162
5163 /* Socket shut down? */
5164 if (sk->sk_shutdown & RCV_SHUTDOWN)
5165 goto out;
5166
5167 /* Sequenced packets can come disconnected. If so we report the
5168 * problem.
5169 */
5170 error = -ENOTCONN;
5171
5172 /* Is there a good reason to think that we may receive some data? */
5173 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5174 goto out;
5175
5176 /* Handle signals. */
5177 if (signal_pending(current))
5178 goto interrupted;
5179
5180 /* Let another process have a go. Since we are going to sleep
5181 * anyway. Note: This may cause odd behaviors if the message
5182 * does not fit in the user's buffer, but this seems to be the
5183 * only way to honor MSG_DONTWAIT realistically.
5184 */
5185 sctp_release_sock(sk);
5186 *timeo_p = schedule_timeout(*timeo_p);
5187 sctp_lock_sock(sk);
5188
5189ready:
5190 finish_wait(sk->sk_sleep, &wait);
5191 return 0;
5192
5193interrupted:
5194 error = sock_intr_errno(*timeo_p);
5195
5196out:
5197 finish_wait(sk->sk_sleep, &wait);
5198 *err = error;
5199 return error;
5200}
5201
5202/* Receive a datagram.
5203 * Note: This is pretty much the same routine as in core/datagram.c
5204 * with a few changes to make lksctp work.
5205 */
5206static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5207 int noblock, int *err)
5208{
5209 int error;
5210 struct sk_buff *skb;
5211 long timeo;
5212
Linus Torvalds1da177e2005-04-16 15:20:36 -07005213 timeo = sock_rcvtimeo(sk, noblock);
5214
5215 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5216 timeo, MAX_SCHEDULE_TIMEOUT);
5217
5218 do {
5219 /* Again only user level code calls this function,
5220 * so nothing interrupt level
5221 * will suddenly eat the receive_queue.
5222 *
5223 * Look at current nfs client by the way...
5224 * However, this function was corrent in any case. 8)
5225 */
5226 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07005227 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005228 skb = skb_peek(&sk->sk_receive_queue);
5229 if (skb)
5230 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07005231 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005232 } else {
5233 skb = skb_dequeue(&sk->sk_receive_queue);
5234 }
5235
5236 if (skb)
5237 return skb;
5238
Neil Horman6736dc32005-12-02 20:30:06 -08005239 /* Caller is allowed not to check sk->sk_err before calling. */
5240 error = sock_error(sk);
5241 if (error)
5242 goto no_packet;
5243
Linus Torvalds1da177e2005-04-16 15:20:36 -07005244 if (sk->sk_shutdown & RCV_SHUTDOWN)
5245 break;
5246
5247 /* User doesn't want to wait. */
5248 error = -EAGAIN;
5249 if (!timeo)
5250 goto no_packet;
5251 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5252
5253 return NULL;
5254
5255no_packet:
5256 *err = error;
5257 return NULL;
5258}
5259
5260/* If sndbuf has changed, wake up per association sndbuf waiters. */
5261static void __sctp_write_space(struct sctp_association *asoc)
5262{
5263 struct sock *sk = asoc->base.sk;
5264 struct socket *sock = sk->sk_socket;
5265
5266 if ((sctp_wspace(asoc) > 0) && sock) {
5267 if (waitqueue_active(&asoc->wait))
5268 wake_up_interruptible(&asoc->wait);
5269
5270 if (sctp_writeable(sk)) {
5271 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5272 wake_up_interruptible(sk->sk_sleep);
5273
5274 /* Note that we try to include the Async I/O support
5275 * here by modeling from the current TCP/UDP code.
5276 * We have not tested with it yet.
5277 */
5278 if (sock->fasync_list &&
5279 !(sk->sk_shutdown & SEND_SHUTDOWN))
5280 sock_wake_async(sock, 2, POLL_OUT);
5281 }
5282 }
5283}
5284
5285/* Do accounting for the sndbuf space.
5286 * Decrement the used sndbuf space of the corresponding association by the
5287 * data size which was just transmitted(freed).
5288 */
5289static void sctp_wfree(struct sk_buff *skb)
5290{
5291 struct sctp_association *asoc;
5292 struct sctp_chunk *chunk;
5293 struct sock *sk;
5294
5295 /* Get the saved chunk pointer. */
5296 chunk = *((struct sctp_chunk **)(skb->cb));
5297 asoc = chunk->asoc;
5298 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07005299 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5300 sizeof(struct sk_buff) +
5301 sizeof(struct sctp_chunk);
5302
Neil Horman4eb701d2005-04-28 12:02:04 -07005303 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5304
5305 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005306 __sctp_write_space(asoc);
5307
5308 sctp_association_put(asoc);
5309}
5310
5311/* Helper function to wait for space in the sndbuf. */
5312static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5313 size_t msg_len)
5314{
5315 struct sock *sk = asoc->base.sk;
5316 int err = 0;
5317 long current_timeo = *timeo_p;
5318 DEFINE_WAIT(wait);
5319
5320 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5321 asoc, (long)(*timeo_p), msg_len);
5322
5323 /* Increment the association's refcnt. */
5324 sctp_association_hold(asoc);
5325
5326 /* Wait on the association specific sndbuf space. */
5327 for (;;) {
5328 prepare_to_wait_exclusive(&asoc->wait, &wait,
5329 TASK_INTERRUPTIBLE);
5330 if (!*timeo_p)
5331 goto do_nonblock;
5332 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5333 asoc->base.dead)
5334 goto do_error;
5335 if (signal_pending(current))
5336 goto do_interrupted;
5337 if (msg_len <= sctp_wspace(asoc))
5338 break;
5339
5340 /* Let another process have a go. Since we are going
5341 * to sleep anyway.
5342 */
5343 sctp_release_sock(sk);
5344 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005345 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005346 sctp_lock_sock(sk);
5347
5348 *timeo_p = current_timeo;
5349 }
5350
5351out:
5352 finish_wait(&asoc->wait, &wait);
5353
5354 /* Release the association's refcnt. */
5355 sctp_association_put(asoc);
5356
5357 return err;
5358
5359do_error:
5360 err = -EPIPE;
5361 goto out;
5362
5363do_interrupted:
5364 err = sock_intr_errno(*timeo_p);
5365 goto out;
5366
5367do_nonblock:
5368 err = -EAGAIN;
5369 goto out;
5370}
5371
5372/* If socket sndbuf has changed, wake up all per association waiters. */
5373void sctp_write_space(struct sock *sk)
5374{
5375 struct sctp_association *asoc;
5376 struct list_head *pos;
5377
5378 /* Wake up the tasks in each wait queue. */
5379 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5380 asoc = list_entry(pos, struct sctp_association, asocs);
5381 __sctp_write_space(asoc);
5382 }
5383}
5384
5385/* Is there any sndbuf space available on the socket?
5386 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08005387 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07005388 * associations on the same socket. For a UDP-style socket with
5389 * multiple associations, it is possible for it to be "unwriteable"
5390 * prematurely. I assume that this is acceptable because
5391 * a premature "unwriteable" is better than an accidental "writeable" which
5392 * would cause an unwanted block under certain circumstances. For the 1-1
5393 * UDP-style sockets or TCP-style sockets, this code should work.
5394 * - Daisy
5395 */
5396static int sctp_writeable(struct sock *sk)
5397{
5398 int amt = 0;
5399
Neil Horman9bffc4a2005-12-19 14:24:40 -08005400 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005401 if (amt < 0)
5402 amt = 0;
5403 return amt;
5404}
5405
5406/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5407 * returns immediately with EINPROGRESS.
5408 */
5409static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5410{
5411 struct sock *sk = asoc->base.sk;
5412 int err = 0;
5413 long current_timeo = *timeo_p;
5414 DEFINE_WAIT(wait);
5415
5416 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5417 (long)(*timeo_p));
5418
5419 /* Increment the association's refcnt. */
5420 sctp_association_hold(asoc);
5421
5422 for (;;) {
5423 prepare_to_wait_exclusive(&asoc->wait, &wait,
5424 TASK_INTERRUPTIBLE);
5425 if (!*timeo_p)
5426 goto do_nonblock;
5427 if (sk->sk_shutdown & RCV_SHUTDOWN)
5428 break;
5429 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5430 asoc->base.dead)
5431 goto do_error;
5432 if (signal_pending(current))
5433 goto do_interrupted;
5434
5435 if (sctp_state(asoc, ESTABLISHED))
5436 break;
5437
5438 /* Let another process have a go. Since we are going
5439 * to sleep anyway.
5440 */
5441 sctp_release_sock(sk);
5442 current_timeo = schedule_timeout(current_timeo);
5443 sctp_lock_sock(sk);
5444
5445 *timeo_p = current_timeo;
5446 }
5447
5448out:
5449 finish_wait(&asoc->wait, &wait);
5450
5451 /* Release the association's refcnt. */
5452 sctp_association_put(asoc);
5453
5454 return err;
5455
5456do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08005457 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005458 err = -ETIMEDOUT;
5459 else
5460 err = -ECONNREFUSED;
5461 goto out;
5462
5463do_interrupted:
5464 err = sock_intr_errno(*timeo_p);
5465 goto out;
5466
5467do_nonblock:
5468 err = -EINPROGRESS;
5469 goto out;
5470}
5471
5472static int sctp_wait_for_accept(struct sock *sk, long timeo)
5473{
5474 struct sctp_endpoint *ep;
5475 int err = 0;
5476 DEFINE_WAIT(wait);
5477
5478 ep = sctp_sk(sk)->ep;
5479
5480
5481 for (;;) {
5482 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5483 TASK_INTERRUPTIBLE);
5484
5485 if (list_empty(&ep->asocs)) {
5486 sctp_release_sock(sk);
5487 timeo = schedule_timeout(timeo);
5488 sctp_lock_sock(sk);
5489 }
5490
5491 err = -EINVAL;
5492 if (!sctp_sstate(sk, LISTENING))
5493 break;
5494
5495 err = 0;
5496 if (!list_empty(&ep->asocs))
5497 break;
5498
5499 err = sock_intr_errno(timeo);
5500 if (signal_pending(current))
5501 break;
5502
5503 err = -EAGAIN;
5504 if (!timeo)
5505 break;
5506 }
5507
5508 finish_wait(sk->sk_sleep, &wait);
5509
5510 return err;
5511}
5512
5513void sctp_wait_for_close(struct sock *sk, long timeout)
5514{
5515 DEFINE_WAIT(wait);
5516
5517 do {
5518 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5519 if (list_empty(&sctp_sk(sk)->ep->asocs))
5520 break;
5521 sctp_release_sock(sk);
5522 timeout = schedule_timeout(timeout);
5523 sctp_lock_sock(sk);
5524 } while (!signal_pending(current) && timeout);
5525
5526 finish_wait(sk->sk_sleep, &wait);
5527}
5528
5529/* Populate the fields of the newsk from the oldsk and migrate the assoc
5530 * and its messages to the newsk.
5531 */
5532static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5533 struct sctp_association *assoc,
5534 sctp_socket_type_t type)
5535{
5536 struct sctp_sock *oldsp = sctp_sk(oldsk);
5537 struct sctp_sock *newsp = sctp_sk(newsk);
5538 struct sctp_bind_bucket *pp; /* hash list port iterator */
5539 struct sctp_endpoint *newep = newsp->ep;
5540 struct sk_buff *skb, *tmp;
5541 struct sctp_ulpevent *event;
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005542 int flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005543
5544 /* Migrate socket buffer sizes and all the socket level options to the
5545 * new socket.
5546 */
5547 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5548 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5549 /* Brute force copy old sctp opt. */
5550 inet_sk_copy_descendant(newsk, oldsk);
5551
5552 /* Restore the ep value that was overwritten with the above structure
5553 * copy.
5554 */
5555 newsp->ep = newep;
5556 newsp->hmac = NULL;
5557
5558 /* Hook this new socket in to the bind_hash list. */
5559 pp = sctp_sk(oldsk)->bind_hash;
5560 sk_add_bind_node(newsk, &pp->owner);
5561 sctp_sk(newsk)->bind_hash = pp;
5562 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5563
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005564 /* Copy the bind_addr list from the original endpoint to the new
5565 * endpoint so that we can handle restarts properly
5566 */
5567 if (assoc->peer.ipv4_address)
5568 flags |= SCTP_ADDR4_PEERSUPP;
5569 if (assoc->peer.ipv6_address)
5570 flags |= SCTP_ADDR6_PEERSUPP;
5571 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5572 &oldsp->ep->base.bind_addr,
5573 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5574
Linus Torvalds1da177e2005-04-16 15:20:36 -07005575 /* Move any messages in the old socket's receive queue that are for the
5576 * peeled off association to the new socket's receive queue.
5577 */
5578 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5579 event = sctp_skb2event(skb);
5580 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005581 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005582 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005583 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005584 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005585 }
5586 }
5587
5588 /* Clean up any messages pending delivery due to partial
5589 * delivery. Three cases:
5590 * 1) No partial deliver; no work.
5591 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5592 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5593 */
5594 skb_queue_head_init(&newsp->pd_lobby);
5595 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5596
5597 if (sctp_sk(oldsk)->pd_mode) {
5598 struct sk_buff_head *queue;
5599
5600 /* Decide which queue to move pd_lobby skbs to. */
5601 if (assoc->ulpq.pd_mode) {
5602 queue = &newsp->pd_lobby;
5603 } else
5604 queue = &newsk->sk_receive_queue;
5605
5606 /* Walk through the pd_lobby, looking for skbs that
5607 * need moved to the new socket.
5608 */
5609 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5610 event = sctp_skb2event(skb);
5611 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005612 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005613 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005614 __skb_queue_tail(queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005615 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005616 }
5617 }
5618
5619 /* Clear up any skbs waiting for the partial
5620 * delivery to finish.
5621 */
5622 if (assoc->ulpq.pd_mode)
5623 sctp_clear_pd(oldsk);
5624
5625 }
5626
5627 /* Set the type of socket to indicate that it is peeled off from the
5628 * original UDP-style socket or created with the accept() call on a
5629 * TCP-style socket..
5630 */
5631 newsp->type = type;
5632
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005633 /* Mark the new socket "in-use" by the user so that any packets
5634 * that may arrive on the association after we've moved it are
5635 * queued to the backlog. This prevents a potential race between
5636 * backlog processing on the old socket and new-packet processing
5637 * on the new socket.
5638 */
5639 sctp_lock_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005640 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005641
5642 /* If the association on the newsk is already closed before accept()
5643 * is called, set RCV_SHUTDOWN flag.
5644 */
5645 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5646 newsk->sk_shutdown |= RCV_SHUTDOWN;
5647
5648 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005649 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005650}
5651
5652/* This proto struct describes the ULP interface for SCTP. */
5653struct proto sctp_prot = {
5654 .name = "SCTP",
5655 .owner = THIS_MODULE,
5656 .close = sctp_close,
5657 .connect = sctp_connect,
5658 .disconnect = sctp_disconnect,
5659 .accept = sctp_accept,
5660 .ioctl = sctp_ioctl,
5661 .init = sctp_init_sock,
5662 .destroy = sctp_destroy_sock,
5663 .shutdown = sctp_shutdown,
5664 .setsockopt = sctp_setsockopt,
5665 .getsockopt = sctp_getsockopt,
5666 .sendmsg = sctp_sendmsg,
5667 .recvmsg = sctp_recvmsg,
5668 .bind = sctp_bind,
5669 .backlog_rcv = sctp_backlog_rcv,
5670 .hash = sctp_hash,
5671 .unhash = sctp_unhash,
5672 .get_port = sctp_get_port,
5673 .obj_size = sizeof(struct sctp_sock),
5674};
5675
5676#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5677struct proto sctpv6_prot = {
5678 .name = "SCTPv6",
5679 .owner = THIS_MODULE,
5680 .close = sctp_close,
5681 .connect = sctp_connect,
5682 .disconnect = sctp_disconnect,
5683 .accept = sctp_accept,
5684 .ioctl = sctp_ioctl,
5685 .init = sctp_init_sock,
5686 .destroy = sctp_destroy_sock,
5687 .shutdown = sctp_shutdown,
5688 .setsockopt = sctp_setsockopt,
5689 .getsockopt = sctp_getsockopt,
5690 .sendmsg = sctp_sendmsg,
5691 .recvmsg = sctp_recvmsg,
5692 .bind = sctp_bind,
5693 .backlog_rcv = sctp_backlog_rcv,
5694 .hash = sctp_hash,
5695 .unhash = sctp_unhash,
5696 .get_port = sctp_get_port,
5697 .obj_size = sizeof(struct sctp6_sock),
5698};
5699#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */