blob: 4454afe4727ef2bb95778a5a8193c8b74df6dcd9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel reference Implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * The SCTP reference implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60#include <linux/config.h>
61#include <linux/types.h>
62#include <linux/kernel.h>
63#include <linux/wait.h>
64#include <linux/time.h>
65#include <linux/ip.h>
66#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
159 sk->sk_wmem_queued += SCTP_DATA_SNDSIZE(chunk) +
160 sizeof(struct sk_buff) +
161 sizeof(struct sctp_chunk);
162
163 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164}
165
166/* Verify that this is a valid address. */
167static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
168 int len)
169{
170 struct sctp_af *af;
171
172 /* Verify basic sockaddr. */
173 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
174 if (!af)
175 return -EINVAL;
176
177 /* Is this a valid SCTP address? */
178 if (!af->addr_valid(addr, sctp_sk(sk)))
179 return -EINVAL;
180
181 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
182 return -EINVAL;
183
184 return 0;
185}
186
187/* Look up the association by its id. If this is not a UDP-style
188 * socket, the ID field is always ignored.
189 */
190struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
191{
192 struct sctp_association *asoc = NULL;
193
194 /* If this is not a UDP-style socket, assoc id should be ignored. */
195 if (!sctp_style(sk, UDP)) {
196 /* Return NULL if the socket state is not ESTABLISHED. It
197 * could be a TCP-style listening socket or a socket which
198 * hasn't yet called connect() to establish an association.
199 */
200 if (!sctp_sstate(sk, ESTABLISHED))
201 return NULL;
202
203 /* Get the first and the only association from the list. */
204 if (!list_empty(&sctp_sk(sk)->ep->asocs))
205 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
206 struct sctp_association, asocs);
207 return asoc;
208 }
209
210 /* Otherwise this is a UDP-style socket. */
211 if (!id || (id == (sctp_assoc_t)-1))
212 return NULL;
213
214 spin_lock_bh(&sctp_assocs_id_lock);
215 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
216 spin_unlock_bh(&sctp_assocs_id_lock);
217
218 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
219 return NULL;
220
221 return asoc;
222}
223
224/* Look up the transport from an address and an assoc id. If both address and
225 * id are specified, the associations matching the address and the id should be
226 * the same.
227 */
228static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
229 struct sockaddr_storage *addr,
230 sctp_assoc_t id)
231{
232 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
233 struct sctp_transport *transport;
234 union sctp_addr *laddr = (union sctp_addr *)addr;
235
236 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
237 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
238 (union sctp_addr *)addr,
239 &transport);
240 laddr->v4.sin_port = htons(laddr->v4.sin_port);
241
242 if (!addr_asoc)
243 return NULL;
244
245 id_asoc = sctp_id2assoc(sk, id);
246 if (id_asoc && (id_asoc != addr_asoc))
247 return NULL;
248
249 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
250 (union sctp_addr *)addr);
251
252 return transport;
253}
254
255/* API 3.1.2 bind() - UDP Style Syntax
256 * The syntax of bind() is,
257 *
258 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
259 *
260 * sd - the socket descriptor returned by socket().
261 * addr - the address structure (struct sockaddr_in or struct
262 * sockaddr_in6 [RFC 2553]),
263 * addr_len - the size of the address structure.
264 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700265SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266{
267 int retval = 0;
268
269 sctp_lock_sock(sk);
270
Frank Filz3f7a87d2005-06-20 13:14:57 -0700271 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
272 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273
274 /* Disallow binding twice. */
275 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700276 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 addr_len);
278 else
279 retval = -EINVAL;
280
281 sctp_release_sock(sk);
282
283 return retval;
284}
285
286static long sctp_get_port_local(struct sock *, union sctp_addr *);
287
288/* Verify this is a valid sockaddr. */
289static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
290 union sctp_addr *addr, int len)
291{
292 struct sctp_af *af;
293
294 /* Check minimum size. */
295 if (len < sizeof (struct sockaddr))
296 return NULL;
297
298 /* Does this PF support this AF? */
299 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
300 return NULL;
301
302 /* If we get this far, af is valid. */
303 af = sctp_get_af_specific(addr->sa.sa_family);
304
305 if (len < af->sockaddr_len)
306 return NULL;
307
308 return af;
309}
310
311/* Bind a local address either to an endpoint or to an association. */
312SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
313{
314 struct sctp_sock *sp = sctp_sk(sk);
315 struct sctp_endpoint *ep = sp->ep;
316 struct sctp_bind_addr *bp = &ep->base.bind_addr;
317 struct sctp_af *af;
318 unsigned short snum;
319 int ret = 0;
320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 /* Common sockaddr verification. */
322 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700323 if (!af) {
324 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
325 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700327 }
328
329 snum = ntohs(addr->v4.sin_port);
330
331 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
332 ", port: %d, new port: %d, len: %d)\n",
333 sk,
334 addr,
335 bp->port, snum,
336 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
338 /* PF specific bind() address verification. */
339 if (!sp->pf->bind_verify(sp, addr))
340 return -EADDRNOTAVAIL;
341
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 /* We must either be unbound, or bind to the same port. */
343 if (bp->port && (snum != bp->port)) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind:"
345 " New port %d does not match existing port "
346 "%d.\n", snum, bp->port);
347 return -EINVAL;
348 }
349
350 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
351 return -EACCES;
352
353 /* Make sure we are allowed to bind here.
354 * The function sctp_get_port_local() does duplicate address
355 * detection.
356 */
357 if ((ret = sctp_get_port_local(sk, addr))) {
358 if (ret == (long) sk) {
359 /* This endpoint has a conflicting address. */
360 return -EINVAL;
361 } else {
362 return -EADDRINUSE;
363 }
364 }
365
366 /* Refresh ephemeral port. */
367 if (!bp->port)
368 bp->port = inet_sk(sk)->num;
369
370 /* Add the address to the bind address list. */
371 sctp_local_bh_disable();
372 sctp_write_lock(&ep->base.addr_lock);
373
374 /* Use GFP_ATOMIC since BHs are disabled. */
375 addr->v4.sin_port = ntohs(addr->v4.sin_port);
376 ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC);
377 addr->v4.sin_port = htons(addr->v4.sin_port);
378 sctp_write_unlock(&ep->base.addr_lock);
379 sctp_local_bh_enable();
380
381 /* Copy back into socket for getsockname() use. */
382 if (!ret) {
383 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
384 af->to_sk_saddr(addr, sk);
385 }
386
387 return ret;
388}
389
390 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
391 *
392 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
393 * at any one time. If a sender, after sending an ASCONF chunk, decides
394 * it needs to transfer another ASCONF Chunk, it MUST wait until the
395 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
396 * subsequent ASCONF. Note this restriction binds each side, so at any
397 * time two ASCONF may be in-transit on any given association (one sent
398 * from each endpoint).
399 */
400static int sctp_send_asconf(struct sctp_association *asoc,
401 struct sctp_chunk *chunk)
402{
403 int retval = 0;
404
405 /* If there is an outstanding ASCONF chunk, queue it for later
406 * transmission.
407 */
408 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700409 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 goto out;
411 }
412
413 /* Hold the chunk until an ASCONF_ACK is received. */
414 sctp_chunk_hold(chunk);
415 retval = sctp_primitive_ASCONF(asoc, chunk);
416 if (retval)
417 sctp_chunk_free(chunk);
418 else
419 asoc->addip_last_asconf = chunk;
420
421out:
422 return retval;
423}
424
425/* Add a list of addresses as bind addresses to local endpoint or
426 * association.
427 *
428 * Basically run through each address specified in the addrs/addrcnt
429 * array/length pair, determine if it is IPv6 or IPv4 and call
430 * sctp_do_bind() on it.
431 *
432 * If any of them fails, then the operation will be reversed and the
433 * ones that were added will be removed.
434 *
435 * Only sctp_setsockopt_bindx() is supposed to call this function.
436 */
437int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
438{
439 int cnt;
440 int retval = 0;
441 void *addr_buf;
442 struct sockaddr *sa_addr;
443 struct sctp_af *af;
444
445 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
446 sk, addrs, addrcnt);
447
448 addr_buf = addrs;
449 for (cnt = 0; cnt < addrcnt; cnt++) {
450 /* The list may contain either IPv4 or IPv6 address;
451 * determine the address length for walking thru the list.
452 */
453 sa_addr = (struct sockaddr *)addr_buf;
454 af = sctp_get_af_specific(sa_addr->sa_family);
455 if (!af) {
456 retval = -EINVAL;
457 goto err_bindx_add;
458 }
459
460 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
461 af->sockaddr_len);
462
463 addr_buf += af->sockaddr_len;
464
465err_bindx_add:
466 if (retval < 0) {
467 /* Failed. Cleanup the ones that have been added */
468 if (cnt > 0)
469 sctp_bindx_rem(sk, addrs, cnt);
470 return retval;
471 }
472 }
473
474 return retval;
475}
476
477/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
478 * associations that are part of the endpoint indicating that a list of local
479 * addresses are added to the endpoint.
480 *
481 * If any of the addresses is already in the bind address list of the
482 * association, we do not send the chunk for that association. But it will not
483 * affect other associations.
484 *
485 * Only sctp_setsockopt_bindx() is supposed to call this function.
486 */
487static int sctp_send_asconf_add_ip(struct sock *sk,
488 struct sockaddr *addrs,
489 int addrcnt)
490{
491 struct sctp_sock *sp;
492 struct sctp_endpoint *ep;
493 struct sctp_association *asoc;
494 struct sctp_bind_addr *bp;
495 struct sctp_chunk *chunk;
496 struct sctp_sockaddr_entry *laddr;
497 union sctp_addr *addr;
498 void *addr_buf;
499 struct sctp_af *af;
500 struct list_head *pos;
501 struct list_head *p;
502 int i;
503 int retval = 0;
504
505 if (!sctp_addip_enable)
506 return retval;
507
508 sp = sctp_sk(sk);
509 ep = sp->ep;
510
511 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
512 __FUNCTION__, sk, addrs, addrcnt);
513
514 list_for_each(pos, &ep->asocs) {
515 asoc = list_entry(pos, struct sctp_association, asocs);
516
517 if (!asoc->peer.asconf_capable)
518 continue;
519
520 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
521 continue;
522
523 if (!sctp_state(asoc, ESTABLISHED))
524 continue;
525
526 /* Check if any address in the packed array of addresses is
527 * in the bind address list of the association. If so,
528 * do not send the asconf chunk to its peer, but continue with
529 * other associations.
530 */
531 addr_buf = addrs;
532 for (i = 0; i < addrcnt; i++) {
533 addr = (union sctp_addr *)addr_buf;
534 af = sctp_get_af_specific(addr->v4.sin_family);
535 if (!af) {
536 retval = -EINVAL;
537 goto out;
538 }
539
540 if (sctp_assoc_lookup_laddr(asoc, addr))
541 break;
542
543 addr_buf += af->sockaddr_len;
544 }
545 if (i < addrcnt)
546 continue;
547
548 /* Use the first address in bind addr list of association as
549 * Address Parameter of ASCONF CHUNK.
550 */
551 sctp_read_lock(&asoc->base.addr_lock);
552 bp = &asoc->base.bind_addr;
553 p = bp->address_list.next;
554 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
555 sctp_read_unlock(&asoc->base.addr_lock);
556
557 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
558 addrcnt, SCTP_PARAM_ADD_IP);
559 if (!chunk) {
560 retval = -ENOMEM;
561 goto out;
562 }
563
564 retval = sctp_send_asconf(asoc, chunk);
565
566 /* FIXME: After sending the add address ASCONF chunk, we
567 * cannot append the address to the association's binding
568 * address list, because the new address may be used as the
569 * source of a message sent to the peer before the ASCONF
570 * chunk is received by the peer. So we should wait until
571 * ASCONF_ACK is received.
572 */
573 }
574
575out:
576 return retval;
577}
578
579/* Remove a list of addresses from bind addresses list. Do not remove the
580 * last address.
581 *
582 * Basically run through each address specified in the addrs/addrcnt
583 * array/length pair, determine if it is IPv6 or IPv4 and call
584 * sctp_del_bind() on it.
585 *
586 * If any of them fails, then the operation will be reversed and the
587 * ones that were removed will be added back.
588 *
589 * At least one address has to be left; if only one address is
590 * available, the operation will return -EBUSY.
591 *
592 * Only sctp_setsockopt_bindx() is supposed to call this function.
593 */
594int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
595{
596 struct sctp_sock *sp = sctp_sk(sk);
597 struct sctp_endpoint *ep = sp->ep;
598 int cnt;
599 struct sctp_bind_addr *bp = &ep->base.bind_addr;
600 int retval = 0;
601 union sctp_addr saveaddr;
602 void *addr_buf;
603 struct sockaddr *sa_addr;
604 struct sctp_af *af;
605
606 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
607 sk, addrs, addrcnt);
608
609 addr_buf = addrs;
610 for (cnt = 0; cnt < addrcnt; cnt++) {
611 /* If the bind address list is empty or if there is only one
612 * bind address, there is nothing more to be removed (we need
613 * at least one address here).
614 */
615 if (list_empty(&bp->address_list) ||
616 (sctp_list_single_entry(&bp->address_list))) {
617 retval = -EBUSY;
618 goto err_bindx_rem;
619 }
620
621 /* The list may contain either IPv4 or IPv6 address;
622 * determine the address length to copy the address to
623 * saveaddr.
624 */
625 sa_addr = (struct sockaddr *)addr_buf;
626 af = sctp_get_af_specific(sa_addr->sa_family);
627 if (!af) {
628 retval = -EINVAL;
629 goto err_bindx_rem;
630 }
631 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
632 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
633 if (saveaddr.v4.sin_port != bp->port) {
634 retval = -EINVAL;
635 goto err_bindx_rem;
636 }
637
638 /* FIXME - There is probably a need to check if sk->sk_saddr and
639 * sk->sk_rcv_addr are currently set to one of the addresses to
640 * be removed. This is something which needs to be looked into
641 * when we are fixing the outstanding issues with multi-homing
642 * socket routing and failover schemes. Refer to comments in
643 * sctp_do_bind(). -daisy
644 */
645 sctp_local_bh_disable();
646 sctp_write_lock(&ep->base.addr_lock);
647
648 retval = sctp_del_bind_addr(bp, &saveaddr);
649
650 sctp_write_unlock(&ep->base.addr_lock);
651 sctp_local_bh_enable();
652
653 addr_buf += af->sockaddr_len;
654err_bindx_rem:
655 if (retval < 0) {
656 /* Failed. Add the ones that has been removed back */
657 if (cnt > 0)
658 sctp_bindx_add(sk, addrs, cnt);
659 return retval;
660 }
661 }
662
663 return retval;
664}
665
666/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
667 * the associations that are part of the endpoint indicating that a list of
668 * local addresses are removed from the endpoint.
669 *
670 * If any of the addresses is already in the bind address list of the
671 * association, we do not send the chunk for that association. But it will not
672 * affect other associations.
673 *
674 * Only sctp_setsockopt_bindx() is supposed to call this function.
675 */
676static int sctp_send_asconf_del_ip(struct sock *sk,
677 struct sockaddr *addrs,
678 int addrcnt)
679{
680 struct sctp_sock *sp;
681 struct sctp_endpoint *ep;
682 struct sctp_association *asoc;
683 struct sctp_bind_addr *bp;
684 struct sctp_chunk *chunk;
685 union sctp_addr *laddr;
686 void *addr_buf;
687 struct sctp_af *af;
688 struct list_head *pos;
689 int i;
690 int retval = 0;
691
692 if (!sctp_addip_enable)
693 return retval;
694
695 sp = sctp_sk(sk);
696 ep = sp->ep;
697
698 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
699 __FUNCTION__, sk, addrs, addrcnt);
700
701 list_for_each(pos, &ep->asocs) {
702 asoc = list_entry(pos, struct sctp_association, asocs);
703
704 if (!asoc->peer.asconf_capable)
705 continue;
706
707 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
708 continue;
709
710 if (!sctp_state(asoc, ESTABLISHED))
711 continue;
712
713 /* Check if any address in the packed array of addresses is
714 * not present in the bind address list of the association.
715 * If so, do not send the asconf chunk to its peer, but
716 * continue with other associations.
717 */
718 addr_buf = addrs;
719 for (i = 0; i < addrcnt; i++) {
720 laddr = (union sctp_addr *)addr_buf;
721 af = sctp_get_af_specific(laddr->v4.sin_family);
722 if (!af) {
723 retval = -EINVAL;
724 goto out;
725 }
726
727 if (!sctp_assoc_lookup_laddr(asoc, laddr))
728 break;
729
730 addr_buf += af->sockaddr_len;
731 }
732 if (i < addrcnt)
733 continue;
734
735 /* Find one address in the association's bind address list
736 * that is not in the packed array of addresses. This is to
737 * make sure that we do not delete all the addresses in the
738 * association.
739 */
740 sctp_read_lock(&asoc->base.addr_lock);
741 bp = &asoc->base.bind_addr;
742 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
743 addrcnt, sp);
744 sctp_read_unlock(&asoc->base.addr_lock);
745 if (!laddr)
746 continue;
747
748 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
749 SCTP_PARAM_DEL_IP);
750 if (!chunk) {
751 retval = -ENOMEM;
752 goto out;
753 }
754
755 retval = sctp_send_asconf(asoc, chunk);
756
757 /* FIXME: After sending the delete address ASCONF chunk, we
758 * cannot remove the addresses from the association's bind
759 * address list, because there maybe some packet send to
760 * the delete addresses, so we should wait until ASCONF_ACK
761 * packet is received.
762 */
763 }
764out:
765 return retval;
766}
767
768/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
769 *
770 * API 8.1
771 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
772 * int flags);
773 *
774 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
775 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
776 * or IPv6 addresses.
777 *
778 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
779 * Section 3.1.2 for this usage.
780 *
781 * addrs is a pointer to an array of one or more socket addresses. Each
782 * address is contained in its appropriate structure (i.e. struct
783 * sockaddr_in or struct sockaddr_in6) the family of the address type
784 * must be used to distengish the address length (note that this
785 * representation is termed a "packed array" of addresses). The caller
786 * specifies the number of addresses in the array with addrcnt.
787 *
788 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
789 * -1, and sets errno to the appropriate error code.
790 *
791 * For SCTP, the port given in each socket address must be the same, or
792 * sctp_bindx() will fail, setting errno to EINVAL.
793 *
794 * The flags parameter is formed from the bitwise OR of zero or more of
795 * the following currently defined flags:
796 *
797 * SCTP_BINDX_ADD_ADDR
798 *
799 * SCTP_BINDX_REM_ADDR
800 *
801 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
802 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
803 * addresses from the association. The two flags are mutually exclusive;
804 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
805 * not remove all addresses from an association; sctp_bindx() will
806 * reject such an attempt with EINVAL.
807 *
808 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
809 * additional addresses with an endpoint after calling bind(). Or use
810 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
811 * socket is associated with so that no new association accepted will be
812 * associated with those addresses. If the endpoint supports dynamic
813 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
814 * endpoint to send the appropriate message to the peer to change the
815 * peers address lists.
816 *
817 * Adding and removing addresses from a connected association is
818 * optional functionality. Implementations that do not support this
819 * functionality should return EOPNOTSUPP.
820 *
821 * Basically do nothing but copying the addresses from user to kernel
822 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700823 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
824 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 *
826 * We don't use copy_from_user() for optimization: we first do the
827 * sanity checks (buffer size -fast- and access check-healthy
828 * pointer); if all of those succeed, then we can alloc the memory
829 * (expensive operation) needed to copy the data to kernel. Then we do
830 * the copying without checking the user space area
831 * (__copy_from_user()).
832 *
833 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
834 * it.
835 *
836 * sk The sk of the socket
837 * addrs The pointer to the addresses in user land
838 * addrssize Size of the addrs buffer
839 * op Operation to perform (add or remove, see the flags of
840 * sctp_bindx)
841 *
842 * Returns 0 if ok, <0 errno code on error.
843 */
844SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
845 struct sockaddr __user *addrs,
846 int addrs_size, int op)
847{
848 struct sockaddr *kaddrs;
849 int err;
850 int addrcnt = 0;
851 int walk_size = 0;
852 struct sockaddr *sa_addr;
853 void *addr_buf;
854 struct sctp_af *af;
855
856 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
857 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
858
859 if (unlikely(addrs_size <= 0))
860 return -EINVAL;
861
862 /* Check the user passed a healthy pointer. */
863 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
864 return -EFAULT;
865
866 /* Alloc space for the address array in kernel memory. */
867 kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL);
868 if (unlikely(!kaddrs))
869 return -ENOMEM;
870
871 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
872 kfree(kaddrs);
873 return -EFAULT;
874 }
875
876 /* Walk through the addrs buffer and count the number of addresses. */
877 addr_buf = kaddrs;
878 while (walk_size < addrs_size) {
879 sa_addr = (struct sockaddr *)addr_buf;
880 af = sctp_get_af_specific(sa_addr->sa_family);
881
882 /* If the address family is not supported or if this address
883 * causes the address buffer to overflow return EINVAL.
884 */
885 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
886 kfree(kaddrs);
887 return -EINVAL;
888 }
889 addrcnt++;
890 addr_buf += af->sockaddr_len;
891 walk_size += af->sockaddr_len;
892 }
893
894 /* Do the work. */
895 switch (op) {
896 case SCTP_BINDX_ADD_ADDR:
897 err = sctp_bindx_add(sk, kaddrs, addrcnt);
898 if (err)
899 goto out;
900 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
901 break;
902
903 case SCTP_BINDX_REM_ADDR:
904 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
905 if (err)
906 goto out;
907 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
908 break;
909
910 default:
911 err = -EINVAL;
912 break;
913 };
914
915out:
916 kfree(kaddrs);
917
918 return err;
919}
920
Frank Filz3f7a87d2005-06-20 13:14:57 -0700921/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
922 *
923 * Common routine for handling connect() and sctp_connectx().
924 * Connect will come in with just a single address.
925 */
926static int __sctp_connect(struct sock* sk,
927 struct sockaddr *kaddrs,
928 int addrs_size)
929{
930 struct sctp_sock *sp;
931 struct sctp_endpoint *ep;
932 struct sctp_association *asoc = NULL;
933 struct sctp_association *asoc2;
934 struct sctp_transport *transport;
935 union sctp_addr to;
936 struct sctp_af *af;
937 sctp_scope_t scope;
938 long timeo;
939 int err = 0;
940 int addrcnt = 0;
941 int walk_size = 0;
942 struct sockaddr *sa_addr;
943 void *addr_buf;
944
945 sp = sctp_sk(sk);
946 ep = sp->ep;
947
948 /* connect() cannot be done on a socket that is already in ESTABLISHED
949 * state - UDP-style peeled off socket or a TCP-style socket that
950 * is already connected.
951 * It cannot be done even on a TCP-style listening socket.
952 */
953 if (sctp_sstate(sk, ESTABLISHED) ||
954 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
955 err = -EISCONN;
956 goto out_free;
957 }
958
959 /* Walk through the addrs buffer and count the number of addresses. */
960 addr_buf = kaddrs;
961 while (walk_size < addrs_size) {
962 sa_addr = (struct sockaddr *)addr_buf;
963 af = sctp_get_af_specific(sa_addr->sa_family);
964
965 /* If the address family is not supported or if this address
966 * causes the address buffer to overflow return EINVAL.
967 */
968 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
969 err = -EINVAL;
970 goto out_free;
971 }
972
973 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
974 af->sockaddr_len);
975 if (err)
976 goto out_free;
977
978 memcpy(&to, sa_addr, af->sockaddr_len);
979 to.v4.sin_port = ntohs(to.v4.sin_port);
980
981 /* Check if there already is a matching association on the
982 * endpoint (other than the one created here).
983 */
984 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
985 if (asoc2 && asoc2 != asoc) {
986 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
987 err = -EISCONN;
988 else
989 err = -EALREADY;
990 goto out_free;
991 }
992
993 /* If we could not find a matching association on the endpoint,
994 * make sure that there is no peeled-off association matching
995 * the peer address even on another socket.
996 */
997 if (sctp_endpoint_is_peeled_off(ep, &to)) {
998 err = -EADDRNOTAVAIL;
999 goto out_free;
1000 }
1001
1002 if (!asoc) {
1003 /* If a bind() or sctp_bindx() is not called prior to
1004 * an sctp_connectx() call, the system picks an
1005 * ephemeral port and will choose an address set
1006 * equivalent to binding with a wildcard address.
1007 */
1008 if (!ep->base.bind_addr.port) {
1009 if (sctp_autobind(sk)) {
1010 err = -EAGAIN;
1011 goto out_free;
1012 }
1013 }
1014
1015 scope = sctp_scope(&to);
1016 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1017 if (!asoc) {
1018 err = -ENOMEM;
1019 goto out_free;
1020 }
1021 }
1022
1023 /* Prime the peer's transport structures. */
1024 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1025 SCTP_UNKNOWN);
1026 if (!transport) {
1027 err = -ENOMEM;
1028 goto out_free;
1029 }
1030
1031 addrcnt++;
1032 addr_buf += af->sockaddr_len;
1033 walk_size += af->sockaddr_len;
1034 }
1035
1036 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1037 if (err < 0) {
1038 goto out_free;
1039 }
1040
1041 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1042 if (err < 0) {
1043 goto out_free;
1044 }
1045
1046 /* Initialize sk's dport and daddr for getpeername() */
1047 inet_sk(sk)->dport = htons(asoc->peer.port);
1048 af = sctp_get_af_specific(to.sa.sa_family);
1049 af->to_sk_daddr(&to, sk);
1050
1051 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1052 err = sctp_wait_for_connect(asoc, &timeo);
1053
1054 /* Don't free association on exit. */
1055 asoc = NULL;
1056
1057out_free:
1058
1059 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1060 " kaddrs: %p err: %d\n",
1061 asoc, kaddrs, err);
1062 if (asoc)
1063 sctp_association_free(asoc);
1064 return err;
1065}
1066
1067/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1068 *
1069 * API 8.9
1070 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1071 *
1072 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1073 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1074 * or IPv6 addresses.
1075 *
1076 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1077 * Section 3.1.2 for this usage.
1078 *
1079 * addrs is a pointer to an array of one or more socket addresses. Each
1080 * address is contained in its appropriate structure (i.e. struct
1081 * sockaddr_in or struct sockaddr_in6) the family of the address type
1082 * must be used to distengish the address length (note that this
1083 * representation is termed a "packed array" of addresses). The caller
1084 * specifies the number of addresses in the array with addrcnt.
1085 *
1086 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1087 * -1, and sets errno to the appropriate error code.
1088 *
1089 * For SCTP, the port given in each socket address must be the same, or
1090 * sctp_connectx() will fail, setting errno to EINVAL.
1091 *
1092 * An application can use sctp_connectx to initiate an association with
1093 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1094 * allows a caller to specify multiple addresses at which a peer can be
1095 * reached. The way the SCTP stack uses the list of addresses to set up
1096 * the association is implementation dependant. This function only
1097 * specifies that the stack will try to make use of all the addresses in
1098 * the list when needed.
1099 *
1100 * Note that the list of addresses passed in is only used for setting up
1101 * the association. It does not necessarily equal the set of addresses
1102 * the peer uses for the resulting association. If the caller wants to
1103 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1104 * retrieve them after the association has been set up.
1105 *
1106 * Basically do nothing but copying the addresses from user to kernel
1107 * land and invoking either sctp_connectx(). This is used for tunneling
1108 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1109 *
1110 * We don't use copy_from_user() for optimization: we first do the
1111 * sanity checks (buffer size -fast- and access check-healthy
1112 * pointer); if all of those succeed, then we can alloc the memory
1113 * (expensive operation) needed to copy the data to kernel. Then we do
1114 * the copying without checking the user space area
1115 * (__copy_from_user()).
1116 *
1117 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1118 * it.
1119 *
1120 * sk The sk of the socket
1121 * addrs The pointer to the addresses in user land
1122 * addrssize Size of the addrs buffer
1123 *
1124 * Returns 0 if ok, <0 errno code on error.
1125 */
1126SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1127 struct sockaddr __user *addrs,
1128 int addrs_size)
1129{
1130 int err = 0;
1131 struct sockaddr *kaddrs;
1132
1133 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1134 __FUNCTION__, sk, addrs, addrs_size);
1135
1136 if (unlikely(addrs_size <= 0))
1137 return -EINVAL;
1138
1139 /* Check the user passed a healthy pointer. */
1140 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1141 return -EFAULT;
1142
1143 /* Alloc space for the address array in kernel memory. */
1144 kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL);
1145 if (unlikely(!kaddrs))
1146 return -ENOMEM;
1147
1148 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1149 err = -EFAULT;
1150 } else {
1151 err = __sctp_connect(sk, kaddrs, addrs_size);
1152 }
1153
1154 kfree(kaddrs);
1155 return err;
1156}
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158/* API 3.1.4 close() - UDP Style Syntax
1159 * Applications use close() to perform graceful shutdown (as described in
1160 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1161 * by a UDP-style socket.
1162 *
1163 * The syntax is
1164 *
1165 * ret = close(int sd);
1166 *
1167 * sd - the socket descriptor of the associations to be closed.
1168 *
1169 * To gracefully shutdown a specific association represented by the
1170 * UDP-style socket, an application should use the sendmsg() call,
1171 * passing no user data, but including the appropriate flag in the
1172 * ancillary data (see Section xxxx).
1173 *
1174 * If sd in the close() call is a branched-off socket representing only
1175 * one association, the shutdown is performed on that association only.
1176 *
1177 * 4.1.6 close() - TCP Style Syntax
1178 *
1179 * Applications use close() to gracefully close down an association.
1180 *
1181 * The syntax is:
1182 *
1183 * int close(int sd);
1184 *
1185 * sd - the socket descriptor of the association to be closed.
1186 *
1187 * After an application calls close() on a socket descriptor, no further
1188 * socket operations will succeed on that descriptor.
1189 *
1190 * API 7.1.4 SO_LINGER
1191 *
1192 * An application using the TCP-style socket can use this option to
1193 * perform the SCTP ABORT primitive. The linger option structure is:
1194 *
1195 * struct linger {
1196 * int l_onoff; // option on/off
1197 * int l_linger; // linger time
1198 * };
1199 *
1200 * To enable the option, set l_onoff to 1. If the l_linger value is set
1201 * to 0, calling close() is the same as the ABORT primitive. If the
1202 * value is set to a negative value, the setsockopt() call will return
1203 * an error. If the value is set to a positive value linger_time, the
1204 * close() can be blocked for at most linger_time ms. If the graceful
1205 * shutdown phase does not finish during this period, close() will
1206 * return but the graceful shutdown phase continues in the system.
1207 */
1208SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1209{
1210 struct sctp_endpoint *ep;
1211 struct sctp_association *asoc;
1212 struct list_head *pos, *temp;
1213
1214 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1215
1216 sctp_lock_sock(sk);
1217 sk->sk_shutdown = SHUTDOWN_MASK;
1218
1219 ep = sctp_sk(sk)->ep;
1220
1221 /* Walk all associations on a socket, not on an endpoint. */
1222 list_for_each_safe(pos, temp, &ep->asocs) {
1223 asoc = list_entry(pos, struct sctp_association, asocs);
1224
1225 if (sctp_style(sk, TCP)) {
1226 /* A closed association can still be in the list if
1227 * it belongs to a TCP-style listening socket that is
1228 * not yet accepted. If so, free it. If not, send an
1229 * ABORT or SHUTDOWN based on the linger options.
1230 */
1231 if (sctp_state(asoc, CLOSED)) {
1232 sctp_unhash_established(asoc);
1233 sctp_association_free(asoc);
1234
1235 } else if (sock_flag(sk, SOCK_LINGER) &&
1236 !sk->sk_lingertime)
1237 sctp_primitive_ABORT(asoc, NULL);
1238 else
1239 sctp_primitive_SHUTDOWN(asoc, NULL);
1240 } else
1241 sctp_primitive_SHUTDOWN(asoc, NULL);
1242 }
1243
1244 /* Clean up any skbs sitting on the receive queue. */
1245 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1246 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1247
1248 /* On a TCP-style socket, block for at most linger_time if set. */
1249 if (sctp_style(sk, TCP) && timeout)
1250 sctp_wait_for_close(sk, timeout);
1251
1252 /* This will run the backlog queue. */
1253 sctp_release_sock(sk);
1254
1255 /* Supposedly, no process has access to the socket, but
1256 * the net layers still may.
1257 */
1258 sctp_local_bh_disable();
1259 sctp_bh_lock_sock(sk);
1260
1261 /* Hold the sock, since sk_common_release() will put sock_put()
1262 * and we have just a little more cleanup.
1263 */
1264 sock_hold(sk);
1265 sk_common_release(sk);
1266
1267 sctp_bh_unlock_sock(sk);
1268 sctp_local_bh_enable();
1269
1270 sock_put(sk);
1271
1272 SCTP_DBG_OBJCNT_DEC(sock);
1273}
1274
1275/* Handle EPIPE error. */
1276static int sctp_error(struct sock *sk, int flags, int err)
1277{
1278 if (err == -EPIPE)
1279 err = sock_error(sk) ? : -EPIPE;
1280 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1281 send_sig(SIGPIPE, current, 0);
1282 return err;
1283}
1284
1285/* API 3.1.3 sendmsg() - UDP Style Syntax
1286 *
1287 * An application uses sendmsg() and recvmsg() calls to transmit data to
1288 * and receive data from its peer.
1289 *
1290 * ssize_t sendmsg(int socket, const struct msghdr *message,
1291 * int flags);
1292 *
1293 * socket - the socket descriptor of the endpoint.
1294 * message - pointer to the msghdr structure which contains a single
1295 * user message and possibly some ancillary data.
1296 *
1297 * See Section 5 for complete description of the data
1298 * structures.
1299 *
1300 * flags - flags sent or received with the user message, see Section
1301 * 5 for complete description of the flags.
1302 *
1303 * Note: This function could use a rewrite especially when explicit
1304 * connect support comes in.
1305 */
1306/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1307
1308SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1309
1310SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1311 struct msghdr *msg, size_t msg_len)
1312{
1313 struct sctp_sock *sp;
1314 struct sctp_endpoint *ep;
1315 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1316 struct sctp_transport *transport, *chunk_tp;
1317 struct sctp_chunk *chunk;
1318 union sctp_addr to;
1319 struct sockaddr *msg_name = NULL;
1320 struct sctp_sndrcvinfo default_sinfo = { 0 };
1321 struct sctp_sndrcvinfo *sinfo;
1322 struct sctp_initmsg *sinit;
1323 sctp_assoc_t associd = 0;
1324 sctp_cmsgs_t cmsgs = { NULL };
1325 int err;
1326 sctp_scope_t scope;
1327 long timeo;
1328 __u16 sinfo_flags = 0;
1329 struct sctp_datamsg *datamsg;
1330 struct list_head *pos;
1331 int msg_flags = msg->msg_flags;
1332
1333 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1334 sk, msg, msg_len);
1335
1336 err = 0;
1337 sp = sctp_sk(sk);
1338 ep = sp->ep;
1339
Frank Filz3f7a87d2005-06-20 13:14:57 -07001340 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341
1342 /* We cannot send a message over a TCP-style listening socket. */
1343 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1344 err = -EPIPE;
1345 goto out_nounlock;
1346 }
1347
1348 /* Parse out the SCTP CMSGs. */
1349 err = sctp_msghdr_parse(msg, &cmsgs);
1350
1351 if (err) {
1352 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1353 goto out_nounlock;
1354 }
1355
1356 /* Fetch the destination address for this packet. This
1357 * address only selects the association--it is not necessarily
1358 * the address we will send to.
1359 * For a peeled-off socket, msg_name is ignored.
1360 */
1361 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1362 int msg_namelen = msg->msg_namelen;
1363
1364 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1365 msg_namelen);
1366 if (err)
1367 return err;
1368
1369 if (msg_namelen > sizeof(to))
1370 msg_namelen = sizeof(to);
1371 memcpy(&to, msg->msg_name, msg_namelen);
1372 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1373 "0x%x:%u.\n",
1374 to.v4.sin_addr.s_addr, to.v4.sin_port);
1375
1376 to.v4.sin_port = ntohs(to.v4.sin_port);
1377 msg_name = msg->msg_name;
1378 }
1379
1380 sinfo = cmsgs.info;
1381 sinit = cmsgs.init;
1382
1383 /* Did the user specify SNDRCVINFO? */
1384 if (sinfo) {
1385 sinfo_flags = sinfo->sinfo_flags;
1386 associd = sinfo->sinfo_assoc_id;
1387 }
1388
1389 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1390 msg_len, sinfo_flags);
1391
1392 /* MSG_EOF or MSG_ABORT cannot be set on a TCP-style socket. */
1393 if (sctp_style(sk, TCP) && (sinfo_flags & (MSG_EOF | MSG_ABORT))) {
1394 err = -EINVAL;
1395 goto out_nounlock;
1396 }
1397
1398 /* If MSG_EOF is set, no data can be sent. Disallow sending zero
1399 * length messages when MSG_EOF|MSG_ABORT is not set.
1400 * If MSG_ABORT is set, the message length could be non zero with
1401 * the msg_iov set to the user abort reason.
1402 */
1403 if (((sinfo_flags & MSG_EOF) && (msg_len > 0)) ||
1404 (!(sinfo_flags & (MSG_EOF|MSG_ABORT)) && (msg_len == 0))) {
1405 err = -EINVAL;
1406 goto out_nounlock;
1407 }
1408
1409 /* If MSG_ADDR_OVER is set, there must be an address
1410 * specified in msg_name.
1411 */
1412 if ((sinfo_flags & MSG_ADDR_OVER) && (!msg->msg_name)) {
1413 err = -EINVAL;
1414 goto out_nounlock;
1415 }
1416
1417 transport = NULL;
1418
1419 SCTP_DEBUG_PRINTK("About to look up association.\n");
1420
1421 sctp_lock_sock(sk);
1422
1423 /* If a msg_name has been specified, assume this is to be used. */
1424 if (msg_name) {
1425 /* Look for a matching association on the endpoint. */
1426 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1427 if (!asoc) {
1428 /* If we could not find a matching association on the
1429 * endpoint, make sure that it is not a TCP-style
1430 * socket that already has an association or there is
1431 * no peeled-off association on another socket.
1432 */
1433 if ((sctp_style(sk, TCP) &&
1434 sctp_sstate(sk, ESTABLISHED)) ||
1435 sctp_endpoint_is_peeled_off(ep, &to)) {
1436 err = -EADDRNOTAVAIL;
1437 goto out_unlock;
1438 }
1439 }
1440 } else {
1441 asoc = sctp_id2assoc(sk, associd);
1442 if (!asoc) {
1443 err = -EPIPE;
1444 goto out_unlock;
1445 }
1446 }
1447
1448 if (asoc) {
1449 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1450
1451 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1452 * socket that has an association in CLOSED state. This can
1453 * happen when an accepted socket has an association that is
1454 * already CLOSED.
1455 */
1456 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1457 err = -EPIPE;
1458 goto out_unlock;
1459 }
1460
1461 if (sinfo_flags & MSG_EOF) {
1462 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1463 asoc);
1464 sctp_primitive_SHUTDOWN(asoc, NULL);
1465 err = 0;
1466 goto out_unlock;
1467 }
1468 if (sinfo_flags & MSG_ABORT) {
1469 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1470 sctp_primitive_ABORT(asoc, msg);
1471 err = 0;
1472 goto out_unlock;
1473 }
1474 }
1475
1476 /* Do we need to create the association? */
1477 if (!asoc) {
1478 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1479
1480 if (sinfo_flags & (MSG_EOF | MSG_ABORT)) {
1481 err = -EINVAL;
1482 goto out_unlock;
1483 }
1484
1485 /* Check for invalid stream against the stream counts,
1486 * either the default or the user specified stream counts.
1487 */
1488 if (sinfo) {
1489 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1490 /* Check against the defaults. */
1491 if (sinfo->sinfo_stream >=
1492 sp->initmsg.sinit_num_ostreams) {
1493 err = -EINVAL;
1494 goto out_unlock;
1495 }
1496 } else {
1497 /* Check against the requested. */
1498 if (sinfo->sinfo_stream >=
1499 sinit->sinit_num_ostreams) {
1500 err = -EINVAL;
1501 goto out_unlock;
1502 }
1503 }
1504 }
1505
1506 /*
1507 * API 3.1.2 bind() - UDP Style Syntax
1508 * If a bind() or sctp_bindx() is not called prior to a
1509 * sendmsg() call that initiates a new association, the
1510 * system picks an ephemeral port and will choose an address
1511 * set equivalent to binding with a wildcard address.
1512 */
1513 if (!ep->base.bind_addr.port) {
1514 if (sctp_autobind(sk)) {
1515 err = -EAGAIN;
1516 goto out_unlock;
1517 }
1518 }
1519
1520 scope = sctp_scope(&to);
1521 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1522 if (!new_asoc) {
1523 err = -ENOMEM;
1524 goto out_unlock;
1525 }
1526 asoc = new_asoc;
1527
1528 /* If the SCTP_INIT ancillary data is specified, set all
1529 * the association init values accordingly.
1530 */
1531 if (sinit) {
1532 if (sinit->sinit_num_ostreams) {
1533 asoc->c.sinit_num_ostreams =
1534 sinit->sinit_num_ostreams;
1535 }
1536 if (sinit->sinit_max_instreams) {
1537 asoc->c.sinit_max_instreams =
1538 sinit->sinit_max_instreams;
1539 }
1540 if (sinit->sinit_max_attempts) {
1541 asoc->max_init_attempts
1542 = sinit->sinit_max_attempts;
1543 }
1544 if (sinit->sinit_max_init_timeo) {
1545 asoc->max_init_timeo =
1546 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1547 }
1548 }
1549
1550 /* Prime the peer's transport structures. */
Frank Filz3f7a87d2005-06-20 13:14:57 -07001551 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 if (!transport) {
1553 err = -ENOMEM;
1554 goto out_free;
1555 }
1556 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1557 if (err < 0) {
1558 err = -ENOMEM;
1559 goto out_free;
1560 }
1561 }
1562
1563 /* ASSERT: we have a valid association at this point. */
1564 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1565
1566 if (!sinfo) {
1567 /* If the user didn't specify SNDRCVINFO, make up one with
1568 * some defaults.
1569 */
1570 default_sinfo.sinfo_stream = asoc->default_stream;
1571 default_sinfo.sinfo_flags = asoc->default_flags;
1572 default_sinfo.sinfo_ppid = asoc->default_ppid;
1573 default_sinfo.sinfo_context = asoc->default_context;
1574 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1575 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1576 sinfo = &default_sinfo;
1577 }
1578
1579 /* API 7.1.7, the sndbuf size per association bounds the
1580 * maximum size of data that can be sent in a single send call.
1581 */
1582 if (msg_len > sk->sk_sndbuf) {
1583 err = -EMSGSIZE;
1584 goto out_free;
1585 }
1586
1587 /* If fragmentation is disabled and the message length exceeds the
1588 * association fragmentation point, return EMSGSIZE. The I-D
1589 * does not specify what this error is, but this looks like
1590 * a great fit.
1591 */
1592 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1593 err = -EMSGSIZE;
1594 goto out_free;
1595 }
1596
1597 if (sinfo) {
1598 /* Check for invalid stream. */
1599 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1600 err = -EINVAL;
1601 goto out_free;
1602 }
1603 }
1604
1605 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1606 if (!sctp_wspace(asoc)) {
1607 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1608 if (err)
1609 goto out_free;
1610 }
1611
1612 /* If an address is passed with the sendto/sendmsg call, it is used
1613 * to override the primary destination address in the TCP model, or
1614 * when MSG_ADDR_OVER flag is set in the UDP model.
1615 */
1616 if ((sctp_style(sk, TCP) && msg_name) ||
1617 (sinfo_flags & MSG_ADDR_OVER)) {
1618 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1619 if (!chunk_tp) {
1620 err = -EINVAL;
1621 goto out_free;
1622 }
1623 } else
1624 chunk_tp = NULL;
1625
1626 /* Auto-connect, if we aren't connected already. */
1627 if (sctp_state(asoc, CLOSED)) {
1628 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1629 if (err < 0)
1630 goto out_free;
1631 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1632 }
1633
1634 /* Break the message into multiple chunks of maximum size. */
1635 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1636 if (!datamsg) {
1637 err = -ENOMEM;
1638 goto out_free;
1639 }
1640
1641 /* Now send the (possibly) fragmented message. */
1642 list_for_each(pos, &datamsg->chunks) {
1643 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1644 sctp_datamsg_track(chunk);
1645
1646 /* Do accounting for the write space. */
1647 sctp_set_owner_w(chunk);
1648
1649 chunk->transport = chunk_tp;
1650
1651 /* Send it to the lower layers. Note: all chunks
1652 * must either fail or succeed. The lower layer
1653 * works that way today. Keep it that way or this
1654 * breaks.
1655 */
1656 err = sctp_primitive_SEND(asoc, chunk);
1657 /* Did the lower layer accept the chunk? */
1658 if (err)
1659 sctp_chunk_free(chunk);
1660 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1661 }
1662
1663 sctp_datamsg_free(datamsg);
1664 if (err)
1665 goto out_free;
1666 else
1667 err = msg_len;
1668
1669 /* If we are already past ASSOCIATE, the lower
1670 * layers are responsible for association cleanup.
1671 */
1672 goto out_unlock;
1673
1674out_free:
1675 if (new_asoc)
1676 sctp_association_free(asoc);
1677out_unlock:
1678 sctp_release_sock(sk);
1679
1680out_nounlock:
1681 return sctp_error(sk, msg_flags, err);
1682
1683#if 0
1684do_sock_err:
1685 if (msg_len)
1686 err = msg_len;
1687 else
1688 err = sock_error(sk);
1689 goto out;
1690
1691do_interrupted:
1692 if (msg_len)
1693 err = msg_len;
1694 goto out;
1695#endif /* 0 */
1696}
1697
1698/* This is an extended version of skb_pull() that removes the data from the
1699 * start of a skb even when data is spread across the list of skb's in the
1700 * frag_list. len specifies the total amount of data that needs to be removed.
1701 * when 'len' bytes could be removed from the skb, it returns 0.
1702 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1703 * could not be removed.
1704 */
1705static int sctp_skb_pull(struct sk_buff *skb, int len)
1706{
1707 struct sk_buff *list;
1708 int skb_len = skb_headlen(skb);
1709 int rlen;
1710
1711 if (len <= skb_len) {
1712 __skb_pull(skb, len);
1713 return 0;
1714 }
1715 len -= skb_len;
1716 __skb_pull(skb, skb_len);
1717
1718 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1719 rlen = sctp_skb_pull(list, len);
1720 skb->len -= (len-rlen);
1721 skb->data_len -= (len-rlen);
1722
1723 if (!rlen)
1724 return 0;
1725
1726 len = rlen;
1727 }
1728
1729 return len;
1730}
1731
1732/* API 3.1.3 recvmsg() - UDP Style Syntax
1733 *
1734 * ssize_t recvmsg(int socket, struct msghdr *message,
1735 * int flags);
1736 *
1737 * socket - the socket descriptor of the endpoint.
1738 * message - pointer to the msghdr structure which contains a single
1739 * user message and possibly some ancillary data.
1740 *
1741 * See Section 5 for complete description of the data
1742 * structures.
1743 *
1744 * flags - flags sent or received with the user message, see Section
1745 * 5 for complete description of the flags.
1746 */
1747static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1748
1749SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1750 struct msghdr *msg, size_t len, int noblock,
1751 int flags, int *addr_len)
1752{
1753 struct sctp_ulpevent *event = NULL;
1754 struct sctp_sock *sp = sctp_sk(sk);
1755 struct sk_buff *skb;
1756 int copied;
1757 int err = 0;
1758 int skb_len;
1759
1760 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1761 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1762 "len", len, "knoblauch", noblock,
1763 "flags", flags, "addr_len", addr_len);
1764
1765 sctp_lock_sock(sk);
1766
1767 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1768 err = -ENOTCONN;
1769 goto out;
1770 }
1771
1772 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1773 if (!skb)
1774 goto out;
1775
1776 /* Get the total length of the skb including any skb's in the
1777 * frag_list.
1778 */
1779 skb_len = skb->len;
1780
1781 copied = skb_len;
1782 if (copied > len)
1783 copied = len;
1784
1785 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1786
1787 event = sctp_skb2event(skb);
1788
1789 if (err)
1790 goto out_free;
1791
1792 sock_recv_timestamp(msg, sk, skb);
1793 if (sctp_ulpevent_is_notification(event)) {
1794 msg->msg_flags |= MSG_NOTIFICATION;
1795 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1796 } else {
1797 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1798 }
1799
1800 /* Check if we allow SCTP_SNDRCVINFO. */
1801 if (sp->subscribe.sctp_data_io_event)
1802 sctp_ulpevent_read_sndrcvinfo(event, msg);
1803#if 0
1804 /* FIXME: we should be calling IP/IPv6 layers. */
1805 if (sk->sk_protinfo.af_inet.cmsg_flags)
1806 ip_cmsg_recv(msg, skb);
1807#endif
1808
1809 err = copied;
1810
1811 /* If skb's length exceeds the user's buffer, update the skb and
1812 * push it back to the receive_queue so that the next call to
1813 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1814 */
1815 if (skb_len > copied) {
1816 msg->msg_flags &= ~MSG_EOR;
1817 if (flags & MSG_PEEK)
1818 goto out_free;
1819 sctp_skb_pull(skb, copied);
1820 skb_queue_head(&sk->sk_receive_queue, skb);
1821
1822 /* When only partial message is copied to the user, increase
1823 * rwnd by that amount. If all the data in the skb is read,
1824 * rwnd is updated when the event is freed.
1825 */
1826 sctp_assoc_rwnd_increase(event->asoc, copied);
1827 goto out;
1828 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1829 (event->msg_flags & MSG_EOR))
1830 msg->msg_flags |= MSG_EOR;
1831 else
1832 msg->msg_flags &= ~MSG_EOR;
1833
1834out_free:
1835 if (flags & MSG_PEEK) {
1836 /* Release the skb reference acquired after peeking the skb in
1837 * sctp_skb_recv_datagram().
1838 */
1839 kfree_skb(skb);
1840 } else {
1841 /* Free the event which includes releasing the reference to
1842 * the owner of the skb, freeing the skb and updating the
1843 * rwnd.
1844 */
1845 sctp_ulpevent_free(event);
1846 }
1847out:
1848 sctp_release_sock(sk);
1849 return err;
1850}
1851
1852/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1853 *
1854 * This option is a on/off flag. If enabled no SCTP message
1855 * fragmentation will be performed. Instead if a message being sent
1856 * exceeds the current PMTU size, the message will NOT be sent and
1857 * instead a error will be indicated to the user.
1858 */
1859static int sctp_setsockopt_disable_fragments(struct sock *sk,
1860 char __user *optval, int optlen)
1861{
1862 int val;
1863
1864 if (optlen < sizeof(int))
1865 return -EINVAL;
1866
1867 if (get_user(val, (int __user *)optval))
1868 return -EFAULT;
1869
1870 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1871
1872 return 0;
1873}
1874
1875static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1876 int optlen)
1877{
1878 if (optlen != sizeof(struct sctp_event_subscribe))
1879 return -EINVAL;
1880 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1881 return -EFAULT;
1882 return 0;
1883}
1884
1885/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1886 *
1887 * This socket option is applicable to the UDP-style socket only. When
1888 * set it will cause associations that are idle for more than the
1889 * specified number of seconds to automatically close. An association
1890 * being idle is defined an association that has NOT sent or received
1891 * user data. The special value of '0' indicates that no automatic
1892 * close of any associations should be performed. The option expects an
1893 * integer defining the number of seconds of idle time before an
1894 * association is closed.
1895 */
1896static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1897 int optlen)
1898{
1899 struct sctp_sock *sp = sctp_sk(sk);
1900
1901 /* Applicable to UDP-style socket only */
1902 if (sctp_style(sk, TCP))
1903 return -EOPNOTSUPP;
1904 if (optlen != sizeof(int))
1905 return -EINVAL;
1906 if (copy_from_user(&sp->autoclose, optval, optlen))
1907 return -EFAULT;
1908
1909 sp->ep->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE] = sp->autoclose * HZ;
1910 return 0;
1911}
1912
1913/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1914 *
1915 * Applications can enable or disable heartbeats for any peer address of
1916 * an association, modify an address's heartbeat interval, force a
1917 * heartbeat to be sent immediately, and adjust the address's maximum
1918 * number of retransmissions sent before an address is considered
1919 * unreachable. The following structure is used to access and modify an
1920 * address's parameters:
1921 *
1922 * struct sctp_paddrparams {
1923 * sctp_assoc_t spp_assoc_id;
1924 * struct sockaddr_storage spp_address;
1925 * uint32_t spp_hbinterval;
1926 * uint16_t spp_pathmaxrxt;
1927 * };
1928 *
1929 * spp_assoc_id - (UDP style socket) This is filled in the application,
1930 * and identifies the association for this query.
1931 * spp_address - This specifies which address is of interest.
1932 * spp_hbinterval - This contains the value of the heartbeat interval,
1933 * in milliseconds. A value of 0, when modifying the
1934 * parameter, specifies that the heartbeat on this
1935 * address should be disabled. A value of UINT32_MAX
1936 * (4294967295), when modifying the parameter,
1937 * specifies that a heartbeat should be sent
1938 * immediately to the peer address, and the current
1939 * interval should remain unchanged.
1940 * spp_pathmaxrxt - This contains the maximum number of
1941 * retransmissions before this address shall be
1942 * considered unreachable.
1943 */
1944static int sctp_setsockopt_peer_addr_params(struct sock *sk,
1945 char __user *optval, int optlen)
1946{
1947 struct sctp_paddrparams params;
1948 struct sctp_transport *trans;
1949 int error;
1950
1951 if (optlen != sizeof(struct sctp_paddrparams))
1952 return -EINVAL;
1953 if (copy_from_user(&params, optval, optlen))
1954 return -EFAULT;
1955
1956 /*
1957 * API 7. Socket Options (setting the default value for the endpoint)
1958 * All options that support specific settings on an association by
1959 * filling in either an association id variable or a sockaddr_storage
1960 * SHOULD also support setting of the same value for the entire endpoint
1961 * (i.e. future associations). To accomplish this the following logic is
1962 * used when setting one of these options:
1963
1964 * c) If neither the sockaddr_storage or association identification is
1965 * set i.e. the sockaddr_storage is set to all 0's (INADDR_ANY) and
1966 * the association identification is 0, the settings are a default
1967 * and to be applied to the endpoint (all future associations).
1968 */
1969
1970 /* update default value for endpoint (all future associations) */
1971 if (!params.spp_assoc_id &&
1972 sctp_is_any(( union sctp_addr *)&params.spp_address)) {
1973 /* Manual heartbeat on an endpoint is invalid. */
1974 if (0xffffffff == params.spp_hbinterval)
1975 return -EINVAL;
1976 else if (params.spp_hbinterval)
1977 sctp_sk(sk)->paddrparam.spp_hbinterval =
1978 params.spp_hbinterval;
1979 if (params.spp_pathmaxrxt)
1980 sctp_sk(sk)->paddrparam.spp_pathmaxrxt =
1981 params.spp_pathmaxrxt;
1982 return 0;
1983 }
1984
1985 trans = sctp_addr_id2transport(sk, &params.spp_address,
1986 params.spp_assoc_id);
1987 if (!trans)
1988 return -EINVAL;
1989
1990 /* Applications can enable or disable heartbeats for any peer address
1991 * of an association, modify an address's heartbeat interval, force a
1992 * heartbeat to be sent immediately, and adjust the address's maximum
1993 * number of retransmissions sent before an address is considered
1994 * unreachable.
1995 *
1996 * The value of the heartbeat interval, in milliseconds. A value of
1997 * UINT32_MAX (4294967295), when modifying the parameter, specifies
1998 * that a heartbeat should be sent immediately to the peer address,
1999 * and the current interval should remain unchanged.
2000 */
2001 if (0xffffffff == params.spp_hbinterval) {
2002 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2003 if (error)
2004 return error;
2005 } else {
2006 /* The value of the heartbeat interval, in milliseconds. A value of 0,
2007 * when modifying the parameter, specifies that the heartbeat on this
2008 * address should be disabled.
2009 */
2010 if (params.spp_hbinterval) {
2011 trans->hb_allowed = 1;
2012 trans->hb_interval =
2013 msecs_to_jiffies(params.spp_hbinterval);
2014 } else
2015 trans->hb_allowed = 0;
2016 }
2017
2018 /* spp_pathmaxrxt contains the maximum number of retransmissions
2019 * before this address shall be considered unreachable.
2020 */
2021 if (params.spp_pathmaxrxt)
2022 trans->max_retrans = params.spp_pathmaxrxt;
2023
2024 return 0;
2025}
2026
2027/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2028 *
2029 * Applications can specify protocol parameters for the default association
2030 * initialization. The option name argument to setsockopt() and getsockopt()
2031 * is SCTP_INITMSG.
2032 *
2033 * Setting initialization parameters is effective only on an unconnected
2034 * socket (for UDP-style sockets only future associations are effected
2035 * by the change). With TCP-style sockets, this option is inherited by
2036 * sockets derived from a listener socket.
2037 */
2038static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2039{
2040 struct sctp_initmsg sinit;
2041 struct sctp_sock *sp = sctp_sk(sk);
2042
2043 if (optlen != sizeof(struct sctp_initmsg))
2044 return -EINVAL;
2045 if (copy_from_user(&sinit, optval, optlen))
2046 return -EFAULT;
2047
2048 if (sinit.sinit_num_ostreams)
2049 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2050 if (sinit.sinit_max_instreams)
2051 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2052 if (sinit.sinit_max_attempts)
2053 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2054 if (sinit.sinit_max_init_timeo)
2055 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2056
2057 return 0;
2058}
2059
2060/*
2061 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2062 *
2063 * Applications that wish to use the sendto() system call may wish to
2064 * specify a default set of parameters that would normally be supplied
2065 * through the inclusion of ancillary data. This socket option allows
2066 * such an application to set the default sctp_sndrcvinfo structure.
2067 * The application that wishes to use this socket option simply passes
2068 * in to this call the sctp_sndrcvinfo structure defined in Section
2069 * 5.2.2) The input parameters accepted by this call include
2070 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2071 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2072 * to this call if the caller is using the UDP model.
2073 */
2074static int sctp_setsockopt_default_send_param(struct sock *sk,
2075 char __user *optval, int optlen)
2076{
2077 struct sctp_sndrcvinfo info;
2078 struct sctp_association *asoc;
2079 struct sctp_sock *sp = sctp_sk(sk);
2080
2081 if (optlen != sizeof(struct sctp_sndrcvinfo))
2082 return -EINVAL;
2083 if (copy_from_user(&info, optval, optlen))
2084 return -EFAULT;
2085
2086 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2087 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2088 return -EINVAL;
2089
2090 if (asoc) {
2091 asoc->default_stream = info.sinfo_stream;
2092 asoc->default_flags = info.sinfo_flags;
2093 asoc->default_ppid = info.sinfo_ppid;
2094 asoc->default_context = info.sinfo_context;
2095 asoc->default_timetolive = info.sinfo_timetolive;
2096 } else {
2097 sp->default_stream = info.sinfo_stream;
2098 sp->default_flags = info.sinfo_flags;
2099 sp->default_ppid = info.sinfo_ppid;
2100 sp->default_context = info.sinfo_context;
2101 sp->default_timetolive = info.sinfo_timetolive;
2102 }
2103
2104 return 0;
2105}
2106
2107/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2108 *
2109 * Requests that the local SCTP stack use the enclosed peer address as
2110 * the association primary. The enclosed address must be one of the
2111 * association peer's addresses.
2112 */
2113static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2114 int optlen)
2115{
2116 struct sctp_prim prim;
2117 struct sctp_transport *trans;
2118
2119 if (optlen != sizeof(struct sctp_prim))
2120 return -EINVAL;
2121
2122 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2123 return -EFAULT;
2124
2125 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2126 if (!trans)
2127 return -EINVAL;
2128
2129 sctp_assoc_set_primary(trans->asoc, trans);
2130
2131 return 0;
2132}
2133
2134/*
2135 * 7.1.5 SCTP_NODELAY
2136 *
2137 * Turn on/off any Nagle-like algorithm. This means that packets are
2138 * generally sent as soon as possible and no unnecessary delays are
2139 * introduced, at the cost of more packets in the network. Expects an
2140 * integer boolean flag.
2141 */
2142static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2143 int optlen)
2144{
2145 int val;
2146
2147 if (optlen < sizeof(int))
2148 return -EINVAL;
2149 if (get_user(val, (int __user *)optval))
2150 return -EFAULT;
2151
2152 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2153 return 0;
2154}
2155
2156/*
2157 *
2158 * 7.1.1 SCTP_RTOINFO
2159 *
2160 * The protocol parameters used to initialize and bound retransmission
2161 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2162 * and modify these parameters.
2163 * All parameters are time values, in milliseconds. A value of 0, when
2164 * modifying the parameters, indicates that the current value should not
2165 * be changed.
2166 *
2167 */
2168static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2169 struct sctp_rtoinfo rtoinfo;
2170 struct sctp_association *asoc;
2171
2172 if (optlen != sizeof (struct sctp_rtoinfo))
2173 return -EINVAL;
2174
2175 if (copy_from_user(&rtoinfo, optval, optlen))
2176 return -EFAULT;
2177
2178 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2179
2180 /* Set the values to the specific association */
2181 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2182 return -EINVAL;
2183
2184 if (asoc) {
2185 if (rtoinfo.srto_initial != 0)
2186 asoc->rto_initial =
2187 msecs_to_jiffies(rtoinfo.srto_initial);
2188 if (rtoinfo.srto_max != 0)
2189 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2190 if (rtoinfo.srto_min != 0)
2191 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2192 } else {
2193 /* If there is no association or the association-id = 0
2194 * set the values to the endpoint.
2195 */
2196 struct sctp_sock *sp = sctp_sk(sk);
2197
2198 if (rtoinfo.srto_initial != 0)
2199 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2200 if (rtoinfo.srto_max != 0)
2201 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2202 if (rtoinfo.srto_min != 0)
2203 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2204 }
2205
2206 return 0;
2207}
2208
2209/*
2210 *
2211 * 7.1.2 SCTP_ASSOCINFO
2212 *
2213 * This option is used to tune the the maximum retransmission attempts
2214 * of the association.
2215 * Returns an error if the new association retransmission value is
2216 * greater than the sum of the retransmission value of the peer.
2217 * See [SCTP] for more information.
2218 *
2219 */
2220static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2221{
2222
2223 struct sctp_assocparams assocparams;
2224 struct sctp_association *asoc;
2225
2226 if (optlen != sizeof(struct sctp_assocparams))
2227 return -EINVAL;
2228 if (copy_from_user(&assocparams, optval, optlen))
2229 return -EFAULT;
2230
2231 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2232
2233 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2234 return -EINVAL;
2235
2236 /* Set the values to the specific association */
2237 if (asoc) {
2238 if (assocparams.sasoc_asocmaxrxt != 0)
2239 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2240 if (assocparams.sasoc_cookie_life != 0) {
2241 asoc->cookie_life.tv_sec =
2242 assocparams.sasoc_cookie_life / 1000;
2243 asoc->cookie_life.tv_usec =
2244 (assocparams.sasoc_cookie_life % 1000)
2245 * 1000;
2246 }
2247 } else {
2248 /* Set the values to the endpoint */
2249 struct sctp_sock *sp = sctp_sk(sk);
2250
2251 if (assocparams.sasoc_asocmaxrxt != 0)
2252 sp->assocparams.sasoc_asocmaxrxt =
2253 assocparams.sasoc_asocmaxrxt;
2254 if (assocparams.sasoc_cookie_life != 0)
2255 sp->assocparams.sasoc_cookie_life =
2256 assocparams.sasoc_cookie_life;
2257 }
2258 return 0;
2259}
2260
2261/*
2262 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2263 *
2264 * This socket option is a boolean flag which turns on or off mapped V4
2265 * addresses. If this option is turned on and the socket is type
2266 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2267 * If this option is turned off, then no mapping will be done of V4
2268 * addresses and a user will receive both PF_INET6 and PF_INET type
2269 * addresses on the socket.
2270 */
2271static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2272{
2273 int val;
2274 struct sctp_sock *sp = sctp_sk(sk);
2275
2276 if (optlen < sizeof(int))
2277 return -EINVAL;
2278 if (get_user(val, (int __user *)optval))
2279 return -EFAULT;
2280 if (val)
2281 sp->v4mapped = 1;
2282 else
2283 sp->v4mapped = 0;
2284
2285 return 0;
2286}
2287
2288/*
2289 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2290 *
2291 * This socket option specifies the maximum size to put in any outgoing
2292 * SCTP chunk. If a message is larger than this size it will be
2293 * fragmented by SCTP into the specified size. Note that the underlying
2294 * SCTP implementation may fragment into smaller sized chunks when the
2295 * PMTU of the underlying association is smaller than the value set by
2296 * the user.
2297 */
2298static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2299{
2300 struct sctp_association *asoc;
2301 struct list_head *pos;
2302 struct sctp_sock *sp = sctp_sk(sk);
2303 int val;
2304
2305 if (optlen < sizeof(int))
2306 return -EINVAL;
2307 if (get_user(val, (int __user *)optval))
2308 return -EFAULT;
2309 if ((val < 8) || (val > SCTP_MAX_CHUNK_LEN))
2310 return -EINVAL;
2311 sp->user_frag = val;
2312
2313 if (val) {
2314 /* Update the frag_point of the existing associations. */
2315 list_for_each(pos, &(sp->ep->asocs)) {
2316 asoc = list_entry(pos, struct sctp_association, asocs);
2317 asoc->frag_point = sctp_frag_point(sp, asoc->pmtu);
2318 }
2319 }
2320
2321 return 0;
2322}
2323
2324
2325/*
2326 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2327 *
2328 * Requests that the peer mark the enclosed address as the association
2329 * primary. The enclosed address must be one of the association's
2330 * locally bound addresses. The following structure is used to make a
2331 * set primary request:
2332 */
2333static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2334 int optlen)
2335{
2336 struct sctp_sock *sp;
2337 struct sctp_endpoint *ep;
2338 struct sctp_association *asoc = NULL;
2339 struct sctp_setpeerprim prim;
2340 struct sctp_chunk *chunk;
2341 int err;
2342
2343 sp = sctp_sk(sk);
2344 ep = sp->ep;
2345
2346 if (!sctp_addip_enable)
2347 return -EPERM;
2348
2349 if (optlen != sizeof(struct sctp_setpeerprim))
2350 return -EINVAL;
2351
2352 if (copy_from_user(&prim, optval, optlen))
2353 return -EFAULT;
2354
2355 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2356 if (!asoc)
2357 return -EINVAL;
2358
2359 if (!asoc->peer.asconf_capable)
2360 return -EPERM;
2361
2362 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2363 return -EPERM;
2364
2365 if (!sctp_state(asoc, ESTABLISHED))
2366 return -ENOTCONN;
2367
2368 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2369 return -EADDRNOTAVAIL;
2370
2371 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2372 chunk = sctp_make_asconf_set_prim(asoc,
2373 (union sctp_addr *)&prim.sspp_addr);
2374 if (!chunk)
2375 return -ENOMEM;
2376
2377 err = sctp_send_asconf(asoc, chunk);
2378
2379 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2380
2381 return err;
2382}
2383
2384static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2385 int optlen)
2386{
2387 __u32 val;
2388
2389 if (optlen < sizeof(__u32))
2390 return -EINVAL;
2391 if (copy_from_user(&val, optval, sizeof(__u32)))
2392 return -EFAULT;
2393
2394 sctp_sk(sk)->adaption_ind = val;
2395
2396 return 0;
2397}
2398
2399/* API 6.2 setsockopt(), getsockopt()
2400 *
2401 * Applications use setsockopt() and getsockopt() to set or retrieve
2402 * socket options. Socket options are used to change the default
2403 * behavior of sockets calls. They are described in Section 7.
2404 *
2405 * The syntax is:
2406 *
2407 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2408 * int __user *optlen);
2409 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2410 * int optlen);
2411 *
2412 * sd - the socket descript.
2413 * level - set to IPPROTO_SCTP for all SCTP options.
2414 * optname - the option name.
2415 * optval - the buffer to store the value of the option.
2416 * optlen - the size of the buffer.
2417 */
2418SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2419 char __user *optval, int optlen)
2420{
2421 int retval = 0;
2422
2423 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2424 sk, optname);
2425
2426 /* I can hardly begin to describe how wrong this is. This is
2427 * so broken as to be worse than useless. The API draft
2428 * REALLY is NOT helpful here... I am not convinced that the
2429 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2430 * are at all well-founded.
2431 */
2432 if (level != SOL_SCTP) {
2433 struct sctp_af *af = sctp_sk(sk)->pf->af;
2434 retval = af->setsockopt(sk, level, optname, optval, optlen);
2435 goto out_nounlock;
2436 }
2437
2438 sctp_lock_sock(sk);
2439
2440 switch (optname) {
2441 case SCTP_SOCKOPT_BINDX_ADD:
2442 /* 'optlen' is the size of the addresses buffer. */
2443 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2444 optlen, SCTP_BINDX_ADD_ADDR);
2445 break;
2446
2447 case SCTP_SOCKOPT_BINDX_REM:
2448 /* 'optlen' is the size of the addresses buffer. */
2449 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2450 optlen, SCTP_BINDX_REM_ADDR);
2451 break;
2452
Frank Filz3f7a87d2005-06-20 13:14:57 -07002453 case SCTP_SOCKOPT_CONNECTX:
2454 /* 'optlen' is the size of the addresses buffer. */
2455 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2456 optlen);
2457 break;
2458
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459 case SCTP_DISABLE_FRAGMENTS:
2460 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2461 break;
2462
2463 case SCTP_EVENTS:
2464 retval = sctp_setsockopt_events(sk, optval, optlen);
2465 break;
2466
2467 case SCTP_AUTOCLOSE:
2468 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2469 break;
2470
2471 case SCTP_PEER_ADDR_PARAMS:
2472 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2473 break;
2474
2475 case SCTP_INITMSG:
2476 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2477 break;
2478 case SCTP_DEFAULT_SEND_PARAM:
2479 retval = sctp_setsockopt_default_send_param(sk, optval,
2480 optlen);
2481 break;
2482 case SCTP_PRIMARY_ADDR:
2483 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2484 break;
2485 case SCTP_SET_PEER_PRIMARY_ADDR:
2486 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2487 break;
2488 case SCTP_NODELAY:
2489 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2490 break;
2491 case SCTP_RTOINFO:
2492 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2493 break;
2494 case SCTP_ASSOCINFO:
2495 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2496 break;
2497 case SCTP_I_WANT_MAPPED_V4_ADDR:
2498 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2499 break;
2500 case SCTP_MAXSEG:
2501 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2502 break;
2503 case SCTP_ADAPTION_LAYER:
2504 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2505 break;
2506
2507 default:
2508 retval = -ENOPROTOOPT;
2509 break;
2510 };
2511
2512 sctp_release_sock(sk);
2513
2514out_nounlock:
2515 return retval;
2516}
2517
2518/* API 3.1.6 connect() - UDP Style Syntax
2519 *
2520 * An application may use the connect() call in the UDP model to initiate an
2521 * association without sending data.
2522 *
2523 * The syntax is:
2524 *
2525 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2526 *
2527 * sd: the socket descriptor to have a new association added to.
2528 *
2529 * nam: the address structure (either struct sockaddr_in or struct
2530 * sockaddr_in6 defined in RFC2553 [7]).
2531 *
2532 * len: the size of the address.
2533 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07002534SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535 int addr_len)
2536{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07002538 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002539
2540 sctp_lock_sock(sk);
2541
Frank Filz3f7a87d2005-06-20 13:14:57 -07002542 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2543 __FUNCTION__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002544
Frank Filz3f7a87d2005-06-20 13:14:57 -07002545 /* Validate addr_len before calling common connect/connectx routine. */
2546 af = sctp_get_af_specific(addr->sa_family);
2547 if (!af || addr_len < af->sockaddr_len) {
2548 err = -EINVAL;
2549 } else {
2550 /* Pass correct addr len to common routine (so it knows there
2551 * is only one address being passed.
2552 */
2553 err = __sctp_connect(sk, addr, af->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554 }
2555
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 return err;
2558}
2559
2560/* FIXME: Write comments. */
2561SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2562{
2563 return -EOPNOTSUPP; /* STUB */
2564}
2565
2566/* 4.1.4 accept() - TCP Style Syntax
2567 *
2568 * Applications use accept() call to remove an established SCTP
2569 * association from the accept queue of the endpoint. A new socket
2570 * descriptor will be returned from accept() to represent the newly
2571 * formed association.
2572 */
2573SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2574{
2575 struct sctp_sock *sp;
2576 struct sctp_endpoint *ep;
2577 struct sock *newsk = NULL;
2578 struct sctp_association *asoc;
2579 long timeo;
2580 int error = 0;
2581
2582 sctp_lock_sock(sk);
2583
2584 sp = sctp_sk(sk);
2585 ep = sp->ep;
2586
2587 if (!sctp_style(sk, TCP)) {
2588 error = -EOPNOTSUPP;
2589 goto out;
2590 }
2591
2592 if (!sctp_sstate(sk, LISTENING)) {
2593 error = -EINVAL;
2594 goto out;
2595 }
2596
2597 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2598
2599 error = sctp_wait_for_accept(sk, timeo);
2600 if (error)
2601 goto out;
2602
2603 /* We treat the list of associations on the endpoint as the accept
2604 * queue and pick the first association on the list.
2605 */
2606 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2607
2608 newsk = sp->pf->create_accept_sk(sk, asoc);
2609 if (!newsk) {
2610 error = -ENOMEM;
2611 goto out;
2612 }
2613
2614 /* Populate the fields of the newsk from the oldsk and migrate the
2615 * asoc to the newsk.
2616 */
2617 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2618
2619out:
2620 sctp_release_sock(sk);
2621 *err = error;
2622 return newsk;
2623}
2624
2625/* The SCTP ioctl handler. */
2626SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2627{
2628 return -ENOIOCTLCMD;
2629}
2630
2631/* This is the function which gets called during socket creation to
2632 * initialized the SCTP-specific portion of the sock.
2633 * The sock structure should already be zero-filled memory.
2634 */
2635SCTP_STATIC int sctp_init_sock(struct sock *sk)
2636{
2637 struct sctp_endpoint *ep;
2638 struct sctp_sock *sp;
2639
2640 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
2641
2642 sp = sctp_sk(sk);
2643
2644 /* Initialize the SCTP per socket area. */
2645 switch (sk->sk_type) {
2646 case SOCK_SEQPACKET:
2647 sp->type = SCTP_SOCKET_UDP;
2648 break;
2649 case SOCK_STREAM:
2650 sp->type = SCTP_SOCKET_TCP;
2651 break;
2652 default:
2653 return -ESOCKTNOSUPPORT;
2654 }
2655
2656 /* Initialize default send parameters. These parameters can be
2657 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2658 */
2659 sp->default_stream = 0;
2660 sp->default_ppid = 0;
2661 sp->default_flags = 0;
2662 sp->default_context = 0;
2663 sp->default_timetolive = 0;
2664
2665 /* Initialize default setup parameters. These parameters
2666 * can be modified with the SCTP_INITMSG socket option or
2667 * overridden by the SCTP_INIT CMSG.
2668 */
2669 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
2670 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
2671 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
2672 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
2673
2674 /* Initialize default RTO related parameters. These parameters can
2675 * be modified for with the SCTP_RTOINFO socket option.
2676 */
2677 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
2678 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
2679 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
2680
2681 /* Initialize default association related parameters. These parameters
2682 * can be modified with the SCTP_ASSOCINFO socket option.
2683 */
2684 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
2685 sp->assocparams.sasoc_number_peer_destinations = 0;
2686 sp->assocparams.sasoc_peer_rwnd = 0;
2687 sp->assocparams.sasoc_local_rwnd = 0;
2688 sp->assocparams.sasoc_cookie_life =
2689 jiffies_to_msecs(sctp_valid_cookie_life);
2690
2691 /* Initialize default event subscriptions. By default, all the
2692 * options are off.
2693 */
2694 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
2695
2696 /* Default Peer Address Parameters. These defaults can
2697 * be modified via SCTP_PEER_ADDR_PARAMS
2698 */
2699 sp->paddrparam.spp_hbinterval = jiffies_to_msecs(sctp_hb_interval);
2700 sp->paddrparam.spp_pathmaxrxt = sctp_max_retrans_path;
2701
2702 /* If enabled no SCTP message fragmentation will be performed.
2703 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
2704 */
2705 sp->disable_fragments = 0;
2706
2707 /* Turn on/off any Nagle-like algorithm. */
2708 sp->nodelay = 1;
2709
2710 /* Enable by default. */
2711 sp->v4mapped = 1;
2712
2713 /* Auto-close idle associations after the configured
2714 * number of seconds. A value of 0 disables this
2715 * feature. Configure through the SCTP_AUTOCLOSE socket option,
2716 * for UDP-style sockets only.
2717 */
2718 sp->autoclose = 0;
2719
2720 /* User specified fragmentation limit. */
2721 sp->user_frag = 0;
2722
2723 sp->adaption_ind = 0;
2724
2725 sp->pf = sctp_get_pf_specific(sk->sk_family);
2726
2727 /* Control variables for partial data delivery. */
2728 sp->pd_mode = 0;
2729 skb_queue_head_init(&sp->pd_lobby);
2730
2731 /* Create a per socket endpoint structure. Even if we
2732 * change the data structure relationships, this may still
2733 * be useful for storing pre-connect address information.
2734 */
2735 ep = sctp_endpoint_new(sk, GFP_KERNEL);
2736 if (!ep)
2737 return -ENOMEM;
2738
2739 sp->ep = ep;
2740 sp->hmac = NULL;
2741
2742 SCTP_DBG_OBJCNT_INC(sock);
2743 return 0;
2744}
2745
2746/* Cleanup any SCTP per socket resources. */
2747SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
2748{
2749 struct sctp_endpoint *ep;
2750
2751 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
2752
2753 /* Release our hold on the endpoint. */
2754 ep = sctp_sk(sk)->ep;
2755 sctp_endpoint_free(ep);
2756
2757 return 0;
2758}
2759
2760/* API 4.1.7 shutdown() - TCP Style Syntax
2761 * int shutdown(int socket, int how);
2762 *
2763 * sd - the socket descriptor of the association to be closed.
2764 * how - Specifies the type of shutdown. The values are
2765 * as follows:
2766 * SHUT_RD
2767 * Disables further receive operations. No SCTP
2768 * protocol action is taken.
2769 * SHUT_WR
2770 * Disables further send operations, and initiates
2771 * the SCTP shutdown sequence.
2772 * SHUT_RDWR
2773 * Disables further send and receive operations
2774 * and initiates the SCTP shutdown sequence.
2775 */
2776SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
2777{
2778 struct sctp_endpoint *ep;
2779 struct sctp_association *asoc;
2780
2781 if (!sctp_style(sk, TCP))
2782 return;
2783
2784 if (how & SEND_SHUTDOWN) {
2785 ep = sctp_sk(sk)->ep;
2786 if (!list_empty(&ep->asocs)) {
2787 asoc = list_entry(ep->asocs.next,
2788 struct sctp_association, asocs);
2789 sctp_primitive_SHUTDOWN(asoc, NULL);
2790 }
2791 }
2792}
2793
2794/* 7.2.1 Association Status (SCTP_STATUS)
2795
2796 * Applications can retrieve current status information about an
2797 * association, including association state, peer receiver window size,
2798 * number of unacked data chunks, and number of data chunks pending
2799 * receipt. This information is read-only.
2800 */
2801static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
2802 char __user *optval,
2803 int __user *optlen)
2804{
2805 struct sctp_status status;
2806 struct sctp_association *asoc = NULL;
2807 struct sctp_transport *transport;
2808 sctp_assoc_t associd;
2809 int retval = 0;
2810
2811 if (len != sizeof(status)) {
2812 retval = -EINVAL;
2813 goto out;
2814 }
2815
2816 if (copy_from_user(&status, optval, sizeof(status))) {
2817 retval = -EFAULT;
2818 goto out;
2819 }
2820
2821 associd = status.sstat_assoc_id;
2822 asoc = sctp_id2assoc(sk, associd);
2823 if (!asoc) {
2824 retval = -EINVAL;
2825 goto out;
2826 }
2827
2828 transport = asoc->peer.primary_path;
2829
2830 status.sstat_assoc_id = sctp_assoc2id(asoc);
2831 status.sstat_state = asoc->state;
2832 status.sstat_rwnd = asoc->peer.rwnd;
2833 status.sstat_unackdata = asoc->unack_data;
2834
2835 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
2836 status.sstat_instrms = asoc->c.sinit_max_instreams;
2837 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
2838 status.sstat_fragmentation_point = asoc->frag_point;
2839 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
2840 memcpy(&status.sstat_primary.spinfo_address,
2841 &(transport->ipaddr), sizeof(union sctp_addr));
2842 /* Map ipv4 address into v4-mapped-on-v6 address. */
2843 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
2844 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07002845 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002846 status.sstat_primary.spinfo_cwnd = transport->cwnd;
2847 status.sstat_primary.spinfo_srtt = transport->srtt;
2848 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
2849 status.sstat_primary.spinfo_mtu = transport->pmtu;
2850
Frank Filz3f7a87d2005-06-20 13:14:57 -07002851 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
2852 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
2853
Linus Torvalds1da177e2005-04-16 15:20:36 -07002854 if (put_user(len, optlen)) {
2855 retval = -EFAULT;
2856 goto out;
2857 }
2858
2859 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
2860 len, status.sstat_state, status.sstat_rwnd,
2861 status.sstat_assoc_id);
2862
2863 if (copy_to_user(optval, &status, len)) {
2864 retval = -EFAULT;
2865 goto out;
2866 }
2867
2868out:
2869 return (retval);
2870}
2871
2872
2873/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
2874 *
2875 * Applications can retrieve information about a specific peer address
2876 * of an association, including its reachability state, congestion
2877 * window, and retransmission timer values. This information is
2878 * read-only.
2879 */
2880static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
2881 char __user *optval,
2882 int __user *optlen)
2883{
2884 struct sctp_paddrinfo pinfo;
2885 struct sctp_transport *transport;
2886 int retval = 0;
2887
2888 if (len != sizeof(pinfo)) {
2889 retval = -EINVAL;
2890 goto out;
2891 }
2892
2893 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
2894 retval = -EFAULT;
2895 goto out;
2896 }
2897
2898 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
2899 pinfo.spinfo_assoc_id);
2900 if (!transport)
2901 return -EINVAL;
2902
2903 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07002904 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905 pinfo.spinfo_cwnd = transport->cwnd;
2906 pinfo.spinfo_srtt = transport->srtt;
2907 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
2908 pinfo.spinfo_mtu = transport->pmtu;
2909
Frank Filz3f7a87d2005-06-20 13:14:57 -07002910 if (pinfo.spinfo_state == SCTP_UNKNOWN)
2911 pinfo.spinfo_state = SCTP_ACTIVE;
2912
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913 if (put_user(len, optlen)) {
2914 retval = -EFAULT;
2915 goto out;
2916 }
2917
2918 if (copy_to_user(optval, &pinfo, len)) {
2919 retval = -EFAULT;
2920 goto out;
2921 }
2922
2923out:
2924 return (retval);
2925}
2926
2927/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2928 *
2929 * This option is a on/off flag. If enabled no SCTP message
2930 * fragmentation will be performed. Instead if a message being sent
2931 * exceeds the current PMTU size, the message will NOT be sent and
2932 * instead a error will be indicated to the user.
2933 */
2934static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
2935 char __user *optval, int __user *optlen)
2936{
2937 int val;
2938
2939 if (len < sizeof(int))
2940 return -EINVAL;
2941
2942 len = sizeof(int);
2943 val = (sctp_sk(sk)->disable_fragments == 1);
2944 if (put_user(len, optlen))
2945 return -EFAULT;
2946 if (copy_to_user(optval, &val, len))
2947 return -EFAULT;
2948 return 0;
2949}
2950
2951/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
2952 *
2953 * This socket option is used to specify various notifications and
2954 * ancillary data the user wishes to receive.
2955 */
2956static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
2957 int __user *optlen)
2958{
2959 if (len != sizeof(struct sctp_event_subscribe))
2960 return -EINVAL;
2961 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
2962 return -EFAULT;
2963 return 0;
2964}
2965
2966/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2967 *
2968 * This socket option is applicable to the UDP-style socket only. When
2969 * set it will cause associations that are idle for more than the
2970 * specified number of seconds to automatically close. An association
2971 * being idle is defined an association that has NOT sent or received
2972 * user data. The special value of '0' indicates that no automatic
2973 * close of any associations should be performed. The option expects an
2974 * integer defining the number of seconds of idle time before an
2975 * association is closed.
2976 */
2977static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
2978{
2979 /* Applicable to UDP-style socket only */
2980 if (sctp_style(sk, TCP))
2981 return -EOPNOTSUPP;
2982 if (len != sizeof(int))
2983 return -EINVAL;
2984 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
2985 return -EFAULT;
2986 return 0;
2987}
2988
2989/* Helper routine to branch off an association to a new socket. */
2990SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
2991 struct socket **sockp)
2992{
2993 struct sock *sk = asoc->base.sk;
2994 struct socket *sock;
2995 int err = 0;
2996
2997 /* An association cannot be branched off from an already peeled-off
2998 * socket, nor is this supported for tcp style sockets.
2999 */
3000 if (!sctp_style(sk, UDP))
3001 return -EINVAL;
3002
3003 /* Create a new socket. */
3004 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3005 if (err < 0)
3006 return err;
3007
3008 /* Populate the fields of the newsk from the oldsk and migrate the
3009 * asoc to the newsk.
3010 */
3011 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3012 *sockp = sock;
3013
3014 return err;
3015}
3016
3017static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3018{
3019 sctp_peeloff_arg_t peeloff;
3020 struct socket *newsock;
3021 int retval = 0;
3022 struct sctp_association *asoc;
3023
3024 if (len != sizeof(sctp_peeloff_arg_t))
3025 return -EINVAL;
3026 if (copy_from_user(&peeloff, optval, len))
3027 return -EFAULT;
3028
3029 asoc = sctp_id2assoc(sk, peeloff.associd);
3030 if (!asoc) {
3031 retval = -EINVAL;
3032 goto out;
3033 }
3034
3035 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3036
3037 retval = sctp_do_peeloff(asoc, &newsock);
3038 if (retval < 0)
3039 goto out;
3040
3041 /* Map the socket to an unused fd that can be returned to the user. */
3042 retval = sock_map_fd(newsock);
3043 if (retval < 0) {
3044 sock_release(newsock);
3045 goto out;
3046 }
3047
3048 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3049 __FUNCTION__, sk, asoc, newsock->sk, retval);
3050
3051 /* Return the fd mapped to the new socket. */
3052 peeloff.sd = retval;
3053 if (copy_to_user(optval, &peeloff, len))
3054 retval = -EFAULT;
3055
3056out:
3057 return retval;
3058}
3059
3060/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3061 *
3062 * Applications can enable or disable heartbeats for any peer address of
3063 * an association, modify an address's heartbeat interval, force a
3064 * heartbeat to be sent immediately, and adjust the address's maximum
3065 * number of retransmissions sent before an address is considered
3066 * unreachable. The following structure is used to access and modify an
3067 * address's parameters:
3068 *
3069 * struct sctp_paddrparams {
3070 * sctp_assoc_t spp_assoc_id;
3071 * struct sockaddr_storage spp_address;
3072 * uint32_t spp_hbinterval;
3073 * uint16_t spp_pathmaxrxt;
3074 * };
3075 *
3076 * spp_assoc_id - (UDP style socket) This is filled in the application,
3077 * and identifies the association for this query.
3078 * spp_address - This specifies which address is of interest.
3079 * spp_hbinterval - This contains the value of the heartbeat interval,
3080 * in milliseconds. A value of 0, when modifying the
3081 * parameter, specifies that the heartbeat on this
3082 * address should be disabled. A value of UINT32_MAX
3083 * (4294967295), when modifying the parameter,
3084 * specifies that a heartbeat should be sent
3085 * immediately to the peer address, and the current
3086 * interval should remain unchanged.
3087 * spp_pathmaxrxt - This contains the maximum number of
3088 * retransmissions before this address shall be
3089 * considered unreachable.
3090 */
3091static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3092 char __user *optval, int __user *optlen)
3093{
3094 struct sctp_paddrparams params;
3095 struct sctp_transport *trans;
3096
3097 if (len != sizeof(struct sctp_paddrparams))
3098 return -EINVAL;
3099 if (copy_from_user(&params, optval, len))
3100 return -EFAULT;
3101
3102 /* If no association id is specified retrieve the default value
3103 * for the endpoint that will be used for all future associations
3104 */
3105 if (!params.spp_assoc_id &&
3106 sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3107 params.spp_hbinterval = sctp_sk(sk)->paddrparam.spp_hbinterval;
3108 params.spp_pathmaxrxt = sctp_sk(sk)->paddrparam.spp_pathmaxrxt;
3109
3110 goto done;
3111 }
3112
3113 trans = sctp_addr_id2transport(sk, &params.spp_address,
3114 params.spp_assoc_id);
3115 if (!trans)
3116 return -EINVAL;
3117
3118 /* The value of the heartbeat interval, in milliseconds. A value of 0,
3119 * when modifying the parameter, specifies that the heartbeat on this
3120 * address should be disabled.
3121 */
3122 if (!trans->hb_allowed)
3123 params.spp_hbinterval = 0;
3124 else
3125 params.spp_hbinterval = jiffies_to_msecs(trans->hb_interval);
3126
3127 /* spp_pathmaxrxt contains the maximum number of retransmissions
3128 * before this address shall be considered unreachable.
3129 */
3130 params.spp_pathmaxrxt = trans->max_retrans;
3131
3132done:
3133 if (copy_to_user(optval, &params, len))
3134 return -EFAULT;
3135
3136 if (put_user(len, optlen))
3137 return -EFAULT;
3138
3139 return 0;
3140}
3141
3142/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3143 *
3144 * Applications can specify protocol parameters for the default association
3145 * initialization. The option name argument to setsockopt() and getsockopt()
3146 * is SCTP_INITMSG.
3147 *
3148 * Setting initialization parameters is effective only on an unconnected
3149 * socket (for UDP-style sockets only future associations are effected
3150 * by the change). With TCP-style sockets, this option is inherited by
3151 * sockets derived from a listener socket.
3152 */
3153static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3154{
3155 if (len != sizeof(struct sctp_initmsg))
3156 return -EINVAL;
3157 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3158 return -EFAULT;
3159 return 0;
3160}
3161
3162static int sctp_getsockopt_peer_addrs_num(struct sock *sk, int len,
3163 char __user *optval, int __user *optlen)
3164{
3165 sctp_assoc_t id;
3166 struct sctp_association *asoc;
3167 struct list_head *pos;
3168 int cnt = 0;
3169
3170 if (len != sizeof(sctp_assoc_t))
3171 return -EINVAL;
3172
3173 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3174 return -EFAULT;
3175
3176 /* For UDP-style sockets, id specifies the association to query. */
3177 asoc = sctp_id2assoc(sk, id);
3178 if (!asoc)
3179 return -EINVAL;
3180
3181 list_for_each(pos, &asoc->peer.transport_addr_list) {
3182 cnt ++;
3183 }
3184
3185 return cnt;
3186}
3187
3188static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3189 char __user *optval, int __user *optlen)
3190{
3191 struct sctp_association *asoc;
3192 struct list_head *pos;
3193 int cnt = 0;
3194 struct sctp_getaddrs getaddrs;
3195 struct sctp_transport *from;
3196 void __user *to;
3197 union sctp_addr temp;
3198 struct sctp_sock *sp = sctp_sk(sk);
3199 int addrlen;
3200
3201 if (len != sizeof(struct sctp_getaddrs))
3202 return -EINVAL;
3203
3204 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3205 return -EFAULT;
3206
3207 if (getaddrs.addr_num <= 0) return -EINVAL;
3208
3209 /* For UDP-style sockets, id specifies the association to query. */
3210 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3211 if (!asoc)
3212 return -EINVAL;
3213
3214 to = (void __user *)getaddrs.addrs;
3215 list_for_each(pos, &asoc->peer.transport_addr_list) {
3216 from = list_entry(pos, struct sctp_transport, transports);
3217 memcpy(&temp, &from->ipaddr, sizeof(temp));
3218 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3219 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3220 temp.v4.sin_port = htons(temp.v4.sin_port);
3221 if (copy_to_user(to, &temp, addrlen))
3222 return -EFAULT;
3223 to += addrlen ;
3224 cnt ++;
3225 if (cnt >= getaddrs.addr_num) break;
3226 }
3227 getaddrs.addr_num = cnt;
3228 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs)))
3229 return -EFAULT;
3230
3231 return 0;
3232}
3233
3234static int sctp_getsockopt_local_addrs_num(struct sock *sk, int len,
3235 char __user *optval,
3236 int __user *optlen)
3237{
3238 sctp_assoc_t id;
3239 struct sctp_bind_addr *bp;
3240 struct sctp_association *asoc;
3241 struct list_head *pos;
3242 struct sctp_sockaddr_entry *addr;
3243 rwlock_t *addr_lock;
3244 unsigned long flags;
3245 int cnt = 0;
3246
3247 if (len != sizeof(sctp_assoc_t))
3248 return -EINVAL;
3249
3250 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3251 return -EFAULT;
3252
3253 /*
3254 * For UDP-style sockets, id specifies the association to query.
3255 * If the id field is set to the value '0' then the locally bound
3256 * addresses are returned without regard to any particular
3257 * association.
3258 */
3259 if (0 == id) {
3260 bp = &sctp_sk(sk)->ep->base.bind_addr;
3261 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3262 } else {
3263 asoc = sctp_id2assoc(sk, id);
3264 if (!asoc)
3265 return -EINVAL;
3266 bp = &asoc->base.bind_addr;
3267 addr_lock = &asoc->base.addr_lock;
3268 }
3269
3270 sctp_read_lock(addr_lock);
3271
3272 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3273 * addresses from the global local address list.
3274 */
3275 if (sctp_list_single_entry(&bp->address_list)) {
3276 addr = list_entry(bp->address_list.next,
3277 struct sctp_sockaddr_entry, list);
3278 if (sctp_is_any(&addr->a)) {
3279 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3280 list_for_each(pos, &sctp_local_addr_list) {
3281 addr = list_entry(pos,
3282 struct sctp_sockaddr_entry,
3283 list);
3284 if ((PF_INET == sk->sk_family) &&
3285 (AF_INET6 == addr->a.sa.sa_family))
3286 continue;
3287 cnt++;
3288 }
3289 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3290 flags);
3291 } else {
3292 cnt = 1;
3293 }
3294 goto done;
3295 }
3296
3297 list_for_each(pos, &bp->address_list) {
3298 cnt ++;
3299 }
3300
3301done:
3302 sctp_read_unlock(addr_lock);
3303 return cnt;
3304}
3305
3306/* Helper function that copies local addresses to user and returns the number
3307 * of addresses copied.
3308 */
3309static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port, int max_addrs,
3310 void __user *to)
3311{
3312 struct list_head *pos;
3313 struct sctp_sockaddr_entry *addr;
3314 unsigned long flags;
3315 union sctp_addr temp;
3316 int cnt = 0;
3317 int addrlen;
3318
3319 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3320 list_for_each(pos, &sctp_local_addr_list) {
3321 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3322 if ((PF_INET == sk->sk_family) &&
3323 (AF_INET6 == addr->a.sa.sa_family))
3324 continue;
3325 memcpy(&temp, &addr->a, sizeof(temp));
3326 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3327 &temp);
3328 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3329 temp.v4.sin_port = htons(port);
3330 if (copy_to_user(to, &temp, addrlen)) {
3331 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3332 flags);
3333 return -EFAULT;
3334 }
3335 to += addrlen;
3336 cnt ++;
3337 if (cnt >= max_addrs) break;
3338 }
3339 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3340
3341 return cnt;
3342}
3343
3344static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
3345 char __user *optval, int __user *optlen)
3346{
3347 struct sctp_bind_addr *bp;
3348 struct sctp_association *asoc;
3349 struct list_head *pos;
3350 int cnt = 0;
3351 struct sctp_getaddrs getaddrs;
3352 struct sctp_sockaddr_entry *addr;
3353 void __user *to;
3354 union sctp_addr temp;
3355 struct sctp_sock *sp = sctp_sk(sk);
3356 int addrlen;
3357 rwlock_t *addr_lock;
3358 int err = 0;
3359
3360 if (len != sizeof(struct sctp_getaddrs))
3361 return -EINVAL;
3362
3363 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3364 return -EFAULT;
3365
3366 if (getaddrs.addr_num <= 0) return -EINVAL;
3367 /*
3368 * For UDP-style sockets, id specifies the association to query.
3369 * If the id field is set to the value '0' then the locally bound
3370 * addresses are returned without regard to any particular
3371 * association.
3372 */
3373 if (0 == getaddrs.assoc_id) {
3374 bp = &sctp_sk(sk)->ep->base.bind_addr;
3375 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3376 } else {
3377 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3378 if (!asoc)
3379 return -EINVAL;
3380 bp = &asoc->base.bind_addr;
3381 addr_lock = &asoc->base.addr_lock;
3382 }
3383
3384 to = getaddrs.addrs;
3385
3386 sctp_read_lock(addr_lock);
3387
3388 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3389 * addresses from the global local address list.
3390 */
3391 if (sctp_list_single_entry(&bp->address_list)) {
3392 addr = list_entry(bp->address_list.next,
3393 struct sctp_sockaddr_entry, list);
3394 if (sctp_is_any(&addr->a)) {
3395 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
3396 getaddrs.addr_num, to);
3397 if (cnt < 0) {
3398 err = cnt;
3399 goto unlock;
3400 }
3401 goto copy_getaddrs;
3402 }
3403 }
3404
3405 list_for_each(pos, &bp->address_list) {
3406 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3407 memcpy(&temp, &addr->a, sizeof(temp));
3408 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3409 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3410 temp.v4.sin_port = htons(temp.v4.sin_port);
3411 if (copy_to_user(to, &temp, addrlen)) {
3412 err = -EFAULT;
3413 goto unlock;
3414 }
3415 to += addrlen;
3416 cnt ++;
3417 if (cnt >= getaddrs.addr_num) break;
3418 }
3419
3420copy_getaddrs:
3421 getaddrs.addr_num = cnt;
3422 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs)))
3423 err = -EFAULT;
3424
3425unlock:
3426 sctp_read_unlock(addr_lock);
3427 return err;
3428}
3429
3430/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3431 *
3432 * Requests that the local SCTP stack use the enclosed peer address as
3433 * the association primary. The enclosed address must be one of the
3434 * association peer's addresses.
3435 */
3436static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
3437 char __user *optval, int __user *optlen)
3438{
3439 struct sctp_prim prim;
3440 struct sctp_association *asoc;
3441 struct sctp_sock *sp = sctp_sk(sk);
3442
3443 if (len != sizeof(struct sctp_prim))
3444 return -EINVAL;
3445
3446 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3447 return -EFAULT;
3448
3449 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
3450 if (!asoc)
3451 return -EINVAL;
3452
3453 if (!asoc->peer.primary_path)
3454 return -ENOTCONN;
3455
3456 asoc->peer.primary_path->ipaddr.v4.sin_port =
3457 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
3458 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
3459 sizeof(union sctp_addr));
3460 asoc->peer.primary_path->ipaddr.v4.sin_port =
3461 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
3462
3463 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
3464 (union sctp_addr *)&prim.ssp_addr);
3465
3466 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
3467 return -EFAULT;
3468
3469 return 0;
3470}
3471
3472/*
3473 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
3474 *
3475 * Requests that the local endpoint set the specified Adaption Layer
3476 * Indication parameter for all future INIT and INIT-ACK exchanges.
3477 */
3478static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
3479 char __user *optval, int __user *optlen)
3480{
3481 __u32 val;
3482
3483 if (len < sizeof(__u32))
3484 return -EINVAL;
3485
3486 len = sizeof(__u32);
3487 val = sctp_sk(sk)->adaption_ind;
3488 if (put_user(len, optlen))
3489 return -EFAULT;
3490 if (copy_to_user(optval, &val, len))
3491 return -EFAULT;
3492 return 0;
3493}
3494
3495/*
3496 *
3497 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
3498 *
3499 * Applications that wish to use the sendto() system call may wish to
3500 * specify a default set of parameters that would normally be supplied
3501 * through the inclusion of ancillary data. This socket option allows
3502 * such an application to set the default sctp_sndrcvinfo structure.
3503
3504
3505 * The application that wishes to use this socket option simply passes
3506 * in to this call the sctp_sndrcvinfo structure defined in Section
3507 * 5.2.2) The input parameters accepted by this call include
3508 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
3509 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
3510 * to this call if the caller is using the UDP model.
3511 *
3512 * For getsockopt, it get the default sctp_sndrcvinfo structure.
3513 */
3514static int sctp_getsockopt_default_send_param(struct sock *sk,
3515 int len, char __user *optval,
3516 int __user *optlen)
3517{
3518 struct sctp_sndrcvinfo info;
3519 struct sctp_association *asoc;
3520 struct sctp_sock *sp = sctp_sk(sk);
3521
3522 if (len != sizeof(struct sctp_sndrcvinfo))
3523 return -EINVAL;
3524 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
3525 return -EFAULT;
3526
3527 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3528 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
3529 return -EINVAL;
3530
3531 if (asoc) {
3532 info.sinfo_stream = asoc->default_stream;
3533 info.sinfo_flags = asoc->default_flags;
3534 info.sinfo_ppid = asoc->default_ppid;
3535 info.sinfo_context = asoc->default_context;
3536 info.sinfo_timetolive = asoc->default_timetolive;
3537 } else {
3538 info.sinfo_stream = sp->default_stream;
3539 info.sinfo_flags = sp->default_flags;
3540 info.sinfo_ppid = sp->default_ppid;
3541 info.sinfo_context = sp->default_context;
3542 info.sinfo_timetolive = sp->default_timetolive;
3543 }
3544
3545 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
3546 return -EFAULT;
3547
3548 return 0;
3549}
3550
3551/*
3552 *
3553 * 7.1.5 SCTP_NODELAY
3554 *
3555 * Turn on/off any Nagle-like algorithm. This means that packets are
3556 * generally sent as soon as possible and no unnecessary delays are
3557 * introduced, at the cost of more packets in the network. Expects an
3558 * integer boolean flag.
3559 */
3560
3561static int sctp_getsockopt_nodelay(struct sock *sk, int len,
3562 char __user *optval, int __user *optlen)
3563{
3564 int val;
3565
3566 if (len < sizeof(int))
3567 return -EINVAL;
3568
3569 len = sizeof(int);
3570 val = (sctp_sk(sk)->nodelay == 1);
3571 if (put_user(len, optlen))
3572 return -EFAULT;
3573 if (copy_to_user(optval, &val, len))
3574 return -EFAULT;
3575 return 0;
3576}
3577
3578/*
3579 *
3580 * 7.1.1 SCTP_RTOINFO
3581 *
3582 * The protocol parameters used to initialize and bound retransmission
3583 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3584 * and modify these parameters.
3585 * All parameters are time values, in milliseconds. A value of 0, when
3586 * modifying the parameters, indicates that the current value should not
3587 * be changed.
3588 *
3589 */
3590static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
3591 char __user *optval,
3592 int __user *optlen) {
3593 struct sctp_rtoinfo rtoinfo;
3594 struct sctp_association *asoc;
3595
3596 if (len != sizeof (struct sctp_rtoinfo))
3597 return -EINVAL;
3598
3599 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
3600 return -EFAULT;
3601
3602 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3603
3604 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3605 return -EINVAL;
3606
3607 /* Values corresponding to the specific association. */
3608 if (asoc) {
3609 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
3610 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
3611 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
3612 } else {
3613 /* Values corresponding to the endpoint. */
3614 struct sctp_sock *sp = sctp_sk(sk);
3615
3616 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
3617 rtoinfo.srto_max = sp->rtoinfo.srto_max;
3618 rtoinfo.srto_min = sp->rtoinfo.srto_min;
3619 }
3620
3621 if (put_user(len, optlen))
3622 return -EFAULT;
3623
3624 if (copy_to_user(optval, &rtoinfo, len))
3625 return -EFAULT;
3626
3627 return 0;
3628}
3629
3630/*
3631 *
3632 * 7.1.2 SCTP_ASSOCINFO
3633 *
3634 * This option is used to tune the the maximum retransmission attempts
3635 * of the association.
3636 * Returns an error if the new association retransmission value is
3637 * greater than the sum of the retransmission value of the peer.
3638 * See [SCTP] for more information.
3639 *
3640 */
3641static int sctp_getsockopt_associnfo(struct sock *sk, int len,
3642 char __user *optval,
3643 int __user *optlen)
3644{
3645
3646 struct sctp_assocparams assocparams;
3647 struct sctp_association *asoc;
3648 struct list_head *pos;
3649 int cnt = 0;
3650
3651 if (len != sizeof (struct sctp_assocparams))
3652 return -EINVAL;
3653
3654 if (copy_from_user(&assocparams, optval,
3655 sizeof (struct sctp_assocparams)))
3656 return -EFAULT;
3657
3658 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3659
3660 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3661 return -EINVAL;
3662
3663 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07003664 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003665 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
3666 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
3667 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
3668 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
3669 * 1000) +
3670 (asoc->cookie_life.tv_usec
3671 / 1000);
3672
3673 list_for_each(pos, &asoc->peer.transport_addr_list) {
3674 cnt ++;
3675 }
3676
3677 assocparams.sasoc_number_peer_destinations = cnt;
3678 } else {
3679 /* Values corresponding to the endpoint */
3680 struct sctp_sock *sp = sctp_sk(sk);
3681
3682 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
3683 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
3684 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
3685 assocparams.sasoc_cookie_life =
3686 sp->assocparams.sasoc_cookie_life;
3687 assocparams.sasoc_number_peer_destinations =
3688 sp->assocparams.
3689 sasoc_number_peer_destinations;
3690 }
3691
3692 if (put_user(len, optlen))
3693 return -EFAULT;
3694
3695 if (copy_to_user(optval, &assocparams, len))
3696 return -EFAULT;
3697
3698 return 0;
3699}
3700
3701/*
3702 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3703 *
3704 * This socket option is a boolean flag which turns on or off mapped V4
3705 * addresses. If this option is turned on and the socket is type
3706 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3707 * If this option is turned off, then no mapping will be done of V4
3708 * addresses and a user will receive both PF_INET6 and PF_INET type
3709 * addresses on the socket.
3710 */
3711static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
3712 char __user *optval, int __user *optlen)
3713{
3714 int val;
3715 struct sctp_sock *sp = sctp_sk(sk);
3716
3717 if (len < sizeof(int))
3718 return -EINVAL;
3719
3720 len = sizeof(int);
3721 val = sp->v4mapped;
3722 if (put_user(len, optlen))
3723 return -EFAULT;
3724 if (copy_to_user(optval, &val, len))
3725 return -EFAULT;
3726
3727 return 0;
3728}
3729
3730/*
3731 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
3732 *
3733 * This socket option specifies the maximum size to put in any outgoing
3734 * SCTP chunk. If a message is larger than this size it will be
3735 * fragmented by SCTP into the specified size. Note that the underlying
3736 * SCTP implementation may fragment into smaller sized chunks when the
3737 * PMTU of the underlying association is smaller than the value set by
3738 * the user.
3739 */
3740static int sctp_getsockopt_maxseg(struct sock *sk, int len,
3741 char __user *optval, int __user *optlen)
3742{
3743 int val;
3744
3745 if (len < sizeof(int))
3746 return -EINVAL;
3747
3748 len = sizeof(int);
3749
3750 val = sctp_sk(sk)->user_frag;
3751 if (put_user(len, optlen))
3752 return -EFAULT;
3753 if (copy_to_user(optval, &val, len))
3754 return -EFAULT;
3755
3756 return 0;
3757}
3758
3759SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
3760 char __user *optval, int __user *optlen)
3761{
3762 int retval = 0;
3763 int len;
3764
Frank Filz3f7a87d2005-06-20 13:14:57 -07003765 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
3766 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003767
3768 /* I can hardly begin to describe how wrong this is. This is
3769 * so broken as to be worse than useless. The API draft
3770 * REALLY is NOT helpful here... I am not convinced that the
3771 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
3772 * are at all well-founded.
3773 */
3774 if (level != SOL_SCTP) {
3775 struct sctp_af *af = sctp_sk(sk)->pf->af;
3776
3777 retval = af->getsockopt(sk, level, optname, optval, optlen);
3778 return retval;
3779 }
3780
3781 if (get_user(len, optlen))
3782 return -EFAULT;
3783
3784 sctp_lock_sock(sk);
3785
3786 switch (optname) {
3787 case SCTP_STATUS:
3788 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
3789 break;
3790 case SCTP_DISABLE_FRAGMENTS:
3791 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
3792 optlen);
3793 break;
3794 case SCTP_EVENTS:
3795 retval = sctp_getsockopt_events(sk, len, optval, optlen);
3796 break;
3797 case SCTP_AUTOCLOSE:
3798 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
3799 break;
3800 case SCTP_SOCKOPT_PEELOFF:
3801 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
3802 break;
3803 case SCTP_PEER_ADDR_PARAMS:
3804 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
3805 optlen);
3806 break;
3807 case SCTP_INITMSG:
3808 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
3809 break;
3810 case SCTP_GET_PEER_ADDRS_NUM:
3811 retval = sctp_getsockopt_peer_addrs_num(sk, len, optval,
3812 optlen);
3813 break;
3814 case SCTP_GET_LOCAL_ADDRS_NUM:
3815 retval = sctp_getsockopt_local_addrs_num(sk, len, optval,
3816 optlen);
3817 break;
3818 case SCTP_GET_PEER_ADDRS:
3819 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
3820 optlen);
3821 break;
3822 case SCTP_GET_LOCAL_ADDRS:
3823 retval = sctp_getsockopt_local_addrs(sk, len, optval,
3824 optlen);
3825 break;
3826 case SCTP_DEFAULT_SEND_PARAM:
3827 retval = sctp_getsockopt_default_send_param(sk, len,
3828 optval, optlen);
3829 break;
3830 case SCTP_PRIMARY_ADDR:
3831 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
3832 break;
3833 case SCTP_NODELAY:
3834 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
3835 break;
3836 case SCTP_RTOINFO:
3837 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
3838 break;
3839 case SCTP_ASSOCINFO:
3840 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
3841 break;
3842 case SCTP_I_WANT_MAPPED_V4_ADDR:
3843 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
3844 break;
3845 case SCTP_MAXSEG:
3846 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
3847 break;
3848 case SCTP_GET_PEER_ADDR_INFO:
3849 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
3850 optlen);
3851 break;
3852 case SCTP_ADAPTION_LAYER:
3853 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
3854 optlen);
3855 break;
3856 default:
3857 retval = -ENOPROTOOPT;
3858 break;
3859 };
3860
3861 sctp_release_sock(sk);
3862 return retval;
3863}
3864
3865static void sctp_hash(struct sock *sk)
3866{
3867 /* STUB */
3868}
3869
3870static void sctp_unhash(struct sock *sk)
3871{
3872 /* STUB */
3873}
3874
3875/* Check if port is acceptable. Possibly find first available port.
3876 *
3877 * The port hash table (contained in the 'global' SCTP protocol storage
3878 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
3879 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
3880 * list (the list number is the port number hashed out, so as you
3881 * would expect from a hash function, all the ports in a given list have
3882 * such a number that hashes out to the same list number; you were
3883 * expecting that, right?); so each list has a set of ports, with a
3884 * link to the socket (struct sock) that uses it, the port number and
3885 * a fastreuse flag (FIXME: NPI ipg).
3886 */
3887static struct sctp_bind_bucket *sctp_bucket_create(
3888 struct sctp_bind_hashbucket *head, unsigned short snum);
3889
3890static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
3891{
3892 struct sctp_bind_hashbucket *head; /* hash list */
3893 struct sctp_bind_bucket *pp; /* hash list port iterator */
3894 unsigned short snum;
3895 int ret;
3896
3897 /* NOTE: Remember to put this back to net order. */
3898 addr->v4.sin_port = ntohs(addr->v4.sin_port);
3899 snum = addr->v4.sin_port;
3900
3901 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
3902 sctp_local_bh_disable();
3903
3904 if (snum == 0) {
3905 /* Search for an available port.
3906 *
3907 * 'sctp_port_rover' was the last port assigned, so
3908 * we start to search from 'sctp_port_rover +
3909 * 1'. What we do is first check if port 'rover' is
3910 * already in the hash table; if not, we use that; if
3911 * it is, we try next.
3912 */
3913 int low = sysctl_local_port_range[0];
3914 int high = sysctl_local_port_range[1];
3915 int remaining = (high - low) + 1;
3916 int rover;
3917 int index;
3918
3919 sctp_spin_lock(&sctp_port_alloc_lock);
3920 rover = sctp_port_rover;
3921 do {
3922 rover++;
3923 if ((rover < low) || (rover > high))
3924 rover = low;
3925 index = sctp_phashfn(rover);
3926 head = &sctp_port_hashtable[index];
3927 sctp_spin_lock(&head->lock);
3928 for (pp = head->chain; pp; pp = pp->next)
3929 if (pp->port == rover)
3930 goto next;
3931 break;
3932 next:
3933 sctp_spin_unlock(&head->lock);
3934 } while (--remaining > 0);
3935 sctp_port_rover = rover;
3936 sctp_spin_unlock(&sctp_port_alloc_lock);
3937
3938 /* Exhausted local port range during search? */
3939 ret = 1;
3940 if (remaining <= 0)
3941 goto fail;
3942
3943 /* OK, here is the one we will use. HEAD (the port
3944 * hash table list entry) is non-NULL and we hold it's
3945 * mutex.
3946 */
3947 snum = rover;
3948 } else {
3949 /* We are given an specific port number; we verify
3950 * that it is not being used. If it is used, we will
3951 * exahust the search in the hash list corresponding
3952 * to the port number (snum) - we detect that with the
3953 * port iterator, pp being NULL.
3954 */
3955 head = &sctp_port_hashtable[sctp_phashfn(snum)];
3956 sctp_spin_lock(&head->lock);
3957 for (pp = head->chain; pp; pp = pp->next) {
3958 if (pp->port == snum)
3959 goto pp_found;
3960 }
3961 }
3962 pp = NULL;
3963 goto pp_not_found;
3964pp_found:
3965 if (!hlist_empty(&pp->owner)) {
3966 /* We had a port hash table hit - there is an
3967 * available port (pp != NULL) and it is being
3968 * used by other socket (pp->owner not empty); that other
3969 * socket is going to be sk2.
3970 */
3971 int reuse = sk->sk_reuse;
3972 struct sock *sk2;
3973 struct hlist_node *node;
3974
3975 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
3976 if (pp->fastreuse && sk->sk_reuse)
3977 goto success;
3978
3979 /* Run through the list of sockets bound to the port
3980 * (pp->port) [via the pointers bind_next and
3981 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
3982 * we get the endpoint they describe and run through
3983 * the endpoint's list of IP (v4 or v6) addresses,
3984 * comparing each of the addresses with the address of
3985 * the socket sk. If we find a match, then that means
3986 * that this port/socket (sk) combination are already
3987 * in an endpoint.
3988 */
3989 sk_for_each_bound(sk2, node, &pp->owner) {
3990 struct sctp_endpoint *ep2;
3991 ep2 = sctp_sk(sk2)->ep;
3992
3993 if (reuse && sk2->sk_reuse)
3994 continue;
3995
3996 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
3997 sctp_sk(sk))) {
3998 ret = (long)sk2;
3999 goto fail_unlock;
4000 }
4001 }
4002 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4003 }
4004pp_not_found:
4005 /* If there was a hash table miss, create a new port. */
4006 ret = 1;
4007 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4008 goto fail_unlock;
4009
4010 /* In either case (hit or miss), make sure fastreuse is 1 only
4011 * if sk->sk_reuse is too (that is, if the caller requested
4012 * SO_REUSEADDR on this socket -sk-).
4013 */
4014 if (hlist_empty(&pp->owner))
4015 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4016 else if (pp->fastreuse && !sk->sk_reuse)
4017 pp->fastreuse = 0;
4018
4019 /* We are set, so fill up all the data in the hash table
4020 * entry, tie the socket list information with the rest of the
4021 * sockets FIXME: Blurry, NPI (ipg).
4022 */
4023success:
4024 inet_sk(sk)->num = snum;
4025 if (!sctp_sk(sk)->bind_hash) {
4026 sk_add_bind_node(sk, &pp->owner);
4027 sctp_sk(sk)->bind_hash = pp;
4028 }
4029 ret = 0;
4030
4031fail_unlock:
4032 sctp_spin_unlock(&head->lock);
4033
4034fail:
4035 sctp_local_bh_enable();
4036 addr->v4.sin_port = htons(addr->v4.sin_port);
4037 return ret;
4038}
4039
4040/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4041 * port is requested.
4042 */
4043static int sctp_get_port(struct sock *sk, unsigned short snum)
4044{
4045 long ret;
4046 union sctp_addr addr;
4047 struct sctp_af *af = sctp_sk(sk)->pf->af;
4048
4049 /* Set up a dummy address struct from the sk. */
4050 af->from_sk(&addr, sk);
4051 addr.v4.sin_port = htons(snum);
4052
4053 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4054 ret = sctp_get_port_local(sk, &addr);
4055
4056 return (ret ? 1 : 0);
4057}
4058
4059/*
4060 * 3.1.3 listen() - UDP Style Syntax
4061 *
4062 * By default, new associations are not accepted for UDP style sockets.
4063 * An application uses listen() to mark a socket as being able to
4064 * accept new associations.
4065 */
4066SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4067{
4068 struct sctp_sock *sp = sctp_sk(sk);
4069 struct sctp_endpoint *ep = sp->ep;
4070
4071 /* Only UDP style sockets that are not peeled off are allowed to
4072 * listen().
4073 */
4074 if (!sctp_style(sk, UDP))
4075 return -EINVAL;
4076
4077 /* If backlog is zero, disable listening. */
4078 if (!backlog) {
4079 if (sctp_sstate(sk, CLOSED))
4080 return 0;
4081
4082 sctp_unhash_endpoint(ep);
4083 sk->sk_state = SCTP_SS_CLOSED;
4084 }
4085
4086 /* Return if we are already listening. */
4087 if (sctp_sstate(sk, LISTENING))
4088 return 0;
4089
4090 /*
4091 * If a bind() or sctp_bindx() is not called prior to a listen()
4092 * call that allows new associations to be accepted, the system
4093 * picks an ephemeral port and will choose an address set equivalent
4094 * to binding with a wildcard address.
4095 *
4096 * This is not currently spelled out in the SCTP sockets
4097 * extensions draft, but follows the practice as seen in TCP
4098 * sockets.
4099 */
4100 if (!ep->base.bind_addr.port) {
4101 if (sctp_autobind(sk))
4102 return -EAGAIN;
4103 }
4104 sk->sk_state = SCTP_SS_LISTENING;
4105 sctp_hash_endpoint(ep);
4106 return 0;
4107}
4108
4109/*
4110 * 4.1.3 listen() - TCP Style Syntax
4111 *
4112 * Applications uses listen() to ready the SCTP endpoint for accepting
4113 * inbound associations.
4114 */
4115SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4116{
4117 struct sctp_sock *sp = sctp_sk(sk);
4118 struct sctp_endpoint *ep = sp->ep;
4119
4120 /* If backlog is zero, disable listening. */
4121 if (!backlog) {
4122 if (sctp_sstate(sk, CLOSED))
4123 return 0;
4124
4125 sctp_unhash_endpoint(ep);
4126 sk->sk_state = SCTP_SS_CLOSED;
4127 }
4128
4129 if (sctp_sstate(sk, LISTENING))
4130 return 0;
4131
4132 /*
4133 * If a bind() or sctp_bindx() is not called prior to a listen()
4134 * call that allows new associations to be accepted, the system
4135 * picks an ephemeral port and will choose an address set equivalent
4136 * to binding with a wildcard address.
4137 *
4138 * This is not currently spelled out in the SCTP sockets
4139 * extensions draft, but follows the practice as seen in TCP
4140 * sockets.
4141 */
4142 if (!ep->base.bind_addr.port) {
4143 if (sctp_autobind(sk))
4144 return -EAGAIN;
4145 }
4146 sk->sk_state = SCTP_SS_LISTENING;
4147 sk->sk_max_ack_backlog = backlog;
4148 sctp_hash_endpoint(ep);
4149 return 0;
4150}
4151
4152/*
4153 * Move a socket to LISTENING state.
4154 */
4155int sctp_inet_listen(struct socket *sock, int backlog)
4156{
4157 struct sock *sk = sock->sk;
4158 struct crypto_tfm *tfm=NULL;
4159 int err = -EINVAL;
4160
4161 if (unlikely(backlog < 0))
4162 goto out;
4163
4164 sctp_lock_sock(sk);
4165
4166 if (sock->state != SS_UNCONNECTED)
4167 goto out;
4168
4169 /* Allocate HMAC for generating cookie. */
4170 if (sctp_hmac_alg) {
4171 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4172 if (!tfm) {
4173 err = -ENOSYS;
4174 goto out;
4175 }
4176 }
4177
4178 switch (sock->type) {
4179 case SOCK_SEQPACKET:
4180 err = sctp_seqpacket_listen(sk, backlog);
4181 break;
4182 case SOCK_STREAM:
4183 err = sctp_stream_listen(sk, backlog);
4184 break;
4185 default:
4186 break;
4187 };
4188 if (err)
4189 goto cleanup;
4190
4191 /* Store away the transform reference. */
4192 sctp_sk(sk)->hmac = tfm;
4193out:
4194 sctp_release_sock(sk);
4195 return err;
4196cleanup:
4197 if (tfm)
4198 sctp_crypto_free_tfm(tfm);
4199 goto out;
4200}
4201
4202/*
4203 * This function is done by modeling the current datagram_poll() and the
4204 * tcp_poll(). Note that, based on these implementations, we don't
4205 * lock the socket in this function, even though it seems that,
4206 * ideally, locking or some other mechanisms can be used to ensure
4207 * the integrity of the counters (sndbuf and wmem_queued) used
4208 * in this place. We assume that we don't need locks either until proven
4209 * otherwise.
4210 *
4211 * Another thing to note is that we include the Async I/O support
4212 * here, again, by modeling the current TCP/UDP code. We don't have
4213 * a good way to test with it yet.
4214 */
4215unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4216{
4217 struct sock *sk = sock->sk;
4218 struct sctp_sock *sp = sctp_sk(sk);
4219 unsigned int mask;
4220
4221 poll_wait(file, sk->sk_sleep, wait);
4222
4223 /* A TCP-style listening socket becomes readable when the accept queue
4224 * is not empty.
4225 */
4226 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4227 return (!list_empty(&sp->ep->asocs)) ?
4228 (POLLIN | POLLRDNORM) : 0;
4229
4230 mask = 0;
4231
4232 /* Is there any exceptional events? */
4233 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4234 mask |= POLLERR;
4235 if (sk->sk_shutdown == SHUTDOWN_MASK)
4236 mask |= POLLHUP;
4237
4238 /* Is it readable? Reconsider this code with TCP-style support. */
4239 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4240 (sk->sk_shutdown & RCV_SHUTDOWN))
4241 mask |= POLLIN | POLLRDNORM;
4242
4243 /* The association is either gone or not ready. */
4244 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4245 return mask;
4246
4247 /* Is it writable? */
4248 if (sctp_writeable(sk)) {
4249 mask |= POLLOUT | POLLWRNORM;
4250 } else {
4251 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4252 /*
4253 * Since the socket is not locked, the buffer
4254 * might be made available after the writeable check and
4255 * before the bit is set. This could cause a lost I/O
4256 * signal. tcp_poll() has a race breaker for this race
4257 * condition. Based on their implementation, we put
4258 * in the following code to cover it as well.
4259 */
4260 if (sctp_writeable(sk))
4261 mask |= POLLOUT | POLLWRNORM;
4262 }
4263 return mask;
4264}
4265
4266/********************************************************************
4267 * 2nd Level Abstractions
4268 ********************************************************************/
4269
4270static struct sctp_bind_bucket *sctp_bucket_create(
4271 struct sctp_bind_hashbucket *head, unsigned short snum)
4272{
4273 struct sctp_bind_bucket *pp;
4274
4275 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
4276 SCTP_DBG_OBJCNT_INC(bind_bucket);
4277 if (pp) {
4278 pp->port = snum;
4279 pp->fastreuse = 0;
4280 INIT_HLIST_HEAD(&pp->owner);
4281 if ((pp->next = head->chain) != NULL)
4282 pp->next->pprev = &pp->next;
4283 head->chain = pp;
4284 pp->pprev = &head->chain;
4285 }
4286 return pp;
4287}
4288
4289/* Caller must hold hashbucket lock for this tb with local BH disabled */
4290static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
4291{
4292 if (hlist_empty(&pp->owner)) {
4293 if (pp->next)
4294 pp->next->pprev = pp->pprev;
4295 *(pp->pprev) = pp->next;
4296 kmem_cache_free(sctp_bucket_cachep, pp);
4297 SCTP_DBG_OBJCNT_DEC(bind_bucket);
4298 }
4299}
4300
4301/* Release this socket's reference to a local port. */
4302static inline void __sctp_put_port(struct sock *sk)
4303{
4304 struct sctp_bind_hashbucket *head =
4305 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
4306 struct sctp_bind_bucket *pp;
4307
4308 sctp_spin_lock(&head->lock);
4309 pp = sctp_sk(sk)->bind_hash;
4310 __sk_del_bind_node(sk);
4311 sctp_sk(sk)->bind_hash = NULL;
4312 inet_sk(sk)->num = 0;
4313 sctp_bucket_destroy(pp);
4314 sctp_spin_unlock(&head->lock);
4315}
4316
4317void sctp_put_port(struct sock *sk)
4318{
4319 sctp_local_bh_disable();
4320 __sctp_put_port(sk);
4321 sctp_local_bh_enable();
4322}
4323
4324/*
4325 * The system picks an ephemeral port and choose an address set equivalent
4326 * to binding with a wildcard address.
4327 * One of those addresses will be the primary address for the association.
4328 * This automatically enables the multihoming capability of SCTP.
4329 */
4330static int sctp_autobind(struct sock *sk)
4331{
4332 union sctp_addr autoaddr;
4333 struct sctp_af *af;
4334 unsigned short port;
4335
4336 /* Initialize a local sockaddr structure to INADDR_ANY. */
4337 af = sctp_sk(sk)->pf->af;
4338
4339 port = htons(inet_sk(sk)->num);
4340 af->inaddr_any(&autoaddr, port);
4341
4342 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
4343}
4344
4345/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
4346 *
4347 * From RFC 2292
4348 * 4.2 The cmsghdr Structure *
4349 *
4350 * When ancillary data is sent or received, any number of ancillary data
4351 * objects can be specified by the msg_control and msg_controllen members of
4352 * the msghdr structure, because each object is preceded by
4353 * a cmsghdr structure defining the object's length (the cmsg_len member).
4354 * Historically Berkeley-derived implementations have passed only one object
4355 * at a time, but this API allows multiple objects to be
4356 * passed in a single call to sendmsg() or recvmsg(). The following example
4357 * shows two ancillary data objects in a control buffer.
4358 *
4359 * |<--------------------------- msg_controllen -------------------------->|
4360 * | |
4361 *
4362 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
4363 *
4364 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
4365 * | | |
4366 *
4367 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
4368 *
4369 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
4370 * | | | | |
4371 *
4372 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4373 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
4374 *
4375 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
4376 *
4377 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4378 * ^
4379 * |
4380 *
4381 * msg_control
4382 * points here
4383 */
4384SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
4385 sctp_cmsgs_t *cmsgs)
4386{
4387 struct cmsghdr *cmsg;
4388
4389 for (cmsg = CMSG_FIRSTHDR(msg);
4390 cmsg != NULL;
4391 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
4392 if (!CMSG_OK(msg, cmsg))
4393 return -EINVAL;
4394
4395 /* Should we parse this header or ignore? */
4396 if (cmsg->cmsg_level != IPPROTO_SCTP)
4397 continue;
4398
4399 /* Strictly check lengths following example in SCM code. */
4400 switch (cmsg->cmsg_type) {
4401 case SCTP_INIT:
4402 /* SCTP Socket API Extension
4403 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
4404 *
4405 * This cmsghdr structure provides information for
4406 * initializing new SCTP associations with sendmsg().
4407 * The SCTP_INITMSG socket option uses this same data
4408 * structure. This structure is not used for
4409 * recvmsg().
4410 *
4411 * cmsg_level cmsg_type cmsg_data[]
4412 * ------------ ------------ ----------------------
4413 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
4414 */
4415 if (cmsg->cmsg_len !=
4416 CMSG_LEN(sizeof(struct sctp_initmsg)))
4417 return -EINVAL;
4418 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
4419 break;
4420
4421 case SCTP_SNDRCV:
4422 /* SCTP Socket API Extension
4423 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
4424 *
4425 * This cmsghdr structure specifies SCTP options for
4426 * sendmsg() and describes SCTP header information
4427 * about a received message through recvmsg().
4428 *
4429 * cmsg_level cmsg_type cmsg_data[]
4430 * ------------ ------------ ----------------------
4431 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
4432 */
4433 if (cmsg->cmsg_len !=
4434 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
4435 return -EINVAL;
4436
4437 cmsgs->info =
4438 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
4439
4440 /* Minimally, validate the sinfo_flags. */
4441 if (cmsgs->info->sinfo_flags &
4442 ~(MSG_UNORDERED | MSG_ADDR_OVER |
4443 MSG_ABORT | MSG_EOF))
4444 return -EINVAL;
4445 break;
4446
4447 default:
4448 return -EINVAL;
4449 };
4450 }
4451 return 0;
4452}
4453
4454/*
4455 * Wait for a packet..
4456 * Note: This function is the same function as in core/datagram.c
4457 * with a few modifications to make lksctp work.
4458 */
4459static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
4460{
4461 int error;
4462 DEFINE_WAIT(wait);
4463
4464 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
4465
4466 /* Socket errors? */
4467 error = sock_error(sk);
4468 if (error)
4469 goto out;
4470
4471 if (!skb_queue_empty(&sk->sk_receive_queue))
4472 goto ready;
4473
4474 /* Socket shut down? */
4475 if (sk->sk_shutdown & RCV_SHUTDOWN)
4476 goto out;
4477
4478 /* Sequenced packets can come disconnected. If so we report the
4479 * problem.
4480 */
4481 error = -ENOTCONN;
4482
4483 /* Is there a good reason to think that we may receive some data? */
4484 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
4485 goto out;
4486
4487 /* Handle signals. */
4488 if (signal_pending(current))
4489 goto interrupted;
4490
4491 /* Let another process have a go. Since we are going to sleep
4492 * anyway. Note: This may cause odd behaviors if the message
4493 * does not fit in the user's buffer, but this seems to be the
4494 * only way to honor MSG_DONTWAIT realistically.
4495 */
4496 sctp_release_sock(sk);
4497 *timeo_p = schedule_timeout(*timeo_p);
4498 sctp_lock_sock(sk);
4499
4500ready:
4501 finish_wait(sk->sk_sleep, &wait);
4502 return 0;
4503
4504interrupted:
4505 error = sock_intr_errno(*timeo_p);
4506
4507out:
4508 finish_wait(sk->sk_sleep, &wait);
4509 *err = error;
4510 return error;
4511}
4512
4513/* Receive a datagram.
4514 * Note: This is pretty much the same routine as in core/datagram.c
4515 * with a few changes to make lksctp work.
4516 */
4517static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
4518 int noblock, int *err)
4519{
4520 int error;
4521 struct sk_buff *skb;
4522 long timeo;
4523
4524 /* Caller is allowed not to check sk->sk_err before calling. */
4525 error = sock_error(sk);
4526 if (error)
4527 goto no_packet;
4528
4529 timeo = sock_rcvtimeo(sk, noblock);
4530
4531 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
4532 timeo, MAX_SCHEDULE_TIMEOUT);
4533
4534 do {
4535 /* Again only user level code calls this function,
4536 * so nothing interrupt level
4537 * will suddenly eat the receive_queue.
4538 *
4539 * Look at current nfs client by the way...
4540 * However, this function was corrent in any case. 8)
4541 */
4542 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07004543 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004544 skb = skb_peek(&sk->sk_receive_queue);
4545 if (skb)
4546 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07004547 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004548 } else {
4549 skb = skb_dequeue(&sk->sk_receive_queue);
4550 }
4551
4552 if (skb)
4553 return skb;
4554
4555 if (sk->sk_shutdown & RCV_SHUTDOWN)
4556 break;
4557
4558 /* User doesn't want to wait. */
4559 error = -EAGAIN;
4560 if (!timeo)
4561 goto no_packet;
4562 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
4563
4564 return NULL;
4565
4566no_packet:
4567 *err = error;
4568 return NULL;
4569}
4570
4571/* If sndbuf has changed, wake up per association sndbuf waiters. */
4572static void __sctp_write_space(struct sctp_association *asoc)
4573{
4574 struct sock *sk = asoc->base.sk;
4575 struct socket *sock = sk->sk_socket;
4576
4577 if ((sctp_wspace(asoc) > 0) && sock) {
4578 if (waitqueue_active(&asoc->wait))
4579 wake_up_interruptible(&asoc->wait);
4580
4581 if (sctp_writeable(sk)) {
4582 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
4583 wake_up_interruptible(sk->sk_sleep);
4584
4585 /* Note that we try to include the Async I/O support
4586 * here by modeling from the current TCP/UDP code.
4587 * We have not tested with it yet.
4588 */
4589 if (sock->fasync_list &&
4590 !(sk->sk_shutdown & SEND_SHUTDOWN))
4591 sock_wake_async(sock, 2, POLL_OUT);
4592 }
4593 }
4594}
4595
4596/* Do accounting for the sndbuf space.
4597 * Decrement the used sndbuf space of the corresponding association by the
4598 * data size which was just transmitted(freed).
4599 */
4600static void sctp_wfree(struct sk_buff *skb)
4601{
4602 struct sctp_association *asoc;
4603 struct sctp_chunk *chunk;
4604 struct sock *sk;
4605
4606 /* Get the saved chunk pointer. */
4607 chunk = *((struct sctp_chunk **)(skb->cb));
4608 asoc = chunk->asoc;
4609 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07004610 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
4611 sizeof(struct sk_buff) +
4612 sizeof(struct sctp_chunk);
4613
4614 sk->sk_wmem_queued -= SCTP_DATA_SNDSIZE(chunk) +
4615 sizeof(struct sk_buff) +
4616 sizeof(struct sctp_chunk);
4617
4618 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
4619
4620 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004621 __sctp_write_space(asoc);
4622
4623 sctp_association_put(asoc);
4624}
4625
4626/* Helper function to wait for space in the sndbuf. */
4627static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
4628 size_t msg_len)
4629{
4630 struct sock *sk = asoc->base.sk;
4631 int err = 0;
4632 long current_timeo = *timeo_p;
4633 DEFINE_WAIT(wait);
4634
4635 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
4636 asoc, (long)(*timeo_p), msg_len);
4637
4638 /* Increment the association's refcnt. */
4639 sctp_association_hold(asoc);
4640
4641 /* Wait on the association specific sndbuf space. */
4642 for (;;) {
4643 prepare_to_wait_exclusive(&asoc->wait, &wait,
4644 TASK_INTERRUPTIBLE);
4645 if (!*timeo_p)
4646 goto do_nonblock;
4647 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
4648 asoc->base.dead)
4649 goto do_error;
4650 if (signal_pending(current))
4651 goto do_interrupted;
4652 if (msg_len <= sctp_wspace(asoc))
4653 break;
4654
4655 /* Let another process have a go. Since we are going
4656 * to sleep anyway.
4657 */
4658 sctp_release_sock(sk);
4659 current_timeo = schedule_timeout(current_timeo);
4660 sctp_lock_sock(sk);
4661
4662 *timeo_p = current_timeo;
4663 }
4664
4665out:
4666 finish_wait(&asoc->wait, &wait);
4667
4668 /* Release the association's refcnt. */
4669 sctp_association_put(asoc);
4670
4671 return err;
4672
4673do_error:
4674 err = -EPIPE;
4675 goto out;
4676
4677do_interrupted:
4678 err = sock_intr_errno(*timeo_p);
4679 goto out;
4680
4681do_nonblock:
4682 err = -EAGAIN;
4683 goto out;
4684}
4685
4686/* If socket sndbuf has changed, wake up all per association waiters. */
4687void sctp_write_space(struct sock *sk)
4688{
4689 struct sctp_association *asoc;
4690 struct list_head *pos;
4691
4692 /* Wake up the tasks in each wait queue. */
4693 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
4694 asoc = list_entry(pos, struct sctp_association, asocs);
4695 __sctp_write_space(asoc);
4696 }
4697}
4698
4699/* Is there any sndbuf space available on the socket?
4700 *
4701 * Note that wmem_queued is the sum of the send buffers on all of the
4702 * associations on the same socket. For a UDP-style socket with
4703 * multiple associations, it is possible for it to be "unwriteable"
4704 * prematurely. I assume that this is acceptable because
4705 * a premature "unwriteable" is better than an accidental "writeable" which
4706 * would cause an unwanted block under certain circumstances. For the 1-1
4707 * UDP-style sockets or TCP-style sockets, this code should work.
4708 * - Daisy
4709 */
4710static int sctp_writeable(struct sock *sk)
4711{
4712 int amt = 0;
4713
4714 amt = sk->sk_sndbuf - sk->sk_wmem_queued;
4715 if (amt < 0)
4716 amt = 0;
4717 return amt;
4718}
4719
4720/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
4721 * returns immediately with EINPROGRESS.
4722 */
4723static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
4724{
4725 struct sock *sk = asoc->base.sk;
4726 int err = 0;
4727 long current_timeo = *timeo_p;
4728 DEFINE_WAIT(wait);
4729
4730 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
4731 (long)(*timeo_p));
4732
4733 /* Increment the association's refcnt. */
4734 sctp_association_hold(asoc);
4735
4736 for (;;) {
4737 prepare_to_wait_exclusive(&asoc->wait, &wait,
4738 TASK_INTERRUPTIBLE);
4739 if (!*timeo_p)
4740 goto do_nonblock;
4741 if (sk->sk_shutdown & RCV_SHUTDOWN)
4742 break;
4743 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
4744 asoc->base.dead)
4745 goto do_error;
4746 if (signal_pending(current))
4747 goto do_interrupted;
4748
4749 if (sctp_state(asoc, ESTABLISHED))
4750 break;
4751
4752 /* Let another process have a go. Since we are going
4753 * to sleep anyway.
4754 */
4755 sctp_release_sock(sk);
4756 current_timeo = schedule_timeout(current_timeo);
4757 sctp_lock_sock(sk);
4758
4759 *timeo_p = current_timeo;
4760 }
4761
4762out:
4763 finish_wait(&asoc->wait, &wait);
4764
4765 /* Release the association's refcnt. */
4766 sctp_association_put(asoc);
4767
4768 return err;
4769
4770do_error:
Frank Filz3f7a87d2005-06-20 13:14:57 -07004771 if (asoc->init_err_counter + 1 >= asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004772 err = -ETIMEDOUT;
4773 else
4774 err = -ECONNREFUSED;
4775 goto out;
4776
4777do_interrupted:
4778 err = sock_intr_errno(*timeo_p);
4779 goto out;
4780
4781do_nonblock:
4782 err = -EINPROGRESS;
4783 goto out;
4784}
4785
4786static int sctp_wait_for_accept(struct sock *sk, long timeo)
4787{
4788 struct sctp_endpoint *ep;
4789 int err = 0;
4790 DEFINE_WAIT(wait);
4791
4792 ep = sctp_sk(sk)->ep;
4793
4794
4795 for (;;) {
4796 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
4797 TASK_INTERRUPTIBLE);
4798
4799 if (list_empty(&ep->asocs)) {
4800 sctp_release_sock(sk);
4801 timeo = schedule_timeout(timeo);
4802 sctp_lock_sock(sk);
4803 }
4804
4805 err = -EINVAL;
4806 if (!sctp_sstate(sk, LISTENING))
4807 break;
4808
4809 err = 0;
4810 if (!list_empty(&ep->asocs))
4811 break;
4812
4813 err = sock_intr_errno(timeo);
4814 if (signal_pending(current))
4815 break;
4816
4817 err = -EAGAIN;
4818 if (!timeo)
4819 break;
4820 }
4821
4822 finish_wait(sk->sk_sleep, &wait);
4823
4824 return err;
4825}
4826
4827void sctp_wait_for_close(struct sock *sk, long timeout)
4828{
4829 DEFINE_WAIT(wait);
4830
4831 do {
4832 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
4833 if (list_empty(&sctp_sk(sk)->ep->asocs))
4834 break;
4835 sctp_release_sock(sk);
4836 timeout = schedule_timeout(timeout);
4837 sctp_lock_sock(sk);
4838 } while (!signal_pending(current) && timeout);
4839
4840 finish_wait(sk->sk_sleep, &wait);
4841}
4842
4843/* Populate the fields of the newsk from the oldsk and migrate the assoc
4844 * and its messages to the newsk.
4845 */
4846static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
4847 struct sctp_association *assoc,
4848 sctp_socket_type_t type)
4849{
4850 struct sctp_sock *oldsp = sctp_sk(oldsk);
4851 struct sctp_sock *newsp = sctp_sk(newsk);
4852 struct sctp_bind_bucket *pp; /* hash list port iterator */
4853 struct sctp_endpoint *newep = newsp->ep;
4854 struct sk_buff *skb, *tmp;
4855 struct sctp_ulpevent *event;
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07004856 int flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004857
4858 /* Migrate socket buffer sizes and all the socket level options to the
4859 * new socket.
4860 */
4861 newsk->sk_sndbuf = oldsk->sk_sndbuf;
4862 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
4863 /* Brute force copy old sctp opt. */
4864 inet_sk_copy_descendant(newsk, oldsk);
4865
4866 /* Restore the ep value that was overwritten with the above structure
4867 * copy.
4868 */
4869 newsp->ep = newep;
4870 newsp->hmac = NULL;
4871
4872 /* Hook this new socket in to the bind_hash list. */
4873 pp = sctp_sk(oldsk)->bind_hash;
4874 sk_add_bind_node(newsk, &pp->owner);
4875 sctp_sk(newsk)->bind_hash = pp;
4876 inet_sk(newsk)->num = inet_sk(oldsk)->num;
4877
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07004878 /* Copy the bind_addr list from the original endpoint to the new
4879 * endpoint so that we can handle restarts properly
4880 */
4881 if (assoc->peer.ipv4_address)
4882 flags |= SCTP_ADDR4_PEERSUPP;
4883 if (assoc->peer.ipv6_address)
4884 flags |= SCTP_ADDR6_PEERSUPP;
4885 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
4886 &oldsp->ep->base.bind_addr,
4887 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
4888
Linus Torvalds1da177e2005-04-16 15:20:36 -07004889 /* Move any messages in the old socket's receive queue that are for the
4890 * peeled off association to the new socket's receive queue.
4891 */
4892 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
4893 event = sctp_skb2event(skb);
4894 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07004895 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004896 __skb_queue_tail(&newsk->sk_receive_queue, skb);
4897 }
4898 }
4899
4900 /* Clean up any messages pending delivery due to partial
4901 * delivery. Three cases:
4902 * 1) No partial deliver; no work.
4903 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
4904 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
4905 */
4906 skb_queue_head_init(&newsp->pd_lobby);
4907 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
4908
4909 if (sctp_sk(oldsk)->pd_mode) {
4910 struct sk_buff_head *queue;
4911
4912 /* Decide which queue to move pd_lobby skbs to. */
4913 if (assoc->ulpq.pd_mode) {
4914 queue = &newsp->pd_lobby;
4915 } else
4916 queue = &newsk->sk_receive_queue;
4917
4918 /* Walk through the pd_lobby, looking for skbs that
4919 * need moved to the new socket.
4920 */
4921 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
4922 event = sctp_skb2event(skb);
4923 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07004924 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004925 __skb_queue_tail(queue, skb);
4926 }
4927 }
4928
4929 /* Clear up any skbs waiting for the partial
4930 * delivery to finish.
4931 */
4932 if (assoc->ulpq.pd_mode)
4933 sctp_clear_pd(oldsk);
4934
4935 }
4936
4937 /* Set the type of socket to indicate that it is peeled off from the
4938 * original UDP-style socket or created with the accept() call on a
4939 * TCP-style socket..
4940 */
4941 newsp->type = type;
4942
4943 /* Migrate the association to the new socket. */
4944 sctp_assoc_migrate(assoc, newsk);
4945
4946 /* If the association on the newsk is already closed before accept()
4947 * is called, set RCV_SHUTDOWN flag.
4948 */
4949 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
4950 newsk->sk_shutdown |= RCV_SHUTDOWN;
4951
4952 newsk->sk_state = SCTP_SS_ESTABLISHED;
4953}
4954
4955/* This proto struct describes the ULP interface for SCTP. */
4956struct proto sctp_prot = {
4957 .name = "SCTP",
4958 .owner = THIS_MODULE,
4959 .close = sctp_close,
4960 .connect = sctp_connect,
4961 .disconnect = sctp_disconnect,
4962 .accept = sctp_accept,
4963 .ioctl = sctp_ioctl,
4964 .init = sctp_init_sock,
4965 .destroy = sctp_destroy_sock,
4966 .shutdown = sctp_shutdown,
4967 .setsockopt = sctp_setsockopt,
4968 .getsockopt = sctp_getsockopt,
4969 .sendmsg = sctp_sendmsg,
4970 .recvmsg = sctp_recvmsg,
4971 .bind = sctp_bind,
4972 .backlog_rcv = sctp_backlog_rcv,
4973 .hash = sctp_hash,
4974 .unhash = sctp_unhash,
4975 .get_port = sctp_get_port,
4976 .obj_size = sizeof(struct sctp_sock),
4977};
4978
4979#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
4980struct proto sctpv6_prot = {
4981 .name = "SCTPv6",
4982 .owner = THIS_MODULE,
4983 .close = sctp_close,
4984 .connect = sctp_connect,
4985 .disconnect = sctp_disconnect,
4986 .accept = sctp_accept,
4987 .ioctl = sctp_ioctl,
4988 .init = sctp_init_sock,
4989 .destroy = sctp_destroy_sock,
4990 .shutdown = sctp_shutdown,
4991 .setsockopt = sctp_setsockopt,
4992 .getsockopt = sctp_getsockopt,
4993 .sendmsg = sctp_sendmsg,
4994 .recvmsg = sctp_recvmsg,
4995 .bind = sctp_bind,
4996 .backlog_rcv = sctp_backlog_rcv,
4997 .hash = sctp_hash,
4998 .unhash = sctp_unhash,
4999 .get_port = sctp_get_port,
5000 .obj_size = sizeof(struct sctp6_sock),
5001};
5002#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */