blob: 0a2c71d0d8aad858eb3408591c1af594d1d51587 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#include <linux/types.h>
61#include <linux/kernel.h>
62#include <linux/wait.h>
63#include <linux/time.h>
64#include <linux/ip.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080065#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066#include <linux/fcntl.h>
67#include <linux/poll.h>
68#include <linux/init.h>
69#include <linux/crypto.h>
70
71#include <net/ip.h>
72#include <net/icmp.h>
73#include <net/route.h>
74#include <net/ipv6.h>
75#include <net/inet_common.h>
76
77#include <linux/socket.h> /* for sa_family_t */
78#include <net/sock.h>
79#include <net/sctp/sctp.h>
80#include <net/sctp/sm.h>
81
82/* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
85 */
86
87/* Forward declarations for internal helper functions. */
88static int sctp_writeable(struct sock *sk);
89static void sctp_wfree(struct sk_buff *skb);
90static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94static int sctp_wait_for_accept(struct sock *sk, long timeo);
95static void sctp_wait_for_close(struct sock *sk, long timeo);
96static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105static int sctp_autobind(struct sock *sk);
106static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
109
110extern kmem_cache_t *sctp_bucket_cachep;
111
112/* Get the sndbuf space available at the time on the association. */
113static inline int sctp_wspace(struct sctp_association *asoc)
114{
115 struct sock *sk = asoc->base.sk;
116 int amt = 0;
117
Neil Horman4eb701d2005-04-28 12:02:04 -0700118 if (asoc->ep->sndbuf_policy) {
119 /* make sure that no association uses more than sk_sndbuf */
120 amt = sk->sk_sndbuf - asoc->sndbuf_used;
121 } else {
122 /* do socket level accounting */
123 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
124 }
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 if (amt < 0)
127 amt = 0;
Neil Horman4eb701d2005-04-28 12:02:04 -0700128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 return amt;
130}
131
132/* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
135 *
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
139 * tracking.
140 */
141static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
142{
143 struct sctp_association *asoc = chunk->asoc;
144 struct sock *sk = asoc->base.sk;
145
146 /* The sndbuf space is tracked per association. */
147 sctp_association_hold(asoc);
148
Neil Horman4eb701d2005-04-28 12:02:04 -0700149 skb_set_owner_w(chunk->skb, sk);
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 chunk->skb->destructor = sctp_wfree;
152 /* Save the chunk pointer in skb for sctp_wfree to use later. */
153 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
154
Neil Horman4eb701d2005-04-28 12:02:04 -0700155 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
156 sizeof(struct sk_buff) +
157 sizeof(struct sctp_chunk);
158
Neil Horman4eb701d2005-04-28 12:02:04 -0700159 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160}
161
162/* Verify that this is a valid address. */
163static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
164 int len)
165{
166 struct sctp_af *af;
167
168 /* Verify basic sockaddr. */
169 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
170 if (!af)
171 return -EINVAL;
172
173 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700174 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 return -EINVAL;
176
177 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
178 return -EINVAL;
179
180 return 0;
181}
182
183/* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
185 */
186struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
187{
188 struct sctp_association *asoc = NULL;
189
190 /* If this is not a UDP-style socket, assoc id should be ignored. */
191 if (!sctp_style(sk, UDP)) {
192 /* Return NULL if the socket state is not ESTABLISHED. It
193 * could be a TCP-style listening socket or a socket which
194 * hasn't yet called connect() to establish an association.
195 */
196 if (!sctp_sstate(sk, ESTABLISHED))
197 return NULL;
198
199 /* Get the first and the only association from the list. */
200 if (!list_empty(&sctp_sk(sk)->ep->asocs))
201 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
202 struct sctp_association, asocs);
203 return asoc;
204 }
205
206 /* Otherwise this is a UDP-style socket. */
207 if (!id || (id == (sctp_assoc_t)-1))
208 return NULL;
209
210 spin_lock_bh(&sctp_assocs_id_lock);
211 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
212 spin_unlock_bh(&sctp_assocs_id_lock);
213
214 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
215 return NULL;
216
217 return asoc;
218}
219
220/* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
222 * the same.
223 */
224static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
226 sctp_assoc_t id)
227{
228 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
229 struct sctp_transport *transport;
230 union sctp_addr *laddr = (union sctp_addr *)addr;
231
232 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
233 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
234 (union sctp_addr *)addr,
235 &transport);
236 laddr->v4.sin_port = htons(laddr->v4.sin_port);
237
238 if (!addr_asoc)
239 return NULL;
240
241 id_asoc = sctp_id2assoc(sk, id);
242 if (id_asoc && (id_asoc != addr_asoc))
243 return NULL;
244
245 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
246 (union sctp_addr *)addr);
247
248 return transport;
249}
250
251/* API 3.1.2 bind() - UDP Style Syntax
252 * The syntax of bind() is,
253 *
254 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
255 *
256 * sd - the socket descriptor returned by socket().
257 * addr - the address structure (struct sockaddr_in or struct
258 * sockaddr_in6 [RFC 2553]),
259 * addr_len - the size of the address structure.
260 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700261SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262{
263 int retval = 0;
264
265 sctp_lock_sock(sk);
266
Frank Filz3f7a87d2005-06-20 13:14:57 -0700267 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
268 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270 /* Disallow binding twice. */
271 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700272 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 addr_len);
274 else
275 retval = -EINVAL;
276
277 sctp_release_sock(sk);
278
279 return retval;
280}
281
282static long sctp_get_port_local(struct sock *, union sctp_addr *);
283
284/* Verify this is a valid sockaddr. */
285static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
286 union sctp_addr *addr, int len)
287{
288 struct sctp_af *af;
289
290 /* Check minimum size. */
291 if (len < sizeof (struct sockaddr))
292 return NULL;
293
294 /* Does this PF support this AF? */
295 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
296 return NULL;
297
298 /* If we get this far, af is valid. */
299 af = sctp_get_af_specific(addr->sa.sa_family);
300
301 if (len < af->sockaddr_len)
302 return NULL;
303
304 return af;
305}
306
307/* Bind a local address either to an endpoint or to an association. */
308SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
309{
310 struct sctp_sock *sp = sctp_sk(sk);
311 struct sctp_endpoint *ep = sp->ep;
312 struct sctp_bind_addr *bp = &ep->base.bind_addr;
313 struct sctp_af *af;
314 unsigned short snum;
315 int ret = 0;
316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 /* Common sockaddr verification. */
318 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700319 if (!af) {
320 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
321 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700323 }
324
325 snum = ntohs(addr->v4.sin_port);
326
327 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
328 ", port: %d, new port: %d, len: %d)\n",
329 sk,
330 addr,
331 bp->port, snum,
332 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334 /* PF specific bind() address verification. */
335 if (!sp->pf->bind_verify(sp, addr))
336 return -EADDRNOTAVAIL;
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 /* We must either be unbound, or bind to the same port. */
339 if (bp->port && (snum != bp->port)) {
340 SCTP_DEBUG_PRINTK("sctp_do_bind:"
341 " New port %d does not match existing port "
342 "%d.\n", snum, bp->port);
343 return -EINVAL;
344 }
345
346 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
347 return -EACCES;
348
349 /* Make sure we are allowed to bind here.
350 * The function sctp_get_port_local() does duplicate address
351 * detection.
352 */
353 if ((ret = sctp_get_port_local(sk, addr))) {
354 if (ret == (long) sk) {
355 /* This endpoint has a conflicting address. */
356 return -EINVAL;
357 } else {
358 return -EADDRINUSE;
359 }
360 }
361
362 /* Refresh ephemeral port. */
363 if (!bp->port)
364 bp->port = inet_sk(sk)->num;
365
366 /* Add the address to the bind address list. */
367 sctp_local_bh_disable();
368 sctp_write_lock(&ep->base.addr_lock);
369
370 /* Use GFP_ATOMIC since BHs are disabled. */
371 addr->v4.sin_port = ntohs(addr->v4.sin_port);
372 ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC);
373 addr->v4.sin_port = htons(addr->v4.sin_port);
374 sctp_write_unlock(&ep->base.addr_lock);
375 sctp_local_bh_enable();
376
377 /* Copy back into socket for getsockname() use. */
378 if (!ret) {
379 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
380 af->to_sk_saddr(addr, sk);
381 }
382
383 return ret;
384}
385
386 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
387 *
388 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
389 * at any one time. If a sender, after sending an ASCONF chunk, decides
390 * it needs to transfer another ASCONF Chunk, it MUST wait until the
391 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
392 * subsequent ASCONF. Note this restriction binds each side, so at any
393 * time two ASCONF may be in-transit on any given association (one sent
394 * from each endpoint).
395 */
396static int sctp_send_asconf(struct sctp_association *asoc,
397 struct sctp_chunk *chunk)
398{
399 int retval = 0;
400
401 /* If there is an outstanding ASCONF chunk, queue it for later
402 * transmission.
403 */
404 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700405 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 goto out;
407 }
408
409 /* Hold the chunk until an ASCONF_ACK is received. */
410 sctp_chunk_hold(chunk);
411 retval = sctp_primitive_ASCONF(asoc, chunk);
412 if (retval)
413 sctp_chunk_free(chunk);
414 else
415 asoc->addip_last_asconf = chunk;
416
417out:
418 return retval;
419}
420
421/* Add a list of addresses as bind addresses to local endpoint or
422 * association.
423 *
424 * Basically run through each address specified in the addrs/addrcnt
425 * array/length pair, determine if it is IPv6 or IPv4 and call
426 * sctp_do_bind() on it.
427 *
428 * If any of them fails, then the operation will be reversed and the
429 * ones that were added will be removed.
430 *
431 * Only sctp_setsockopt_bindx() is supposed to call this function.
432 */
433int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
434{
435 int cnt;
436 int retval = 0;
437 void *addr_buf;
438 struct sockaddr *sa_addr;
439 struct sctp_af *af;
440
441 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
442 sk, addrs, addrcnt);
443
444 addr_buf = addrs;
445 for (cnt = 0; cnt < addrcnt; cnt++) {
446 /* The list may contain either IPv4 or IPv6 address;
447 * determine the address length for walking thru the list.
448 */
449 sa_addr = (struct sockaddr *)addr_buf;
450 af = sctp_get_af_specific(sa_addr->sa_family);
451 if (!af) {
452 retval = -EINVAL;
453 goto err_bindx_add;
454 }
455
456 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
457 af->sockaddr_len);
458
459 addr_buf += af->sockaddr_len;
460
461err_bindx_add:
462 if (retval < 0) {
463 /* Failed. Cleanup the ones that have been added */
464 if (cnt > 0)
465 sctp_bindx_rem(sk, addrs, cnt);
466 return retval;
467 }
468 }
469
470 return retval;
471}
472
473/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
474 * associations that are part of the endpoint indicating that a list of local
475 * addresses are added to the endpoint.
476 *
477 * If any of the addresses is already in the bind address list of the
478 * association, we do not send the chunk for that association. But it will not
479 * affect other associations.
480 *
481 * Only sctp_setsockopt_bindx() is supposed to call this function.
482 */
483static int sctp_send_asconf_add_ip(struct sock *sk,
484 struct sockaddr *addrs,
485 int addrcnt)
486{
487 struct sctp_sock *sp;
488 struct sctp_endpoint *ep;
489 struct sctp_association *asoc;
490 struct sctp_bind_addr *bp;
491 struct sctp_chunk *chunk;
492 struct sctp_sockaddr_entry *laddr;
493 union sctp_addr *addr;
494 void *addr_buf;
495 struct sctp_af *af;
496 struct list_head *pos;
497 struct list_head *p;
498 int i;
499 int retval = 0;
500
501 if (!sctp_addip_enable)
502 return retval;
503
504 sp = sctp_sk(sk);
505 ep = sp->ep;
506
507 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
508 __FUNCTION__, sk, addrs, addrcnt);
509
510 list_for_each(pos, &ep->asocs) {
511 asoc = list_entry(pos, struct sctp_association, asocs);
512
513 if (!asoc->peer.asconf_capable)
514 continue;
515
516 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
517 continue;
518
519 if (!sctp_state(asoc, ESTABLISHED))
520 continue;
521
522 /* Check if any address in the packed array of addresses is
523 * in the bind address list of the association. If so,
524 * do not send the asconf chunk to its peer, but continue with
525 * other associations.
526 */
527 addr_buf = addrs;
528 for (i = 0; i < addrcnt; i++) {
529 addr = (union sctp_addr *)addr_buf;
530 af = sctp_get_af_specific(addr->v4.sin_family);
531 if (!af) {
532 retval = -EINVAL;
533 goto out;
534 }
535
536 if (sctp_assoc_lookup_laddr(asoc, addr))
537 break;
538
539 addr_buf += af->sockaddr_len;
540 }
541 if (i < addrcnt)
542 continue;
543
544 /* Use the first address in bind addr list of association as
545 * Address Parameter of ASCONF CHUNK.
546 */
547 sctp_read_lock(&asoc->base.addr_lock);
548 bp = &asoc->base.bind_addr;
549 p = bp->address_list.next;
550 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
551 sctp_read_unlock(&asoc->base.addr_lock);
552
553 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
554 addrcnt, SCTP_PARAM_ADD_IP);
555 if (!chunk) {
556 retval = -ENOMEM;
557 goto out;
558 }
559
560 retval = sctp_send_asconf(asoc, chunk);
561
562 /* FIXME: After sending the add address ASCONF chunk, we
563 * cannot append the address to the association's binding
564 * address list, because the new address may be used as the
565 * source of a message sent to the peer before the ASCONF
566 * chunk is received by the peer. So we should wait until
567 * ASCONF_ACK is received.
568 */
569 }
570
571out:
572 return retval;
573}
574
575/* Remove a list of addresses from bind addresses list. Do not remove the
576 * last address.
577 *
578 * Basically run through each address specified in the addrs/addrcnt
579 * array/length pair, determine if it is IPv6 or IPv4 and call
580 * sctp_del_bind() on it.
581 *
582 * If any of them fails, then the operation will be reversed and the
583 * ones that were removed will be added back.
584 *
585 * At least one address has to be left; if only one address is
586 * available, the operation will return -EBUSY.
587 *
588 * Only sctp_setsockopt_bindx() is supposed to call this function.
589 */
590int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
591{
592 struct sctp_sock *sp = sctp_sk(sk);
593 struct sctp_endpoint *ep = sp->ep;
594 int cnt;
595 struct sctp_bind_addr *bp = &ep->base.bind_addr;
596 int retval = 0;
597 union sctp_addr saveaddr;
598 void *addr_buf;
599 struct sockaddr *sa_addr;
600 struct sctp_af *af;
601
602 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
603 sk, addrs, addrcnt);
604
605 addr_buf = addrs;
606 for (cnt = 0; cnt < addrcnt; cnt++) {
607 /* If the bind address list is empty or if there is only one
608 * bind address, there is nothing more to be removed (we need
609 * at least one address here).
610 */
611 if (list_empty(&bp->address_list) ||
612 (sctp_list_single_entry(&bp->address_list))) {
613 retval = -EBUSY;
614 goto err_bindx_rem;
615 }
616
617 /* The list may contain either IPv4 or IPv6 address;
618 * determine the address length to copy the address to
619 * saveaddr.
620 */
621 sa_addr = (struct sockaddr *)addr_buf;
622 af = sctp_get_af_specific(sa_addr->sa_family);
623 if (!af) {
624 retval = -EINVAL;
625 goto err_bindx_rem;
626 }
627 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
628 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
629 if (saveaddr.v4.sin_port != bp->port) {
630 retval = -EINVAL;
631 goto err_bindx_rem;
632 }
633
634 /* FIXME - There is probably a need to check if sk->sk_saddr and
635 * sk->sk_rcv_addr are currently set to one of the addresses to
636 * be removed. This is something which needs to be looked into
637 * when we are fixing the outstanding issues with multi-homing
638 * socket routing and failover schemes. Refer to comments in
639 * sctp_do_bind(). -daisy
640 */
641 sctp_local_bh_disable();
642 sctp_write_lock(&ep->base.addr_lock);
643
644 retval = sctp_del_bind_addr(bp, &saveaddr);
645
646 sctp_write_unlock(&ep->base.addr_lock);
647 sctp_local_bh_enable();
648
649 addr_buf += af->sockaddr_len;
650err_bindx_rem:
651 if (retval < 0) {
652 /* Failed. Add the ones that has been removed back */
653 if (cnt > 0)
654 sctp_bindx_add(sk, addrs, cnt);
655 return retval;
656 }
657 }
658
659 return retval;
660}
661
662/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
663 * the associations that are part of the endpoint indicating that a list of
664 * local addresses are removed from the endpoint.
665 *
666 * If any of the addresses is already in the bind address list of the
667 * association, we do not send the chunk for that association. But it will not
668 * affect other associations.
669 *
670 * Only sctp_setsockopt_bindx() is supposed to call this function.
671 */
672static int sctp_send_asconf_del_ip(struct sock *sk,
673 struct sockaddr *addrs,
674 int addrcnt)
675{
676 struct sctp_sock *sp;
677 struct sctp_endpoint *ep;
678 struct sctp_association *asoc;
679 struct sctp_bind_addr *bp;
680 struct sctp_chunk *chunk;
681 union sctp_addr *laddr;
682 void *addr_buf;
683 struct sctp_af *af;
684 struct list_head *pos;
685 int i;
686 int retval = 0;
687
688 if (!sctp_addip_enable)
689 return retval;
690
691 sp = sctp_sk(sk);
692 ep = sp->ep;
693
694 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
695 __FUNCTION__, sk, addrs, addrcnt);
696
697 list_for_each(pos, &ep->asocs) {
698 asoc = list_entry(pos, struct sctp_association, asocs);
699
700 if (!asoc->peer.asconf_capable)
701 continue;
702
703 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
704 continue;
705
706 if (!sctp_state(asoc, ESTABLISHED))
707 continue;
708
709 /* Check if any address in the packed array of addresses is
710 * not present in the bind address list of the association.
711 * If so, do not send the asconf chunk to its peer, but
712 * continue with other associations.
713 */
714 addr_buf = addrs;
715 for (i = 0; i < addrcnt; i++) {
716 laddr = (union sctp_addr *)addr_buf;
717 af = sctp_get_af_specific(laddr->v4.sin_family);
718 if (!af) {
719 retval = -EINVAL;
720 goto out;
721 }
722
723 if (!sctp_assoc_lookup_laddr(asoc, laddr))
724 break;
725
726 addr_buf += af->sockaddr_len;
727 }
728 if (i < addrcnt)
729 continue;
730
731 /* Find one address in the association's bind address list
732 * that is not in the packed array of addresses. This is to
733 * make sure that we do not delete all the addresses in the
734 * association.
735 */
736 sctp_read_lock(&asoc->base.addr_lock);
737 bp = &asoc->base.bind_addr;
738 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
739 addrcnt, sp);
740 sctp_read_unlock(&asoc->base.addr_lock);
741 if (!laddr)
742 continue;
743
744 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
745 SCTP_PARAM_DEL_IP);
746 if (!chunk) {
747 retval = -ENOMEM;
748 goto out;
749 }
750
751 retval = sctp_send_asconf(asoc, chunk);
752
753 /* FIXME: After sending the delete address ASCONF chunk, we
754 * cannot remove the addresses from the association's bind
755 * address list, because there maybe some packet send to
756 * the delete addresses, so we should wait until ASCONF_ACK
757 * packet is received.
758 */
759 }
760out:
761 return retval;
762}
763
764/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
765 *
766 * API 8.1
767 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
768 * int flags);
769 *
770 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
771 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
772 * or IPv6 addresses.
773 *
774 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
775 * Section 3.1.2 for this usage.
776 *
777 * addrs is a pointer to an array of one or more socket addresses. Each
778 * address is contained in its appropriate structure (i.e. struct
779 * sockaddr_in or struct sockaddr_in6) the family of the address type
780 * must be used to distengish the address length (note that this
781 * representation is termed a "packed array" of addresses). The caller
782 * specifies the number of addresses in the array with addrcnt.
783 *
784 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
785 * -1, and sets errno to the appropriate error code.
786 *
787 * For SCTP, the port given in each socket address must be the same, or
788 * sctp_bindx() will fail, setting errno to EINVAL.
789 *
790 * The flags parameter is formed from the bitwise OR of zero or more of
791 * the following currently defined flags:
792 *
793 * SCTP_BINDX_ADD_ADDR
794 *
795 * SCTP_BINDX_REM_ADDR
796 *
797 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
798 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
799 * addresses from the association. The two flags are mutually exclusive;
800 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
801 * not remove all addresses from an association; sctp_bindx() will
802 * reject such an attempt with EINVAL.
803 *
804 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
805 * additional addresses with an endpoint after calling bind(). Or use
806 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
807 * socket is associated with so that no new association accepted will be
808 * associated with those addresses. If the endpoint supports dynamic
809 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
810 * endpoint to send the appropriate message to the peer to change the
811 * peers address lists.
812 *
813 * Adding and removing addresses from a connected association is
814 * optional functionality. Implementations that do not support this
815 * functionality should return EOPNOTSUPP.
816 *
817 * Basically do nothing but copying the addresses from user to kernel
818 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700819 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
820 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 *
822 * We don't use copy_from_user() for optimization: we first do the
823 * sanity checks (buffer size -fast- and access check-healthy
824 * pointer); if all of those succeed, then we can alloc the memory
825 * (expensive operation) needed to copy the data to kernel. Then we do
826 * the copying without checking the user space area
827 * (__copy_from_user()).
828 *
829 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
830 * it.
831 *
832 * sk The sk of the socket
833 * addrs The pointer to the addresses in user land
834 * addrssize Size of the addrs buffer
835 * op Operation to perform (add or remove, see the flags of
836 * sctp_bindx)
837 *
838 * Returns 0 if ok, <0 errno code on error.
839 */
840SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
841 struct sockaddr __user *addrs,
842 int addrs_size, int op)
843{
844 struct sockaddr *kaddrs;
845 int err;
846 int addrcnt = 0;
847 int walk_size = 0;
848 struct sockaddr *sa_addr;
849 void *addr_buf;
850 struct sctp_af *af;
851
852 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
853 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
854
855 if (unlikely(addrs_size <= 0))
856 return -EINVAL;
857
858 /* Check the user passed a healthy pointer. */
859 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
860 return -EFAULT;
861
862 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800863 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 if (unlikely(!kaddrs))
865 return -ENOMEM;
866
867 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
868 kfree(kaddrs);
869 return -EFAULT;
870 }
871
872 /* Walk through the addrs buffer and count the number of addresses. */
873 addr_buf = kaddrs;
874 while (walk_size < addrs_size) {
875 sa_addr = (struct sockaddr *)addr_buf;
876 af = sctp_get_af_specific(sa_addr->sa_family);
877
878 /* If the address family is not supported or if this address
879 * causes the address buffer to overflow return EINVAL.
880 */
881 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
882 kfree(kaddrs);
883 return -EINVAL;
884 }
885 addrcnt++;
886 addr_buf += af->sockaddr_len;
887 walk_size += af->sockaddr_len;
888 }
889
890 /* Do the work. */
891 switch (op) {
892 case SCTP_BINDX_ADD_ADDR:
893 err = sctp_bindx_add(sk, kaddrs, addrcnt);
894 if (err)
895 goto out;
896 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
897 break;
898
899 case SCTP_BINDX_REM_ADDR:
900 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
901 if (err)
902 goto out;
903 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
904 break;
905
906 default:
907 err = -EINVAL;
908 break;
909 };
910
911out:
912 kfree(kaddrs);
913
914 return err;
915}
916
Frank Filz3f7a87d2005-06-20 13:14:57 -0700917/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
918 *
919 * Common routine for handling connect() and sctp_connectx().
920 * Connect will come in with just a single address.
921 */
922static int __sctp_connect(struct sock* sk,
923 struct sockaddr *kaddrs,
924 int addrs_size)
925{
926 struct sctp_sock *sp;
927 struct sctp_endpoint *ep;
928 struct sctp_association *asoc = NULL;
929 struct sctp_association *asoc2;
930 struct sctp_transport *transport;
931 union sctp_addr to;
932 struct sctp_af *af;
933 sctp_scope_t scope;
934 long timeo;
935 int err = 0;
936 int addrcnt = 0;
937 int walk_size = 0;
938 struct sockaddr *sa_addr;
939 void *addr_buf;
940
941 sp = sctp_sk(sk);
942 ep = sp->ep;
943
944 /* connect() cannot be done on a socket that is already in ESTABLISHED
945 * state - UDP-style peeled off socket or a TCP-style socket that
946 * is already connected.
947 * It cannot be done even on a TCP-style listening socket.
948 */
949 if (sctp_sstate(sk, ESTABLISHED) ||
950 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
951 err = -EISCONN;
952 goto out_free;
953 }
954
955 /* Walk through the addrs buffer and count the number of addresses. */
956 addr_buf = kaddrs;
957 while (walk_size < addrs_size) {
958 sa_addr = (struct sockaddr *)addr_buf;
959 af = sctp_get_af_specific(sa_addr->sa_family);
960
961 /* If the address family is not supported or if this address
962 * causes the address buffer to overflow return EINVAL.
963 */
964 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
965 err = -EINVAL;
966 goto out_free;
967 }
968
969 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
970 af->sockaddr_len);
971 if (err)
972 goto out_free;
973
974 memcpy(&to, sa_addr, af->sockaddr_len);
975 to.v4.sin_port = ntohs(to.v4.sin_port);
976
977 /* Check if there already is a matching association on the
978 * endpoint (other than the one created here).
979 */
980 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
981 if (asoc2 && asoc2 != asoc) {
982 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
983 err = -EISCONN;
984 else
985 err = -EALREADY;
986 goto out_free;
987 }
988
989 /* If we could not find a matching association on the endpoint,
990 * make sure that there is no peeled-off association matching
991 * the peer address even on another socket.
992 */
993 if (sctp_endpoint_is_peeled_off(ep, &to)) {
994 err = -EADDRNOTAVAIL;
995 goto out_free;
996 }
997
998 if (!asoc) {
999 /* If a bind() or sctp_bindx() is not called prior to
1000 * an sctp_connectx() call, the system picks an
1001 * ephemeral port and will choose an address set
1002 * equivalent to binding with a wildcard address.
1003 */
1004 if (!ep->base.bind_addr.port) {
1005 if (sctp_autobind(sk)) {
1006 err = -EAGAIN;
1007 goto out_free;
1008 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001009 } else {
1010 /*
1011 * If an unprivileged user inherits a 1-many
1012 * style socket with open associations on a
1013 * privileged port, it MAY be permitted to
1014 * accept new associations, but it SHOULD NOT
1015 * be permitted to open new associations.
1016 */
1017 if (ep->base.bind_addr.port < PROT_SOCK &&
1018 !capable(CAP_NET_BIND_SERVICE)) {
1019 err = -EACCES;
1020 goto out_free;
1021 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001022 }
1023
1024 scope = sctp_scope(&to);
1025 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1026 if (!asoc) {
1027 err = -ENOMEM;
1028 goto out_free;
1029 }
1030 }
1031
1032 /* Prime the peer's transport structures. */
1033 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1034 SCTP_UNKNOWN);
1035 if (!transport) {
1036 err = -ENOMEM;
1037 goto out_free;
1038 }
1039
1040 addrcnt++;
1041 addr_buf += af->sockaddr_len;
1042 walk_size += af->sockaddr_len;
1043 }
1044
1045 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1046 if (err < 0) {
1047 goto out_free;
1048 }
1049
1050 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1051 if (err < 0) {
1052 goto out_free;
1053 }
1054
1055 /* Initialize sk's dport and daddr for getpeername() */
1056 inet_sk(sk)->dport = htons(asoc->peer.port);
1057 af = sctp_get_af_specific(to.sa.sa_family);
1058 af->to_sk_daddr(&to, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001059 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001060
1061 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1062 err = sctp_wait_for_connect(asoc, &timeo);
1063
1064 /* Don't free association on exit. */
1065 asoc = NULL;
1066
1067out_free:
1068
1069 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1070 " kaddrs: %p err: %d\n",
1071 asoc, kaddrs, err);
1072 if (asoc)
1073 sctp_association_free(asoc);
1074 return err;
1075}
1076
1077/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1078 *
1079 * API 8.9
1080 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1081 *
1082 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1083 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1084 * or IPv6 addresses.
1085 *
1086 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1087 * Section 3.1.2 for this usage.
1088 *
1089 * addrs is a pointer to an array of one or more socket addresses. Each
1090 * address is contained in its appropriate structure (i.e. struct
1091 * sockaddr_in or struct sockaddr_in6) the family of the address type
1092 * must be used to distengish the address length (note that this
1093 * representation is termed a "packed array" of addresses). The caller
1094 * specifies the number of addresses in the array with addrcnt.
1095 *
1096 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1097 * -1, and sets errno to the appropriate error code.
1098 *
1099 * For SCTP, the port given in each socket address must be the same, or
1100 * sctp_connectx() will fail, setting errno to EINVAL.
1101 *
1102 * An application can use sctp_connectx to initiate an association with
1103 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1104 * allows a caller to specify multiple addresses at which a peer can be
1105 * reached. The way the SCTP stack uses the list of addresses to set up
1106 * the association is implementation dependant. This function only
1107 * specifies that the stack will try to make use of all the addresses in
1108 * the list when needed.
1109 *
1110 * Note that the list of addresses passed in is only used for setting up
1111 * the association. It does not necessarily equal the set of addresses
1112 * the peer uses for the resulting association. If the caller wants to
1113 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1114 * retrieve them after the association has been set up.
1115 *
1116 * Basically do nothing but copying the addresses from user to kernel
1117 * land and invoking either sctp_connectx(). This is used for tunneling
1118 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1119 *
1120 * We don't use copy_from_user() for optimization: we first do the
1121 * sanity checks (buffer size -fast- and access check-healthy
1122 * pointer); if all of those succeed, then we can alloc the memory
1123 * (expensive operation) needed to copy the data to kernel. Then we do
1124 * the copying without checking the user space area
1125 * (__copy_from_user()).
1126 *
1127 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1128 * it.
1129 *
1130 * sk The sk of the socket
1131 * addrs The pointer to the addresses in user land
1132 * addrssize Size of the addrs buffer
1133 *
1134 * Returns 0 if ok, <0 errno code on error.
1135 */
1136SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1137 struct sockaddr __user *addrs,
1138 int addrs_size)
1139{
1140 int err = 0;
1141 struct sockaddr *kaddrs;
1142
1143 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1144 __FUNCTION__, sk, addrs, addrs_size);
1145
1146 if (unlikely(addrs_size <= 0))
1147 return -EINVAL;
1148
1149 /* Check the user passed a healthy pointer. */
1150 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1151 return -EFAULT;
1152
1153 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001154 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001155 if (unlikely(!kaddrs))
1156 return -ENOMEM;
1157
1158 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1159 err = -EFAULT;
1160 } else {
1161 err = __sctp_connect(sk, kaddrs, addrs_size);
1162 }
1163
1164 kfree(kaddrs);
1165 return err;
1166}
1167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168/* API 3.1.4 close() - UDP Style Syntax
1169 * Applications use close() to perform graceful shutdown (as described in
1170 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1171 * by a UDP-style socket.
1172 *
1173 * The syntax is
1174 *
1175 * ret = close(int sd);
1176 *
1177 * sd - the socket descriptor of the associations to be closed.
1178 *
1179 * To gracefully shutdown a specific association represented by the
1180 * UDP-style socket, an application should use the sendmsg() call,
1181 * passing no user data, but including the appropriate flag in the
1182 * ancillary data (see Section xxxx).
1183 *
1184 * If sd in the close() call is a branched-off socket representing only
1185 * one association, the shutdown is performed on that association only.
1186 *
1187 * 4.1.6 close() - TCP Style Syntax
1188 *
1189 * Applications use close() to gracefully close down an association.
1190 *
1191 * The syntax is:
1192 *
1193 * int close(int sd);
1194 *
1195 * sd - the socket descriptor of the association to be closed.
1196 *
1197 * After an application calls close() on a socket descriptor, no further
1198 * socket operations will succeed on that descriptor.
1199 *
1200 * API 7.1.4 SO_LINGER
1201 *
1202 * An application using the TCP-style socket can use this option to
1203 * perform the SCTP ABORT primitive. The linger option structure is:
1204 *
1205 * struct linger {
1206 * int l_onoff; // option on/off
1207 * int l_linger; // linger time
1208 * };
1209 *
1210 * To enable the option, set l_onoff to 1. If the l_linger value is set
1211 * to 0, calling close() is the same as the ABORT primitive. If the
1212 * value is set to a negative value, the setsockopt() call will return
1213 * an error. If the value is set to a positive value linger_time, the
1214 * close() can be blocked for at most linger_time ms. If the graceful
1215 * shutdown phase does not finish during this period, close() will
1216 * return but the graceful shutdown phase continues in the system.
1217 */
1218SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1219{
1220 struct sctp_endpoint *ep;
1221 struct sctp_association *asoc;
1222 struct list_head *pos, *temp;
1223
1224 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1225
1226 sctp_lock_sock(sk);
1227 sk->sk_shutdown = SHUTDOWN_MASK;
1228
1229 ep = sctp_sk(sk)->ep;
1230
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001231 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 list_for_each_safe(pos, temp, &ep->asocs) {
1233 asoc = list_entry(pos, struct sctp_association, asocs);
1234
1235 if (sctp_style(sk, TCP)) {
1236 /* A closed association can still be in the list if
1237 * it belongs to a TCP-style listening socket that is
1238 * not yet accepted. If so, free it. If not, send an
1239 * ABORT or SHUTDOWN based on the linger options.
1240 */
1241 if (sctp_state(asoc, CLOSED)) {
1242 sctp_unhash_established(asoc);
1243 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001244 continue;
1245 }
1246 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001248 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)
1249 sctp_primitive_ABORT(asoc, NULL);
1250 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 sctp_primitive_SHUTDOWN(asoc, NULL);
1252 }
1253
1254 /* Clean up any skbs sitting on the receive queue. */
1255 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1256 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1257
1258 /* On a TCP-style socket, block for at most linger_time if set. */
1259 if (sctp_style(sk, TCP) && timeout)
1260 sctp_wait_for_close(sk, timeout);
1261
1262 /* This will run the backlog queue. */
1263 sctp_release_sock(sk);
1264
1265 /* Supposedly, no process has access to the socket, but
1266 * the net layers still may.
1267 */
1268 sctp_local_bh_disable();
1269 sctp_bh_lock_sock(sk);
1270
1271 /* Hold the sock, since sk_common_release() will put sock_put()
1272 * and we have just a little more cleanup.
1273 */
1274 sock_hold(sk);
1275 sk_common_release(sk);
1276
1277 sctp_bh_unlock_sock(sk);
1278 sctp_local_bh_enable();
1279
1280 sock_put(sk);
1281
1282 SCTP_DBG_OBJCNT_DEC(sock);
1283}
1284
1285/* Handle EPIPE error. */
1286static int sctp_error(struct sock *sk, int flags, int err)
1287{
1288 if (err == -EPIPE)
1289 err = sock_error(sk) ? : -EPIPE;
1290 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1291 send_sig(SIGPIPE, current, 0);
1292 return err;
1293}
1294
1295/* API 3.1.3 sendmsg() - UDP Style Syntax
1296 *
1297 * An application uses sendmsg() and recvmsg() calls to transmit data to
1298 * and receive data from its peer.
1299 *
1300 * ssize_t sendmsg(int socket, const struct msghdr *message,
1301 * int flags);
1302 *
1303 * socket - the socket descriptor of the endpoint.
1304 * message - pointer to the msghdr structure which contains a single
1305 * user message and possibly some ancillary data.
1306 *
1307 * See Section 5 for complete description of the data
1308 * structures.
1309 *
1310 * flags - flags sent or received with the user message, see Section
1311 * 5 for complete description of the flags.
1312 *
1313 * Note: This function could use a rewrite especially when explicit
1314 * connect support comes in.
1315 */
1316/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1317
1318SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1319
1320SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1321 struct msghdr *msg, size_t msg_len)
1322{
1323 struct sctp_sock *sp;
1324 struct sctp_endpoint *ep;
1325 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1326 struct sctp_transport *transport, *chunk_tp;
1327 struct sctp_chunk *chunk;
1328 union sctp_addr to;
1329 struct sockaddr *msg_name = NULL;
1330 struct sctp_sndrcvinfo default_sinfo = { 0 };
1331 struct sctp_sndrcvinfo *sinfo;
1332 struct sctp_initmsg *sinit;
1333 sctp_assoc_t associd = 0;
1334 sctp_cmsgs_t cmsgs = { NULL };
1335 int err;
1336 sctp_scope_t scope;
1337 long timeo;
1338 __u16 sinfo_flags = 0;
1339 struct sctp_datamsg *datamsg;
1340 struct list_head *pos;
1341 int msg_flags = msg->msg_flags;
1342
1343 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1344 sk, msg, msg_len);
1345
1346 err = 0;
1347 sp = sctp_sk(sk);
1348 ep = sp->ep;
1349
Frank Filz3f7a87d2005-06-20 13:14:57 -07001350 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
1352 /* We cannot send a message over a TCP-style listening socket. */
1353 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1354 err = -EPIPE;
1355 goto out_nounlock;
1356 }
1357
1358 /* Parse out the SCTP CMSGs. */
1359 err = sctp_msghdr_parse(msg, &cmsgs);
1360
1361 if (err) {
1362 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1363 goto out_nounlock;
1364 }
1365
1366 /* Fetch the destination address for this packet. This
1367 * address only selects the association--it is not necessarily
1368 * the address we will send to.
1369 * For a peeled-off socket, msg_name is ignored.
1370 */
1371 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1372 int msg_namelen = msg->msg_namelen;
1373
1374 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1375 msg_namelen);
1376 if (err)
1377 return err;
1378
1379 if (msg_namelen > sizeof(to))
1380 msg_namelen = sizeof(to);
1381 memcpy(&to, msg->msg_name, msg_namelen);
1382 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1383 "0x%x:%u.\n",
1384 to.v4.sin_addr.s_addr, to.v4.sin_port);
1385
1386 to.v4.sin_port = ntohs(to.v4.sin_port);
1387 msg_name = msg->msg_name;
1388 }
1389
1390 sinfo = cmsgs.info;
1391 sinit = cmsgs.init;
1392
1393 /* Did the user specify SNDRCVINFO? */
1394 if (sinfo) {
1395 sinfo_flags = sinfo->sinfo_flags;
1396 associd = sinfo->sinfo_assoc_id;
1397 }
1398
1399 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1400 msg_len, sinfo_flags);
1401
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001402 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1403 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 err = -EINVAL;
1405 goto out_nounlock;
1406 }
1407
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001408 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1409 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1410 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 * the msg_iov set to the user abort reason.
1412 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001413 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1414 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 err = -EINVAL;
1416 goto out_nounlock;
1417 }
1418
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001419 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 * specified in msg_name.
1421 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001422 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 err = -EINVAL;
1424 goto out_nounlock;
1425 }
1426
1427 transport = NULL;
1428
1429 SCTP_DEBUG_PRINTK("About to look up association.\n");
1430
1431 sctp_lock_sock(sk);
1432
1433 /* If a msg_name has been specified, assume this is to be used. */
1434 if (msg_name) {
1435 /* Look for a matching association on the endpoint. */
1436 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1437 if (!asoc) {
1438 /* If we could not find a matching association on the
1439 * endpoint, make sure that it is not a TCP-style
1440 * socket that already has an association or there is
1441 * no peeled-off association on another socket.
1442 */
1443 if ((sctp_style(sk, TCP) &&
1444 sctp_sstate(sk, ESTABLISHED)) ||
1445 sctp_endpoint_is_peeled_off(ep, &to)) {
1446 err = -EADDRNOTAVAIL;
1447 goto out_unlock;
1448 }
1449 }
1450 } else {
1451 asoc = sctp_id2assoc(sk, associd);
1452 if (!asoc) {
1453 err = -EPIPE;
1454 goto out_unlock;
1455 }
1456 }
1457
1458 if (asoc) {
1459 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1460
1461 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1462 * socket that has an association in CLOSED state. This can
1463 * happen when an accepted socket has an association that is
1464 * already CLOSED.
1465 */
1466 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1467 err = -EPIPE;
1468 goto out_unlock;
1469 }
1470
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001471 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1473 asoc);
1474 sctp_primitive_SHUTDOWN(asoc, NULL);
1475 err = 0;
1476 goto out_unlock;
1477 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001478 if (sinfo_flags & SCTP_ABORT) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1480 sctp_primitive_ABORT(asoc, msg);
1481 err = 0;
1482 goto out_unlock;
1483 }
1484 }
1485
1486 /* Do we need to create the association? */
1487 if (!asoc) {
1488 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1489
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001490 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 err = -EINVAL;
1492 goto out_unlock;
1493 }
1494
1495 /* Check for invalid stream against the stream counts,
1496 * either the default or the user specified stream counts.
1497 */
1498 if (sinfo) {
1499 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1500 /* Check against the defaults. */
1501 if (sinfo->sinfo_stream >=
1502 sp->initmsg.sinit_num_ostreams) {
1503 err = -EINVAL;
1504 goto out_unlock;
1505 }
1506 } else {
1507 /* Check against the requested. */
1508 if (sinfo->sinfo_stream >=
1509 sinit->sinit_num_ostreams) {
1510 err = -EINVAL;
1511 goto out_unlock;
1512 }
1513 }
1514 }
1515
1516 /*
1517 * API 3.1.2 bind() - UDP Style Syntax
1518 * If a bind() or sctp_bindx() is not called prior to a
1519 * sendmsg() call that initiates a new association, the
1520 * system picks an ephemeral port and will choose an address
1521 * set equivalent to binding with a wildcard address.
1522 */
1523 if (!ep->base.bind_addr.port) {
1524 if (sctp_autobind(sk)) {
1525 err = -EAGAIN;
1526 goto out_unlock;
1527 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001528 } else {
1529 /*
1530 * If an unprivileged user inherits a one-to-many
1531 * style socket with open associations on a privileged
1532 * port, it MAY be permitted to accept new associations,
1533 * but it SHOULD NOT be permitted to open new
1534 * associations.
1535 */
1536 if (ep->base.bind_addr.port < PROT_SOCK &&
1537 !capable(CAP_NET_BIND_SERVICE)) {
1538 err = -EACCES;
1539 goto out_unlock;
1540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 }
1542
1543 scope = sctp_scope(&to);
1544 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1545 if (!new_asoc) {
1546 err = -ENOMEM;
1547 goto out_unlock;
1548 }
1549 asoc = new_asoc;
1550
1551 /* If the SCTP_INIT ancillary data is specified, set all
1552 * the association init values accordingly.
1553 */
1554 if (sinit) {
1555 if (sinit->sinit_num_ostreams) {
1556 asoc->c.sinit_num_ostreams =
1557 sinit->sinit_num_ostreams;
1558 }
1559 if (sinit->sinit_max_instreams) {
1560 asoc->c.sinit_max_instreams =
1561 sinit->sinit_max_instreams;
1562 }
1563 if (sinit->sinit_max_attempts) {
1564 asoc->max_init_attempts
1565 = sinit->sinit_max_attempts;
1566 }
1567 if (sinit->sinit_max_init_timeo) {
1568 asoc->max_init_timeo =
1569 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1570 }
1571 }
1572
1573 /* Prime the peer's transport structures. */
Frank Filz3f7a87d2005-06-20 13:14:57 -07001574 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 if (!transport) {
1576 err = -ENOMEM;
1577 goto out_free;
1578 }
1579 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1580 if (err < 0) {
1581 err = -ENOMEM;
1582 goto out_free;
1583 }
1584 }
1585
1586 /* ASSERT: we have a valid association at this point. */
1587 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1588
1589 if (!sinfo) {
1590 /* If the user didn't specify SNDRCVINFO, make up one with
1591 * some defaults.
1592 */
1593 default_sinfo.sinfo_stream = asoc->default_stream;
1594 default_sinfo.sinfo_flags = asoc->default_flags;
1595 default_sinfo.sinfo_ppid = asoc->default_ppid;
1596 default_sinfo.sinfo_context = asoc->default_context;
1597 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1598 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1599 sinfo = &default_sinfo;
1600 }
1601
1602 /* API 7.1.7, the sndbuf size per association bounds the
1603 * maximum size of data that can be sent in a single send call.
1604 */
1605 if (msg_len > sk->sk_sndbuf) {
1606 err = -EMSGSIZE;
1607 goto out_free;
1608 }
1609
1610 /* If fragmentation is disabled and the message length exceeds the
1611 * association fragmentation point, return EMSGSIZE. The I-D
1612 * does not specify what this error is, but this looks like
1613 * a great fit.
1614 */
1615 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1616 err = -EMSGSIZE;
1617 goto out_free;
1618 }
1619
1620 if (sinfo) {
1621 /* Check for invalid stream. */
1622 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1623 err = -EINVAL;
1624 goto out_free;
1625 }
1626 }
1627
1628 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1629 if (!sctp_wspace(asoc)) {
1630 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1631 if (err)
1632 goto out_free;
1633 }
1634
1635 /* If an address is passed with the sendto/sendmsg call, it is used
1636 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001637 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 */
1639 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001640 (sinfo_flags & SCTP_ADDR_OVER)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1642 if (!chunk_tp) {
1643 err = -EINVAL;
1644 goto out_free;
1645 }
1646 } else
1647 chunk_tp = NULL;
1648
1649 /* Auto-connect, if we aren't connected already. */
1650 if (sctp_state(asoc, CLOSED)) {
1651 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1652 if (err < 0)
1653 goto out_free;
1654 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1655 }
1656
1657 /* Break the message into multiple chunks of maximum size. */
1658 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1659 if (!datamsg) {
1660 err = -ENOMEM;
1661 goto out_free;
1662 }
1663
1664 /* Now send the (possibly) fragmented message. */
1665 list_for_each(pos, &datamsg->chunks) {
1666 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1667 sctp_datamsg_track(chunk);
1668
1669 /* Do accounting for the write space. */
1670 sctp_set_owner_w(chunk);
1671
1672 chunk->transport = chunk_tp;
1673
1674 /* Send it to the lower layers. Note: all chunks
1675 * must either fail or succeed. The lower layer
1676 * works that way today. Keep it that way or this
1677 * breaks.
1678 */
1679 err = sctp_primitive_SEND(asoc, chunk);
1680 /* Did the lower layer accept the chunk? */
1681 if (err)
1682 sctp_chunk_free(chunk);
1683 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1684 }
1685
1686 sctp_datamsg_free(datamsg);
1687 if (err)
1688 goto out_free;
1689 else
1690 err = msg_len;
1691
1692 /* If we are already past ASSOCIATE, the lower
1693 * layers are responsible for association cleanup.
1694 */
1695 goto out_unlock;
1696
1697out_free:
1698 if (new_asoc)
1699 sctp_association_free(asoc);
1700out_unlock:
1701 sctp_release_sock(sk);
1702
1703out_nounlock:
1704 return sctp_error(sk, msg_flags, err);
1705
1706#if 0
1707do_sock_err:
1708 if (msg_len)
1709 err = msg_len;
1710 else
1711 err = sock_error(sk);
1712 goto out;
1713
1714do_interrupted:
1715 if (msg_len)
1716 err = msg_len;
1717 goto out;
1718#endif /* 0 */
1719}
1720
1721/* This is an extended version of skb_pull() that removes the data from the
1722 * start of a skb even when data is spread across the list of skb's in the
1723 * frag_list. len specifies the total amount of data that needs to be removed.
1724 * when 'len' bytes could be removed from the skb, it returns 0.
1725 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1726 * could not be removed.
1727 */
1728static int sctp_skb_pull(struct sk_buff *skb, int len)
1729{
1730 struct sk_buff *list;
1731 int skb_len = skb_headlen(skb);
1732 int rlen;
1733
1734 if (len <= skb_len) {
1735 __skb_pull(skb, len);
1736 return 0;
1737 }
1738 len -= skb_len;
1739 __skb_pull(skb, skb_len);
1740
1741 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1742 rlen = sctp_skb_pull(list, len);
1743 skb->len -= (len-rlen);
1744 skb->data_len -= (len-rlen);
1745
1746 if (!rlen)
1747 return 0;
1748
1749 len = rlen;
1750 }
1751
1752 return len;
1753}
1754
1755/* API 3.1.3 recvmsg() - UDP Style Syntax
1756 *
1757 * ssize_t recvmsg(int socket, struct msghdr *message,
1758 * int flags);
1759 *
1760 * socket - the socket descriptor of the endpoint.
1761 * message - pointer to the msghdr structure which contains a single
1762 * user message and possibly some ancillary data.
1763 *
1764 * See Section 5 for complete description of the data
1765 * structures.
1766 *
1767 * flags - flags sent or received with the user message, see Section
1768 * 5 for complete description of the flags.
1769 */
1770static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1771
1772SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1773 struct msghdr *msg, size_t len, int noblock,
1774 int flags, int *addr_len)
1775{
1776 struct sctp_ulpevent *event = NULL;
1777 struct sctp_sock *sp = sctp_sk(sk);
1778 struct sk_buff *skb;
1779 int copied;
1780 int err = 0;
1781 int skb_len;
1782
1783 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1784 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1785 "len", len, "knoblauch", noblock,
1786 "flags", flags, "addr_len", addr_len);
1787
1788 sctp_lock_sock(sk);
1789
1790 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1791 err = -ENOTCONN;
1792 goto out;
1793 }
1794
1795 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1796 if (!skb)
1797 goto out;
1798
1799 /* Get the total length of the skb including any skb's in the
1800 * frag_list.
1801 */
1802 skb_len = skb->len;
1803
1804 copied = skb_len;
1805 if (copied > len)
1806 copied = len;
1807
1808 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1809
1810 event = sctp_skb2event(skb);
1811
1812 if (err)
1813 goto out_free;
1814
1815 sock_recv_timestamp(msg, sk, skb);
1816 if (sctp_ulpevent_is_notification(event)) {
1817 msg->msg_flags |= MSG_NOTIFICATION;
1818 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1819 } else {
1820 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1821 }
1822
1823 /* Check if we allow SCTP_SNDRCVINFO. */
1824 if (sp->subscribe.sctp_data_io_event)
1825 sctp_ulpevent_read_sndrcvinfo(event, msg);
1826#if 0
1827 /* FIXME: we should be calling IP/IPv6 layers. */
1828 if (sk->sk_protinfo.af_inet.cmsg_flags)
1829 ip_cmsg_recv(msg, skb);
1830#endif
1831
1832 err = copied;
1833
1834 /* If skb's length exceeds the user's buffer, update the skb and
1835 * push it back to the receive_queue so that the next call to
1836 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1837 */
1838 if (skb_len > copied) {
1839 msg->msg_flags &= ~MSG_EOR;
1840 if (flags & MSG_PEEK)
1841 goto out_free;
1842 sctp_skb_pull(skb, copied);
1843 skb_queue_head(&sk->sk_receive_queue, skb);
1844
1845 /* When only partial message is copied to the user, increase
1846 * rwnd by that amount. If all the data in the skb is read,
1847 * rwnd is updated when the event is freed.
1848 */
1849 sctp_assoc_rwnd_increase(event->asoc, copied);
1850 goto out;
1851 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1852 (event->msg_flags & MSG_EOR))
1853 msg->msg_flags |= MSG_EOR;
1854 else
1855 msg->msg_flags &= ~MSG_EOR;
1856
1857out_free:
1858 if (flags & MSG_PEEK) {
1859 /* Release the skb reference acquired after peeking the skb in
1860 * sctp_skb_recv_datagram().
1861 */
1862 kfree_skb(skb);
1863 } else {
1864 /* Free the event which includes releasing the reference to
1865 * the owner of the skb, freeing the skb and updating the
1866 * rwnd.
1867 */
1868 sctp_ulpevent_free(event);
1869 }
1870out:
1871 sctp_release_sock(sk);
1872 return err;
1873}
1874
1875/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1876 *
1877 * This option is a on/off flag. If enabled no SCTP message
1878 * fragmentation will be performed. Instead if a message being sent
1879 * exceeds the current PMTU size, the message will NOT be sent and
1880 * instead a error will be indicated to the user.
1881 */
1882static int sctp_setsockopt_disable_fragments(struct sock *sk,
1883 char __user *optval, int optlen)
1884{
1885 int val;
1886
1887 if (optlen < sizeof(int))
1888 return -EINVAL;
1889
1890 if (get_user(val, (int __user *)optval))
1891 return -EFAULT;
1892
1893 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1894
1895 return 0;
1896}
1897
1898static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1899 int optlen)
1900{
1901 if (optlen != sizeof(struct sctp_event_subscribe))
1902 return -EINVAL;
1903 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1904 return -EFAULT;
1905 return 0;
1906}
1907
1908/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1909 *
1910 * This socket option is applicable to the UDP-style socket only. When
1911 * set it will cause associations that are idle for more than the
1912 * specified number of seconds to automatically close. An association
1913 * being idle is defined an association that has NOT sent or received
1914 * user data. The special value of '0' indicates that no automatic
1915 * close of any associations should be performed. The option expects an
1916 * integer defining the number of seconds of idle time before an
1917 * association is closed.
1918 */
1919static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1920 int optlen)
1921{
1922 struct sctp_sock *sp = sctp_sk(sk);
1923
1924 /* Applicable to UDP-style socket only */
1925 if (sctp_style(sk, TCP))
1926 return -EOPNOTSUPP;
1927 if (optlen != sizeof(int))
1928 return -EINVAL;
1929 if (copy_from_user(&sp->autoclose, optval, optlen))
1930 return -EFAULT;
1931
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 return 0;
1933}
1934
1935/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1936 *
1937 * Applications can enable or disable heartbeats for any peer address of
1938 * an association, modify an address's heartbeat interval, force a
1939 * heartbeat to be sent immediately, and adjust the address's maximum
1940 * number of retransmissions sent before an address is considered
1941 * unreachable. The following structure is used to access and modify an
1942 * address's parameters:
1943 *
1944 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08001945 * sctp_assoc_t spp_assoc_id;
1946 * struct sockaddr_storage spp_address;
1947 * uint32_t spp_hbinterval;
1948 * uint16_t spp_pathmaxrxt;
1949 * uint32_t spp_pathmtu;
1950 * uint32_t spp_sackdelay;
1951 * uint32_t spp_flags;
1952 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08001954 * spp_assoc_id - (one-to-many style socket) This is filled in the
1955 * application, and identifies the association for
1956 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 * spp_address - This specifies which address is of interest.
1958 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08001959 * in milliseconds. If a value of zero
1960 * is present in this field then no changes are to
1961 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 * spp_pathmaxrxt - This contains the maximum number of
1963 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08001964 * considered unreachable. If a value of zero
1965 * is present in this field then no changes are to
1966 * be made to this parameter.
1967 * spp_pathmtu - When Path MTU discovery is disabled the value
1968 * specified here will be the "fixed" path mtu.
1969 * Note that if the spp_address field is empty
1970 * then all associations on this address will
1971 * have this fixed path mtu set upon them.
1972 *
1973 * spp_sackdelay - When delayed sack is enabled, this value specifies
1974 * the number of milliseconds that sacks will be delayed
1975 * for. This value will apply to all addresses of an
1976 * association if the spp_address field is empty. Note
1977 * also, that if delayed sack is enabled and this
1978 * value is set to 0, no change is made to the last
1979 * recorded delayed sack timer value.
1980 *
1981 * spp_flags - These flags are used to control various features
1982 * on an association. The flag field may contain
1983 * zero or more of the following options.
1984 *
1985 * SPP_HB_ENABLE - Enable heartbeats on the
1986 * specified address. Note that if the address
1987 * field is empty all addresses for the association
1988 * have heartbeats enabled upon them.
1989 *
1990 * SPP_HB_DISABLE - Disable heartbeats on the
1991 * speicifed address. Note that if the address
1992 * field is empty all addresses for the association
1993 * will have their heartbeats disabled. Note also
1994 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
1995 * mutually exclusive, only one of these two should
1996 * be specified. Enabling both fields will have
1997 * undetermined results.
1998 *
1999 * SPP_HB_DEMAND - Request a user initiated heartbeat
2000 * to be made immediately.
2001 *
2002 * SPP_PMTUD_ENABLE - This field will enable PMTU
2003 * discovery upon the specified address. Note that
2004 * if the address feild is empty then all addresses
2005 * on the association are effected.
2006 *
2007 * SPP_PMTUD_DISABLE - This field will disable PMTU
2008 * discovery upon the specified address. Note that
2009 * if the address feild is empty then all addresses
2010 * on the association are effected. Not also that
2011 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2012 * exclusive. Enabling both will have undetermined
2013 * results.
2014 *
2015 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2016 * on delayed sack. The time specified in spp_sackdelay
2017 * is used to specify the sack delay for this address. Note
2018 * that if spp_address is empty then all addresses will
2019 * enable delayed sack and take on the sack delay
2020 * value specified in spp_sackdelay.
2021 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2022 * off delayed sack. If the spp_address field is blank then
2023 * delayed sack is disabled for the entire association. Note
2024 * also that this field is mutually exclusive to
2025 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2026 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08002028int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2029 struct sctp_transport *trans,
2030 struct sctp_association *asoc,
2031 struct sctp_sock *sp,
2032 int hb_change,
2033 int pmtud_change,
2034 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 int error;
2037
Frank Filz52ccb8e2005-12-22 11:36:46 -08002038 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2040 if (error)
2041 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 }
2043
Frank Filz52ccb8e2005-12-22 11:36:46 -08002044 if (params->spp_hbinterval) {
2045 if (trans) {
2046 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2047 } else if (asoc) {
2048 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2049 } else {
2050 sp->hbinterval = params->spp_hbinterval;
2051 }
2052 }
2053
2054 if (hb_change) {
2055 if (trans) {
2056 trans->param_flags =
2057 (trans->param_flags & ~SPP_HB) | hb_change;
2058 } else if (asoc) {
2059 asoc->param_flags =
2060 (asoc->param_flags & ~SPP_HB) | hb_change;
2061 } else {
2062 sp->param_flags =
2063 (sp->param_flags & ~SPP_HB) | hb_change;
2064 }
2065 }
2066
2067 if (params->spp_pathmtu) {
2068 if (trans) {
2069 trans->pathmtu = params->spp_pathmtu;
2070 sctp_assoc_sync_pmtu(asoc);
2071 } else if (asoc) {
2072 asoc->pathmtu = params->spp_pathmtu;
2073 sctp_frag_point(sp, params->spp_pathmtu);
2074 } else {
2075 sp->pathmtu = params->spp_pathmtu;
2076 }
2077 }
2078
2079 if (pmtud_change) {
2080 if (trans) {
2081 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2082 (params->spp_flags & SPP_PMTUD_ENABLE);
2083 trans->param_flags =
2084 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2085 if (update) {
2086 sctp_transport_pmtu(trans);
2087 sctp_assoc_sync_pmtu(asoc);
2088 }
2089 } else if (asoc) {
2090 asoc->param_flags =
2091 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2092 } else {
2093 sp->param_flags =
2094 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2095 }
2096 }
2097
2098 if (params->spp_sackdelay) {
2099 if (trans) {
2100 trans->sackdelay =
2101 msecs_to_jiffies(params->spp_sackdelay);
2102 } else if (asoc) {
2103 asoc->sackdelay =
2104 msecs_to_jiffies(params->spp_sackdelay);
2105 } else {
2106 sp->sackdelay = params->spp_sackdelay;
2107 }
2108 }
2109
2110 if (sackdelay_change) {
2111 if (trans) {
2112 trans->param_flags =
2113 (trans->param_flags & ~SPP_SACKDELAY) |
2114 sackdelay_change;
2115 } else if (asoc) {
2116 asoc->param_flags =
2117 (asoc->param_flags & ~SPP_SACKDELAY) |
2118 sackdelay_change;
2119 } else {
2120 sp->param_flags =
2121 (sp->param_flags & ~SPP_SACKDELAY) |
2122 sackdelay_change;
2123 }
2124 }
2125
2126 if (params->spp_pathmaxrxt) {
2127 if (trans) {
2128 trans->pathmaxrxt = params->spp_pathmaxrxt;
2129 } else if (asoc) {
2130 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2131 } else {
2132 sp->pathmaxrxt = params->spp_pathmaxrxt;
2133 }
2134 }
2135
2136 return 0;
2137}
2138
2139static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2140 char __user *optval, int optlen)
2141{
2142 struct sctp_paddrparams params;
2143 struct sctp_transport *trans = NULL;
2144 struct sctp_association *asoc = NULL;
2145 struct sctp_sock *sp = sctp_sk(sk);
2146 int error;
2147 int hb_change, pmtud_change, sackdelay_change;
2148
2149 if (optlen != sizeof(struct sctp_paddrparams))
2150 return - EINVAL;
2151
2152 if (copy_from_user(&params, optval, optlen))
2153 return -EFAULT;
2154
2155 /* Validate flags and value parameters. */
2156 hb_change = params.spp_flags & SPP_HB;
2157 pmtud_change = params.spp_flags & SPP_PMTUD;
2158 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2159
2160 if (hb_change == SPP_HB ||
2161 pmtud_change == SPP_PMTUD ||
2162 sackdelay_change == SPP_SACKDELAY ||
2163 params.spp_sackdelay > 500 ||
2164 (params.spp_pathmtu
2165 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2166 return -EINVAL;
2167
2168 /* If an address other than INADDR_ANY is specified, and
2169 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08002171 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2172 trans = sctp_addr_id2transport(sk, &params.spp_address,
2173 params.spp_assoc_id);
2174 if (!trans)
2175 return -EINVAL;
2176 }
2177
2178 /* Get association, if assoc_id != 0 and the socket is a one
2179 * to many style socket, and an association was not found, then
2180 * the id was invalid.
2181 */
2182 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2183 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2184 return -EINVAL;
2185
2186 /* Heartbeat demand can only be sent on a transport or
2187 * association, but not a socket.
2188 */
2189 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2190 return -EINVAL;
2191
2192 /* Process parameters. */
2193 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2194 hb_change, pmtud_change,
2195 sackdelay_change);
2196
2197 if (error)
2198 return error;
2199
2200 /* If changes are for association, also apply parameters to each
2201 * transport.
2202 */
2203 if (!trans && asoc) {
2204 struct list_head *pos;
2205
2206 list_for_each(pos, &asoc->peer.transport_addr_list) {
2207 trans = list_entry(pos, struct sctp_transport,
2208 transports);
2209 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2210 hb_change, pmtud_change,
2211 sackdelay_change);
2212 }
2213 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214
2215 return 0;
2216}
2217
Frank Filz77086102005-12-22 11:37:30 -08002218/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2219 *
2220 * This options will get or set the delayed ack timer. The time is set
2221 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2222 * endpoints default delayed ack timer value. If the assoc_id field is
2223 * non-zero, then the set or get effects the specified association.
2224 *
2225 * struct sctp_assoc_value {
2226 * sctp_assoc_t assoc_id;
2227 * uint32_t assoc_value;
2228 * };
2229 *
2230 * assoc_id - This parameter, indicates which association the
2231 * user is preforming an action upon. Note that if
2232 * this field's value is zero then the endpoints
2233 * default value is changed (effecting future
2234 * associations only).
2235 *
2236 * assoc_value - This parameter contains the number of milliseconds
2237 * that the user is requesting the delayed ACK timer
2238 * be set to. Note that this value is defined in
2239 * the standard to be between 200 and 500 milliseconds.
2240 *
2241 * Note: a value of zero will leave the value alone,
2242 * but disable SACK delay. A non-zero value will also
2243 * enable SACK delay.
2244 */
2245
2246static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2247 char __user *optval, int optlen)
2248{
2249 struct sctp_assoc_value params;
2250 struct sctp_transport *trans = NULL;
2251 struct sctp_association *asoc = NULL;
2252 struct sctp_sock *sp = sctp_sk(sk);
2253
2254 if (optlen != sizeof(struct sctp_assoc_value))
2255 return - EINVAL;
2256
2257 if (copy_from_user(&params, optval, optlen))
2258 return -EFAULT;
2259
2260 /* Validate value parameter. */
2261 if (params.assoc_value > 500)
2262 return -EINVAL;
2263
2264 /* Get association, if assoc_id != 0 and the socket is a one
2265 * to many style socket, and an association was not found, then
2266 * the id was invalid.
2267 */
2268 asoc = sctp_id2assoc(sk, params.assoc_id);
2269 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2270 return -EINVAL;
2271
2272 if (params.assoc_value) {
2273 if (asoc) {
2274 asoc->sackdelay =
2275 msecs_to_jiffies(params.assoc_value);
2276 asoc->param_flags =
2277 (asoc->param_flags & ~SPP_SACKDELAY) |
2278 SPP_SACKDELAY_ENABLE;
2279 } else {
2280 sp->sackdelay = params.assoc_value;
2281 sp->param_flags =
2282 (sp->param_flags & ~SPP_SACKDELAY) |
2283 SPP_SACKDELAY_ENABLE;
2284 }
2285 } else {
2286 if (asoc) {
2287 asoc->param_flags =
2288 (asoc->param_flags & ~SPP_SACKDELAY) |
2289 SPP_SACKDELAY_DISABLE;
2290 } else {
2291 sp->param_flags =
2292 (sp->param_flags & ~SPP_SACKDELAY) |
2293 SPP_SACKDELAY_DISABLE;
2294 }
2295 }
2296
2297 /* If change is for association, also apply to each transport. */
2298 if (asoc) {
2299 struct list_head *pos;
2300
2301 list_for_each(pos, &asoc->peer.transport_addr_list) {
2302 trans = list_entry(pos, struct sctp_transport,
2303 transports);
2304 if (params.assoc_value) {
2305 trans->sackdelay =
2306 msecs_to_jiffies(params.assoc_value);
2307 trans->param_flags =
2308 (trans->param_flags & ~SPP_SACKDELAY) |
2309 SPP_SACKDELAY_ENABLE;
2310 } else {
2311 trans->param_flags =
2312 (trans->param_flags & ~SPP_SACKDELAY) |
2313 SPP_SACKDELAY_DISABLE;
2314 }
2315 }
2316 }
2317
2318 return 0;
2319}
2320
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2322 *
2323 * Applications can specify protocol parameters for the default association
2324 * initialization. The option name argument to setsockopt() and getsockopt()
2325 * is SCTP_INITMSG.
2326 *
2327 * Setting initialization parameters is effective only on an unconnected
2328 * socket (for UDP-style sockets only future associations are effected
2329 * by the change). With TCP-style sockets, this option is inherited by
2330 * sockets derived from a listener socket.
2331 */
2332static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2333{
2334 struct sctp_initmsg sinit;
2335 struct sctp_sock *sp = sctp_sk(sk);
2336
2337 if (optlen != sizeof(struct sctp_initmsg))
2338 return -EINVAL;
2339 if (copy_from_user(&sinit, optval, optlen))
2340 return -EFAULT;
2341
2342 if (sinit.sinit_num_ostreams)
2343 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2344 if (sinit.sinit_max_instreams)
2345 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2346 if (sinit.sinit_max_attempts)
2347 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2348 if (sinit.sinit_max_init_timeo)
2349 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2350
2351 return 0;
2352}
2353
2354/*
2355 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2356 *
2357 * Applications that wish to use the sendto() system call may wish to
2358 * specify a default set of parameters that would normally be supplied
2359 * through the inclusion of ancillary data. This socket option allows
2360 * such an application to set the default sctp_sndrcvinfo structure.
2361 * The application that wishes to use this socket option simply passes
2362 * in to this call the sctp_sndrcvinfo structure defined in Section
2363 * 5.2.2) The input parameters accepted by this call include
2364 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2365 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2366 * to this call if the caller is using the UDP model.
2367 */
2368static int sctp_setsockopt_default_send_param(struct sock *sk,
2369 char __user *optval, int optlen)
2370{
2371 struct sctp_sndrcvinfo info;
2372 struct sctp_association *asoc;
2373 struct sctp_sock *sp = sctp_sk(sk);
2374
2375 if (optlen != sizeof(struct sctp_sndrcvinfo))
2376 return -EINVAL;
2377 if (copy_from_user(&info, optval, optlen))
2378 return -EFAULT;
2379
2380 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2381 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2382 return -EINVAL;
2383
2384 if (asoc) {
2385 asoc->default_stream = info.sinfo_stream;
2386 asoc->default_flags = info.sinfo_flags;
2387 asoc->default_ppid = info.sinfo_ppid;
2388 asoc->default_context = info.sinfo_context;
2389 asoc->default_timetolive = info.sinfo_timetolive;
2390 } else {
2391 sp->default_stream = info.sinfo_stream;
2392 sp->default_flags = info.sinfo_flags;
2393 sp->default_ppid = info.sinfo_ppid;
2394 sp->default_context = info.sinfo_context;
2395 sp->default_timetolive = info.sinfo_timetolive;
2396 }
2397
2398 return 0;
2399}
2400
2401/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2402 *
2403 * Requests that the local SCTP stack use the enclosed peer address as
2404 * the association primary. The enclosed address must be one of the
2405 * association peer's addresses.
2406 */
2407static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2408 int optlen)
2409{
2410 struct sctp_prim prim;
2411 struct sctp_transport *trans;
2412
2413 if (optlen != sizeof(struct sctp_prim))
2414 return -EINVAL;
2415
2416 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2417 return -EFAULT;
2418
2419 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2420 if (!trans)
2421 return -EINVAL;
2422
2423 sctp_assoc_set_primary(trans->asoc, trans);
2424
2425 return 0;
2426}
2427
2428/*
2429 * 7.1.5 SCTP_NODELAY
2430 *
2431 * Turn on/off any Nagle-like algorithm. This means that packets are
2432 * generally sent as soon as possible and no unnecessary delays are
2433 * introduced, at the cost of more packets in the network. Expects an
2434 * integer boolean flag.
2435 */
2436static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2437 int optlen)
2438{
2439 int val;
2440
2441 if (optlen < sizeof(int))
2442 return -EINVAL;
2443 if (get_user(val, (int __user *)optval))
2444 return -EFAULT;
2445
2446 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2447 return 0;
2448}
2449
2450/*
2451 *
2452 * 7.1.1 SCTP_RTOINFO
2453 *
2454 * The protocol parameters used to initialize and bound retransmission
2455 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2456 * and modify these parameters.
2457 * All parameters are time values, in milliseconds. A value of 0, when
2458 * modifying the parameters, indicates that the current value should not
2459 * be changed.
2460 *
2461 */
2462static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2463 struct sctp_rtoinfo rtoinfo;
2464 struct sctp_association *asoc;
2465
2466 if (optlen != sizeof (struct sctp_rtoinfo))
2467 return -EINVAL;
2468
2469 if (copy_from_user(&rtoinfo, optval, optlen))
2470 return -EFAULT;
2471
2472 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2473
2474 /* Set the values to the specific association */
2475 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2476 return -EINVAL;
2477
2478 if (asoc) {
2479 if (rtoinfo.srto_initial != 0)
2480 asoc->rto_initial =
2481 msecs_to_jiffies(rtoinfo.srto_initial);
2482 if (rtoinfo.srto_max != 0)
2483 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2484 if (rtoinfo.srto_min != 0)
2485 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2486 } else {
2487 /* If there is no association or the association-id = 0
2488 * set the values to the endpoint.
2489 */
2490 struct sctp_sock *sp = sctp_sk(sk);
2491
2492 if (rtoinfo.srto_initial != 0)
2493 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2494 if (rtoinfo.srto_max != 0)
2495 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2496 if (rtoinfo.srto_min != 0)
2497 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2498 }
2499
2500 return 0;
2501}
2502
2503/*
2504 *
2505 * 7.1.2 SCTP_ASSOCINFO
2506 *
2507 * This option is used to tune the the maximum retransmission attempts
2508 * of the association.
2509 * Returns an error if the new association retransmission value is
2510 * greater than the sum of the retransmission value of the peer.
2511 * See [SCTP] for more information.
2512 *
2513 */
2514static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2515{
2516
2517 struct sctp_assocparams assocparams;
2518 struct sctp_association *asoc;
2519
2520 if (optlen != sizeof(struct sctp_assocparams))
2521 return -EINVAL;
2522 if (copy_from_user(&assocparams, optval, optlen))
2523 return -EFAULT;
2524
2525 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2526
2527 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2528 return -EINVAL;
2529
2530 /* Set the values to the specific association */
2531 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002532 if (assocparams.sasoc_asocmaxrxt != 0) {
2533 __u32 path_sum = 0;
2534 int paths = 0;
2535 struct list_head *pos;
2536 struct sctp_transport *peer_addr;
2537
2538 list_for_each(pos, &asoc->peer.transport_addr_list) {
2539 peer_addr = list_entry(pos,
2540 struct sctp_transport,
2541 transports);
2542 path_sum += peer_addr->pathmaxrxt;
2543 paths++;
2544 }
2545
2546 /* Only validate asocmaxrxt if we have more then
2547 * one path/transport. We do this because path
2548 * retransmissions are only counted when we have more
2549 * then one path.
2550 */
2551 if (paths > 1 &&
2552 assocparams.sasoc_asocmaxrxt > path_sum)
2553 return -EINVAL;
2554
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002556 }
2557
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558 if (assocparams.sasoc_cookie_life != 0) {
2559 asoc->cookie_life.tv_sec =
2560 assocparams.sasoc_cookie_life / 1000;
2561 asoc->cookie_life.tv_usec =
2562 (assocparams.sasoc_cookie_life % 1000)
2563 * 1000;
2564 }
2565 } else {
2566 /* Set the values to the endpoint */
2567 struct sctp_sock *sp = sctp_sk(sk);
2568
2569 if (assocparams.sasoc_asocmaxrxt != 0)
2570 sp->assocparams.sasoc_asocmaxrxt =
2571 assocparams.sasoc_asocmaxrxt;
2572 if (assocparams.sasoc_cookie_life != 0)
2573 sp->assocparams.sasoc_cookie_life =
2574 assocparams.sasoc_cookie_life;
2575 }
2576 return 0;
2577}
2578
2579/*
2580 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2581 *
2582 * This socket option is a boolean flag which turns on or off mapped V4
2583 * addresses. If this option is turned on and the socket is type
2584 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2585 * If this option is turned off, then no mapping will be done of V4
2586 * addresses and a user will receive both PF_INET6 and PF_INET type
2587 * addresses on the socket.
2588 */
2589static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2590{
2591 int val;
2592 struct sctp_sock *sp = sctp_sk(sk);
2593
2594 if (optlen < sizeof(int))
2595 return -EINVAL;
2596 if (get_user(val, (int __user *)optval))
2597 return -EFAULT;
2598 if (val)
2599 sp->v4mapped = 1;
2600 else
2601 sp->v4mapped = 0;
2602
2603 return 0;
2604}
2605
2606/*
2607 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2608 *
2609 * This socket option specifies the maximum size to put in any outgoing
2610 * SCTP chunk. If a message is larger than this size it will be
2611 * fragmented by SCTP into the specified size. Note that the underlying
2612 * SCTP implementation may fragment into smaller sized chunks when the
2613 * PMTU of the underlying association is smaller than the value set by
2614 * the user.
2615 */
2616static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2617{
2618 struct sctp_association *asoc;
2619 struct list_head *pos;
2620 struct sctp_sock *sp = sctp_sk(sk);
2621 int val;
2622
2623 if (optlen < sizeof(int))
2624 return -EINVAL;
2625 if (get_user(val, (int __user *)optval))
2626 return -EFAULT;
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002627 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002628 return -EINVAL;
2629 sp->user_frag = val;
2630
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002631 /* Update the frag_point of the existing associations. */
2632 list_for_each(pos, &(sp->ep->asocs)) {
2633 asoc = list_entry(pos, struct sctp_association, asocs);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002634 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 }
2636
2637 return 0;
2638}
2639
2640
2641/*
2642 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2643 *
2644 * Requests that the peer mark the enclosed address as the association
2645 * primary. The enclosed address must be one of the association's
2646 * locally bound addresses. The following structure is used to make a
2647 * set primary request:
2648 */
2649static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2650 int optlen)
2651{
2652 struct sctp_sock *sp;
2653 struct sctp_endpoint *ep;
2654 struct sctp_association *asoc = NULL;
2655 struct sctp_setpeerprim prim;
2656 struct sctp_chunk *chunk;
2657 int err;
2658
2659 sp = sctp_sk(sk);
2660 ep = sp->ep;
2661
2662 if (!sctp_addip_enable)
2663 return -EPERM;
2664
2665 if (optlen != sizeof(struct sctp_setpeerprim))
2666 return -EINVAL;
2667
2668 if (copy_from_user(&prim, optval, optlen))
2669 return -EFAULT;
2670
2671 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2672 if (!asoc)
2673 return -EINVAL;
2674
2675 if (!asoc->peer.asconf_capable)
2676 return -EPERM;
2677
2678 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2679 return -EPERM;
2680
2681 if (!sctp_state(asoc, ESTABLISHED))
2682 return -ENOTCONN;
2683
2684 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2685 return -EADDRNOTAVAIL;
2686
2687 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2688 chunk = sctp_make_asconf_set_prim(asoc,
2689 (union sctp_addr *)&prim.sspp_addr);
2690 if (!chunk)
2691 return -ENOMEM;
2692
2693 err = sctp_send_asconf(asoc, chunk);
2694
2695 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2696
2697 return err;
2698}
2699
2700static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2701 int optlen)
2702{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002703 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002704
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002705 if (optlen != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706 return -EINVAL;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002707 if (copy_from_user(&adaption, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708 return -EFAULT;
2709
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002710 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002711
2712 return 0;
2713}
2714
2715/* API 6.2 setsockopt(), getsockopt()
2716 *
2717 * Applications use setsockopt() and getsockopt() to set or retrieve
2718 * socket options. Socket options are used to change the default
2719 * behavior of sockets calls. They are described in Section 7.
2720 *
2721 * The syntax is:
2722 *
2723 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2724 * int __user *optlen);
2725 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2726 * int optlen);
2727 *
2728 * sd - the socket descript.
2729 * level - set to IPPROTO_SCTP for all SCTP options.
2730 * optname - the option name.
2731 * optval - the buffer to store the value of the option.
2732 * optlen - the size of the buffer.
2733 */
2734SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2735 char __user *optval, int optlen)
2736{
2737 int retval = 0;
2738
2739 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2740 sk, optname);
2741
2742 /* I can hardly begin to describe how wrong this is. This is
2743 * so broken as to be worse than useless. The API draft
2744 * REALLY is NOT helpful here... I am not convinced that the
2745 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2746 * are at all well-founded.
2747 */
2748 if (level != SOL_SCTP) {
2749 struct sctp_af *af = sctp_sk(sk)->pf->af;
2750 retval = af->setsockopt(sk, level, optname, optval, optlen);
2751 goto out_nounlock;
2752 }
2753
2754 sctp_lock_sock(sk);
2755
2756 switch (optname) {
2757 case SCTP_SOCKOPT_BINDX_ADD:
2758 /* 'optlen' is the size of the addresses buffer. */
2759 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2760 optlen, SCTP_BINDX_ADD_ADDR);
2761 break;
2762
2763 case SCTP_SOCKOPT_BINDX_REM:
2764 /* 'optlen' is the size of the addresses buffer. */
2765 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2766 optlen, SCTP_BINDX_REM_ADDR);
2767 break;
2768
Frank Filz3f7a87d2005-06-20 13:14:57 -07002769 case SCTP_SOCKOPT_CONNECTX:
2770 /* 'optlen' is the size of the addresses buffer. */
2771 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2772 optlen);
2773 break;
2774
Linus Torvalds1da177e2005-04-16 15:20:36 -07002775 case SCTP_DISABLE_FRAGMENTS:
2776 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2777 break;
2778
2779 case SCTP_EVENTS:
2780 retval = sctp_setsockopt_events(sk, optval, optlen);
2781 break;
2782
2783 case SCTP_AUTOCLOSE:
2784 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2785 break;
2786
2787 case SCTP_PEER_ADDR_PARAMS:
2788 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2789 break;
2790
Frank Filz77086102005-12-22 11:37:30 -08002791 case SCTP_DELAYED_ACK_TIME:
2792 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2793 break;
2794
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795 case SCTP_INITMSG:
2796 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2797 break;
2798 case SCTP_DEFAULT_SEND_PARAM:
2799 retval = sctp_setsockopt_default_send_param(sk, optval,
2800 optlen);
2801 break;
2802 case SCTP_PRIMARY_ADDR:
2803 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2804 break;
2805 case SCTP_SET_PEER_PRIMARY_ADDR:
2806 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2807 break;
2808 case SCTP_NODELAY:
2809 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2810 break;
2811 case SCTP_RTOINFO:
2812 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2813 break;
2814 case SCTP_ASSOCINFO:
2815 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2816 break;
2817 case SCTP_I_WANT_MAPPED_V4_ADDR:
2818 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2819 break;
2820 case SCTP_MAXSEG:
2821 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2822 break;
2823 case SCTP_ADAPTION_LAYER:
2824 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2825 break;
2826
2827 default:
2828 retval = -ENOPROTOOPT;
2829 break;
2830 };
2831
2832 sctp_release_sock(sk);
2833
2834out_nounlock:
2835 return retval;
2836}
2837
2838/* API 3.1.6 connect() - UDP Style Syntax
2839 *
2840 * An application may use the connect() call in the UDP model to initiate an
2841 * association without sending data.
2842 *
2843 * The syntax is:
2844 *
2845 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2846 *
2847 * sd: the socket descriptor to have a new association added to.
2848 *
2849 * nam: the address structure (either struct sockaddr_in or struct
2850 * sockaddr_in6 defined in RFC2553 [7]).
2851 *
2852 * len: the size of the address.
2853 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07002854SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855 int addr_len)
2856{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07002858 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859
2860 sctp_lock_sock(sk);
2861
Frank Filz3f7a87d2005-06-20 13:14:57 -07002862 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2863 __FUNCTION__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002864
Frank Filz3f7a87d2005-06-20 13:14:57 -07002865 /* Validate addr_len before calling common connect/connectx routine. */
2866 af = sctp_get_af_specific(addr->sa_family);
2867 if (!af || addr_len < af->sockaddr_len) {
2868 err = -EINVAL;
2869 } else {
2870 /* Pass correct addr len to common routine (so it knows there
2871 * is only one address being passed.
2872 */
2873 err = __sctp_connect(sk, addr, af->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874 }
2875
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877 return err;
2878}
2879
2880/* FIXME: Write comments. */
2881SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2882{
2883 return -EOPNOTSUPP; /* STUB */
2884}
2885
2886/* 4.1.4 accept() - TCP Style Syntax
2887 *
2888 * Applications use accept() call to remove an established SCTP
2889 * association from the accept queue of the endpoint. A new socket
2890 * descriptor will be returned from accept() to represent the newly
2891 * formed association.
2892 */
2893SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2894{
2895 struct sctp_sock *sp;
2896 struct sctp_endpoint *ep;
2897 struct sock *newsk = NULL;
2898 struct sctp_association *asoc;
2899 long timeo;
2900 int error = 0;
2901
2902 sctp_lock_sock(sk);
2903
2904 sp = sctp_sk(sk);
2905 ep = sp->ep;
2906
2907 if (!sctp_style(sk, TCP)) {
2908 error = -EOPNOTSUPP;
2909 goto out;
2910 }
2911
2912 if (!sctp_sstate(sk, LISTENING)) {
2913 error = -EINVAL;
2914 goto out;
2915 }
2916
2917 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2918
2919 error = sctp_wait_for_accept(sk, timeo);
2920 if (error)
2921 goto out;
2922
2923 /* We treat the list of associations on the endpoint as the accept
2924 * queue and pick the first association on the list.
2925 */
2926 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2927
2928 newsk = sp->pf->create_accept_sk(sk, asoc);
2929 if (!newsk) {
2930 error = -ENOMEM;
2931 goto out;
2932 }
2933
2934 /* Populate the fields of the newsk from the oldsk and migrate the
2935 * asoc to the newsk.
2936 */
2937 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2938
2939out:
2940 sctp_release_sock(sk);
2941 *err = error;
2942 return newsk;
2943}
2944
2945/* The SCTP ioctl handler. */
2946SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2947{
2948 return -ENOIOCTLCMD;
2949}
2950
2951/* This is the function which gets called during socket creation to
2952 * initialized the SCTP-specific portion of the sock.
2953 * The sock structure should already be zero-filled memory.
2954 */
2955SCTP_STATIC int sctp_init_sock(struct sock *sk)
2956{
2957 struct sctp_endpoint *ep;
2958 struct sctp_sock *sp;
2959
2960 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
2961
2962 sp = sctp_sk(sk);
2963
2964 /* Initialize the SCTP per socket area. */
2965 switch (sk->sk_type) {
2966 case SOCK_SEQPACKET:
2967 sp->type = SCTP_SOCKET_UDP;
2968 break;
2969 case SOCK_STREAM:
2970 sp->type = SCTP_SOCKET_TCP;
2971 break;
2972 default:
2973 return -ESOCKTNOSUPPORT;
2974 }
2975
2976 /* Initialize default send parameters. These parameters can be
2977 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2978 */
2979 sp->default_stream = 0;
2980 sp->default_ppid = 0;
2981 sp->default_flags = 0;
2982 sp->default_context = 0;
2983 sp->default_timetolive = 0;
2984
2985 /* Initialize default setup parameters. These parameters
2986 * can be modified with the SCTP_INITMSG socket option or
2987 * overridden by the SCTP_INIT CMSG.
2988 */
2989 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
2990 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
2991 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
2992 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
2993
2994 /* Initialize default RTO related parameters. These parameters can
2995 * be modified for with the SCTP_RTOINFO socket option.
2996 */
2997 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
2998 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
2999 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
3000
3001 /* Initialize default association related parameters. These parameters
3002 * can be modified with the SCTP_ASSOCINFO socket option.
3003 */
3004 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3005 sp->assocparams.sasoc_number_peer_destinations = 0;
3006 sp->assocparams.sasoc_peer_rwnd = 0;
3007 sp->assocparams.sasoc_local_rwnd = 0;
3008 sp->assocparams.sasoc_cookie_life =
3009 jiffies_to_msecs(sctp_valid_cookie_life);
3010
3011 /* Initialize default event subscriptions. By default, all the
3012 * options are off.
3013 */
3014 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3015
3016 /* Default Peer Address Parameters. These defaults can
3017 * be modified via SCTP_PEER_ADDR_PARAMS
3018 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08003019 sp->hbinterval = jiffies_to_msecs(sctp_hb_interval);
3020 sp->pathmaxrxt = sctp_max_retrans_path;
3021 sp->pathmtu = 0; // allow default discovery
Vlad Yasevich8116ffa2006-01-17 11:55:17 -08003022 sp->sackdelay = jiffies_to_msecs(sctp_sack_timeout);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003023 sp->param_flags = SPP_HB_ENABLE |
3024 SPP_PMTUD_ENABLE |
3025 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003026
3027 /* If enabled no SCTP message fragmentation will be performed.
3028 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3029 */
3030 sp->disable_fragments = 0;
3031
3032 /* Turn on/off any Nagle-like algorithm. */
3033 sp->nodelay = 1;
3034
3035 /* Enable by default. */
3036 sp->v4mapped = 1;
3037
3038 /* Auto-close idle associations after the configured
3039 * number of seconds. A value of 0 disables this
3040 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3041 * for UDP-style sockets only.
3042 */
3043 sp->autoclose = 0;
3044
3045 /* User specified fragmentation limit. */
3046 sp->user_frag = 0;
3047
3048 sp->adaption_ind = 0;
3049
3050 sp->pf = sctp_get_pf_specific(sk->sk_family);
3051
3052 /* Control variables for partial data delivery. */
3053 sp->pd_mode = 0;
3054 skb_queue_head_init(&sp->pd_lobby);
3055
3056 /* Create a per socket endpoint structure. Even if we
3057 * change the data structure relationships, this may still
3058 * be useful for storing pre-connect address information.
3059 */
3060 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3061 if (!ep)
3062 return -ENOMEM;
3063
3064 sp->ep = ep;
3065 sp->hmac = NULL;
3066
3067 SCTP_DBG_OBJCNT_INC(sock);
3068 return 0;
3069}
3070
3071/* Cleanup any SCTP per socket resources. */
3072SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3073{
3074 struct sctp_endpoint *ep;
3075
3076 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3077
3078 /* Release our hold on the endpoint. */
3079 ep = sctp_sk(sk)->ep;
3080 sctp_endpoint_free(ep);
3081
3082 return 0;
3083}
3084
3085/* API 4.1.7 shutdown() - TCP Style Syntax
3086 * int shutdown(int socket, int how);
3087 *
3088 * sd - the socket descriptor of the association to be closed.
3089 * how - Specifies the type of shutdown. The values are
3090 * as follows:
3091 * SHUT_RD
3092 * Disables further receive operations. No SCTP
3093 * protocol action is taken.
3094 * SHUT_WR
3095 * Disables further send operations, and initiates
3096 * the SCTP shutdown sequence.
3097 * SHUT_RDWR
3098 * Disables further send and receive operations
3099 * and initiates the SCTP shutdown sequence.
3100 */
3101SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3102{
3103 struct sctp_endpoint *ep;
3104 struct sctp_association *asoc;
3105
3106 if (!sctp_style(sk, TCP))
3107 return;
3108
3109 if (how & SEND_SHUTDOWN) {
3110 ep = sctp_sk(sk)->ep;
3111 if (!list_empty(&ep->asocs)) {
3112 asoc = list_entry(ep->asocs.next,
3113 struct sctp_association, asocs);
3114 sctp_primitive_SHUTDOWN(asoc, NULL);
3115 }
3116 }
3117}
3118
3119/* 7.2.1 Association Status (SCTP_STATUS)
3120
3121 * Applications can retrieve current status information about an
3122 * association, including association state, peer receiver window size,
3123 * number of unacked data chunks, and number of data chunks pending
3124 * receipt. This information is read-only.
3125 */
3126static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3127 char __user *optval,
3128 int __user *optlen)
3129{
3130 struct sctp_status status;
3131 struct sctp_association *asoc = NULL;
3132 struct sctp_transport *transport;
3133 sctp_assoc_t associd;
3134 int retval = 0;
3135
3136 if (len != sizeof(status)) {
3137 retval = -EINVAL;
3138 goto out;
3139 }
3140
3141 if (copy_from_user(&status, optval, sizeof(status))) {
3142 retval = -EFAULT;
3143 goto out;
3144 }
3145
3146 associd = status.sstat_assoc_id;
3147 asoc = sctp_id2assoc(sk, associd);
3148 if (!asoc) {
3149 retval = -EINVAL;
3150 goto out;
3151 }
3152
3153 transport = asoc->peer.primary_path;
3154
3155 status.sstat_assoc_id = sctp_assoc2id(asoc);
3156 status.sstat_state = asoc->state;
3157 status.sstat_rwnd = asoc->peer.rwnd;
3158 status.sstat_unackdata = asoc->unack_data;
3159
3160 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3161 status.sstat_instrms = asoc->c.sinit_max_instreams;
3162 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3163 status.sstat_fragmentation_point = asoc->frag_point;
3164 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3165 memcpy(&status.sstat_primary.spinfo_address,
3166 &(transport->ipaddr), sizeof(union sctp_addr));
3167 /* Map ipv4 address into v4-mapped-on-v6 address. */
3168 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3169 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003170 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003171 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3172 status.sstat_primary.spinfo_srtt = transport->srtt;
3173 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003174 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003175
Frank Filz3f7a87d2005-06-20 13:14:57 -07003176 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3177 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3178
Linus Torvalds1da177e2005-04-16 15:20:36 -07003179 if (put_user(len, optlen)) {
3180 retval = -EFAULT;
3181 goto out;
3182 }
3183
3184 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3185 len, status.sstat_state, status.sstat_rwnd,
3186 status.sstat_assoc_id);
3187
3188 if (copy_to_user(optval, &status, len)) {
3189 retval = -EFAULT;
3190 goto out;
3191 }
3192
3193out:
3194 return (retval);
3195}
3196
3197
3198/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3199 *
3200 * Applications can retrieve information about a specific peer address
3201 * of an association, including its reachability state, congestion
3202 * window, and retransmission timer values. This information is
3203 * read-only.
3204 */
3205static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3206 char __user *optval,
3207 int __user *optlen)
3208{
3209 struct sctp_paddrinfo pinfo;
3210 struct sctp_transport *transport;
3211 int retval = 0;
3212
3213 if (len != sizeof(pinfo)) {
3214 retval = -EINVAL;
3215 goto out;
3216 }
3217
3218 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3219 retval = -EFAULT;
3220 goto out;
3221 }
3222
3223 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3224 pinfo.spinfo_assoc_id);
3225 if (!transport)
3226 return -EINVAL;
3227
3228 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003229 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003230 pinfo.spinfo_cwnd = transport->cwnd;
3231 pinfo.spinfo_srtt = transport->srtt;
3232 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003233 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003234
Frank Filz3f7a87d2005-06-20 13:14:57 -07003235 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3236 pinfo.spinfo_state = SCTP_ACTIVE;
3237
Linus Torvalds1da177e2005-04-16 15:20:36 -07003238 if (put_user(len, optlen)) {
3239 retval = -EFAULT;
3240 goto out;
3241 }
3242
3243 if (copy_to_user(optval, &pinfo, len)) {
3244 retval = -EFAULT;
3245 goto out;
3246 }
3247
3248out:
3249 return (retval);
3250}
3251
3252/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3253 *
3254 * This option is a on/off flag. If enabled no SCTP message
3255 * fragmentation will be performed. Instead if a message being sent
3256 * exceeds the current PMTU size, the message will NOT be sent and
3257 * instead a error will be indicated to the user.
3258 */
3259static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3260 char __user *optval, int __user *optlen)
3261{
3262 int val;
3263
3264 if (len < sizeof(int))
3265 return -EINVAL;
3266
3267 len = sizeof(int);
3268 val = (sctp_sk(sk)->disable_fragments == 1);
3269 if (put_user(len, optlen))
3270 return -EFAULT;
3271 if (copy_to_user(optval, &val, len))
3272 return -EFAULT;
3273 return 0;
3274}
3275
3276/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3277 *
3278 * This socket option is used to specify various notifications and
3279 * ancillary data the user wishes to receive.
3280 */
3281static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3282 int __user *optlen)
3283{
3284 if (len != sizeof(struct sctp_event_subscribe))
3285 return -EINVAL;
3286 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3287 return -EFAULT;
3288 return 0;
3289}
3290
3291/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3292 *
3293 * This socket option is applicable to the UDP-style socket only. When
3294 * set it will cause associations that are idle for more than the
3295 * specified number of seconds to automatically close. An association
3296 * being idle is defined an association that has NOT sent or received
3297 * user data. The special value of '0' indicates that no automatic
3298 * close of any associations should be performed. The option expects an
3299 * integer defining the number of seconds of idle time before an
3300 * association is closed.
3301 */
3302static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3303{
3304 /* Applicable to UDP-style socket only */
3305 if (sctp_style(sk, TCP))
3306 return -EOPNOTSUPP;
3307 if (len != sizeof(int))
3308 return -EINVAL;
3309 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3310 return -EFAULT;
3311 return 0;
3312}
3313
3314/* Helper routine to branch off an association to a new socket. */
3315SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3316 struct socket **sockp)
3317{
3318 struct sock *sk = asoc->base.sk;
3319 struct socket *sock;
3320 int err = 0;
3321
3322 /* An association cannot be branched off from an already peeled-off
3323 * socket, nor is this supported for tcp style sockets.
3324 */
3325 if (!sctp_style(sk, UDP))
3326 return -EINVAL;
3327
3328 /* Create a new socket. */
3329 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3330 if (err < 0)
3331 return err;
3332
3333 /* Populate the fields of the newsk from the oldsk and migrate the
3334 * asoc to the newsk.
3335 */
3336 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3337 *sockp = sock;
3338
3339 return err;
3340}
3341
3342static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3343{
3344 sctp_peeloff_arg_t peeloff;
3345 struct socket *newsock;
3346 int retval = 0;
3347 struct sctp_association *asoc;
3348
3349 if (len != sizeof(sctp_peeloff_arg_t))
3350 return -EINVAL;
3351 if (copy_from_user(&peeloff, optval, len))
3352 return -EFAULT;
3353
3354 asoc = sctp_id2assoc(sk, peeloff.associd);
3355 if (!asoc) {
3356 retval = -EINVAL;
3357 goto out;
3358 }
3359
3360 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3361
3362 retval = sctp_do_peeloff(asoc, &newsock);
3363 if (retval < 0)
3364 goto out;
3365
3366 /* Map the socket to an unused fd that can be returned to the user. */
3367 retval = sock_map_fd(newsock);
3368 if (retval < 0) {
3369 sock_release(newsock);
3370 goto out;
3371 }
3372
3373 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3374 __FUNCTION__, sk, asoc, newsock->sk, retval);
3375
3376 /* Return the fd mapped to the new socket. */
3377 peeloff.sd = retval;
3378 if (copy_to_user(optval, &peeloff, len))
3379 retval = -EFAULT;
3380
3381out:
3382 return retval;
3383}
3384
3385/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3386 *
3387 * Applications can enable or disable heartbeats for any peer address of
3388 * an association, modify an address's heartbeat interval, force a
3389 * heartbeat to be sent immediately, and adjust the address's maximum
3390 * number of retransmissions sent before an address is considered
3391 * unreachable. The following structure is used to access and modify an
3392 * address's parameters:
3393 *
3394 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08003395 * sctp_assoc_t spp_assoc_id;
3396 * struct sockaddr_storage spp_address;
3397 * uint32_t spp_hbinterval;
3398 * uint16_t spp_pathmaxrxt;
3399 * uint32_t spp_pathmtu;
3400 * uint32_t spp_sackdelay;
3401 * uint32_t spp_flags;
3402 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07003403 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08003404 * spp_assoc_id - (one-to-many style socket) This is filled in the
3405 * application, and identifies the association for
3406 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003407 * spp_address - This specifies which address is of interest.
3408 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003409 * in milliseconds. If a value of zero
3410 * is present in this field then no changes are to
3411 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003412 * spp_pathmaxrxt - This contains the maximum number of
3413 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08003414 * considered unreachable. If a value of zero
3415 * is present in this field then no changes are to
3416 * be made to this parameter.
3417 * spp_pathmtu - When Path MTU discovery is disabled the value
3418 * specified here will be the "fixed" path mtu.
3419 * Note that if the spp_address field is empty
3420 * then all associations on this address will
3421 * have this fixed path mtu set upon them.
3422 *
3423 * spp_sackdelay - When delayed sack is enabled, this value specifies
3424 * the number of milliseconds that sacks will be delayed
3425 * for. This value will apply to all addresses of an
3426 * association if the spp_address field is empty. Note
3427 * also, that if delayed sack is enabled and this
3428 * value is set to 0, no change is made to the last
3429 * recorded delayed sack timer value.
3430 *
3431 * spp_flags - These flags are used to control various features
3432 * on an association. The flag field may contain
3433 * zero or more of the following options.
3434 *
3435 * SPP_HB_ENABLE - Enable heartbeats on the
3436 * specified address. Note that if the address
3437 * field is empty all addresses for the association
3438 * have heartbeats enabled upon them.
3439 *
3440 * SPP_HB_DISABLE - Disable heartbeats on the
3441 * speicifed address. Note that if the address
3442 * field is empty all addresses for the association
3443 * will have their heartbeats disabled. Note also
3444 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3445 * mutually exclusive, only one of these two should
3446 * be specified. Enabling both fields will have
3447 * undetermined results.
3448 *
3449 * SPP_HB_DEMAND - Request a user initiated heartbeat
3450 * to be made immediately.
3451 *
3452 * SPP_PMTUD_ENABLE - This field will enable PMTU
3453 * discovery upon the specified address. Note that
3454 * if the address feild is empty then all addresses
3455 * on the association are effected.
3456 *
3457 * SPP_PMTUD_DISABLE - This field will disable PMTU
3458 * discovery upon the specified address. Note that
3459 * if the address feild is empty then all addresses
3460 * on the association are effected. Not also that
3461 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3462 * exclusive. Enabling both will have undetermined
3463 * results.
3464 *
3465 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3466 * on delayed sack. The time specified in spp_sackdelay
3467 * is used to specify the sack delay for this address. Note
3468 * that if spp_address is empty then all addresses will
3469 * enable delayed sack and take on the sack delay
3470 * value specified in spp_sackdelay.
3471 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3472 * off delayed sack. If the spp_address field is blank then
3473 * delayed sack is disabled for the entire association. Note
3474 * also that this field is mutually exclusive to
3475 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3476 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003477 */
3478static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003479 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003480{
Frank Filz52ccb8e2005-12-22 11:36:46 -08003481 struct sctp_paddrparams params;
3482 struct sctp_transport *trans = NULL;
3483 struct sctp_association *asoc = NULL;
3484 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003485
3486 if (len != sizeof(struct sctp_paddrparams))
3487 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003488
Linus Torvalds1da177e2005-04-16 15:20:36 -07003489 if (copy_from_user(&params, optval, len))
3490 return -EFAULT;
3491
Frank Filz52ccb8e2005-12-22 11:36:46 -08003492 /* If an address other than INADDR_ANY is specified, and
3493 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003494 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08003495 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3496 trans = sctp_addr_id2transport(sk, &params.spp_address,
3497 params.spp_assoc_id);
3498 if (!trans) {
3499 SCTP_DEBUG_PRINTK("Failed no transport\n");
3500 return -EINVAL;
3501 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003502 }
3503
Frank Filz52ccb8e2005-12-22 11:36:46 -08003504 /* Get association, if assoc_id != 0 and the socket is a one
3505 * to many style socket, and an association was not found, then
3506 * the id was invalid.
3507 */
3508 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3509 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3510 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003511 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003512 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003513
Frank Filz52ccb8e2005-12-22 11:36:46 -08003514 if (trans) {
3515 /* Fetch transport values. */
3516 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3517 params.spp_pathmtu = trans->pathmtu;
3518 params.spp_pathmaxrxt = trans->pathmaxrxt;
3519 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003520
Frank Filz52ccb8e2005-12-22 11:36:46 -08003521 /*draft-11 doesn't say what to return in spp_flags*/
3522 params.spp_flags = trans->param_flags;
3523 } else if (asoc) {
3524 /* Fetch association values. */
3525 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3526 params.spp_pathmtu = asoc->pathmtu;
3527 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3528 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003529
Frank Filz52ccb8e2005-12-22 11:36:46 -08003530 /*draft-11 doesn't say what to return in spp_flags*/
3531 params.spp_flags = asoc->param_flags;
3532 } else {
3533 /* Fetch socket values. */
3534 params.spp_hbinterval = sp->hbinterval;
3535 params.spp_pathmtu = sp->pathmtu;
3536 params.spp_sackdelay = sp->sackdelay;
3537 params.spp_pathmaxrxt = sp->pathmaxrxt;
3538
3539 /*draft-11 doesn't say what to return in spp_flags*/
3540 params.spp_flags = sp->param_flags;
3541 }
3542
Linus Torvalds1da177e2005-04-16 15:20:36 -07003543 if (copy_to_user(optval, &params, len))
3544 return -EFAULT;
3545
3546 if (put_user(len, optlen))
3547 return -EFAULT;
3548
3549 return 0;
3550}
3551
Frank Filz77086102005-12-22 11:37:30 -08003552/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3553 *
3554 * This options will get or set the delayed ack timer. The time is set
3555 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3556 * endpoints default delayed ack timer value. If the assoc_id field is
3557 * non-zero, then the set or get effects the specified association.
3558 *
3559 * struct sctp_assoc_value {
3560 * sctp_assoc_t assoc_id;
3561 * uint32_t assoc_value;
3562 * };
3563 *
3564 * assoc_id - This parameter, indicates which association the
3565 * user is preforming an action upon. Note that if
3566 * this field's value is zero then the endpoints
3567 * default value is changed (effecting future
3568 * associations only).
3569 *
3570 * assoc_value - This parameter contains the number of milliseconds
3571 * that the user is requesting the delayed ACK timer
3572 * be set to. Note that this value is defined in
3573 * the standard to be between 200 and 500 milliseconds.
3574 *
3575 * Note: a value of zero will leave the value alone,
3576 * but disable SACK delay. A non-zero value will also
3577 * enable SACK delay.
3578 */
3579static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3580 char __user *optval,
3581 int __user *optlen)
3582{
3583 struct sctp_assoc_value params;
3584 struct sctp_association *asoc = NULL;
3585 struct sctp_sock *sp = sctp_sk(sk);
3586
3587 if (len != sizeof(struct sctp_assoc_value))
3588 return - EINVAL;
3589
3590 if (copy_from_user(&params, optval, len))
3591 return -EFAULT;
3592
3593 /* Get association, if assoc_id != 0 and the socket is a one
3594 * to many style socket, and an association was not found, then
3595 * the id was invalid.
3596 */
3597 asoc = sctp_id2assoc(sk, params.assoc_id);
3598 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3599 return -EINVAL;
3600
3601 if (asoc) {
3602 /* Fetch association values. */
3603 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3604 params.assoc_value = jiffies_to_msecs(
3605 asoc->sackdelay);
3606 else
3607 params.assoc_value = 0;
3608 } else {
3609 /* Fetch socket values. */
3610 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3611 params.assoc_value = sp->sackdelay;
3612 else
3613 params.assoc_value = 0;
3614 }
3615
3616 if (copy_to_user(optval, &params, len))
3617 return -EFAULT;
3618
3619 if (put_user(len, optlen))
3620 return -EFAULT;
3621
3622 return 0;
3623}
3624
Linus Torvalds1da177e2005-04-16 15:20:36 -07003625/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3626 *
3627 * Applications can specify protocol parameters for the default association
3628 * initialization. The option name argument to setsockopt() and getsockopt()
3629 * is SCTP_INITMSG.
3630 *
3631 * Setting initialization parameters is effective only on an unconnected
3632 * socket (for UDP-style sockets only future associations are effected
3633 * by the change). With TCP-style sockets, this option is inherited by
3634 * sockets derived from a listener socket.
3635 */
3636static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3637{
3638 if (len != sizeof(struct sctp_initmsg))
3639 return -EINVAL;
3640 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3641 return -EFAULT;
3642 return 0;
3643}
3644
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003645static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3646 char __user *optval,
3647 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003648{
3649 sctp_assoc_t id;
3650 struct sctp_association *asoc;
3651 struct list_head *pos;
3652 int cnt = 0;
3653
3654 if (len != sizeof(sctp_assoc_t))
3655 return -EINVAL;
3656
3657 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3658 return -EFAULT;
3659
3660 /* For UDP-style sockets, id specifies the association to query. */
3661 asoc = sctp_id2assoc(sk, id);
3662 if (!asoc)
3663 return -EINVAL;
3664
3665 list_for_each(pos, &asoc->peer.transport_addr_list) {
3666 cnt ++;
3667 }
3668
3669 return cnt;
3670}
3671
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003672/*
3673 * Old API for getting list of peer addresses. Does not work for 32-bit
3674 * programs running on a 64-bit kernel
3675 */
3676static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3677 char __user *optval,
3678 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003679{
3680 struct sctp_association *asoc;
3681 struct list_head *pos;
3682 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003683 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003684 struct sctp_transport *from;
3685 void __user *to;
3686 union sctp_addr temp;
3687 struct sctp_sock *sp = sctp_sk(sk);
3688 int addrlen;
3689
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003690 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003691 return -EINVAL;
3692
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003693 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003694 return -EFAULT;
3695
3696 if (getaddrs.addr_num <= 0) return -EINVAL;
3697
3698 /* For UDP-style sockets, id specifies the association to query. */
3699 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3700 if (!asoc)
3701 return -EINVAL;
3702
3703 to = (void __user *)getaddrs.addrs;
3704 list_for_each(pos, &asoc->peer.transport_addr_list) {
3705 from = list_entry(pos, struct sctp_transport, transports);
3706 memcpy(&temp, &from->ipaddr, sizeof(temp));
3707 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3708 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3709 temp.v4.sin_port = htons(temp.v4.sin_port);
3710 if (copy_to_user(to, &temp, addrlen))
3711 return -EFAULT;
3712 to += addrlen ;
3713 cnt ++;
3714 if (cnt >= getaddrs.addr_num) break;
3715 }
3716 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003717 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003718 return -EFAULT;
3719
3720 return 0;
3721}
3722
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003723static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3724 char __user *optval, int __user *optlen)
3725{
3726 struct sctp_association *asoc;
3727 struct list_head *pos;
3728 int cnt = 0;
3729 struct sctp_getaddrs getaddrs;
3730 struct sctp_transport *from;
3731 void __user *to;
3732 union sctp_addr temp;
3733 struct sctp_sock *sp = sctp_sk(sk);
3734 int addrlen;
3735 size_t space_left;
3736 int bytes_copied;
3737
3738 if (len < sizeof(struct sctp_getaddrs))
3739 return -EINVAL;
3740
3741 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3742 return -EFAULT;
3743
3744 /* For UDP-style sockets, id specifies the association to query. */
3745 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3746 if (!asoc)
3747 return -EINVAL;
3748
3749 to = optval + offsetof(struct sctp_getaddrs,addrs);
3750 space_left = len - sizeof(struct sctp_getaddrs) -
3751 offsetof(struct sctp_getaddrs,addrs);
3752
3753 list_for_each(pos, &asoc->peer.transport_addr_list) {
3754 from = list_entry(pos, struct sctp_transport, transports);
3755 memcpy(&temp, &from->ipaddr, sizeof(temp));
3756 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3757 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3758 if(space_left < addrlen)
3759 return -ENOMEM;
3760 temp.v4.sin_port = htons(temp.v4.sin_port);
3761 if (copy_to_user(to, &temp, addrlen))
3762 return -EFAULT;
3763 to += addrlen;
3764 cnt++;
3765 space_left -= addrlen;
3766 }
3767
3768 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3769 return -EFAULT;
3770 bytes_copied = ((char __user *)to) - optval;
3771 if (put_user(bytes_copied, optlen))
3772 return -EFAULT;
3773
3774 return 0;
3775}
3776
3777static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3778 char __user *optval,
3779 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003780{
3781 sctp_assoc_t id;
3782 struct sctp_bind_addr *bp;
3783 struct sctp_association *asoc;
3784 struct list_head *pos;
3785 struct sctp_sockaddr_entry *addr;
3786 rwlock_t *addr_lock;
3787 unsigned long flags;
3788 int cnt = 0;
3789
3790 if (len != sizeof(sctp_assoc_t))
3791 return -EINVAL;
3792
3793 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3794 return -EFAULT;
3795
3796 /*
3797 * For UDP-style sockets, id specifies the association to query.
3798 * If the id field is set to the value '0' then the locally bound
3799 * addresses are returned without regard to any particular
3800 * association.
3801 */
3802 if (0 == id) {
3803 bp = &sctp_sk(sk)->ep->base.bind_addr;
3804 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3805 } else {
3806 asoc = sctp_id2assoc(sk, id);
3807 if (!asoc)
3808 return -EINVAL;
3809 bp = &asoc->base.bind_addr;
3810 addr_lock = &asoc->base.addr_lock;
3811 }
3812
3813 sctp_read_lock(addr_lock);
3814
3815 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3816 * addresses from the global local address list.
3817 */
3818 if (sctp_list_single_entry(&bp->address_list)) {
3819 addr = list_entry(bp->address_list.next,
3820 struct sctp_sockaddr_entry, list);
3821 if (sctp_is_any(&addr->a)) {
3822 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3823 list_for_each(pos, &sctp_local_addr_list) {
3824 addr = list_entry(pos,
3825 struct sctp_sockaddr_entry,
3826 list);
3827 if ((PF_INET == sk->sk_family) &&
3828 (AF_INET6 == addr->a.sa.sa_family))
3829 continue;
3830 cnt++;
3831 }
3832 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3833 flags);
3834 } else {
3835 cnt = 1;
3836 }
3837 goto done;
3838 }
3839
3840 list_for_each(pos, &bp->address_list) {
3841 cnt ++;
3842 }
3843
3844done:
3845 sctp_read_unlock(addr_lock);
3846 return cnt;
3847}
3848
3849/* Helper function that copies local addresses to user and returns the number
3850 * of addresses copied.
3851 */
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003852static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3853 void __user *to)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003854{
3855 struct list_head *pos;
3856 struct sctp_sockaddr_entry *addr;
3857 unsigned long flags;
3858 union sctp_addr temp;
3859 int cnt = 0;
3860 int addrlen;
3861
3862 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3863 list_for_each(pos, &sctp_local_addr_list) {
3864 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3865 if ((PF_INET == sk->sk_family) &&
3866 (AF_INET6 == addr->a.sa.sa_family))
3867 continue;
3868 memcpy(&temp, &addr->a, sizeof(temp));
3869 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3870 &temp);
3871 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3872 temp.v4.sin_port = htons(port);
3873 if (copy_to_user(to, &temp, addrlen)) {
3874 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3875 flags);
3876 return -EFAULT;
3877 }
3878 to += addrlen;
3879 cnt ++;
3880 if (cnt >= max_addrs) break;
3881 }
3882 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3883
3884 return cnt;
3885}
3886
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003887static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
Al Virod3a880e2005-12-15 09:18:30 +00003888 void __user **to, size_t space_left)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003889{
3890 struct list_head *pos;
3891 struct sctp_sockaddr_entry *addr;
3892 unsigned long flags;
3893 union sctp_addr temp;
3894 int cnt = 0;
3895 int addrlen;
3896
3897 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3898 list_for_each(pos, &sctp_local_addr_list) {
3899 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3900 if ((PF_INET == sk->sk_family) &&
3901 (AF_INET6 == addr->a.sa.sa_family))
3902 continue;
3903 memcpy(&temp, &addr->a, sizeof(temp));
3904 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3905 &temp);
3906 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3907 if(space_left<addrlen)
3908 return -ENOMEM;
3909 temp.v4.sin_port = htons(port);
3910 if (copy_to_user(*to, &temp, addrlen)) {
3911 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3912 flags);
3913 return -EFAULT;
3914 }
3915 *to += addrlen;
3916 cnt ++;
3917 space_left -= addrlen;
3918 }
3919 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3920
3921 return cnt;
3922}
3923
3924/* Old API for getting list of local addresses. Does not work for 32-bit
3925 * programs running on a 64-bit kernel
3926 */
3927static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3928 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003929{
3930 struct sctp_bind_addr *bp;
3931 struct sctp_association *asoc;
3932 struct list_head *pos;
3933 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003934 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003935 struct sctp_sockaddr_entry *addr;
3936 void __user *to;
3937 union sctp_addr temp;
3938 struct sctp_sock *sp = sctp_sk(sk);
3939 int addrlen;
3940 rwlock_t *addr_lock;
3941 int err = 0;
3942
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003943 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003944 return -EINVAL;
3945
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003946 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003947 return -EFAULT;
3948
3949 if (getaddrs.addr_num <= 0) return -EINVAL;
3950 /*
3951 * For UDP-style sockets, id specifies the association to query.
3952 * If the id field is set to the value '0' then the locally bound
3953 * addresses are returned without regard to any particular
3954 * association.
3955 */
3956 if (0 == getaddrs.assoc_id) {
3957 bp = &sctp_sk(sk)->ep->base.bind_addr;
3958 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3959 } else {
3960 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3961 if (!asoc)
3962 return -EINVAL;
3963 bp = &asoc->base.bind_addr;
3964 addr_lock = &asoc->base.addr_lock;
3965 }
3966
3967 to = getaddrs.addrs;
3968
3969 sctp_read_lock(addr_lock);
3970
3971 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3972 * addresses from the global local address list.
3973 */
3974 if (sctp_list_single_entry(&bp->address_list)) {
3975 addr = list_entry(bp->address_list.next,
3976 struct sctp_sockaddr_entry, list);
3977 if (sctp_is_any(&addr->a)) {
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003978 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
3979 getaddrs.addr_num,
3980 to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003981 if (cnt < 0) {
3982 err = cnt;
3983 goto unlock;
3984 }
3985 goto copy_getaddrs;
3986 }
3987 }
3988
3989 list_for_each(pos, &bp->address_list) {
3990 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3991 memcpy(&temp, &addr->a, sizeof(temp));
3992 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3993 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3994 temp.v4.sin_port = htons(temp.v4.sin_port);
3995 if (copy_to_user(to, &temp, addrlen)) {
3996 err = -EFAULT;
3997 goto unlock;
3998 }
3999 to += addrlen;
4000 cnt ++;
4001 if (cnt >= getaddrs.addr_num) break;
4002 }
4003
4004copy_getaddrs:
4005 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004006 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004007 err = -EFAULT;
4008
4009unlock:
4010 sctp_read_unlock(addr_lock);
4011 return err;
4012}
4013
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004014static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4015 char __user *optval, int __user *optlen)
4016{
4017 struct sctp_bind_addr *bp;
4018 struct sctp_association *asoc;
4019 struct list_head *pos;
4020 int cnt = 0;
4021 struct sctp_getaddrs getaddrs;
4022 struct sctp_sockaddr_entry *addr;
4023 void __user *to;
4024 union sctp_addr temp;
4025 struct sctp_sock *sp = sctp_sk(sk);
4026 int addrlen;
4027 rwlock_t *addr_lock;
4028 int err = 0;
4029 size_t space_left;
4030 int bytes_copied;
4031
4032 if (len <= sizeof(struct sctp_getaddrs))
4033 return -EINVAL;
4034
4035 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4036 return -EFAULT;
4037
4038 /*
4039 * For UDP-style sockets, id specifies the association to query.
4040 * If the id field is set to the value '0' then the locally bound
4041 * addresses are returned without regard to any particular
4042 * association.
4043 */
4044 if (0 == getaddrs.assoc_id) {
4045 bp = &sctp_sk(sk)->ep->base.bind_addr;
4046 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4047 } else {
4048 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4049 if (!asoc)
4050 return -EINVAL;
4051 bp = &asoc->base.bind_addr;
4052 addr_lock = &asoc->base.addr_lock;
4053 }
4054
4055 to = optval + offsetof(struct sctp_getaddrs,addrs);
4056 space_left = len - sizeof(struct sctp_getaddrs) -
4057 offsetof(struct sctp_getaddrs,addrs);
4058
4059 sctp_read_lock(addr_lock);
4060
4061 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4062 * addresses from the global local address list.
4063 */
4064 if (sctp_list_single_entry(&bp->address_list)) {
4065 addr = list_entry(bp->address_list.next,
4066 struct sctp_sockaddr_entry, list);
4067 if (sctp_is_any(&addr->a)) {
4068 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4069 &to, space_left);
4070 if (cnt < 0) {
4071 err = cnt;
4072 goto unlock;
4073 }
4074 goto copy_getaddrs;
4075 }
4076 }
4077
4078 list_for_each(pos, &bp->address_list) {
4079 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4080 memcpy(&temp, &addr->a, sizeof(temp));
4081 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4082 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4083 if(space_left < addrlen)
4084 return -ENOMEM; /*fixme: right error?*/
4085 temp.v4.sin_port = htons(temp.v4.sin_port);
4086 if (copy_to_user(to, &temp, addrlen)) {
4087 err = -EFAULT;
4088 goto unlock;
4089 }
4090 to += addrlen;
4091 cnt ++;
4092 space_left -= addrlen;
4093 }
4094
4095copy_getaddrs:
4096 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4097 return -EFAULT;
4098 bytes_copied = ((char __user *)to) - optval;
4099 if (put_user(bytes_copied, optlen))
4100 return -EFAULT;
4101
4102unlock:
4103 sctp_read_unlock(addr_lock);
4104 return err;
4105}
4106
Linus Torvalds1da177e2005-04-16 15:20:36 -07004107/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4108 *
4109 * Requests that the local SCTP stack use the enclosed peer address as
4110 * the association primary. The enclosed address must be one of the
4111 * association peer's addresses.
4112 */
4113static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4114 char __user *optval, int __user *optlen)
4115{
4116 struct sctp_prim prim;
4117 struct sctp_association *asoc;
4118 struct sctp_sock *sp = sctp_sk(sk);
4119
4120 if (len != sizeof(struct sctp_prim))
4121 return -EINVAL;
4122
4123 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4124 return -EFAULT;
4125
4126 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4127 if (!asoc)
4128 return -EINVAL;
4129
4130 if (!asoc->peer.primary_path)
4131 return -ENOTCONN;
4132
4133 asoc->peer.primary_path->ipaddr.v4.sin_port =
4134 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
4135 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4136 sizeof(union sctp_addr));
4137 asoc->peer.primary_path->ipaddr.v4.sin_port =
4138 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
4139
4140 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4141 (union sctp_addr *)&prim.ssp_addr);
4142
4143 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4144 return -EFAULT;
4145
4146 return 0;
4147}
4148
4149/*
4150 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4151 *
4152 * Requests that the local endpoint set the specified Adaption Layer
4153 * Indication parameter for all future INIT and INIT-ACK exchanges.
4154 */
4155static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4156 char __user *optval, int __user *optlen)
4157{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004158 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004159
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004160 if (len != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004161 return -EINVAL;
4162
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004163 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4164 if (copy_to_user(optval, &adaption, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004165 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004166
Linus Torvalds1da177e2005-04-16 15:20:36 -07004167 return 0;
4168}
4169
4170/*
4171 *
4172 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4173 *
4174 * Applications that wish to use the sendto() system call may wish to
4175 * specify a default set of parameters that would normally be supplied
4176 * through the inclusion of ancillary data. This socket option allows
4177 * such an application to set the default sctp_sndrcvinfo structure.
4178
4179
4180 * The application that wishes to use this socket option simply passes
4181 * in to this call the sctp_sndrcvinfo structure defined in Section
4182 * 5.2.2) The input parameters accepted by this call include
4183 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4184 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4185 * to this call if the caller is using the UDP model.
4186 *
4187 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4188 */
4189static int sctp_getsockopt_default_send_param(struct sock *sk,
4190 int len, char __user *optval,
4191 int __user *optlen)
4192{
4193 struct sctp_sndrcvinfo info;
4194 struct sctp_association *asoc;
4195 struct sctp_sock *sp = sctp_sk(sk);
4196
4197 if (len != sizeof(struct sctp_sndrcvinfo))
4198 return -EINVAL;
4199 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4200 return -EFAULT;
4201
4202 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4203 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4204 return -EINVAL;
4205
4206 if (asoc) {
4207 info.sinfo_stream = asoc->default_stream;
4208 info.sinfo_flags = asoc->default_flags;
4209 info.sinfo_ppid = asoc->default_ppid;
4210 info.sinfo_context = asoc->default_context;
4211 info.sinfo_timetolive = asoc->default_timetolive;
4212 } else {
4213 info.sinfo_stream = sp->default_stream;
4214 info.sinfo_flags = sp->default_flags;
4215 info.sinfo_ppid = sp->default_ppid;
4216 info.sinfo_context = sp->default_context;
4217 info.sinfo_timetolive = sp->default_timetolive;
4218 }
4219
4220 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4221 return -EFAULT;
4222
4223 return 0;
4224}
4225
4226/*
4227 *
4228 * 7.1.5 SCTP_NODELAY
4229 *
4230 * Turn on/off any Nagle-like algorithm. This means that packets are
4231 * generally sent as soon as possible and no unnecessary delays are
4232 * introduced, at the cost of more packets in the network. Expects an
4233 * integer boolean flag.
4234 */
4235
4236static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4237 char __user *optval, int __user *optlen)
4238{
4239 int val;
4240
4241 if (len < sizeof(int))
4242 return -EINVAL;
4243
4244 len = sizeof(int);
4245 val = (sctp_sk(sk)->nodelay == 1);
4246 if (put_user(len, optlen))
4247 return -EFAULT;
4248 if (copy_to_user(optval, &val, len))
4249 return -EFAULT;
4250 return 0;
4251}
4252
4253/*
4254 *
4255 * 7.1.1 SCTP_RTOINFO
4256 *
4257 * The protocol parameters used to initialize and bound retransmission
4258 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4259 * and modify these parameters.
4260 * All parameters are time values, in milliseconds. A value of 0, when
4261 * modifying the parameters, indicates that the current value should not
4262 * be changed.
4263 *
4264 */
4265static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4266 char __user *optval,
4267 int __user *optlen) {
4268 struct sctp_rtoinfo rtoinfo;
4269 struct sctp_association *asoc;
4270
4271 if (len != sizeof (struct sctp_rtoinfo))
4272 return -EINVAL;
4273
4274 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4275 return -EFAULT;
4276
4277 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4278
4279 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4280 return -EINVAL;
4281
4282 /* Values corresponding to the specific association. */
4283 if (asoc) {
4284 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4285 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4286 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4287 } else {
4288 /* Values corresponding to the endpoint. */
4289 struct sctp_sock *sp = sctp_sk(sk);
4290
4291 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4292 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4293 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4294 }
4295
4296 if (put_user(len, optlen))
4297 return -EFAULT;
4298
4299 if (copy_to_user(optval, &rtoinfo, len))
4300 return -EFAULT;
4301
4302 return 0;
4303}
4304
4305/*
4306 *
4307 * 7.1.2 SCTP_ASSOCINFO
4308 *
4309 * This option is used to tune the the maximum retransmission attempts
4310 * of the association.
4311 * Returns an error if the new association retransmission value is
4312 * greater than the sum of the retransmission value of the peer.
4313 * See [SCTP] for more information.
4314 *
4315 */
4316static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4317 char __user *optval,
4318 int __user *optlen)
4319{
4320
4321 struct sctp_assocparams assocparams;
4322 struct sctp_association *asoc;
4323 struct list_head *pos;
4324 int cnt = 0;
4325
4326 if (len != sizeof (struct sctp_assocparams))
4327 return -EINVAL;
4328
4329 if (copy_from_user(&assocparams, optval,
4330 sizeof (struct sctp_assocparams)))
4331 return -EFAULT;
4332
4333 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4334
4335 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4336 return -EINVAL;
4337
4338 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07004339 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004340 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4341 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4342 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4343 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4344 * 1000) +
4345 (asoc->cookie_life.tv_usec
4346 / 1000);
4347
4348 list_for_each(pos, &asoc->peer.transport_addr_list) {
4349 cnt ++;
4350 }
4351
4352 assocparams.sasoc_number_peer_destinations = cnt;
4353 } else {
4354 /* Values corresponding to the endpoint */
4355 struct sctp_sock *sp = sctp_sk(sk);
4356
4357 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4358 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4359 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4360 assocparams.sasoc_cookie_life =
4361 sp->assocparams.sasoc_cookie_life;
4362 assocparams.sasoc_number_peer_destinations =
4363 sp->assocparams.
4364 sasoc_number_peer_destinations;
4365 }
4366
4367 if (put_user(len, optlen))
4368 return -EFAULT;
4369
4370 if (copy_to_user(optval, &assocparams, len))
4371 return -EFAULT;
4372
4373 return 0;
4374}
4375
4376/*
4377 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4378 *
4379 * This socket option is a boolean flag which turns on or off mapped V4
4380 * addresses. If this option is turned on and the socket is type
4381 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4382 * If this option is turned off, then no mapping will be done of V4
4383 * addresses and a user will receive both PF_INET6 and PF_INET type
4384 * addresses on the socket.
4385 */
4386static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4387 char __user *optval, int __user *optlen)
4388{
4389 int val;
4390 struct sctp_sock *sp = sctp_sk(sk);
4391
4392 if (len < sizeof(int))
4393 return -EINVAL;
4394
4395 len = sizeof(int);
4396 val = sp->v4mapped;
4397 if (put_user(len, optlen))
4398 return -EFAULT;
4399 if (copy_to_user(optval, &val, len))
4400 return -EFAULT;
4401
4402 return 0;
4403}
4404
4405/*
4406 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4407 *
4408 * This socket option specifies the maximum size to put in any outgoing
4409 * SCTP chunk. If a message is larger than this size it will be
4410 * fragmented by SCTP into the specified size. Note that the underlying
4411 * SCTP implementation may fragment into smaller sized chunks when the
4412 * PMTU of the underlying association is smaller than the value set by
4413 * the user.
4414 */
4415static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4416 char __user *optval, int __user *optlen)
4417{
4418 int val;
4419
4420 if (len < sizeof(int))
4421 return -EINVAL;
4422
4423 len = sizeof(int);
4424
4425 val = sctp_sk(sk)->user_frag;
4426 if (put_user(len, optlen))
4427 return -EFAULT;
4428 if (copy_to_user(optval, &val, len))
4429 return -EFAULT;
4430
4431 return 0;
4432}
4433
4434SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4435 char __user *optval, int __user *optlen)
4436{
4437 int retval = 0;
4438 int len;
4439
Frank Filz3f7a87d2005-06-20 13:14:57 -07004440 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4441 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004442
4443 /* I can hardly begin to describe how wrong this is. This is
4444 * so broken as to be worse than useless. The API draft
4445 * REALLY is NOT helpful here... I am not convinced that the
4446 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4447 * are at all well-founded.
4448 */
4449 if (level != SOL_SCTP) {
4450 struct sctp_af *af = sctp_sk(sk)->pf->af;
4451
4452 retval = af->getsockopt(sk, level, optname, optval, optlen);
4453 return retval;
4454 }
4455
4456 if (get_user(len, optlen))
4457 return -EFAULT;
4458
4459 sctp_lock_sock(sk);
4460
4461 switch (optname) {
4462 case SCTP_STATUS:
4463 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4464 break;
4465 case SCTP_DISABLE_FRAGMENTS:
4466 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4467 optlen);
4468 break;
4469 case SCTP_EVENTS:
4470 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4471 break;
4472 case SCTP_AUTOCLOSE:
4473 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4474 break;
4475 case SCTP_SOCKOPT_PEELOFF:
4476 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4477 break;
4478 case SCTP_PEER_ADDR_PARAMS:
4479 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4480 optlen);
4481 break;
Frank Filz77086102005-12-22 11:37:30 -08004482 case SCTP_DELAYED_ACK_TIME:
4483 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4484 optlen);
4485 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004486 case SCTP_INITMSG:
4487 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4488 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004489 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4490 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4491 optlen);
4492 break;
4493 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4494 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4495 optlen);
4496 break;
4497 case SCTP_GET_PEER_ADDRS_OLD:
4498 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004499 optlen);
4500 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004501 case SCTP_GET_LOCAL_ADDRS_OLD:
4502 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004503 optlen);
4504 break;
4505 case SCTP_GET_PEER_ADDRS:
4506 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4507 optlen);
4508 break;
4509 case SCTP_GET_LOCAL_ADDRS:
4510 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4511 optlen);
4512 break;
4513 case SCTP_DEFAULT_SEND_PARAM:
4514 retval = sctp_getsockopt_default_send_param(sk, len,
4515 optval, optlen);
4516 break;
4517 case SCTP_PRIMARY_ADDR:
4518 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4519 break;
4520 case SCTP_NODELAY:
4521 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4522 break;
4523 case SCTP_RTOINFO:
4524 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4525 break;
4526 case SCTP_ASSOCINFO:
4527 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4528 break;
4529 case SCTP_I_WANT_MAPPED_V4_ADDR:
4530 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4531 break;
4532 case SCTP_MAXSEG:
4533 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4534 break;
4535 case SCTP_GET_PEER_ADDR_INFO:
4536 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4537 optlen);
4538 break;
4539 case SCTP_ADAPTION_LAYER:
4540 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4541 optlen);
4542 break;
4543 default:
4544 retval = -ENOPROTOOPT;
4545 break;
4546 };
4547
4548 sctp_release_sock(sk);
4549 return retval;
4550}
4551
4552static void sctp_hash(struct sock *sk)
4553{
4554 /* STUB */
4555}
4556
4557static void sctp_unhash(struct sock *sk)
4558{
4559 /* STUB */
4560}
4561
4562/* Check if port is acceptable. Possibly find first available port.
4563 *
4564 * The port hash table (contained in the 'global' SCTP protocol storage
4565 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4566 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4567 * list (the list number is the port number hashed out, so as you
4568 * would expect from a hash function, all the ports in a given list have
4569 * such a number that hashes out to the same list number; you were
4570 * expecting that, right?); so each list has a set of ports, with a
4571 * link to the socket (struct sock) that uses it, the port number and
4572 * a fastreuse flag (FIXME: NPI ipg).
4573 */
4574static struct sctp_bind_bucket *sctp_bucket_create(
4575 struct sctp_bind_hashbucket *head, unsigned short snum);
4576
4577static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4578{
4579 struct sctp_bind_hashbucket *head; /* hash list */
4580 struct sctp_bind_bucket *pp; /* hash list port iterator */
4581 unsigned short snum;
4582 int ret;
4583
4584 /* NOTE: Remember to put this back to net order. */
4585 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4586 snum = addr->v4.sin_port;
4587
4588 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4589 sctp_local_bh_disable();
4590
4591 if (snum == 0) {
4592 /* Search for an available port.
4593 *
4594 * 'sctp_port_rover' was the last port assigned, so
4595 * we start to search from 'sctp_port_rover +
4596 * 1'. What we do is first check if port 'rover' is
4597 * already in the hash table; if not, we use that; if
4598 * it is, we try next.
4599 */
4600 int low = sysctl_local_port_range[0];
4601 int high = sysctl_local_port_range[1];
4602 int remaining = (high - low) + 1;
4603 int rover;
4604 int index;
4605
4606 sctp_spin_lock(&sctp_port_alloc_lock);
4607 rover = sctp_port_rover;
4608 do {
4609 rover++;
4610 if ((rover < low) || (rover > high))
4611 rover = low;
4612 index = sctp_phashfn(rover);
4613 head = &sctp_port_hashtable[index];
4614 sctp_spin_lock(&head->lock);
4615 for (pp = head->chain; pp; pp = pp->next)
4616 if (pp->port == rover)
4617 goto next;
4618 break;
4619 next:
4620 sctp_spin_unlock(&head->lock);
4621 } while (--remaining > 0);
4622 sctp_port_rover = rover;
4623 sctp_spin_unlock(&sctp_port_alloc_lock);
4624
4625 /* Exhausted local port range during search? */
4626 ret = 1;
4627 if (remaining <= 0)
4628 goto fail;
4629
4630 /* OK, here is the one we will use. HEAD (the port
4631 * hash table list entry) is non-NULL and we hold it's
4632 * mutex.
4633 */
4634 snum = rover;
4635 } else {
4636 /* We are given an specific port number; we verify
4637 * that it is not being used. If it is used, we will
4638 * exahust the search in the hash list corresponding
4639 * to the port number (snum) - we detect that with the
4640 * port iterator, pp being NULL.
4641 */
4642 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4643 sctp_spin_lock(&head->lock);
4644 for (pp = head->chain; pp; pp = pp->next) {
4645 if (pp->port == snum)
4646 goto pp_found;
4647 }
4648 }
4649 pp = NULL;
4650 goto pp_not_found;
4651pp_found:
4652 if (!hlist_empty(&pp->owner)) {
4653 /* We had a port hash table hit - there is an
4654 * available port (pp != NULL) and it is being
4655 * used by other socket (pp->owner not empty); that other
4656 * socket is going to be sk2.
4657 */
4658 int reuse = sk->sk_reuse;
4659 struct sock *sk2;
4660 struct hlist_node *node;
4661
4662 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4663 if (pp->fastreuse && sk->sk_reuse)
4664 goto success;
4665
4666 /* Run through the list of sockets bound to the port
4667 * (pp->port) [via the pointers bind_next and
4668 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4669 * we get the endpoint they describe and run through
4670 * the endpoint's list of IP (v4 or v6) addresses,
4671 * comparing each of the addresses with the address of
4672 * the socket sk. If we find a match, then that means
4673 * that this port/socket (sk) combination are already
4674 * in an endpoint.
4675 */
4676 sk_for_each_bound(sk2, node, &pp->owner) {
4677 struct sctp_endpoint *ep2;
4678 ep2 = sctp_sk(sk2)->ep;
4679
4680 if (reuse && sk2->sk_reuse)
4681 continue;
4682
4683 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4684 sctp_sk(sk))) {
4685 ret = (long)sk2;
4686 goto fail_unlock;
4687 }
4688 }
4689 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4690 }
4691pp_not_found:
4692 /* If there was a hash table miss, create a new port. */
4693 ret = 1;
4694 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4695 goto fail_unlock;
4696
4697 /* In either case (hit or miss), make sure fastreuse is 1 only
4698 * if sk->sk_reuse is too (that is, if the caller requested
4699 * SO_REUSEADDR on this socket -sk-).
4700 */
4701 if (hlist_empty(&pp->owner))
4702 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4703 else if (pp->fastreuse && !sk->sk_reuse)
4704 pp->fastreuse = 0;
4705
4706 /* We are set, so fill up all the data in the hash table
4707 * entry, tie the socket list information with the rest of the
4708 * sockets FIXME: Blurry, NPI (ipg).
4709 */
4710success:
4711 inet_sk(sk)->num = snum;
4712 if (!sctp_sk(sk)->bind_hash) {
4713 sk_add_bind_node(sk, &pp->owner);
4714 sctp_sk(sk)->bind_hash = pp;
4715 }
4716 ret = 0;
4717
4718fail_unlock:
4719 sctp_spin_unlock(&head->lock);
4720
4721fail:
4722 sctp_local_bh_enable();
4723 addr->v4.sin_port = htons(addr->v4.sin_port);
4724 return ret;
4725}
4726
4727/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4728 * port is requested.
4729 */
4730static int sctp_get_port(struct sock *sk, unsigned short snum)
4731{
4732 long ret;
4733 union sctp_addr addr;
4734 struct sctp_af *af = sctp_sk(sk)->pf->af;
4735
4736 /* Set up a dummy address struct from the sk. */
4737 af->from_sk(&addr, sk);
4738 addr.v4.sin_port = htons(snum);
4739
4740 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4741 ret = sctp_get_port_local(sk, &addr);
4742
4743 return (ret ? 1 : 0);
4744}
4745
4746/*
4747 * 3.1.3 listen() - UDP Style Syntax
4748 *
4749 * By default, new associations are not accepted for UDP style sockets.
4750 * An application uses listen() to mark a socket as being able to
4751 * accept new associations.
4752 */
4753SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4754{
4755 struct sctp_sock *sp = sctp_sk(sk);
4756 struct sctp_endpoint *ep = sp->ep;
4757
4758 /* Only UDP style sockets that are not peeled off are allowed to
4759 * listen().
4760 */
4761 if (!sctp_style(sk, UDP))
4762 return -EINVAL;
4763
4764 /* If backlog is zero, disable listening. */
4765 if (!backlog) {
4766 if (sctp_sstate(sk, CLOSED))
4767 return 0;
4768
4769 sctp_unhash_endpoint(ep);
4770 sk->sk_state = SCTP_SS_CLOSED;
4771 }
4772
4773 /* Return if we are already listening. */
4774 if (sctp_sstate(sk, LISTENING))
4775 return 0;
4776
4777 /*
4778 * If a bind() or sctp_bindx() is not called prior to a listen()
4779 * call that allows new associations to be accepted, the system
4780 * picks an ephemeral port and will choose an address set equivalent
4781 * to binding with a wildcard address.
4782 *
4783 * This is not currently spelled out in the SCTP sockets
4784 * extensions draft, but follows the practice as seen in TCP
4785 * sockets.
4786 */
4787 if (!ep->base.bind_addr.port) {
4788 if (sctp_autobind(sk))
4789 return -EAGAIN;
4790 }
4791 sk->sk_state = SCTP_SS_LISTENING;
4792 sctp_hash_endpoint(ep);
4793 return 0;
4794}
4795
4796/*
4797 * 4.1.3 listen() - TCP Style Syntax
4798 *
4799 * Applications uses listen() to ready the SCTP endpoint for accepting
4800 * inbound associations.
4801 */
4802SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4803{
4804 struct sctp_sock *sp = sctp_sk(sk);
4805 struct sctp_endpoint *ep = sp->ep;
4806
4807 /* If backlog is zero, disable listening. */
4808 if (!backlog) {
4809 if (sctp_sstate(sk, CLOSED))
4810 return 0;
4811
4812 sctp_unhash_endpoint(ep);
4813 sk->sk_state = SCTP_SS_CLOSED;
4814 }
4815
4816 if (sctp_sstate(sk, LISTENING))
4817 return 0;
4818
4819 /*
4820 * If a bind() or sctp_bindx() is not called prior to a listen()
4821 * call that allows new associations to be accepted, the system
4822 * picks an ephemeral port and will choose an address set equivalent
4823 * to binding with a wildcard address.
4824 *
4825 * This is not currently spelled out in the SCTP sockets
4826 * extensions draft, but follows the practice as seen in TCP
4827 * sockets.
4828 */
4829 if (!ep->base.bind_addr.port) {
4830 if (sctp_autobind(sk))
4831 return -EAGAIN;
4832 }
4833 sk->sk_state = SCTP_SS_LISTENING;
4834 sk->sk_max_ack_backlog = backlog;
4835 sctp_hash_endpoint(ep);
4836 return 0;
4837}
4838
4839/*
4840 * Move a socket to LISTENING state.
4841 */
4842int sctp_inet_listen(struct socket *sock, int backlog)
4843{
4844 struct sock *sk = sock->sk;
4845 struct crypto_tfm *tfm=NULL;
4846 int err = -EINVAL;
4847
4848 if (unlikely(backlog < 0))
4849 goto out;
4850
4851 sctp_lock_sock(sk);
4852
4853 if (sock->state != SS_UNCONNECTED)
4854 goto out;
4855
4856 /* Allocate HMAC for generating cookie. */
4857 if (sctp_hmac_alg) {
4858 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4859 if (!tfm) {
4860 err = -ENOSYS;
4861 goto out;
4862 }
4863 }
4864
4865 switch (sock->type) {
4866 case SOCK_SEQPACKET:
4867 err = sctp_seqpacket_listen(sk, backlog);
4868 break;
4869 case SOCK_STREAM:
4870 err = sctp_stream_listen(sk, backlog);
4871 break;
4872 default:
4873 break;
4874 };
4875 if (err)
4876 goto cleanup;
4877
4878 /* Store away the transform reference. */
4879 sctp_sk(sk)->hmac = tfm;
4880out:
4881 sctp_release_sock(sk);
4882 return err;
4883cleanup:
Jesper Juhl573dbd92005-09-01 17:44:29 -07004884 sctp_crypto_free_tfm(tfm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004885 goto out;
4886}
4887
4888/*
4889 * This function is done by modeling the current datagram_poll() and the
4890 * tcp_poll(). Note that, based on these implementations, we don't
4891 * lock the socket in this function, even though it seems that,
4892 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08004893 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07004894 * in this place. We assume that we don't need locks either until proven
4895 * otherwise.
4896 *
4897 * Another thing to note is that we include the Async I/O support
4898 * here, again, by modeling the current TCP/UDP code. We don't have
4899 * a good way to test with it yet.
4900 */
4901unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4902{
4903 struct sock *sk = sock->sk;
4904 struct sctp_sock *sp = sctp_sk(sk);
4905 unsigned int mask;
4906
4907 poll_wait(file, sk->sk_sleep, wait);
4908
4909 /* A TCP-style listening socket becomes readable when the accept queue
4910 * is not empty.
4911 */
4912 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4913 return (!list_empty(&sp->ep->asocs)) ?
4914 (POLLIN | POLLRDNORM) : 0;
4915
4916 mask = 0;
4917
4918 /* Is there any exceptional events? */
4919 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4920 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08004921 if (sk->sk_shutdown & RCV_SHUTDOWN)
4922 mask |= POLLRDHUP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004923 if (sk->sk_shutdown == SHUTDOWN_MASK)
4924 mask |= POLLHUP;
4925
4926 /* Is it readable? Reconsider this code with TCP-style support. */
4927 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4928 (sk->sk_shutdown & RCV_SHUTDOWN))
4929 mask |= POLLIN | POLLRDNORM;
4930
4931 /* The association is either gone or not ready. */
4932 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4933 return mask;
4934
4935 /* Is it writable? */
4936 if (sctp_writeable(sk)) {
4937 mask |= POLLOUT | POLLWRNORM;
4938 } else {
4939 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4940 /*
4941 * Since the socket is not locked, the buffer
4942 * might be made available after the writeable check and
4943 * before the bit is set. This could cause a lost I/O
4944 * signal. tcp_poll() has a race breaker for this race
4945 * condition. Based on their implementation, we put
4946 * in the following code to cover it as well.
4947 */
4948 if (sctp_writeable(sk))
4949 mask |= POLLOUT | POLLWRNORM;
4950 }
4951 return mask;
4952}
4953
4954/********************************************************************
4955 * 2nd Level Abstractions
4956 ********************************************************************/
4957
4958static struct sctp_bind_bucket *sctp_bucket_create(
4959 struct sctp_bind_hashbucket *head, unsigned short snum)
4960{
4961 struct sctp_bind_bucket *pp;
4962
4963 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
4964 SCTP_DBG_OBJCNT_INC(bind_bucket);
4965 if (pp) {
4966 pp->port = snum;
4967 pp->fastreuse = 0;
4968 INIT_HLIST_HEAD(&pp->owner);
4969 if ((pp->next = head->chain) != NULL)
4970 pp->next->pprev = &pp->next;
4971 head->chain = pp;
4972 pp->pprev = &head->chain;
4973 }
4974 return pp;
4975}
4976
4977/* Caller must hold hashbucket lock for this tb with local BH disabled */
4978static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
4979{
4980 if (hlist_empty(&pp->owner)) {
4981 if (pp->next)
4982 pp->next->pprev = pp->pprev;
4983 *(pp->pprev) = pp->next;
4984 kmem_cache_free(sctp_bucket_cachep, pp);
4985 SCTP_DBG_OBJCNT_DEC(bind_bucket);
4986 }
4987}
4988
4989/* Release this socket's reference to a local port. */
4990static inline void __sctp_put_port(struct sock *sk)
4991{
4992 struct sctp_bind_hashbucket *head =
4993 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
4994 struct sctp_bind_bucket *pp;
4995
4996 sctp_spin_lock(&head->lock);
4997 pp = sctp_sk(sk)->bind_hash;
4998 __sk_del_bind_node(sk);
4999 sctp_sk(sk)->bind_hash = NULL;
5000 inet_sk(sk)->num = 0;
5001 sctp_bucket_destroy(pp);
5002 sctp_spin_unlock(&head->lock);
5003}
5004
5005void sctp_put_port(struct sock *sk)
5006{
5007 sctp_local_bh_disable();
5008 __sctp_put_port(sk);
5009 sctp_local_bh_enable();
5010}
5011
5012/*
5013 * The system picks an ephemeral port and choose an address set equivalent
5014 * to binding with a wildcard address.
5015 * One of those addresses will be the primary address for the association.
5016 * This automatically enables the multihoming capability of SCTP.
5017 */
5018static int sctp_autobind(struct sock *sk)
5019{
5020 union sctp_addr autoaddr;
5021 struct sctp_af *af;
5022 unsigned short port;
5023
5024 /* Initialize a local sockaddr structure to INADDR_ANY. */
5025 af = sctp_sk(sk)->pf->af;
5026
5027 port = htons(inet_sk(sk)->num);
5028 af->inaddr_any(&autoaddr, port);
5029
5030 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5031}
5032
5033/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5034 *
5035 * From RFC 2292
5036 * 4.2 The cmsghdr Structure *
5037 *
5038 * When ancillary data is sent or received, any number of ancillary data
5039 * objects can be specified by the msg_control and msg_controllen members of
5040 * the msghdr structure, because each object is preceded by
5041 * a cmsghdr structure defining the object's length (the cmsg_len member).
5042 * Historically Berkeley-derived implementations have passed only one object
5043 * at a time, but this API allows multiple objects to be
5044 * passed in a single call to sendmsg() or recvmsg(). The following example
5045 * shows two ancillary data objects in a control buffer.
5046 *
5047 * |<--------------------------- msg_controllen -------------------------->|
5048 * | |
5049 *
5050 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5051 *
5052 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5053 * | | |
5054 *
5055 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5056 *
5057 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5058 * | | | | |
5059 *
5060 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5061 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5062 *
5063 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5064 *
5065 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5066 * ^
5067 * |
5068 *
5069 * msg_control
5070 * points here
5071 */
5072SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5073 sctp_cmsgs_t *cmsgs)
5074{
5075 struct cmsghdr *cmsg;
5076
5077 for (cmsg = CMSG_FIRSTHDR(msg);
5078 cmsg != NULL;
5079 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5080 if (!CMSG_OK(msg, cmsg))
5081 return -EINVAL;
5082
5083 /* Should we parse this header or ignore? */
5084 if (cmsg->cmsg_level != IPPROTO_SCTP)
5085 continue;
5086
5087 /* Strictly check lengths following example in SCM code. */
5088 switch (cmsg->cmsg_type) {
5089 case SCTP_INIT:
5090 /* SCTP Socket API Extension
5091 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5092 *
5093 * This cmsghdr structure provides information for
5094 * initializing new SCTP associations with sendmsg().
5095 * The SCTP_INITMSG socket option uses this same data
5096 * structure. This structure is not used for
5097 * recvmsg().
5098 *
5099 * cmsg_level cmsg_type cmsg_data[]
5100 * ------------ ------------ ----------------------
5101 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5102 */
5103 if (cmsg->cmsg_len !=
5104 CMSG_LEN(sizeof(struct sctp_initmsg)))
5105 return -EINVAL;
5106 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5107 break;
5108
5109 case SCTP_SNDRCV:
5110 /* SCTP Socket API Extension
5111 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5112 *
5113 * This cmsghdr structure specifies SCTP options for
5114 * sendmsg() and describes SCTP header information
5115 * about a received message through recvmsg().
5116 *
5117 * cmsg_level cmsg_type cmsg_data[]
5118 * ------------ ------------ ----------------------
5119 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5120 */
5121 if (cmsg->cmsg_len !=
5122 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5123 return -EINVAL;
5124
5125 cmsgs->info =
5126 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5127
5128 /* Minimally, validate the sinfo_flags. */
5129 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07005130 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5131 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005132 return -EINVAL;
5133 break;
5134
5135 default:
5136 return -EINVAL;
5137 };
5138 }
5139 return 0;
5140}
5141
5142/*
5143 * Wait for a packet..
5144 * Note: This function is the same function as in core/datagram.c
5145 * with a few modifications to make lksctp work.
5146 */
5147static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5148{
5149 int error;
5150 DEFINE_WAIT(wait);
5151
5152 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5153
5154 /* Socket errors? */
5155 error = sock_error(sk);
5156 if (error)
5157 goto out;
5158
5159 if (!skb_queue_empty(&sk->sk_receive_queue))
5160 goto ready;
5161
5162 /* Socket shut down? */
5163 if (sk->sk_shutdown & RCV_SHUTDOWN)
5164 goto out;
5165
5166 /* Sequenced packets can come disconnected. If so we report the
5167 * problem.
5168 */
5169 error = -ENOTCONN;
5170
5171 /* Is there a good reason to think that we may receive some data? */
5172 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5173 goto out;
5174
5175 /* Handle signals. */
5176 if (signal_pending(current))
5177 goto interrupted;
5178
5179 /* Let another process have a go. Since we are going to sleep
5180 * anyway. Note: This may cause odd behaviors if the message
5181 * does not fit in the user's buffer, but this seems to be the
5182 * only way to honor MSG_DONTWAIT realistically.
5183 */
5184 sctp_release_sock(sk);
5185 *timeo_p = schedule_timeout(*timeo_p);
5186 sctp_lock_sock(sk);
5187
5188ready:
5189 finish_wait(sk->sk_sleep, &wait);
5190 return 0;
5191
5192interrupted:
5193 error = sock_intr_errno(*timeo_p);
5194
5195out:
5196 finish_wait(sk->sk_sleep, &wait);
5197 *err = error;
5198 return error;
5199}
5200
5201/* Receive a datagram.
5202 * Note: This is pretty much the same routine as in core/datagram.c
5203 * with a few changes to make lksctp work.
5204 */
5205static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5206 int noblock, int *err)
5207{
5208 int error;
5209 struct sk_buff *skb;
5210 long timeo;
5211
Linus Torvalds1da177e2005-04-16 15:20:36 -07005212 timeo = sock_rcvtimeo(sk, noblock);
5213
5214 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5215 timeo, MAX_SCHEDULE_TIMEOUT);
5216
5217 do {
5218 /* Again only user level code calls this function,
5219 * so nothing interrupt level
5220 * will suddenly eat the receive_queue.
5221 *
5222 * Look at current nfs client by the way...
5223 * However, this function was corrent in any case. 8)
5224 */
5225 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07005226 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005227 skb = skb_peek(&sk->sk_receive_queue);
5228 if (skb)
5229 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07005230 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005231 } else {
5232 skb = skb_dequeue(&sk->sk_receive_queue);
5233 }
5234
5235 if (skb)
5236 return skb;
5237
Neil Horman6736dc32005-12-02 20:30:06 -08005238 /* Caller is allowed not to check sk->sk_err before calling. */
5239 error = sock_error(sk);
5240 if (error)
5241 goto no_packet;
5242
Linus Torvalds1da177e2005-04-16 15:20:36 -07005243 if (sk->sk_shutdown & RCV_SHUTDOWN)
5244 break;
5245
5246 /* User doesn't want to wait. */
5247 error = -EAGAIN;
5248 if (!timeo)
5249 goto no_packet;
5250 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5251
5252 return NULL;
5253
5254no_packet:
5255 *err = error;
5256 return NULL;
5257}
5258
5259/* If sndbuf has changed, wake up per association sndbuf waiters. */
5260static void __sctp_write_space(struct sctp_association *asoc)
5261{
5262 struct sock *sk = asoc->base.sk;
5263 struct socket *sock = sk->sk_socket;
5264
5265 if ((sctp_wspace(asoc) > 0) && sock) {
5266 if (waitqueue_active(&asoc->wait))
5267 wake_up_interruptible(&asoc->wait);
5268
5269 if (sctp_writeable(sk)) {
5270 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5271 wake_up_interruptible(sk->sk_sleep);
5272
5273 /* Note that we try to include the Async I/O support
5274 * here by modeling from the current TCP/UDP code.
5275 * We have not tested with it yet.
5276 */
5277 if (sock->fasync_list &&
5278 !(sk->sk_shutdown & SEND_SHUTDOWN))
5279 sock_wake_async(sock, 2, POLL_OUT);
5280 }
5281 }
5282}
5283
5284/* Do accounting for the sndbuf space.
5285 * Decrement the used sndbuf space of the corresponding association by the
5286 * data size which was just transmitted(freed).
5287 */
5288static void sctp_wfree(struct sk_buff *skb)
5289{
5290 struct sctp_association *asoc;
5291 struct sctp_chunk *chunk;
5292 struct sock *sk;
5293
5294 /* Get the saved chunk pointer. */
5295 chunk = *((struct sctp_chunk **)(skb->cb));
5296 asoc = chunk->asoc;
5297 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07005298 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5299 sizeof(struct sk_buff) +
5300 sizeof(struct sctp_chunk);
5301
Neil Horman4eb701d2005-04-28 12:02:04 -07005302 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5303
5304 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005305 __sctp_write_space(asoc);
5306
5307 sctp_association_put(asoc);
5308}
5309
5310/* Helper function to wait for space in the sndbuf. */
5311static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5312 size_t msg_len)
5313{
5314 struct sock *sk = asoc->base.sk;
5315 int err = 0;
5316 long current_timeo = *timeo_p;
5317 DEFINE_WAIT(wait);
5318
5319 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5320 asoc, (long)(*timeo_p), msg_len);
5321
5322 /* Increment the association's refcnt. */
5323 sctp_association_hold(asoc);
5324
5325 /* Wait on the association specific sndbuf space. */
5326 for (;;) {
5327 prepare_to_wait_exclusive(&asoc->wait, &wait,
5328 TASK_INTERRUPTIBLE);
5329 if (!*timeo_p)
5330 goto do_nonblock;
5331 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5332 asoc->base.dead)
5333 goto do_error;
5334 if (signal_pending(current))
5335 goto do_interrupted;
5336 if (msg_len <= sctp_wspace(asoc))
5337 break;
5338
5339 /* Let another process have a go. Since we are going
5340 * to sleep anyway.
5341 */
5342 sctp_release_sock(sk);
5343 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005344 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005345 sctp_lock_sock(sk);
5346
5347 *timeo_p = current_timeo;
5348 }
5349
5350out:
5351 finish_wait(&asoc->wait, &wait);
5352
5353 /* Release the association's refcnt. */
5354 sctp_association_put(asoc);
5355
5356 return err;
5357
5358do_error:
5359 err = -EPIPE;
5360 goto out;
5361
5362do_interrupted:
5363 err = sock_intr_errno(*timeo_p);
5364 goto out;
5365
5366do_nonblock:
5367 err = -EAGAIN;
5368 goto out;
5369}
5370
5371/* If socket sndbuf has changed, wake up all per association waiters. */
5372void sctp_write_space(struct sock *sk)
5373{
5374 struct sctp_association *asoc;
5375 struct list_head *pos;
5376
5377 /* Wake up the tasks in each wait queue. */
5378 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5379 asoc = list_entry(pos, struct sctp_association, asocs);
5380 __sctp_write_space(asoc);
5381 }
5382}
5383
5384/* Is there any sndbuf space available on the socket?
5385 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08005386 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07005387 * associations on the same socket. For a UDP-style socket with
5388 * multiple associations, it is possible for it to be "unwriteable"
5389 * prematurely. I assume that this is acceptable because
5390 * a premature "unwriteable" is better than an accidental "writeable" which
5391 * would cause an unwanted block under certain circumstances. For the 1-1
5392 * UDP-style sockets or TCP-style sockets, this code should work.
5393 * - Daisy
5394 */
5395static int sctp_writeable(struct sock *sk)
5396{
5397 int amt = 0;
5398
Neil Horman9bffc4a2005-12-19 14:24:40 -08005399 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005400 if (amt < 0)
5401 amt = 0;
5402 return amt;
5403}
5404
5405/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5406 * returns immediately with EINPROGRESS.
5407 */
5408static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5409{
5410 struct sock *sk = asoc->base.sk;
5411 int err = 0;
5412 long current_timeo = *timeo_p;
5413 DEFINE_WAIT(wait);
5414
5415 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5416 (long)(*timeo_p));
5417
5418 /* Increment the association's refcnt. */
5419 sctp_association_hold(asoc);
5420
5421 for (;;) {
5422 prepare_to_wait_exclusive(&asoc->wait, &wait,
5423 TASK_INTERRUPTIBLE);
5424 if (!*timeo_p)
5425 goto do_nonblock;
5426 if (sk->sk_shutdown & RCV_SHUTDOWN)
5427 break;
5428 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5429 asoc->base.dead)
5430 goto do_error;
5431 if (signal_pending(current))
5432 goto do_interrupted;
5433
5434 if (sctp_state(asoc, ESTABLISHED))
5435 break;
5436
5437 /* Let another process have a go. Since we are going
5438 * to sleep anyway.
5439 */
5440 sctp_release_sock(sk);
5441 current_timeo = schedule_timeout(current_timeo);
5442 sctp_lock_sock(sk);
5443
5444 *timeo_p = current_timeo;
5445 }
5446
5447out:
5448 finish_wait(&asoc->wait, &wait);
5449
5450 /* Release the association's refcnt. */
5451 sctp_association_put(asoc);
5452
5453 return err;
5454
5455do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08005456 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005457 err = -ETIMEDOUT;
5458 else
5459 err = -ECONNREFUSED;
5460 goto out;
5461
5462do_interrupted:
5463 err = sock_intr_errno(*timeo_p);
5464 goto out;
5465
5466do_nonblock:
5467 err = -EINPROGRESS;
5468 goto out;
5469}
5470
5471static int sctp_wait_for_accept(struct sock *sk, long timeo)
5472{
5473 struct sctp_endpoint *ep;
5474 int err = 0;
5475 DEFINE_WAIT(wait);
5476
5477 ep = sctp_sk(sk)->ep;
5478
5479
5480 for (;;) {
5481 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5482 TASK_INTERRUPTIBLE);
5483
5484 if (list_empty(&ep->asocs)) {
5485 sctp_release_sock(sk);
5486 timeo = schedule_timeout(timeo);
5487 sctp_lock_sock(sk);
5488 }
5489
5490 err = -EINVAL;
5491 if (!sctp_sstate(sk, LISTENING))
5492 break;
5493
5494 err = 0;
5495 if (!list_empty(&ep->asocs))
5496 break;
5497
5498 err = sock_intr_errno(timeo);
5499 if (signal_pending(current))
5500 break;
5501
5502 err = -EAGAIN;
5503 if (!timeo)
5504 break;
5505 }
5506
5507 finish_wait(sk->sk_sleep, &wait);
5508
5509 return err;
5510}
5511
5512void sctp_wait_for_close(struct sock *sk, long timeout)
5513{
5514 DEFINE_WAIT(wait);
5515
5516 do {
5517 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5518 if (list_empty(&sctp_sk(sk)->ep->asocs))
5519 break;
5520 sctp_release_sock(sk);
5521 timeout = schedule_timeout(timeout);
5522 sctp_lock_sock(sk);
5523 } while (!signal_pending(current) && timeout);
5524
5525 finish_wait(sk->sk_sleep, &wait);
5526}
5527
5528/* Populate the fields of the newsk from the oldsk and migrate the assoc
5529 * and its messages to the newsk.
5530 */
5531static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5532 struct sctp_association *assoc,
5533 sctp_socket_type_t type)
5534{
5535 struct sctp_sock *oldsp = sctp_sk(oldsk);
5536 struct sctp_sock *newsp = sctp_sk(newsk);
5537 struct sctp_bind_bucket *pp; /* hash list port iterator */
5538 struct sctp_endpoint *newep = newsp->ep;
5539 struct sk_buff *skb, *tmp;
5540 struct sctp_ulpevent *event;
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005541 int flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005542
5543 /* Migrate socket buffer sizes and all the socket level options to the
5544 * new socket.
5545 */
5546 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5547 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5548 /* Brute force copy old sctp opt. */
5549 inet_sk_copy_descendant(newsk, oldsk);
5550
5551 /* Restore the ep value that was overwritten with the above structure
5552 * copy.
5553 */
5554 newsp->ep = newep;
5555 newsp->hmac = NULL;
5556
5557 /* Hook this new socket in to the bind_hash list. */
5558 pp = sctp_sk(oldsk)->bind_hash;
5559 sk_add_bind_node(newsk, &pp->owner);
5560 sctp_sk(newsk)->bind_hash = pp;
5561 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5562
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005563 /* Copy the bind_addr list from the original endpoint to the new
5564 * endpoint so that we can handle restarts properly
5565 */
5566 if (assoc->peer.ipv4_address)
5567 flags |= SCTP_ADDR4_PEERSUPP;
5568 if (assoc->peer.ipv6_address)
5569 flags |= SCTP_ADDR6_PEERSUPP;
5570 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5571 &oldsp->ep->base.bind_addr,
5572 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5573
Linus Torvalds1da177e2005-04-16 15:20:36 -07005574 /* Move any messages in the old socket's receive queue that are for the
5575 * peeled off association to the new socket's receive queue.
5576 */
5577 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5578 event = sctp_skb2event(skb);
5579 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005580 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005581 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005582 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005583 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005584 }
5585 }
5586
5587 /* Clean up any messages pending delivery due to partial
5588 * delivery. Three cases:
5589 * 1) No partial deliver; no work.
5590 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5591 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5592 */
5593 skb_queue_head_init(&newsp->pd_lobby);
5594 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5595
5596 if (sctp_sk(oldsk)->pd_mode) {
5597 struct sk_buff_head *queue;
5598
5599 /* Decide which queue to move pd_lobby skbs to. */
5600 if (assoc->ulpq.pd_mode) {
5601 queue = &newsp->pd_lobby;
5602 } else
5603 queue = &newsk->sk_receive_queue;
5604
5605 /* Walk through the pd_lobby, looking for skbs that
5606 * need moved to the new socket.
5607 */
5608 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5609 event = sctp_skb2event(skb);
5610 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005611 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005612 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005613 __skb_queue_tail(queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005614 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005615 }
5616 }
5617
5618 /* Clear up any skbs waiting for the partial
5619 * delivery to finish.
5620 */
5621 if (assoc->ulpq.pd_mode)
5622 sctp_clear_pd(oldsk);
5623
5624 }
5625
5626 /* Set the type of socket to indicate that it is peeled off from the
5627 * original UDP-style socket or created with the accept() call on a
5628 * TCP-style socket..
5629 */
5630 newsp->type = type;
5631
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005632 /* Mark the new socket "in-use" by the user so that any packets
5633 * that may arrive on the association after we've moved it are
5634 * queued to the backlog. This prevents a potential race between
5635 * backlog processing on the old socket and new-packet processing
5636 * on the new socket.
5637 */
5638 sctp_lock_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005639 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005640
5641 /* If the association on the newsk is already closed before accept()
5642 * is called, set RCV_SHUTDOWN flag.
5643 */
5644 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5645 newsk->sk_shutdown |= RCV_SHUTDOWN;
5646
5647 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005648 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005649}
5650
5651/* This proto struct describes the ULP interface for SCTP. */
5652struct proto sctp_prot = {
5653 .name = "SCTP",
5654 .owner = THIS_MODULE,
5655 .close = sctp_close,
5656 .connect = sctp_connect,
5657 .disconnect = sctp_disconnect,
5658 .accept = sctp_accept,
5659 .ioctl = sctp_ioctl,
5660 .init = sctp_init_sock,
5661 .destroy = sctp_destroy_sock,
5662 .shutdown = sctp_shutdown,
5663 .setsockopt = sctp_setsockopt,
5664 .getsockopt = sctp_getsockopt,
5665 .sendmsg = sctp_sendmsg,
5666 .recvmsg = sctp_recvmsg,
5667 .bind = sctp_bind,
5668 .backlog_rcv = sctp_backlog_rcv,
5669 .hash = sctp_hash,
5670 .unhash = sctp_unhash,
5671 .get_port = sctp_get_port,
5672 .obj_size = sizeof(struct sctp_sock),
5673};
5674
5675#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5676struct proto sctpv6_prot = {
5677 .name = "SCTPv6",
5678 .owner = THIS_MODULE,
5679 .close = sctp_close,
5680 .connect = sctp_connect,
5681 .disconnect = sctp_disconnect,
5682 .accept = sctp_accept,
5683 .ioctl = sctp_ioctl,
5684 .init = sctp_init_sock,
5685 .destroy = sctp_destroy_sock,
5686 .shutdown = sctp_shutdown,
5687 .setsockopt = sctp_setsockopt,
5688 .getsockopt = sctp_getsockopt,
5689 .sendmsg = sctp_sendmsg,
5690 .recvmsg = sctp_recvmsg,
5691 .bind = sctp_bind,
5692 .backlog_rcv = sctp_backlog_rcv,
5693 .hash = sctp_hash,
5694 .unhash = sctp_unhash,
5695 .get_port = sctp_get_port,
5696 .obj_size = sizeof(struct sctp6_sock),
5697};
5698#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */