blob: 79c3e072cf282027f3988c17d90bbc2c6c2143b8 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel reference Implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * The SCTP reference implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#include <linux/types.h>
61#include <linux/kernel.h>
62#include <linux/wait.h>
63#include <linux/time.h>
64#include <linux/ip.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080065#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066#include <linux/fcntl.h>
67#include <linux/poll.h>
68#include <linux/init.h>
69#include <linux/crypto.h>
70
71#include <net/ip.h>
72#include <net/icmp.h>
73#include <net/route.h>
74#include <net/ipv6.h>
75#include <net/inet_common.h>
76
77#include <linux/socket.h> /* for sa_family_t */
78#include <net/sock.h>
79#include <net/sctp/sctp.h>
80#include <net/sctp/sm.h>
81
82/* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
85 */
86
87/* Forward declarations for internal helper functions. */
88static int sctp_writeable(struct sock *sk);
89static void sctp_wfree(struct sk_buff *skb);
90static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94static int sctp_wait_for_accept(struct sock *sk, long timeo);
95static void sctp_wait_for_close(struct sock *sk, long timeo);
96static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105static int sctp_autobind(struct sock *sk);
106static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
109
110extern kmem_cache_t *sctp_bucket_cachep;
111
112/* Get the sndbuf space available at the time on the association. */
113static inline int sctp_wspace(struct sctp_association *asoc)
114{
115 struct sock *sk = asoc->base.sk;
116 int amt = 0;
117
Neil Horman4eb701d2005-04-28 12:02:04 -0700118 if (asoc->ep->sndbuf_policy) {
119 /* make sure that no association uses more than sk_sndbuf */
120 amt = sk->sk_sndbuf - asoc->sndbuf_used;
121 } else {
122 /* do socket level accounting */
123 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
124 }
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 if (amt < 0)
127 amt = 0;
Neil Horman4eb701d2005-04-28 12:02:04 -0700128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 return amt;
130}
131
132/* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
135 *
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
139 * tracking.
140 */
141static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
142{
143 struct sctp_association *asoc = chunk->asoc;
144 struct sock *sk = asoc->base.sk;
145
146 /* The sndbuf space is tracked per association. */
147 sctp_association_hold(asoc);
148
Neil Horman4eb701d2005-04-28 12:02:04 -0700149 skb_set_owner_w(chunk->skb, sk);
150
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 chunk->skb->destructor = sctp_wfree;
152 /* Save the chunk pointer in skb for sctp_wfree to use later. */
153 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
154
Neil Horman4eb701d2005-04-28 12:02:04 -0700155 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
156 sizeof(struct sk_buff) +
157 sizeof(struct sctp_chunk);
158
Neil Horman4eb701d2005-04-28 12:02:04 -0700159 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160}
161
162/* Verify that this is a valid address. */
163static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
164 int len)
165{
166 struct sctp_af *af;
167
168 /* Verify basic sockaddr. */
169 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
170 if (!af)
171 return -EINVAL;
172
173 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700174 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 return -EINVAL;
176
177 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
178 return -EINVAL;
179
180 return 0;
181}
182
183/* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
185 */
186struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
187{
188 struct sctp_association *asoc = NULL;
189
190 /* If this is not a UDP-style socket, assoc id should be ignored. */
191 if (!sctp_style(sk, UDP)) {
192 /* Return NULL if the socket state is not ESTABLISHED. It
193 * could be a TCP-style listening socket or a socket which
194 * hasn't yet called connect() to establish an association.
195 */
196 if (!sctp_sstate(sk, ESTABLISHED))
197 return NULL;
198
199 /* Get the first and the only association from the list. */
200 if (!list_empty(&sctp_sk(sk)->ep->asocs))
201 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
202 struct sctp_association, asocs);
203 return asoc;
204 }
205
206 /* Otherwise this is a UDP-style socket. */
207 if (!id || (id == (sctp_assoc_t)-1))
208 return NULL;
209
210 spin_lock_bh(&sctp_assocs_id_lock);
211 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
212 spin_unlock_bh(&sctp_assocs_id_lock);
213
214 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
215 return NULL;
216
217 return asoc;
218}
219
220/* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
222 * the same.
223 */
224static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
226 sctp_assoc_t id)
227{
228 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
229 struct sctp_transport *transport;
230 union sctp_addr *laddr = (union sctp_addr *)addr;
231
232 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
233 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
234 (union sctp_addr *)addr,
235 &transport);
236 laddr->v4.sin_port = htons(laddr->v4.sin_port);
237
238 if (!addr_asoc)
239 return NULL;
240
241 id_asoc = sctp_id2assoc(sk, id);
242 if (id_asoc && (id_asoc != addr_asoc))
243 return NULL;
244
245 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
246 (union sctp_addr *)addr);
247
248 return transport;
249}
250
251/* API 3.1.2 bind() - UDP Style Syntax
252 * The syntax of bind() is,
253 *
254 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
255 *
256 * sd - the socket descriptor returned by socket().
257 * addr - the address structure (struct sockaddr_in or struct
258 * sockaddr_in6 [RFC 2553]),
259 * addr_len - the size of the address structure.
260 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700261SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262{
263 int retval = 0;
264
265 sctp_lock_sock(sk);
266
Frank Filz3f7a87d2005-06-20 13:14:57 -0700267 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
268 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270 /* Disallow binding twice. */
271 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700272 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 addr_len);
274 else
275 retval = -EINVAL;
276
277 sctp_release_sock(sk);
278
279 return retval;
280}
281
282static long sctp_get_port_local(struct sock *, union sctp_addr *);
283
284/* Verify this is a valid sockaddr. */
285static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
286 union sctp_addr *addr, int len)
287{
288 struct sctp_af *af;
289
290 /* Check minimum size. */
291 if (len < sizeof (struct sockaddr))
292 return NULL;
293
294 /* Does this PF support this AF? */
295 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
296 return NULL;
297
298 /* If we get this far, af is valid. */
299 af = sctp_get_af_specific(addr->sa.sa_family);
300
301 if (len < af->sockaddr_len)
302 return NULL;
303
304 return af;
305}
306
307/* Bind a local address either to an endpoint or to an association. */
308SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
309{
310 struct sctp_sock *sp = sctp_sk(sk);
311 struct sctp_endpoint *ep = sp->ep;
312 struct sctp_bind_addr *bp = &ep->base.bind_addr;
313 struct sctp_af *af;
314 unsigned short snum;
315 int ret = 0;
316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 /* Common sockaddr verification. */
318 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700319 if (!af) {
320 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
321 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700323 }
324
325 snum = ntohs(addr->v4.sin_port);
326
327 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
328 ", port: %d, new port: %d, len: %d)\n",
329 sk,
330 addr,
331 bp->port, snum,
332 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334 /* PF specific bind() address verification. */
335 if (!sp->pf->bind_verify(sp, addr))
336 return -EADDRNOTAVAIL;
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 /* We must either be unbound, or bind to the same port. */
339 if (bp->port && (snum != bp->port)) {
340 SCTP_DEBUG_PRINTK("sctp_do_bind:"
341 " New port %d does not match existing port "
342 "%d.\n", snum, bp->port);
343 return -EINVAL;
344 }
345
346 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
347 return -EACCES;
348
349 /* Make sure we are allowed to bind here.
350 * The function sctp_get_port_local() does duplicate address
351 * detection.
352 */
353 if ((ret = sctp_get_port_local(sk, addr))) {
354 if (ret == (long) sk) {
355 /* This endpoint has a conflicting address. */
356 return -EINVAL;
357 } else {
358 return -EADDRINUSE;
359 }
360 }
361
362 /* Refresh ephemeral port. */
363 if (!bp->port)
364 bp->port = inet_sk(sk)->num;
365
366 /* Add the address to the bind address list. */
367 sctp_local_bh_disable();
368 sctp_write_lock(&ep->base.addr_lock);
369
370 /* Use GFP_ATOMIC since BHs are disabled. */
371 addr->v4.sin_port = ntohs(addr->v4.sin_port);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700372 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 addr->v4.sin_port = htons(addr->v4.sin_port);
374 sctp_write_unlock(&ep->base.addr_lock);
375 sctp_local_bh_enable();
376
377 /* Copy back into socket for getsockname() use. */
378 if (!ret) {
379 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
380 af->to_sk_saddr(addr, sk);
381 }
382
383 return ret;
384}
385
386 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
387 *
388 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
389 * at any one time. If a sender, after sending an ASCONF chunk, decides
390 * it needs to transfer another ASCONF Chunk, it MUST wait until the
391 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
392 * subsequent ASCONF. Note this restriction binds each side, so at any
393 * time two ASCONF may be in-transit on any given association (one sent
394 * from each endpoint).
395 */
396static int sctp_send_asconf(struct sctp_association *asoc,
397 struct sctp_chunk *chunk)
398{
399 int retval = 0;
400
401 /* If there is an outstanding ASCONF chunk, queue it for later
402 * transmission.
403 */
404 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700405 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 goto out;
407 }
408
409 /* Hold the chunk until an ASCONF_ACK is received. */
410 sctp_chunk_hold(chunk);
411 retval = sctp_primitive_ASCONF(asoc, chunk);
412 if (retval)
413 sctp_chunk_free(chunk);
414 else
415 asoc->addip_last_asconf = chunk;
416
417out:
418 return retval;
419}
420
421/* Add a list of addresses as bind addresses to local endpoint or
422 * association.
423 *
424 * Basically run through each address specified in the addrs/addrcnt
425 * array/length pair, determine if it is IPv6 or IPv4 and call
426 * sctp_do_bind() on it.
427 *
428 * If any of them fails, then the operation will be reversed and the
429 * ones that were added will be removed.
430 *
431 * Only sctp_setsockopt_bindx() is supposed to call this function.
432 */
433int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
434{
435 int cnt;
436 int retval = 0;
437 void *addr_buf;
438 struct sockaddr *sa_addr;
439 struct sctp_af *af;
440
441 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
442 sk, addrs, addrcnt);
443
444 addr_buf = addrs;
445 for (cnt = 0; cnt < addrcnt; cnt++) {
446 /* The list may contain either IPv4 or IPv6 address;
447 * determine the address length for walking thru the list.
448 */
449 sa_addr = (struct sockaddr *)addr_buf;
450 af = sctp_get_af_specific(sa_addr->sa_family);
451 if (!af) {
452 retval = -EINVAL;
453 goto err_bindx_add;
454 }
455
456 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
457 af->sockaddr_len);
458
459 addr_buf += af->sockaddr_len;
460
461err_bindx_add:
462 if (retval < 0) {
463 /* Failed. Cleanup the ones that have been added */
464 if (cnt > 0)
465 sctp_bindx_rem(sk, addrs, cnt);
466 return retval;
467 }
468 }
469
470 return retval;
471}
472
473/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
474 * associations that are part of the endpoint indicating that a list of local
475 * addresses are added to the endpoint.
476 *
477 * If any of the addresses is already in the bind address list of the
478 * association, we do not send the chunk for that association. But it will not
479 * affect other associations.
480 *
481 * Only sctp_setsockopt_bindx() is supposed to call this function.
482 */
483static int sctp_send_asconf_add_ip(struct sock *sk,
484 struct sockaddr *addrs,
485 int addrcnt)
486{
487 struct sctp_sock *sp;
488 struct sctp_endpoint *ep;
489 struct sctp_association *asoc;
490 struct sctp_bind_addr *bp;
491 struct sctp_chunk *chunk;
492 struct sctp_sockaddr_entry *laddr;
493 union sctp_addr *addr;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700494 union sctp_addr saveaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 void *addr_buf;
496 struct sctp_af *af;
497 struct list_head *pos;
498 struct list_head *p;
499 int i;
500 int retval = 0;
501
502 if (!sctp_addip_enable)
503 return retval;
504
505 sp = sctp_sk(sk);
506 ep = sp->ep;
507
508 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
509 __FUNCTION__, sk, addrs, addrcnt);
510
511 list_for_each(pos, &ep->asocs) {
512 asoc = list_entry(pos, struct sctp_association, asocs);
513
514 if (!asoc->peer.asconf_capable)
515 continue;
516
517 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
518 continue;
519
520 if (!sctp_state(asoc, ESTABLISHED))
521 continue;
522
523 /* Check if any address in the packed array of addresses is
524 * in the bind address list of the association. If so,
525 * do not send the asconf chunk to its peer, but continue with
526 * other associations.
527 */
528 addr_buf = addrs;
529 for (i = 0; i < addrcnt; i++) {
530 addr = (union sctp_addr *)addr_buf;
531 af = sctp_get_af_specific(addr->v4.sin_family);
532 if (!af) {
533 retval = -EINVAL;
534 goto out;
535 }
536
537 if (sctp_assoc_lookup_laddr(asoc, addr))
538 break;
539
540 addr_buf += af->sockaddr_len;
541 }
542 if (i < addrcnt)
543 continue;
544
545 /* Use the first address in bind addr list of association as
546 * Address Parameter of ASCONF CHUNK.
547 */
548 sctp_read_lock(&asoc->base.addr_lock);
549 bp = &asoc->base.bind_addr;
550 p = bp->address_list.next;
551 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
552 sctp_read_unlock(&asoc->base.addr_lock);
553
554 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
555 addrcnt, SCTP_PARAM_ADD_IP);
556 if (!chunk) {
557 retval = -ENOMEM;
558 goto out;
559 }
560
561 retval = sctp_send_asconf(asoc, chunk);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700562 if (retval)
563 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700565 /* Add the new addresses to the bind address list with
566 * use_as_src set to 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700568 sctp_local_bh_disable();
569 sctp_write_lock(&asoc->base.addr_lock);
570 addr_buf = addrs;
571 for (i = 0; i < addrcnt; i++) {
572 addr = (union sctp_addr *)addr_buf;
573 af = sctp_get_af_specific(addr->v4.sin_family);
574 memcpy(&saveaddr, addr, af->sockaddr_len);
575 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
576 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
577 GFP_ATOMIC);
578 addr_buf += af->sockaddr_len;
579 }
580 sctp_write_unlock(&asoc->base.addr_lock);
581 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 }
583
584out:
585 return retval;
586}
587
588/* Remove a list of addresses from bind addresses list. Do not remove the
589 * last address.
590 *
591 * Basically run through each address specified in the addrs/addrcnt
592 * array/length pair, determine if it is IPv6 or IPv4 and call
593 * sctp_del_bind() on it.
594 *
595 * If any of them fails, then the operation will be reversed and the
596 * ones that were removed will be added back.
597 *
598 * At least one address has to be left; if only one address is
599 * available, the operation will return -EBUSY.
600 *
601 * Only sctp_setsockopt_bindx() is supposed to call this function.
602 */
603int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
604{
605 struct sctp_sock *sp = sctp_sk(sk);
606 struct sctp_endpoint *ep = sp->ep;
607 int cnt;
608 struct sctp_bind_addr *bp = &ep->base.bind_addr;
609 int retval = 0;
610 union sctp_addr saveaddr;
611 void *addr_buf;
612 struct sockaddr *sa_addr;
613 struct sctp_af *af;
614
615 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
616 sk, addrs, addrcnt);
617
618 addr_buf = addrs;
619 for (cnt = 0; cnt < addrcnt; cnt++) {
620 /* If the bind address list is empty or if there is only one
621 * bind address, there is nothing more to be removed (we need
622 * at least one address here).
623 */
624 if (list_empty(&bp->address_list) ||
625 (sctp_list_single_entry(&bp->address_list))) {
626 retval = -EBUSY;
627 goto err_bindx_rem;
628 }
629
630 /* The list may contain either IPv4 or IPv6 address;
631 * determine the address length to copy the address to
632 * saveaddr.
633 */
634 sa_addr = (struct sockaddr *)addr_buf;
635 af = sctp_get_af_specific(sa_addr->sa_family);
636 if (!af) {
637 retval = -EINVAL;
638 goto err_bindx_rem;
639 }
640 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
641 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
642 if (saveaddr.v4.sin_port != bp->port) {
643 retval = -EINVAL;
644 goto err_bindx_rem;
645 }
646
647 /* FIXME - There is probably a need to check if sk->sk_saddr and
648 * sk->sk_rcv_addr are currently set to one of the addresses to
649 * be removed. This is something which needs to be looked into
650 * when we are fixing the outstanding issues with multi-homing
651 * socket routing and failover schemes. Refer to comments in
652 * sctp_do_bind(). -daisy
653 */
654 sctp_local_bh_disable();
655 sctp_write_lock(&ep->base.addr_lock);
656
657 retval = sctp_del_bind_addr(bp, &saveaddr);
658
659 sctp_write_unlock(&ep->base.addr_lock);
660 sctp_local_bh_enable();
661
662 addr_buf += af->sockaddr_len;
663err_bindx_rem:
664 if (retval < 0) {
665 /* Failed. Add the ones that has been removed back */
666 if (cnt > 0)
667 sctp_bindx_add(sk, addrs, cnt);
668 return retval;
669 }
670 }
671
672 return retval;
673}
674
675/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
676 * the associations that are part of the endpoint indicating that a list of
677 * local addresses are removed from the endpoint.
678 *
679 * If any of the addresses is already in the bind address list of the
680 * association, we do not send the chunk for that association. But it will not
681 * affect other associations.
682 *
683 * Only sctp_setsockopt_bindx() is supposed to call this function.
684 */
685static int sctp_send_asconf_del_ip(struct sock *sk,
686 struct sockaddr *addrs,
687 int addrcnt)
688{
689 struct sctp_sock *sp;
690 struct sctp_endpoint *ep;
691 struct sctp_association *asoc;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700692 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 struct sctp_bind_addr *bp;
694 struct sctp_chunk *chunk;
695 union sctp_addr *laddr;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700696 union sctp_addr saveaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 void *addr_buf;
698 struct sctp_af *af;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700699 struct list_head *pos, *pos1;
700 struct sctp_sockaddr_entry *saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 int i;
702 int retval = 0;
703
704 if (!sctp_addip_enable)
705 return retval;
706
707 sp = sctp_sk(sk);
708 ep = sp->ep;
709
710 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 __FUNCTION__, sk, addrs, addrcnt);
712
713 list_for_each(pos, &ep->asocs) {
714 asoc = list_entry(pos, struct sctp_association, asocs);
715
716 if (!asoc->peer.asconf_capable)
717 continue;
718
719 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
720 continue;
721
722 if (!sctp_state(asoc, ESTABLISHED))
723 continue;
724
725 /* Check if any address in the packed array of addresses is
726 * not present in the bind address list of the association.
727 * If so, do not send the asconf chunk to its peer, but
728 * continue with other associations.
729 */
730 addr_buf = addrs;
731 for (i = 0; i < addrcnt; i++) {
732 laddr = (union sctp_addr *)addr_buf;
733 af = sctp_get_af_specific(laddr->v4.sin_family);
734 if (!af) {
735 retval = -EINVAL;
736 goto out;
737 }
738
739 if (!sctp_assoc_lookup_laddr(asoc, laddr))
740 break;
741
742 addr_buf += af->sockaddr_len;
743 }
744 if (i < addrcnt)
745 continue;
746
747 /* Find one address in the association's bind address list
748 * that is not in the packed array of addresses. This is to
749 * make sure that we do not delete all the addresses in the
750 * association.
751 */
752 sctp_read_lock(&asoc->base.addr_lock);
753 bp = &asoc->base.bind_addr;
754 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
755 addrcnt, sp);
756 sctp_read_unlock(&asoc->base.addr_lock);
757 if (!laddr)
758 continue;
759
760 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
761 SCTP_PARAM_DEL_IP);
762 if (!chunk) {
763 retval = -ENOMEM;
764 goto out;
765 }
766
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700767 /* Reset use_as_src flag for the addresses in the bind address
768 * list that are to be deleted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700770 sctp_local_bh_disable();
771 sctp_write_lock(&asoc->base.addr_lock);
772 addr_buf = addrs;
773 for (i = 0; i < addrcnt; i++) {
774 laddr = (union sctp_addr *)addr_buf;
775 af = sctp_get_af_specific(laddr->v4.sin_family);
776 memcpy(&saveaddr, laddr, af->sockaddr_len);
777 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
778 list_for_each(pos1, &bp->address_list) {
779 saddr = list_entry(pos1,
780 struct sctp_sockaddr_entry,
781 list);
782 if (sctp_cmp_addr_exact(&saddr->a, &saveaddr))
783 saddr->use_as_src = 0;
784 }
785 addr_buf += af->sockaddr_len;
786 }
787 sctp_write_unlock(&asoc->base.addr_lock);
788 sctp_local_bh_enable();
789
790 /* Update the route and saddr entries for all the transports
791 * as some of the addresses in the bind address list are
792 * about to be deleted and cannot be used as source addresses.
793 */
794 list_for_each(pos1, &asoc->peer.transport_addr_list) {
795 transport = list_entry(pos1, struct sctp_transport,
796 transports);
797 dst_release(transport->dst);
798 sctp_transport_route(transport, NULL,
799 sctp_sk(asoc->base.sk));
800 }
801
802 retval = sctp_send_asconf(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 }
804out:
805 return retval;
806}
807
808/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809 *
810 * API 8.1
811 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
812 * int flags);
813 *
814 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
815 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
816 * or IPv6 addresses.
817 *
818 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
819 * Section 3.1.2 for this usage.
820 *
821 * addrs is a pointer to an array of one or more socket addresses. Each
822 * address is contained in its appropriate structure (i.e. struct
823 * sockaddr_in or struct sockaddr_in6) the family of the address type
824 * must be used to distengish the address length (note that this
825 * representation is termed a "packed array" of addresses). The caller
826 * specifies the number of addresses in the array with addrcnt.
827 *
828 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
829 * -1, and sets errno to the appropriate error code.
830 *
831 * For SCTP, the port given in each socket address must be the same, or
832 * sctp_bindx() will fail, setting errno to EINVAL.
833 *
834 * The flags parameter is formed from the bitwise OR of zero or more of
835 * the following currently defined flags:
836 *
837 * SCTP_BINDX_ADD_ADDR
838 *
839 * SCTP_BINDX_REM_ADDR
840 *
841 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
842 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
843 * addresses from the association. The two flags are mutually exclusive;
844 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
845 * not remove all addresses from an association; sctp_bindx() will
846 * reject such an attempt with EINVAL.
847 *
848 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
849 * additional addresses with an endpoint after calling bind(). Or use
850 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
851 * socket is associated with so that no new association accepted will be
852 * associated with those addresses. If the endpoint supports dynamic
853 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
854 * endpoint to send the appropriate message to the peer to change the
855 * peers address lists.
856 *
857 * Adding and removing addresses from a connected association is
858 * optional functionality. Implementations that do not support this
859 * functionality should return EOPNOTSUPP.
860 *
861 * Basically do nothing but copying the addresses from user to kernel
862 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700863 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
864 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 *
866 * We don't use copy_from_user() for optimization: we first do the
867 * sanity checks (buffer size -fast- and access check-healthy
868 * pointer); if all of those succeed, then we can alloc the memory
869 * (expensive operation) needed to copy the data to kernel. Then we do
870 * the copying without checking the user space area
871 * (__copy_from_user()).
872 *
873 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
874 * it.
875 *
876 * sk The sk of the socket
877 * addrs The pointer to the addresses in user land
878 * addrssize Size of the addrs buffer
879 * op Operation to perform (add or remove, see the flags of
880 * sctp_bindx)
881 *
882 * Returns 0 if ok, <0 errno code on error.
883 */
884SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
885 struct sockaddr __user *addrs,
886 int addrs_size, int op)
887{
888 struct sockaddr *kaddrs;
889 int err;
890 int addrcnt = 0;
891 int walk_size = 0;
892 struct sockaddr *sa_addr;
893 void *addr_buf;
894 struct sctp_af *af;
895
896 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
897 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
898
899 if (unlikely(addrs_size <= 0))
900 return -EINVAL;
901
902 /* Check the user passed a healthy pointer. */
903 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
904 return -EFAULT;
905
906 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800907 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 if (unlikely(!kaddrs))
909 return -ENOMEM;
910
911 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
912 kfree(kaddrs);
913 return -EFAULT;
914 }
915
916 /* Walk through the addrs buffer and count the number of addresses. */
917 addr_buf = kaddrs;
918 while (walk_size < addrs_size) {
919 sa_addr = (struct sockaddr *)addr_buf;
920 af = sctp_get_af_specific(sa_addr->sa_family);
921
922 /* If the address family is not supported or if this address
923 * causes the address buffer to overflow return EINVAL.
924 */
925 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
926 kfree(kaddrs);
927 return -EINVAL;
928 }
929 addrcnt++;
930 addr_buf += af->sockaddr_len;
931 walk_size += af->sockaddr_len;
932 }
933
934 /* Do the work. */
935 switch (op) {
936 case SCTP_BINDX_ADD_ADDR:
937 err = sctp_bindx_add(sk, kaddrs, addrcnt);
938 if (err)
939 goto out;
940 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
941 break;
942
943 case SCTP_BINDX_REM_ADDR:
944 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
945 if (err)
946 goto out;
947 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
948 break;
949
950 default:
951 err = -EINVAL;
952 break;
953 };
954
955out:
956 kfree(kaddrs);
957
958 return err;
959}
960
Frank Filz3f7a87d2005-06-20 13:14:57 -0700961/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962 *
963 * Common routine for handling connect() and sctp_connectx().
964 * Connect will come in with just a single address.
965 */
966static int __sctp_connect(struct sock* sk,
967 struct sockaddr *kaddrs,
968 int addrs_size)
969{
970 struct sctp_sock *sp;
971 struct sctp_endpoint *ep;
972 struct sctp_association *asoc = NULL;
973 struct sctp_association *asoc2;
974 struct sctp_transport *transport;
975 union sctp_addr to;
976 struct sctp_af *af;
977 sctp_scope_t scope;
978 long timeo;
979 int err = 0;
980 int addrcnt = 0;
981 int walk_size = 0;
982 struct sockaddr *sa_addr;
983 void *addr_buf;
984
985 sp = sctp_sk(sk);
986 ep = sp->ep;
987
988 /* connect() cannot be done on a socket that is already in ESTABLISHED
989 * state - UDP-style peeled off socket or a TCP-style socket that
990 * is already connected.
991 * It cannot be done even on a TCP-style listening socket.
992 */
993 if (sctp_sstate(sk, ESTABLISHED) ||
994 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
995 err = -EISCONN;
996 goto out_free;
997 }
998
999 /* Walk through the addrs buffer and count the number of addresses. */
1000 addr_buf = kaddrs;
1001 while (walk_size < addrs_size) {
1002 sa_addr = (struct sockaddr *)addr_buf;
1003 af = sctp_get_af_specific(sa_addr->sa_family);
1004
1005 /* If the address family is not supported or if this address
1006 * causes the address buffer to overflow return EINVAL.
1007 */
1008 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1009 err = -EINVAL;
1010 goto out_free;
1011 }
1012
1013 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
1014 af->sockaddr_len);
1015 if (err)
1016 goto out_free;
1017
1018 memcpy(&to, sa_addr, af->sockaddr_len);
1019 to.v4.sin_port = ntohs(to.v4.sin_port);
1020
1021 /* Check if there already is a matching association on the
1022 * endpoint (other than the one created here).
1023 */
1024 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1025 if (asoc2 && asoc2 != asoc) {
1026 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1027 err = -EISCONN;
1028 else
1029 err = -EALREADY;
1030 goto out_free;
1031 }
1032
1033 /* If we could not find a matching association on the endpoint,
1034 * make sure that there is no peeled-off association matching
1035 * the peer address even on another socket.
1036 */
1037 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1038 err = -EADDRNOTAVAIL;
1039 goto out_free;
1040 }
1041
1042 if (!asoc) {
1043 /* If a bind() or sctp_bindx() is not called prior to
1044 * an sctp_connectx() call, the system picks an
1045 * ephemeral port and will choose an address set
1046 * equivalent to binding with a wildcard address.
1047 */
1048 if (!ep->base.bind_addr.port) {
1049 if (sctp_autobind(sk)) {
1050 err = -EAGAIN;
1051 goto out_free;
1052 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001053 } else {
1054 /*
1055 * If an unprivileged user inherits a 1-many
1056 * style socket with open associations on a
1057 * privileged port, it MAY be permitted to
1058 * accept new associations, but it SHOULD NOT
1059 * be permitted to open new associations.
1060 */
1061 if (ep->base.bind_addr.port < PROT_SOCK &&
1062 !capable(CAP_NET_BIND_SERVICE)) {
1063 err = -EACCES;
1064 goto out_free;
1065 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001066 }
1067
1068 scope = sctp_scope(&to);
1069 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1070 if (!asoc) {
1071 err = -ENOMEM;
1072 goto out_free;
1073 }
1074 }
1075
1076 /* Prime the peer's transport structures. */
1077 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1078 SCTP_UNKNOWN);
1079 if (!transport) {
1080 err = -ENOMEM;
1081 goto out_free;
1082 }
1083
1084 addrcnt++;
1085 addr_buf += af->sockaddr_len;
1086 walk_size += af->sockaddr_len;
1087 }
1088
1089 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1090 if (err < 0) {
1091 goto out_free;
1092 }
1093
1094 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1095 if (err < 0) {
1096 goto out_free;
1097 }
1098
1099 /* Initialize sk's dport and daddr for getpeername() */
1100 inet_sk(sk)->dport = htons(asoc->peer.port);
1101 af = sctp_get_af_specific(to.sa.sa_family);
1102 af->to_sk_daddr(&to, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001103 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001104
1105 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1106 err = sctp_wait_for_connect(asoc, &timeo);
1107
1108 /* Don't free association on exit. */
1109 asoc = NULL;
1110
1111out_free:
1112
1113 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1114 " kaddrs: %p err: %d\n",
1115 asoc, kaddrs, err);
1116 if (asoc)
1117 sctp_association_free(asoc);
1118 return err;
1119}
1120
1121/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1122 *
1123 * API 8.9
1124 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1125 *
1126 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1127 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1128 * or IPv6 addresses.
1129 *
1130 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1131 * Section 3.1.2 for this usage.
1132 *
1133 * addrs is a pointer to an array of one or more socket addresses. Each
1134 * address is contained in its appropriate structure (i.e. struct
1135 * sockaddr_in or struct sockaddr_in6) the family of the address type
1136 * must be used to distengish the address length (note that this
1137 * representation is termed a "packed array" of addresses). The caller
1138 * specifies the number of addresses in the array with addrcnt.
1139 *
1140 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1141 * -1, and sets errno to the appropriate error code.
1142 *
1143 * For SCTP, the port given in each socket address must be the same, or
1144 * sctp_connectx() will fail, setting errno to EINVAL.
1145 *
1146 * An application can use sctp_connectx to initiate an association with
1147 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1148 * allows a caller to specify multiple addresses at which a peer can be
1149 * reached. The way the SCTP stack uses the list of addresses to set up
1150 * the association is implementation dependant. This function only
1151 * specifies that the stack will try to make use of all the addresses in
1152 * the list when needed.
1153 *
1154 * Note that the list of addresses passed in is only used for setting up
1155 * the association. It does not necessarily equal the set of addresses
1156 * the peer uses for the resulting association. If the caller wants to
1157 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1158 * retrieve them after the association has been set up.
1159 *
1160 * Basically do nothing but copying the addresses from user to kernel
1161 * land and invoking either sctp_connectx(). This is used for tunneling
1162 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1163 *
1164 * We don't use copy_from_user() for optimization: we first do the
1165 * sanity checks (buffer size -fast- and access check-healthy
1166 * pointer); if all of those succeed, then we can alloc the memory
1167 * (expensive operation) needed to copy the data to kernel. Then we do
1168 * the copying without checking the user space area
1169 * (__copy_from_user()).
1170 *
1171 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1172 * it.
1173 *
1174 * sk The sk of the socket
1175 * addrs The pointer to the addresses in user land
1176 * addrssize Size of the addrs buffer
1177 *
1178 * Returns 0 if ok, <0 errno code on error.
1179 */
1180SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1181 struct sockaddr __user *addrs,
1182 int addrs_size)
1183{
1184 int err = 0;
1185 struct sockaddr *kaddrs;
1186
1187 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1188 __FUNCTION__, sk, addrs, addrs_size);
1189
1190 if (unlikely(addrs_size <= 0))
1191 return -EINVAL;
1192
1193 /* Check the user passed a healthy pointer. */
1194 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1195 return -EFAULT;
1196
1197 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001198 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001199 if (unlikely(!kaddrs))
1200 return -ENOMEM;
1201
1202 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1203 err = -EFAULT;
1204 } else {
1205 err = __sctp_connect(sk, kaddrs, addrs_size);
1206 }
1207
1208 kfree(kaddrs);
1209 return err;
1210}
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212/* API 3.1.4 close() - UDP Style Syntax
1213 * Applications use close() to perform graceful shutdown (as described in
1214 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1215 * by a UDP-style socket.
1216 *
1217 * The syntax is
1218 *
1219 * ret = close(int sd);
1220 *
1221 * sd - the socket descriptor of the associations to be closed.
1222 *
1223 * To gracefully shutdown a specific association represented by the
1224 * UDP-style socket, an application should use the sendmsg() call,
1225 * passing no user data, but including the appropriate flag in the
1226 * ancillary data (see Section xxxx).
1227 *
1228 * If sd in the close() call is a branched-off socket representing only
1229 * one association, the shutdown is performed on that association only.
1230 *
1231 * 4.1.6 close() - TCP Style Syntax
1232 *
1233 * Applications use close() to gracefully close down an association.
1234 *
1235 * The syntax is:
1236 *
1237 * int close(int sd);
1238 *
1239 * sd - the socket descriptor of the association to be closed.
1240 *
1241 * After an application calls close() on a socket descriptor, no further
1242 * socket operations will succeed on that descriptor.
1243 *
1244 * API 7.1.4 SO_LINGER
1245 *
1246 * An application using the TCP-style socket can use this option to
1247 * perform the SCTP ABORT primitive. The linger option structure is:
1248 *
1249 * struct linger {
1250 * int l_onoff; // option on/off
1251 * int l_linger; // linger time
1252 * };
1253 *
1254 * To enable the option, set l_onoff to 1. If the l_linger value is set
1255 * to 0, calling close() is the same as the ABORT primitive. If the
1256 * value is set to a negative value, the setsockopt() call will return
1257 * an error. If the value is set to a positive value linger_time, the
1258 * close() can be blocked for at most linger_time ms. If the graceful
1259 * shutdown phase does not finish during this period, close() will
1260 * return but the graceful shutdown phase continues in the system.
1261 */
1262SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1263{
1264 struct sctp_endpoint *ep;
1265 struct sctp_association *asoc;
1266 struct list_head *pos, *temp;
1267
1268 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1269
1270 sctp_lock_sock(sk);
1271 sk->sk_shutdown = SHUTDOWN_MASK;
1272
1273 ep = sctp_sk(sk)->ep;
1274
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001275 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 list_for_each_safe(pos, temp, &ep->asocs) {
1277 asoc = list_entry(pos, struct sctp_association, asocs);
1278
1279 if (sctp_style(sk, TCP)) {
1280 /* A closed association can still be in the list if
1281 * it belongs to a TCP-style listening socket that is
1282 * not yet accepted. If so, free it. If not, send an
1283 * ABORT or SHUTDOWN based on the linger options.
1284 */
1285 if (sctp_state(asoc, CLOSED)) {
1286 sctp_unhash_established(asoc);
1287 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001288 continue;
1289 }
1290 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291
Sridhar Samudralab9ac8672006-08-28 13:53:01 -07001292 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1293 struct sctp_chunk *chunk;
1294
1295 chunk = sctp_make_abort_user(asoc, NULL, 0);
1296 if (chunk)
1297 sctp_primitive_ABORT(asoc, chunk);
1298 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 sctp_primitive_SHUTDOWN(asoc, NULL);
1300 }
1301
1302 /* Clean up any skbs sitting on the receive queue. */
1303 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1304 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1305
1306 /* On a TCP-style socket, block for at most linger_time if set. */
1307 if (sctp_style(sk, TCP) && timeout)
1308 sctp_wait_for_close(sk, timeout);
1309
1310 /* This will run the backlog queue. */
1311 sctp_release_sock(sk);
1312
1313 /* Supposedly, no process has access to the socket, but
1314 * the net layers still may.
1315 */
1316 sctp_local_bh_disable();
1317 sctp_bh_lock_sock(sk);
1318
1319 /* Hold the sock, since sk_common_release() will put sock_put()
1320 * and we have just a little more cleanup.
1321 */
1322 sock_hold(sk);
1323 sk_common_release(sk);
1324
1325 sctp_bh_unlock_sock(sk);
1326 sctp_local_bh_enable();
1327
1328 sock_put(sk);
1329
1330 SCTP_DBG_OBJCNT_DEC(sock);
1331}
1332
1333/* Handle EPIPE error. */
1334static int sctp_error(struct sock *sk, int flags, int err)
1335{
1336 if (err == -EPIPE)
1337 err = sock_error(sk) ? : -EPIPE;
1338 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1339 send_sig(SIGPIPE, current, 0);
1340 return err;
1341}
1342
1343/* API 3.1.3 sendmsg() - UDP Style Syntax
1344 *
1345 * An application uses sendmsg() and recvmsg() calls to transmit data to
1346 * and receive data from its peer.
1347 *
1348 * ssize_t sendmsg(int socket, const struct msghdr *message,
1349 * int flags);
1350 *
1351 * socket - the socket descriptor of the endpoint.
1352 * message - pointer to the msghdr structure which contains a single
1353 * user message and possibly some ancillary data.
1354 *
1355 * See Section 5 for complete description of the data
1356 * structures.
1357 *
1358 * flags - flags sent or received with the user message, see Section
1359 * 5 for complete description of the flags.
1360 *
1361 * Note: This function could use a rewrite especially when explicit
1362 * connect support comes in.
1363 */
1364/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1365
1366SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1367
1368SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1369 struct msghdr *msg, size_t msg_len)
1370{
1371 struct sctp_sock *sp;
1372 struct sctp_endpoint *ep;
1373 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1374 struct sctp_transport *transport, *chunk_tp;
1375 struct sctp_chunk *chunk;
1376 union sctp_addr to;
1377 struct sockaddr *msg_name = NULL;
1378 struct sctp_sndrcvinfo default_sinfo = { 0 };
1379 struct sctp_sndrcvinfo *sinfo;
1380 struct sctp_initmsg *sinit;
1381 sctp_assoc_t associd = 0;
1382 sctp_cmsgs_t cmsgs = { NULL };
1383 int err;
1384 sctp_scope_t scope;
1385 long timeo;
1386 __u16 sinfo_flags = 0;
1387 struct sctp_datamsg *datamsg;
1388 struct list_head *pos;
1389 int msg_flags = msg->msg_flags;
1390
1391 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1392 sk, msg, msg_len);
1393
1394 err = 0;
1395 sp = sctp_sk(sk);
1396 ep = sp->ep;
1397
Frank Filz3f7a87d2005-06-20 13:14:57 -07001398 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399
1400 /* We cannot send a message over a TCP-style listening socket. */
1401 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1402 err = -EPIPE;
1403 goto out_nounlock;
1404 }
1405
1406 /* Parse out the SCTP CMSGs. */
1407 err = sctp_msghdr_parse(msg, &cmsgs);
1408
1409 if (err) {
1410 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1411 goto out_nounlock;
1412 }
1413
1414 /* Fetch the destination address for this packet. This
1415 * address only selects the association--it is not necessarily
1416 * the address we will send to.
1417 * For a peeled-off socket, msg_name is ignored.
1418 */
1419 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1420 int msg_namelen = msg->msg_namelen;
1421
1422 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1423 msg_namelen);
1424 if (err)
1425 return err;
1426
1427 if (msg_namelen > sizeof(to))
1428 msg_namelen = sizeof(to);
1429 memcpy(&to, msg->msg_name, msg_namelen);
1430 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1431 "0x%x:%u.\n",
1432 to.v4.sin_addr.s_addr, to.v4.sin_port);
1433
1434 to.v4.sin_port = ntohs(to.v4.sin_port);
1435 msg_name = msg->msg_name;
1436 }
1437
1438 sinfo = cmsgs.info;
1439 sinit = cmsgs.init;
1440
1441 /* Did the user specify SNDRCVINFO? */
1442 if (sinfo) {
1443 sinfo_flags = sinfo->sinfo_flags;
1444 associd = sinfo->sinfo_assoc_id;
1445 }
1446
1447 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1448 msg_len, sinfo_flags);
1449
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001450 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1451 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 err = -EINVAL;
1453 goto out_nounlock;
1454 }
1455
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001456 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1457 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1458 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 * the msg_iov set to the user abort reason.
1460 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001461 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1462 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 err = -EINVAL;
1464 goto out_nounlock;
1465 }
1466
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001467 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 * specified in msg_name.
1469 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001470 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 err = -EINVAL;
1472 goto out_nounlock;
1473 }
1474
1475 transport = NULL;
1476
1477 SCTP_DEBUG_PRINTK("About to look up association.\n");
1478
1479 sctp_lock_sock(sk);
1480
1481 /* If a msg_name has been specified, assume this is to be used. */
1482 if (msg_name) {
1483 /* Look for a matching association on the endpoint. */
1484 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1485 if (!asoc) {
1486 /* If we could not find a matching association on the
1487 * endpoint, make sure that it is not a TCP-style
1488 * socket that already has an association or there is
1489 * no peeled-off association on another socket.
1490 */
1491 if ((sctp_style(sk, TCP) &&
1492 sctp_sstate(sk, ESTABLISHED)) ||
1493 sctp_endpoint_is_peeled_off(ep, &to)) {
1494 err = -EADDRNOTAVAIL;
1495 goto out_unlock;
1496 }
1497 }
1498 } else {
1499 asoc = sctp_id2assoc(sk, associd);
1500 if (!asoc) {
1501 err = -EPIPE;
1502 goto out_unlock;
1503 }
1504 }
1505
1506 if (asoc) {
1507 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1508
1509 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1510 * socket that has an association in CLOSED state. This can
1511 * happen when an accepted socket has an association that is
1512 * already CLOSED.
1513 */
1514 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1515 err = -EPIPE;
1516 goto out_unlock;
1517 }
1518
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001519 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1521 asoc);
1522 sctp_primitive_SHUTDOWN(asoc, NULL);
1523 err = 0;
1524 goto out_unlock;
1525 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001526 if (sinfo_flags & SCTP_ABORT) {
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001527 struct sctp_chunk *chunk;
1528
1529 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1530 if (!chunk) {
1531 err = -ENOMEM;
1532 goto out_unlock;
1533 }
1534
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001536 sctp_primitive_ABORT(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 err = 0;
1538 goto out_unlock;
1539 }
1540 }
1541
1542 /* Do we need to create the association? */
1543 if (!asoc) {
1544 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1545
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001546 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 err = -EINVAL;
1548 goto out_unlock;
1549 }
1550
1551 /* Check for invalid stream against the stream counts,
1552 * either the default or the user specified stream counts.
1553 */
1554 if (sinfo) {
1555 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1556 /* Check against the defaults. */
1557 if (sinfo->sinfo_stream >=
1558 sp->initmsg.sinit_num_ostreams) {
1559 err = -EINVAL;
1560 goto out_unlock;
1561 }
1562 } else {
1563 /* Check against the requested. */
1564 if (sinfo->sinfo_stream >=
1565 sinit->sinit_num_ostreams) {
1566 err = -EINVAL;
1567 goto out_unlock;
1568 }
1569 }
1570 }
1571
1572 /*
1573 * API 3.1.2 bind() - UDP Style Syntax
1574 * If a bind() or sctp_bindx() is not called prior to a
1575 * sendmsg() call that initiates a new association, the
1576 * system picks an ephemeral port and will choose an address
1577 * set equivalent to binding with a wildcard address.
1578 */
1579 if (!ep->base.bind_addr.port) {
1580 if (sctp_autobind(sk)) {
1581 err = -EAGAIN;
1582 goto out_unlock;
1583 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001584 } else {
1585 /*
1586 * If an unprivileged user inherits a one-to-many
1587 * style socket with open associations on a privileged
1588 * port, it MAY be permitted to accept new associations,
1589 * but it SHOULD NOT be permitted to open new
1590 * associations.
1591 */
1592 if (ep->base.bind_addr.port < PROT_SOCK &&
1593 !capable(CAP_NET_BIND_SERVICE)) {
1594 err = -EACCES;
1595 goto out_unlock;
1596 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 }
1598
1599 scope = sctp_scope(&to);
1600 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1601 if (!new_asoc) {
1602 err = -ENOMEM;
1603 goto out_unlock;
1604 }
1605 asoc = new_asoc;
1606
1607 /* If the SCTP_INIT ancillary data is specified, set all
1608 * the association init values accordingly.
1609 */
1610 if (sinit) {
1611 if (sinit->sinit_num_ostreams) {
1612 asoc->c.sinit_num_ostreams =
1613 sinit->sinit_num_ostreams;
1614 }
1615 if (sinit->sinit_max_instreams) {
1616 asoc->c.sinit_max_instreams =
1617 sinit->sinit_max_instreams;
1618 }
1619 if (sinit->sinit_max_attempts) {
1620 asoc->max_init_attempts
1621 = sinit->sinit_max_attempts;
1622 }
1623 if (sinit->sinit_max_init_timeo) {
1624 asoc->max_init_timeo =
1625 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1626 }
1627 }
1628
1629 /* Prime the peer's transport structures. */
Frank Filz3f7a87d2005-06-20 13:14:57 -07001630 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 if (!transport) {
1632 err = -ENOMEM;
1633 goto out_free;
1634 }
1635 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1636 if (err < 0) {
1637 err = -ENOMEM;
1638 goto out_free;
1639 }
1640 }
1641
1642 /* ASSERT: we have a valid association at this point. */
1643 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1644
1645 if (!sinfo) {
1646 /* If the user didn't specify SNDRCVINFO, make up one with
1647 * some defaults.
1648 */
1649 default_sinfo.sinfo_stream = asoc->default_stream;
1650 default_sinfo.sinfo_flags = asoc->default_flags;
1651 default_sinfo.sinfo_ppid = asoc->default_ppid;
1652 default_sinfo.sinfo_context = asoc->default_context;
1653 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1654 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1655 sinfo = &default_sinfo;
1656 }
1657
1658 /* API 7.1.7, the sndbuf size per association bounds the
1659 * maximum size of data that can be sent in a single send call.
1660 */
1661 if (msg_len > sk->sk_sndbuf) {
1662 err = -EMSGSIZE;
1663 goto out_free;
1664 }
1665
1666 /* If fragmentation is disabled and the message length exceeds the
1667 * association fragmentation point, return EMSGSIZE. The I-D
1668 * does not specify what this error is, but this looks like
1669 * a great fit.
1670 */
1671 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1672 err = -EMSGSIZE;
1673 goto out_free;
1674 }
1675
1676 if (sinfo) {
1677 /* Check for invalid stream. */
1678 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1679 err = -EINVAL;
1680 goto out_free;
1681 }
1682 }
1683
1684 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1685 if (!sctp_wspace(asoc)) {
1686 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1687 if (err)
1688 goto out_free;
1689 }
1690
1691 /* If an address is passed with the sendto/sendmsg call, it is used
1692 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001693 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 */
1695 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001696 (sinfo_flags & SCTP_ADDR_OVER)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1698 if (!chunk_tp) {
1699 err = -EINVAL;
1700 goto out_free;
1701 }
1702 } else
1703 chunk_tp = NULL;
1704
1705 /* Auto-connect, if we aren't connected already. */
1706 if (sctp_state(asoc, CLOSED)) {
1707 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1708 if (err < 0)
1709 goto out_free;
1710 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1711 }
1712
1713 /* Break the message into multiple chunks of maximum size. */
1714 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1715 if (!datamsg) {
1716 err = -ENOMEM;
1717 goto out_free;
1718 }
1719
1720 /* Now send the (possibly) fragmented message. */
1721 list_for_each(pos, &datamsg->chunks) {
1722 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1723 sctp_datamsg_track(chunk);
1724
1725 /* Do accounting for the write space. */
1726 sctp_set_owner_w(chunk);
1727
1728 chunk->transport = chunk_tp;
1729
1730 /* Send it to the lower layers. Note: all chunks
1731 * must either fail or succeed. The lower layer
1732 * works that way today. Keep it that way or this
1733 * breaks.
1734 */
1735 err = sctp_primitive_SEND(asoc, chunk);
1736 /* Did the lower layer accept the chunk? */
1737 if (err)
1738 sctp_chunk_free(chunk);
1739 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1740 }
1741
1742 sctp_datamsg_free(datamsg);
1743 if (err)
1744 goto out_free;
1745 else
1746 err = msg_len;
1747
1748 /* If we are already past ASSOCIATE, the lower
1749 * layers are responsible for association cleanup.
1750 */
1751 goto out_unlock;
1752
1753out_free:
1754 if (new_asoc)
1755 sctp_association_free(asoc);
1756out_unlock:
1757 sctp_release_sock(sk);
1758
1759out_nounlock:
1760 return sctp_error(sk, msg_flags, err);
1761
1762#if 0
1763do_sock_err:
1764 if (msg_len)
1765 err = msg_len;
1766 else
1767 err = sock_error(sk);
1768 goto out;
1769
1770do_interrupted:
1771 if (msg_len)
1772 err = msg_len;
1773 goto out;
1774#endif /* 0 */
1775}
1776
1777/* This is an extended version of skb_pull() that removes the data from the
1778 * start of a skb even when data is spread across the list of skb's in the
1779 * frag_list. len specifies the total amount of data that needs to be removed.
1780 * when 'len' bytes could be removed from the skb, it returns 0.
1781 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1782 * could not be removed.
1783 */
1784static int sctp_skb_pull(struct sk_buff *skb, int len)
1785{
1786 struct sk_buff *list;
1787 int skb_len = skb_headlen(skb);
1788 int rlen;
1789
1790 if (len <= skb_len) {
1791 __skb_pull(skb, len);
1792 return 0;
1793 }
1794 len -= skb_len;
1795 __skb_pull(skb, skb_len);
1796
1797 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1798 rlen = sctp_skb_pull(list, len);
1799 skb->len -= (len-rlen);
1800 skb->data_len -= (len-rlen);
1801
1802 if (!rlen)
1803 return 0;
1804
1805 len = rlen;
1806 }
1807
1808 return len;
1809}
1810
1811/* API 3.1.3 recvmsg() - UDP Style Syntax
1812 *
1813 * ssize_t recvmsg(int socket, struct msghdr *message,
1814 * int flags);
1815 *
1816 * socket - the socket descriptor of the endpoint.
1817 * message - pointer to the msghdr structure which contains a single
1818 * user message and possibly some ancillary data.
1819 *
1820 * See Section 5 for complete description of the data
1821 * structures.
1822 *
1823 * flags - flags sent or received with the user message, see Section
1824 * 5 for complete description of the flags.
1825 */
1826static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1827
1828SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1829 struct msghdr *msg, size_t len, int noblock,
1830 int flags, int *addr_len)
1831{
1832 struct sctp_ulpevent *event = NULL;
1833 struct sctp_sock *sp = sctp_sk(sk);
1834 struct sk_buff *skb;
1835 int copied;
1836 int err = 0;
1837 int skb_len;
1838
1839 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1840 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1841 "len", len, "knoblauch", noblock,
1842 "flags", flags, "addr_len", addr_len);
1843
1844 sctp_lock_sock(sk);
1845
1846 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1847 err = -ENOTCONN;
1848 goto out;
1849 }
1850
1851 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1852 if (!skb)
1853 goto out;
1854
1855 /* Get the total length of the skb including any skb's in the
1856 * frag_list.
1857 */
1858 skb_len = skb->len;
1859
1860 copied = skb_len;
1861 if (copied > len)
1862 copied = len;
1863
1864 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1865
1866 event = sctp_skb2event(skb);
1867
1868 if (err)
1869 goto out_free;
1870
1871 sock_recv_timestamp(msg, sk, skb);
1872 if (sctp_ulpevent_is_notification(event)) {
1873 msg->msg_flags |= MSG_NOTIFICATION;
1874 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1875 } else {
1876 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1877 }
1878
1879 /* Check if we allow SCTP_SNDRCVINFO. */
1880 if (sp->subscribe.sctp_data_io_event)
1881 sctp_ulpevent_read_sndrcvinfo(event, msg);
1882#if 0
1883 /* FIXME: we should be calling IP/IPv6 layers. */
1884 if (sk->sk_protinfo.af_inet.cmsg_flags)
1885 ip_cmsg_recv(msg, skb);
1886#endif
1887
1888 err = copied;
1889
1890 /* If skb's length exceeds the user's buffer, update the skb and
1891 * push it back to the receive_queue so that the next call to
1892 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1893 */
1894 if (skb_len > copied) {
1895 msg->msg_flags &= ~MSG_EOR;
1896 if (flags & MSG_PEEK)
1897 goto out_free;
1898 sctp_skb_pull(skb, copied);
1899 skb_queue_head(&sk->sk_receive_queue, skb);
1900
1901 /* When only partial message is copied to the user, increase
1902 * rwnd by that amount. If all the data in the skb is read,
1903 * rwnd is updated when the event is freed.
1904 */
1905 sctp_assoc_rwnd_increase(event->asoc, copied);
1906 goto out;
1907 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1908 (event->msg_flags & MSG_EOR))
1909 msg->msg_flags |= MSG_EOR;
1910 else
1911 msg->msg_flags &= ~MSG_EOR;
1912
1913out_free:
1914 if (flags & MSG_PEEK) {
1915 /* Release the skb reference acquired after peeking the skb in
1916 * sctp_skb_recv_datagram().
1917 */
1918 kfree_skb(skb);
1919 } else {
1920 /* Free the event which includes releasing the reference to
1921 * the owner of the skb, freeing the skb and updating the
1922 * rwnd.
1923 */
1924 sctp_ulpevent_free(event);
1925 }
1926out:
1927 sctp_release_sock(sk);
1928 return err;
1929}
1930
1931/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1932 *
1933 * This option is a on/off flag. If enabled no SCTP message
1934 * fragmentation will be performed. Instead if a message being sent
1935 * exceeds the current PMTU size, the message will NOT be sent and
1936 * instead a error will be indicated to the user.
1937 */
1938static int sctp_setsockopt_disable_fragments(struct sock *sk,
1939 char __user *optval, int optlen)
1940{
1941 int val;
1942
1943 if (optlen < sizeof(int))
1944 return -EINVAL;
1945
1946 if (get_user(val, (int __user *)optval))
1947 return -EFAULT;
1948
1949 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1950
1951 return 0;
1952}
1953
1954static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1955 int optlen)
1956{
1957 if (optlen != sizeof(struct sctp_event_subscribe))
1958 return -EINVAL;
1959 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1960 return -EFAULT;
1961 return 0;
1962}
1963
1964/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1965 *
1966 * This socket option is applicable to the UDP-style socket only. When
1967 * set it will cause associations that are idle for more than the
1968 * specified number of seconds to automatically close. An association
1969 * being idle is defined an association that has NOT sent or received
1970 * user data. The special value of '0' indicates that no automatic
1971 * close of any associations should be performed. The option expects an
1972 * integer defining the number of seconds of idle time before an
1973 * association is closed.
1974 */
1975static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1976 int optlen)
1977{
1978 struct sctp_sock *sp = sctp_sk(sk);
1979
1980 /* Applicable to UDP-style socket only */
1981 if (sctp_style(sk, TCP))
1982 return -EOPNOTSUPP;
1983 if (optlen != sizeof(int))
1984 return -EINVAL;
1985 if (copy_from_user(&sp->autoclose, optval, optlen))
1986 return -EFAULT;
1987
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 return 0;
1989}
1990
1991/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1992 *
1993 * Applications can enable or disable heartbeats for any peer address of
1994 * an association, modify an address's heartbeat interval, force a
1995 * heartbeat to be sent immediately, and adjust the address's maximum
1996 * number of retransmissions sent before an address is considered
1997 * unreachable. The following structure is used to access and modify an
1998 * address's parameters:
1999 *
2000 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002001 * sctp_assoc_t spp_assoc_id;
2002 * struct sockaddr_storage spp_address;
2003 * uint32_t spp_hbinterval;
2004 * uint16_t spp_pathmaxrxt;
2005 * uint32_t spp_pathmtu;
2006 * uint32_t spp_sackdelay;
2007 * uint32_t spp_flags;
2008 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002010 * spp_assoc_id - (one-to-many style socket) This is filled in the
2011 * application, and identifies the association for
2012 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 * spp_address - This specifies which address is of interest.
2014 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08002015 * in milliseconds. If a value of zero
2016 * is present in this field then no changes are to
2017 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 * spp_pathmaxrxt - This contains the maximum number of
2019 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08002020 * considered unreachable. If a value of zero
2021 * is present in this field then no changes are to
2022 * be made to this parameter.
2023 * spp_pathmtu - When Path MTU discovery is disabled the value
2024 * specified here will be the "fixed" path mtu.
2025 * Note that if the spp_address field is empty
2026 * then all associations on this address will
2027 * have this fixed path mtu set upon them.
2028 *
2029 * spp_sackdelay - When delayed sack is enabled, this value specifies
2030 * the number of milliseconds that sacks will be delayed
2031 * for. This value will apply to all addresses of an
2032 * association if the spp_address field is empty. Note
2033 * also, that if delayed sack is enabled and this
2034 * value is set to 0, no change is made to the last
2035 * recorded delayed sack timer value.
2036 *
2037 * spp_flags - These flags are used to control various features
2038 * on an association. The flag field may contain
2039 * zero or more of the following options.
2040 *
2041 * SPP_HB_ENABLE - Enable heartbeats on the
2042 * specified address. Note that if the address
2043 * field is empty all addresses for the association
2044 * have heartbeats enabled upon them.
2045 *
2046 * SPP_HB_DISABLE - Disable heartbeats on the
2047 * speicifed address. Note that if the address
2048 * field is empty all addresses for the association
2049 * will have their heartbeats disabled. Note also
2050 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2051 * mutually exclusive, only one of these two should
2052 * be specified. Enabling both fields will have
2053 * undetermined results.
2054 *
2055 * SPP_HB_DEMAND - Request a user initiated heartbeat
2056 * to be made immediately.
2057 *
2058 * SPP_PMTUD_ENABLE - This field will enable PMTU
2059 * discovery upon the specified address. Note that
2060 * if the address feild is empty then all addresses
2061 * on the association are effected.
2062 *
2063 * SPP_PMTUD_DISABLE - This field will disable PMTU
2064 * discovery upon the specified address. Note that
2065 * if the address feild is empty then all addresses
2066 * on the association are effected. Not also that
2067 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2068 * exclusive. Enabling both will have undetermined
2069 * results.
2070 *
2071 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2072 * on delayed sack. The time specified in spp_sackdelay
2073 * is used to specify the sack delay for this address. Note
2074 * that if spp_address is empty then all addresses will
2075 * enable delayed sack and take on the sack delay
2076 * value specified in spp_sackdelay.
2077 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2078 * off delayed sack. If the spp_address field is blank then
2079 * delayed sack is disabled for the entire association. Note
2080 * also that this field is mutually exclusive to
2081 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2082 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 */
Adrian Bunk16164362006-09-18 00:40:38 -07002084static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2085 struct sctp_transport *trans,
2086 struct sctp_association *asoc,
2087 struct sctp_sock *sp,
2088 int hb_change,
2089 int pmtud_change,
2090 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 int error;
2093
Frank Filz52ccb8e2005-12-22 11:36:46 -08002094 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2096 if (error)
2097 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098 }
2099
Frank Filz52ccb8e2005-12-22 11:36:46 -08002100 if (params->spp_hbinterval) {
2101 if (trans) {
2102 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2103 } else if (asoc) {
2104 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2105 } else {
2106 sp->hbinterval = params->spp_hbinterval;
2107 }
2108 }
2109
2110 if (hb_change) {
2111 if (trans) {
2112 trans->param_flags =
2113 (trans->param_flags & ~SPP_HB) | hb_change;
2114 } else if (asoc) {
2115 asoc->param_flags =
2116 (asoc->param_flags & ~SPP_HB) | hb_change;
2117 } else {
2118 sp->param_flags =
2119 (sp->param_flags & ~SPP_HB) | hb_change;
2120 }
2121 }
2122
2123 if (params->spp_pathmtu) {
2124 if (trans) {
2125 trans->pathmtu = params->spp_pathmtu;
2126 sctp_assoc_sync_pmtu(asoc);
2127 } else if (asoc) {
2128 asoc->pathmtu = params->spp_pathmtu;
2129 sctp_frag_point(sp, params->spp_pathmtu);
2130 } else {
2131 sp->pathmtu = params->spp_pathmtu;
2132 }
2133 }
2134
2135 if (pmtud_change) {
2136 if (trans) {
2137 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2138 (params->spp_flags & SPP_PMTUD_ENABLE);
2139 trans->param_flags =
2140 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2141 if (update) {
2142 sctp_transport_pmtu(trans);
2143 sctp_assoc_sync_pmtu(asoc);
2144 }
2145 } else if (asoc) {
2146 asoc->param_flags =
2147 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2148 } else {
2149 sp->param_flags =
2150 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2151 }
2152 }
2153
2154 if (params->spp_sackdelay) {
2155 if (trans) {
2156 trans->sackdelay =
2157 msecs_to_jiffies(params->spp_sackdelay);
2158 } else if (asoc) {
2159 asoc->sackdelay =
2160 msecs_to_jiffies(params->spp_sackdelay);
2161 } else {
2162 sp->sackdelay = params->spp_sackdelay;
2163 }
2164 }
2165
2166 if (sackdelay_change) {
2167 if (trans) {
2168 trans->param_flags =
2169 (trans->param_flags & ~SPP_SACKDELAY) |
2170 sackdelay_change;
2171 } else if (asoc) {
2172 asoc->param_flags =
2173 (asoc->param_flags & ~SPP_SACKDELAY) |
2174 sackdelay_change;
2175 } else {
2176 sp->param_flags =
2177 (sp->param_flags & ~SPP_SACKDELAY) |
2178 sackdelay_change;
2179 }
2180 }
2181
2182 if (params->spp_pathmaxrxt) {
2183 if (trans) {
2184 trans->pathmaxrxt = params->spp_pathmaxrxt;
2185 } else if (asoc) {
2186 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2187 } else {
2188 sp->pathmaxrxt = params->spp_pathmaxrxt;
2189 }
2190 }
2191
2192 return 0;
2193}
2194
2195static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2196 char __user *optval, int optlen)
2197{
2198 struct sctp_paddrparams params;
2199 struct sctp_transport *trans = NULL;
2200 struct sctp_association *asoc = NULL;
2201 struct sctp_sock *sp = sctp_sk(sk);
2202 int error;
2203 int hb_change, pmtud_change, sackdelay_change;
2204
2205 if (optlen != sizeof(struct sctp_paddrparams))
2206 return - EINVAL;
2207
2208 if (copy_from_user(&params, optval, optlen))
2209 return -EFAULT;
2210
2211 /* Validate flags and value parameters. */
2212 hb_change = params.spp_flags & SPP_HB;
2213 pmtud_change = params.spp_flags & SPP_PMTUD;
2214 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2215
2216 if (hb_change == SPP_HB ||
2217 pmtud_change == SPP_PMTUD ||
2218 sackdelay_change == SPP_SACKDELAY ||
2219 params.spp_sackdelay > 500 ||
2220 (params.spp_pathmtu
2221 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2222 return -EINVAL;
2223
2224 /* If an address other than INADDR_ANY is specified, and
2225 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08002227 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2228 trans = sctp_addr_id2transport(sk, &params.spp_address,
2229 params.spp_assoc_id);
2230 if (!trans)
2231 return -EINVAL;
2232 }
2233
2234 /* Get association, if assoc_id != 0 and the socket is a one
2235 * to many style socket, and an association was not found, then
2236 * the id was invalid.
2237 */
2238 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2239 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2240 return -EINVAL;
2241
2242 /* Heartbeat demand can only be sent on a transport or
2243 * association, but not a socket.
2244 */
2245 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2246 return -EINVAL;
2247
2248 /* Process parameters. */
2249 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2250 hb_change, pmtud_change,
2251 sackdelay_change);
2252
2253 if (error)
2254 return error;
2255
2256 /* If changes are for association, also apply parameters to each
2257 * transport.
2258 */
2259 if (!trans && asoc) {
2260 struct list_head *pos;
2261
2262 list_for_each(pos, &asoc->peer.transport_addr_list) {
2263 trans = list_entry(pos, struct sctp_transport,
2264 transports);
2265 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2266 hb_change, pmtud_change,
2267 sackdelay_change);
2268 }
2269 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270
2271 return 0;
2272}
2273
Frank Filz77086102005-12-22 11:37:30 -08002274/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2275 *
2276 * This options will get or set the delayed ack timer. The time is set
2277 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2278 * endpoints default delayed ack timer value. If the assoc_id field is
2279 * non-zero, then the set or get effects the specified association.
2280 *
2281 * struct sctp_assoc_value {
2282 * sctp_assoc_t assoc_id;
2283 * uint32_t assoc_value;
2284 * };
2285 *
2286 * assoc_id - This parameter, indicates which association the
2287 * user is preforming an action upon. Note that if
2288 * this field's value is zero then the endpoints
2289 * default value is changed (effecting future
2290 * associations only).
2291 *
2292 * assoc_value - This parameter contains the number of milliseconds
2293 * that the user is requesting the delayed ACK timer
2294 * be set to. Note that this value is defined in
2295 * the standard to be between 200 and 500 milliseconds.
2296 *
2297 * Note: a value of zero will leave the value alone,
2298 * but disable SACK delay. A non-zero value will also
2299 * enable SACK delay.
2300 */
2301
2302static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2303 char __user *optval, int optlen)
2304{
2305 struct sctp_assoc_value params;
2306 struct sctp_transport *trans = NULL;
2307 struct sctp_association *asoc = NULL;
2308 struct sctp_sock *sp = sctp_sk(sk);
2309
2310 if (optlen != sizeof(struct sctp_assoc_value))
2311 return - EINVAL;
2312
2313 if (copy_from_user(&params, optval, optlen))
2314 return -EFAULT;
2315
2316 /* Validate value parameter. */
2317 if (params.assoc_value > 500)
2318 return -EINVAL;
2319
2320 /* Get association, if assoc_id != 0 and the socket is a one
2321 * to many style socket, and an association was not found, then
2322 * the id was invalid.
2323 */
2324 asoc = sctp_id2assoc(sk, params.assoc_id);
2325 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2326 return -EINVAL;
2327
2328 if (params.assoc_value) {
2329 if (asoc) {
2330 asoc->sackdelay =
2331 msecs_to_jiffies(params.assoc_value);
2332 asoc->param_flags =
2333 (asoc->param_flags & ~SPP_SACKDELAY) |
2334 SPP_SACKDELAY_ENABLE;
2335 } else {
2336 sp->sackdelay = params.assoc_value;
2337 sp->param_flags =
2338 (sp->param_flags & ~SPP_SACKDELAY) |
2339 SPP_SACKDELAY_ENABLE;
2340 }
2341 } else {
2342 if (asoc) {
2343 asoc->param_flags =
2344 (asoc->param_flags & ~SPP_SACKDELAY) |
2345 SPP_SACKDELAY_DISABLE;
2346 } else {
2347 sp->param_flags =
2348 (sp->param_flags & ~SPP_SACKDELAY) |
2349 SPP_SACKDELAY_DISABLE;
2350 }
2351 }
2352
2353 /* If change is for association, also apply to each transport. */
2354 if (asoc) {
2355 struct list_head *pos;
2356
2357 list_for_each(pos, &asoc->peer.transport_addr_list) {
2358 trans = list_entry(pos, struct sctp_transport,
2359 transports);
2360 if (params.assoc_value) {
2361 trans->sackdelay =
2362 msecs_to_jiffies(params.assoc_value);
2363 trans->param_flags =
2364 (trans->param_flags & ~SPP_SACKDELAY) |
2365 SPP_SACKDELAY_ENABLE;
2366 } else {
2367 trans->param_flags =
2368 (trans->param_flags & ~SPP_SACKDELAY) |
2369 SPP_SACKDELAY_DISABLE;
2370 }
2371 }
2372 }
2373
2374 return 0;
2375}
2376
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2378 *
2379 * Applications can specify protocol parameters for the default association
2380 * initialization. The option name argument to setsockopt() and getsockopt()
2381 * is SCTP_INITMSG.
2382 *
2383 * Setting initialization parameters is effective only on an unconnected
2384 * socket (for UDP-style sockets only future associations are effected
2385 * by the change). With TCP-style sockets, this option is inherited by
2386 * sockets derived from a listener socket.
2387 */
2388static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2389{
2390 struct sctp_initmsg sinit;
2391 struct sctp_sock *sp = sctp_sk(sk);
2392
2393 if (optlen != sizeof(struct sctp_initmsg))
2394 return -EINVAL;
2395 if (copy_from_user(&sinit, optval, optlen))
2396 return -EFAULT;
2397
2398 if (sinit.sinit_num_ostreams)
2399 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2400 if (sinit.sinit_max_instreams)
2401 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2402 if (sinit.sinit_max_attempts)
2403 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2404 if (sinit.sinit_max_init_timeo)
2405 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2406
2407 return 0;
2408}
2409
2410/*
2411 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2412 *
2413 * Applications that wish to use the sendto() system call may wish to
2414 * specify a default set of parameters that would normally be supplied
2415 * through the inclusion of ancillary data. This socket option allows
2416 * such an application to set the default sctp_sndrcvinfo structure.
2417 * The application that wishes to use this socket option simply passes
2418 * in to this call the sctp_sndrcvinfo structure defined in Section
2419 * 5.2.2) The input parameters accepted by this call include
2420 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2421 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2422 * to this call if the caller is using the UDP model.
2423 */
2424static int sctp_setsockopt_default_send_param(struct sock *sk,
2425 char __user *optval, int optlen)
2426{
2427 struct sctp_sndrcvinfo info;
2428 struct sctp_association *asoc;
2429 struct sctp_sock *sp = sctp_sk(sk);
2430
2431 if (optlen != sizeof(struct sctp_sndrcvinfo))
2432 return -EINVAL;
2433 if (copy_from_user(&info, optval, optlen))
2434 return -EFAULT;
2435
2436 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2437 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2438 return -EINVAL;
2439
2440 if (asoc) {
2441 asoc->default_stream = info.sinfo_stream;
2442 asoc->default_flags = info.sinfo_flags;
2443 asoc->default_ppid = info.sinfo_ppid;
2444 asoc->default_context = info.sinfo_context;
2445 asoc->default_timetolive = info.sinfo_timetolive;
2446 } else {
2447 sp->default_stream = info.sinfo_stream;
2448 sp->default_flags = info.sinfo_flags;
2449 sp->default_ppid = info.sinfo_ppid;
2450 sp->default_context = info.sinfo_context;
2451 sp->default_timetolive = info.sinfo_timetolive;
2452 }
2453
2454 return 0;
2455}
2456
2457/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2458 *
2459 * Requests that the local SCTP stack use the enclosed peer address as
2460 * the association primary. The enclosed address must be one of the
2461 * association peer's addresses.
2462 */
2463static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2464 int optlen)
2465{
2466 struct sctp_prim prim;
2467 struct sctp_transport *trans;
2468
2469 if (optlen != sizeof(struct sctp_prim))
2470 return -EINVAL;
2471
2472 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2473 return -EFAULT;
2474
2475 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2476 if (!trans)
2477 return -EINVAL;
2478
2479 sctp_assoc_set_primary(trans->asoc, trans);
2480
2481 return 0;
2482}
2483
2484/*
2485 * 7.1.5 SCTP_NODELAY
2486 *
2487 * Turn on/off any Nagle-like algorithm. This means that packets are
2488 * generally sent as soon as possible and no unnecessary delays are
2489 * introduced, at the cost of more packets in the network. Expects an
2490 * integer boolean flag.
2491 */
2492static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2493 int optlen)
2494{
2495 int val;
2496
2497 if (optlen < sizeof(int))
2498 return -EINVAL;
2499 if (get_user(val, (int __user *)optval))
2500 return -EFAULT;
2501
2502 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2503 return 0;
2504}
2505
2506/*
2507 *
2508 * 7.1.1 SCTP_RTOINFO
2509 *
2510 * The protocol parameters used to initialize and bound retransmission
2511 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2512 * and modify these parameters.
2513 * All parameters are time values, in milliseconds. A value of 0, when
2514 * modifying the parameters, indicates that the current value should not
2515 * be changed.
2516 *
2517 */
2518static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2519 struct sctp_rtoinfo rtoinfo;
2520 struct sctp_association *asoc;
2521
2522 if (optlen != sizeof (struct sctp_rtoinfo))
2523 return -EINVAL;
2524
2525 if (copy_from_user(&rtoinfo, optval, optlen))
2526 return -EFAULT;
2527
2528 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2529
2530 /* Set the values to the specific association */
2531 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2532 return -EINVAL;
2533
2534 if (asoc) {
2535 if (rtoinfo.srto_initial != 0)
2536 asoc->rto_initial =
2537 msecs_to_jiffies(rtoinfo.srto_initial);
2538 if (rtoinfo.srto_max != 0)
2539 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2540 if (rtoinfo.srto_min != 0)
2541 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2542 } else {
2543 /* If there is no association or the association-id = 0
2544 * set the values to the endpoint.
2545 */
2546 struct sctp_sock *sp = sctp_sk(sk);
2547
2548 if (rtoinfo.srto_initial != 0)
2549 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2550 if (rtoinfo.srto_max != 0)
2551 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2552 if (rtoinfo.srto_min != 0)
2553 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2554 }
2555
2556 return 0;
2557}
2558
2559/*
2560 *
2561 * 7.1.2 SCTP_ASSOCINFO
2562 *
2563 * This option is used to tune the the maximum retransmission attempts
2564 * of the association.
2565 * Returns an error if the new association retransmission value is
2566 * greater than the sum of the retransmission value of the peer.
2567 * See [SCTP] for more information.
2568 *
2569 */
2570static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2571{
2572
2573 struct sctp_assocparams assocparams;
2574 struct sctp_association *asoc;
2575
2576 if (optlen != sizeof(struct sctp_assocparams))
2577 return -EINVAL;
2578 if (copy_from_user(&assocparams, optval, optlen))
2579 return -EFAULT;
2580
2581 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2582
2583 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2584 return -EINVAL;
2585
2586 /* Set the values to the specific association */
2587 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002588 if (assocparams.sasoc_asocmaxrxt != 0) {
2589 __u32 path_sum = 0;
2590 int paths = 0;
2591 struct list_head *pos;
2592 struct sctp_transport *peer_addr;
2593
2594 list_for_each(pos, &asoc->peer.transport_addr_list) {
2595 peer_addr = list_entry(pos,
2596 struct sctp_transport,
2597 transports);
2598 path_sum += peer_addr->pathmaxrxt;
2599 paths++;
2600 }
2601
2602 /* Only validate asocmaxrxt if we have more then
2603 * one path/transport. We do this because path
2604 * retransmissions are only counted when we have more
2605 * then one path.
2606 */
2607 if (paths > 1 &&
2608 assocparams.sasoc_asocmaxrxt > path_sum)
2609 return -EINVAL;
2610
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002612 }
2613
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614 if (assocparams.sasoc_cookie_life != 0) {
2615 asoc->cookie_life.tv_sec =
2616 assocparams.sasoc_cookie_life / 1000;
2617 asoc->cookie_life.tv_usec =
2618 (assocparams.sasoc_cookie_life % 1000)
2619 * 1000;
2620 }
2621 } else {
2622 /* Set the values to the endpoint */
2623 struct sctp_sock *sp = sctp_sk(sk);
2624
2625 if (assocparams.sasoc_asocmaxrxt != 0)
2626 sp->assocparams.sasoc_asocmaxrxt =
2627 assocparams.sasoc_asocmaxrxt;
2628 if (assocparams.sasoc_cookie_life != 0)
2629 sp->assocparams.sasoc_cookie_life =
2630 assocparams.sasoc_cookie_life;
2631 }
2632 return 0;
2633}
2634
2635/*
2636 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2637 *
2638 * This socket option is a boolean flag which turns on or off mapped V4
2639 * addresses. If this option is turned on and the socket is type
2640 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2641 * If this option is turned off, then no mapping will be done of V4
2642 * addresses and a user will receive both PF_INET6 and PF_INET type
2643 * addresses on the socket.
2644 */
2645static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2646{
2647 int val;
2648 struct sctp_sock *sp = sctp_sk(sk);
2649
2650 if (optlen < sizeof(int))
2651 return -EINVAL;
2652 if (get_user(val, (int __user *)optval))
2653 return -EFAULT;
2654 if (val)
2655 sp->v4mapped = 1;
2656 else
2657 sp->v4mapped = 0;
2658
2659 return 0;
2660}
2661
2662/*
2663 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2664 *
2665 * This socket option specifies the maximum size to put in any outgoing
2666 * SCTP chunk. If a message is larger than this size it will be
2667 * fragmented by SCTP into the specified size. Note that the underlying
2668 * SCTP implementation may fragment into smaller sized chunks when the
2669 * PMTU of the underlying association is smaller than the value set by
2670 * the user.
2671 */
2672static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2673{
2674 struct sctp_association *asoc;
2675 struct list_head *pos;
2676 struct sctp_sock *sp = sctp_sk(sk);
2677 int val;
2678
2679 if (optlen < sizeof(int))
2680 return -EINVAL;
2681 if (get_user(val, (int __user *)optval))
2682 return -EFAULT;
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002683 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002684 return -EINVAL;
2685 sp->user_frag = val;
2686
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002687 /* Update the frag_point of the existing associations. */
2688 list_for_each(pos, &(sp->ep->asocs)) {
2689 asoc = list_entry(pos, struct sctp_association, asocs);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002690 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691 }
2692
2693 return 0;
2694}
2695
2696
2697/*
2698 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2699 *
2700 * Requests that the peer mark the enclosed address as the association
2701 * primary. The enclosed address must be one of the association's
2702 * locally bound addresses. The following structure is used to make a
2703 * set primary request:
2704 */
2705static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2706 int optlen)
2707{
2708 struct sctp_sock *sp;
2709 struct sctp_endpoint *ep;
2710 struct sctp_association *asoc = NULL;
2711 struct sctp_setpeerprim prim;
2712 struct sctp_chunk *chunk;
2713 int err;
2714
2715 sp = sctp_sk(sk);
2716 ep = sp->ep;
2717
2718 if (!sctp_addip_enable)
2719 return -EPERM;
2720
2721 if (optlen != sizeof(struct sctp_setpeerprim))
2722 return -EINVAL;
2723
2724 if (copy_from_user(&prim, optval, optlen))
2725 return -EFAULT;
2726
2727 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2728 if (!asoc)
2729 return -EINVAL;
2730
2731 if (!asoc->peer.asconf_capable)
2732 return -EPERM;
2733
2734 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2735 return -EPERM;
2736
2737 if (!sctp_state(asoc, ESTABLISHED))
2738 return -ENOTCONN;
2739
2740 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2741 return -EADDRNOTAVAIL;
2742
2743 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2744 chunk = sctp_make_asconf_set_prim(asoc,
2745 (union sctp_addr *)&prim.sspp_addr);
2746 if (!chunk)
2747 return -ENOMEM;
2748
2749 err = sctp_send_asconf(asoc, chunk);
2750
2751 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2752
2753 return err;
2754}
2755
2756static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2757 int optlen)
2758{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002759 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002761 if (optlen != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762 return -EINVAL;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002763 if (copy_from_user(&adaption, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764 return -EFAULT;
2765
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07002766 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002767
2768 return 0;
2769}
2770
2771/* API 6.2 setsockopt(), getsockopt()
2772 *
2773 * Applications use setsockopt() and getsockopt() to set or retrieve
2774 * socket options. Socket options are used to change the default
2775 * behavior of sockets calls. They are described in Section 7.
2776 *
2777 * The syntax is:
2778 *
2779 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2780 * int __user *optlen);
2781 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2782 * int optlen);
2783 *
2784 * sd - the socket descript.
2785 * level - set to IPPROTO_SCTP for all SCTP options.
2786 * optname - the option name.
2787 * optval - the buffer to store the value of the option.
2788 * optlen - the size of the buffer.
2789 */
2790SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2791 char __user *optval, int optlen)
2792{
2793 int retval = 0;
2794
2795 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2796 sk, optname);
2797
2798 /* I can hardly begin to describe how wrong this is. This is
2799 * so broken as to be worse than useless. The API draft
2800 * REALLY is NOT helpful here... I am not convinced that the
2801 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2802 * are at all well-founded.
2803 */
2804 if (level != SOL_SCTP) {
2805 struct sctp_af *af = sctp_sk(sk)->pf->af;
2806 retval = af->setsockopt(sk, level, optname, optval, optlen);
2807 goto out_nounlock;
2808 }
2809
2810 sctp_lock_sock(sk);
2811
2812 switch (optname) {
2813 case SCTP_SOCKOPT_BINDX_ADD:
2814 /* 'optlen' is the size of the addresses buffer. */
2815 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2816 optlen, SCTP_BINDX_ADD_ADDR);
2817 break;
2818
2819 case SCTP_SOCKOPT_BINDX_REM:
2820 /* 'optlen' is the size of the addresses buffer. */
2821 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2822 optlen, SCTP_BINDX_REM_ADDR);
2823 break;
2824
Frank Filz3f7a87d2005-06-20 13:14:57 -07002825 case SCTP_SOCKOPT_CONNECTX:
2826 /* 'optlen' is the size of the addresses buffer. */
2827 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2828 optlen);
2829 break;
2830
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831 case SCTP_DISABLE_FRAGMENTS:
2832 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2833 break;
2834
2835 case SCTP_EVENTS:
2836 retval = sctp_setsockopt_events(sk, optval, optlen);
2837 break;
2838
2839 case SCTP_AUTOCLOSE:
2840 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2841 break;
2842
2843 case SCTP_PEER_ADDR_PARAMS:
2844 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2845 break;
2846
Frank Filz77086102005-12-22 11:37:30 -08002847 case SCTP_DELAYED_ACK_TIME:
2848 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2849 break;
2850
Linus Torvalds1da177e2005-04-16 15:20:36 -07002851 case SCTP_INITMSG:
2852 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2853 break;
2854 case SCTP_DEFAULT_SEND_PARAM:
2855 retval = sctp_setsockopt_default_send_param(sk, optval,
2856 optlen);
2857 break;
2858 case SCTP_PRIMARY_ADDR:
2859 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2860 break;
2861 case SCTP_SET_PEER_PRIMARY_ADDR:
2862 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2863 break;
2864 case SCTP_NODELAY:
2865 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2866 break;
2867 case SCTP_RTOINFO:
2868 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2869 break;
2870 case SCTP_ASSOCINFO:
2871 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2872 break;
2873 case SCTP_I_WANT_MAPPED_V4_ADDR:
2874 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2875 break;
2876 case SCTP_MAXSEG:
2877 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2878 break;
2879 case SCTP_ADAPTION_LAYER:
2880 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2881 break;
2882
2883 default:
2884 retval = -ENOPROTOOPT;
2885 break;
2886 };
2887
2888 sctp_release_sock(sk);
2889
2890out_nounlock:
2891 return retval;
2892}
2893
2894/* API 3.1.6 connect() - UDP Style Syntax
2895 *
2896 * An application may use the connect() call in the UDP model to initiate an
2897 * association without sending data.
2898 *
2899 * The syntax is:
2900 *
2901 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2902 *
2903 * sd: the socket descriptor to have a new association added to.
2904 *
2905 * nam: the address structure (either struct sockaddr_in or struct
2906 * sockaddr_in6 defined in RFC2553 [7]).
2907 *
2908 * len: the size of the address.
2909 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07002910SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911 int addr_len)
2912{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002913 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07002914 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915
2916 sctp_lock_sock(sk);
2917
Frank Filz3f7a87d2005-06-20 13:14:57 -07002918 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2919 __FUNCTION__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002920
Frank Filz3f7a87d2005-06-20 13:14:57 -07002921 /* Validate addr_len before calling common connect/connectx routine. */
2922 af = sctp_get_af_specific(addr->sa_family);
2923 if (!af || addr_len < af->sockaddr_len) {
2924 err = -EINVAL;
2925 } else {
2926 /* Pass correct addr len to common routine (so it knows there
2927 * is only one address being passed.
2928 */
2929 err = __sctp_connect(sk, addr, af->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002930 }
2931
Linus Torvalds1da177e2005-04-16 15:20:36 -07002932 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002933 return err;
2934}
2935
2936/* FIXME: Write comments. */
2937SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2938{
2939 return -EOPNOTSUPP; /* STUB */
2940}
2941
2942/* 4.1.4 accept() - TCP Style Syntax
2943 *
2944 * Applications use accept() call to remove an established SCTP
2945 * association from the accept queue of the endpoint. A new socket
2946 * descriptor will be returned from accept() to represent the newly
2947 * formed association.
2948 */
2949SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2950{
2951 struct sctp_sock *sp;
2952 struct sctp_endpoint *ep;
2953 struct sock *newsk = NULL;
2954 struct sctp_association *asoc;
2955 long timeo;
2956 int error = 0;
2957
2958 sctp_lock_sock(sk);
2959
2960 sp = sctp_sk(sk);
2961 ep = sp->ep;
2962
2963 if (!sctp_style(sk, TCP)) {
2964 error = -EOPNOTSUPP;
2965 goto out;
2966 }
2967
2968 if (!sctp_sstate(sk, LISTENING)) {
2969 error = -EINVAL;
2970 goto out;
2971 }
2972
Sridhar Samudrala8abfedd2006-08-22 00:24:09 -07002973 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002974
2975 error = sctp_wait_for_accept(sk, timeo);
2976 if (error)
2977 goto out;
2978
2979 /* We treat the list of associations on the endpoint as the accept
2980 * queue and pick the first association on the list.
2981 */
2982 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2983
2984 newsk = sp->pf->create_accept_sk(sk, asoc);
2985 if (!newsk) {
2986 error = -ENOMEM;
2987 goto out;
2988 }
2989
2990 /* Populate the fields of the newsk from the oldsk and migrate the
2991 * asoc to the newsk.
2992 */
2993 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2994
2995out:
2996 sctp_release_sock(sk);
2997 *err = error;
2998 return newsk;
2999}
3000
3001/* The SCTP ioctl handler. */
3002SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3003{
3004 return -ENOIOCTLCMD;
3005}
3006
3007/* This is the function which gets called during socket creation to
3008 * initialized the SCTP-specific portion of the sock.
3009 * The sock structure should already be zero-filled memory.
3010 */
3011SCTP_STATIC int sctp_init_sock(struct sock *sk)
3012{
3013 struct sctp_endpoint *ep;
3014 struct sctp_sock *sp;
3015
3016 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3017
3018 sp = sctp_sk(sk);
3019
3020 /* Initialize the SCTP per socket area. */
3021 switch (sk->sk_type) {
3022 case SOCK_SEQPACKET:
3023 sp->type = SCTP_SOCKET_UDP;
3024 break;
3025 case SOCK_STREAM:
3026 sp->type = SCTP_SOCKET_TCP;
3027 break;
3028 default:
3029 return -ESOCKTNOSUPPORT;
3030 }
3031
3032 /* Initialize default send parameters. These parameters can be
3033 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3034 */
3035 sp->default_stream = 0;
3036 sp->default_ppid = 0;
3037 sp->default_flags = 0;
3038 sp->default_context = 0;
3039 sp->default_timetolive = 0;
3040
3041 /* Initialize default setup parameters. These parameters
3042 * can be modified with the SCTP_INITMSG socket option or
3043 * overridden by the SCTP_INIT CMSG.
3044 */
3045 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3046 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3047 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003048 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049
3050 /* Initialize default RTO related parameters. These parameters can
3051 * be modified for with the SCTP_RTOINFO socket option.
3052 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003053 sp->rtoinfo.srto_initial = sctp_rto_initial;
3054 sp->rtoinfo.srto_max = sctp_rto_max;
3055 sp->rtoinfo.srto_min = sctp_rto_min;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003056
3057 /* Initialize default association related parameters. These parameters
3058 * can be modified with the SCTP_ASSOCINFO socket option.
3059 */
3060 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3061 sp->assocparams.sasoc_number_peer_destinations = 0;
3062 sp->assocparams.sasoc_peer_rwnd = 0;
3063 sp->assocparams.sasoc_local_rwnd = 0;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003064 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003065
3066 /* Initialize default event subscriptions. By default, all the
3067 * options are off.
3068 */
3069 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3070
3071 /* Default Peer Address Parameters. These defaults can
3072 * be modified via SCTP_PEER_ADDR_PARAMS
3073 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003074 sp->hbinterval = sctp_hb_interval;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003075 sp->pathmaxrxt = sctp_max_retrans_path;
3076 sp->pathmtu = 0; // allow default discovery
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003077 sp->sackdelay = sctp_sack_timeout;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003078 sp->param_flags = SPP_HB_ENABLE |
3079 SPP_PMTUD_ENABLE |
3080 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003081
3082 /* If enabled no SCTP message fragmentation will be performed.
3083 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3084 */
3085 sp->disable_fragments = 0;
3086
3087 /* Turn on/off any Nagle-like algorithm. */
3088 sp->nodelay = 1;
3089
3090 /* Enable by default. */
3091 sp->v4mapped = 1;
3092
3093 /* Auto-close idle associations after the configured
3094 * number of seconds. A value of 0 disables this
3095 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3096 * for UDP-style sockets only.
3097 */
3098 sp->autoclose = 0;
3099
3100 /* User specified fragmentation limit. */
3101 sp->user_frag = 0;
3102
3103 sp->adaption_ind = 0;
3104
3105 sp->pf = sctp_get_pf_specific(sk->sk_family);
3106
3107 /* Control variables for partial data delivery. */
3108 sp->pd_mode = 0;
3109 skb_queue_head_init(&sp->pd_lobby);
3110
3111 /* Create a per socket endpoint structure. Even if we
3112 * change the data structure relationships, this may still
3113 * be useful for storing pre-connect address information.
3114 */
3115 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3116 if (!ep)
3117 return -ENOMEM;
3118
3119 sp->ep = ep;
3120 sp->hmac = NULL;
3121
3122 SCTP_DBG_OBJCNT_INC(sock);
3123 return 0;
3124}
3125
3126/* Cleanup any SCTP per socket resources. */
3127SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3128{
3129 struct sctp_endpoint *ep;
3130
3131 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3132
3133 /* Release our hold on the endpoint. */
3134 ep = sctp_sk(sk)->ep;
3135 sctp_endpoint_free(ep);
3136
3137 return 0;
3138}
3139
3140/* API 4.1.7 shutdown() - TCP Style Syntax
3141 * int shutdown(int socket, int how);
3142 *
3143 * sd - the socket descriptor of the association to be closed.
3144 * how - Specifies the type of shutdown. The values are
3145 * as follows:
3146 * SHUT_RD
3147 * Disables further receive operations. No SCTP
3148 * protocol action is taken.
3149 * SHUT_WR
3150 * Disables further send operations, and initiates
3151 * the SCTP shutdown sequence.
3152 * SHUT_RDWR
3153 * Disables further send and receive operations
3154 * and initiates the SCTP shutdown sequence.
3155 */
3156SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3157{
3158 struct sctp_endpoint *ep;
3159 struct sctp_association *asoc;
3160
3161 if (!sctp_style(sk, TCP))
3162 return;
3163
3164 if (how & SEND_SHUTDOWN) {
3165 ep = sctp_sk(sk)->ep;
3166 if (!list_empty(&ep->asocs)) {
3167 asoc = list_entry(ep->asocs.next,
3168 struct sctp_association, asocs);
3169 sctp_primitive_SHUTDOWN(asoc, NULL);
3170 }
3171 }
3172}
3173
3174/* 7.2.1 Association Status (SCTP_STATUS)
3175
3176 * Applications can retrieve current status information about an
3177 * association, including association state, peer receiver window size,
3178 * number of unacked data chunks, and number of data chunks pending
3179 * receipt. This information is read-only.
3180 */
3181static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3182 char __user *optval,
3183 int __user *optlen)
3184{
3185 struct sctp_status status;
3186 struct sctp_association *asoc = NULL;
3187 struct sctp_transport *transport;
3188 sctp_assoc_t associd;
3189 int retval = 0;
3190
3191 if (len != sizeof(status)) {
3192 retval = -EINVAL;
3193 goto out;
3194 }
3195
3196 if (copy_from_user(&status, optval, sizeof(status))) {
3197 retval = -EFAULT;
3198 goto out;
3199 }
3200
3201 associd = status.sstat_assoc_id;
3202 asoc = sctp_id2assoc(sk, associd);
3203 if (!asoc) {
3204 retval = -EINVAL;
3205 goto out;
3206 }
3207
3208 transport = asoc->peer.primary_path;
3209
3210 status.sstat_assoc_id = sctp_assoc2id(asoc);
3211 status.sstat_state = asoc->state;
3212 status.sstat_rwnd = asoc->peer.rwnd;
3213 status.sstat_unackdata = asoc->unack_data;
3214
3215 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3216 status.sstat_instrms = asoc->c.sinit_max_instreams;
3217 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3218 status.sstat_fragmentation_point = asoc->frag_point;
3219 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3220 memcpy(&status.sstat_primary.spinfo_address,
3221 &(transport->ipaddr), sizeof(union sctp_addr));
3222 /* Map ipv4 address into v4-mapped-on-v6 address. */
3223 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3224 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003225 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003226 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3227 status.sstat_primary.spinfo_srtt = transport->srtt;
3228 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003229 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003230
Frank Filz3f7a87d2005-06-20 13:14:57 -07003231 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3232 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3233
Linus Torvalds1da177e2005-04-16 15:20:36 -07003234 if (put_user(len, optlen)) {
3235 retval = -EFAULT;
3236 goto out;
3237 }
3238
3239 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3240 len, status.sstat_state, status.sstat_rwnd,
3241 status.sstat_assoc_id);
3242
3243 if (copy_to_user(optval, &status, len)) {
3244 retval = -EFAULT;
3245 goto out;
3246 }
3247
3248out:
3249 return (retval);
3250}
3251
3252
3253/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3254 *
3255 * Applications can retrieve information about a specific peer address
3256 * of an association, including its reachability state, congestion
3257 * window, and retransmission timer values. This information is
3258 * read-only.
3259 */
3260static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3261 char __user *optval,
3262 int __user *optlen)
3263{
3264 struct sctp_paddrinfo pinfo;
3265 struct sctp_transport *transport;
3266 int retval = 0;
3267
3268 if (len != sizeof(pinfo)) {
3269 retval = -EINVAL;
3270 goto out;
3271 }
3272
3273 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3274 retval = -EFAULT;
3275 goto out;
3276 }
3277
3278 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3279 pinfo.spinfo_assoc_id);
3280 if (!transport)
3281 return -EINVAL;
3282
3283 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003284 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003285 pinfo.spinfo_cwnd = transport->cwnd;
3286 pinfo.spinfo_srtt = transport->srtt;
3287 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003288 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003289
Frank Filz3f7a87d2005-06-20 13:14:57 -07003290 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3291 pinfo.spinfo_state = SCTP_ACTIVE;
3292
Linus Torvalds1da177e2005-04-16 15:20:36 -07003293 if (put_user(len, optlen)) {
3294 retval = -EFAULT;
3295 goto out;
3296 }
3297
3298 if (copy_to_user(optval, &pinfo, len)) {
3299 retval = -EFAULT;
3300 goto out;
3301 }
3302
3303out:
3304 return (retval);
3305}
3306
3307/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3308 *
3309 * This option is a on/off flag. If enabled no SCTP message
3310 * fragmentation will be performed. Instead if a message being sent
3311 * exceeds the current PMTU size, the message will NOT be sent and
3312 * instead a error will be indicated to the user.
3313 */
3314static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3315 char __user *optval, int __user *optlen)
3316{
3317 int val;
3318
3319 if (len < sizeof(int))
3320 return -EINVAL;
3321
3322 len = sizeof(int);
3323 val = (sctp_sk(sk)->disable_fragments == 1);
3324 if (put_user(len, optlen))
3325 return -EFAULT;
3326 if (copy_to_user(optval, &val, len))
3327 return -EFAULT;
3328 return 0;
3329}
3330
3331/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3332 *
3333 * This socket option is used to specify various notifications and
3334 * ancillary data the user wishes to receive.
3335 */
3336static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3337 int __user *optlen)
3338{
3339 if (len != sizeof(struct sctp_event_subscribe))
3340 return -EINVAL;
3341 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3342 return -EFAULT;
3343 return 0;
3344}
3345
3346/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3347 *
3348 * This socket option is applicable to the UDP-style socket only. When
3349 * set it will cause associations that are idle for more than the
3350 * specified number of seconds to automatically close. An association
3351 * being idle is defined an association that has NOT sent or received
3352 * user data. The special value of '0' indicates that no automatic
3353 * close of any associations should be performed. The option expects an
3354 * integer defining the number of seconds of idle time before an
3355 * association is closed.
3356 */
3357static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3358{
3359 /* Applicable to UDP-style socket only */
3360 if (sctp_style(sk, TCP))
3361 return -EOPNOTSUPP;
3362 if (len != sizeof(int))
3363 return -EINVAL;
3364 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3365 return -EFAULT;
3366 return 0;
3367}
3368
3369/* Helper routine to branch off an association to a new socket. */
3370SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3371 struct socket **sockp)
3372{
3373 struct sock *sk = asoc->base.sk;
3374 struct socket *sock;
3375 int err = 0;
3376
3377 /* An association cannot be branched off from an already peeled-off
3378 * socket, nor is this supported for tcp style sockets.
3379 */
3380 if (!sctp_style(sk, UDP))
3381 return -EINVAL;
3382
3383 /* Create a new socket. */
3384 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3385 if (err < 0)
3386 return err;
3387
3388 /* Populate the fields of the newsk from the oldsk and migrate the
3389 * asoc to the newsk.
3390 */
3391 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3392 *sockp = sock;
3393
3394 return err;
3395}
3396
3397static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3398{
3399 sctp_peeloff_arg_t peeloff;
3400 struct socket *newsock;
3401 int retval = 0;
3402 struct sctp_association *asoc;
3403
3404 if (len != sizeof(sctp_peeloff_arg_t))
3405 return -EINVAL;
3406 if (copy_from_user(&peeloff, optval, len))
3407 return -EFAULT;
3408
3409 asoc = sctp_id2assoc(sk, peeloff.associd);
3410 if (!asoc) {
3411 retval = -EINVAL;
3412 goto out;
3413 }
3414
3415 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3416
3417 retval = sctp_do_peeloff(asoc, &newsock);
3418 if (retval < 0)
3419 goto out;
3420
3421 /* Map the socket to an unused fd that can be returned to the user. */
3422 retval = sock_map_fd(newsock);
3423 if (retval < 0) {
3424 sock_release(newsock);
3425 goto out;
3426 }
3427
3428 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3429 __FUNCTION__, sk, asoc, newsock->sk, retval);
3430
3431 /* Return the fd mapped to the new socket. */
3432 peeloff.sd = retval;
3433 if (copy_to_user(optval, &peeloff, len))
3434 retval = -EFAULT;
3435
3436out:
3437 return retval;
3438}
3439
3440/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3441 *
3442 * Applications can enable or disable heartbeats for any peer address of
3443 * an association, modify an address's heartbeat interval, force a
3444 * heartbeat to be sent immediately, and adjust the address's maximum
3445 * number of retransmissions sent before an address is considered
3446 * unreachable. The following structure is used to access and modify an
3447 * address's parameters:
3448 *
3449 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08003450 * sctp_assoc_t spp_assoc_id;
3451 * struct sockaddr_storage spp_address;
3452 * uint32_t spp_hbinterval;
3453 * uint16_t spp_pathmaxrxt;
3454 * uint32_t spp_pathmtu;
3455 * uint32_t spp_sackdelay;
3456 * uint32_t spp_flags;
3457 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08003459 * spp_assoc_id - (one-to-many style socket) This is filled in the
3460 * application, and identifies the association for
3461 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003462 * spp_address - This specifies which address is of interest.
3463 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003464 * in milliseconds. If a value of zero
3465 * is present in this field then no changes are to
3466 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003467 * spp_pathmaxrxt - This contains the maximum number of
3468 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08003469 * considered unreachable. If a value of zero
3470 * is present in this field then no changes are to
3471 * be made to this parameter.
3472 * spp_pathmtu - When Path MTU discovery is disabled the value
3473 * specified here will be the "fixed" path mtu.
3474 * Note that if the spp_address field is empty
3475 * then all associations on this address will
3476 * have this fixed path mtu set upon them.
3477 *
3478 * spp_sackdelay - When delayed sack is enabled, this value specifies
3479 * the number of milliseconds that sacks will be delayed
3480 * for. This value will apply to all addresses of an
3481 * association if the spp_address field is empty. Note
3482 * also, that if delayed sack is enabled and this
3483 * value is set to 0, no change is made to the last
3484 * recorded delayed sack timer value.
3485 *
3486 * spp_flags - These flags are used to control various features
3487 * on an association. The flag field may contain
3488 * zero or more of the following options.
3489 *
3490 * SPP_HB_ENABLE - Enable heartbeats on the
3491 * specified address. Note that if the address
3492 * field is empty all addresses for the association
3493 * have heartbeats enabled upon them.
3494 *
3495 * SPP_HB_DISABLE - Disable heartbeats on the
3496 * speicifed address. Note that if the address
3497 * field is empty all addresses for the association
3498 * will have their heartbeats disabled. Note also
3499 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3500 * mutually exclusive, only one of these two should
3501 * be specified. Enabling both fields will have
3502 * undetermined results.
3503 *
3504 * SPP_HB_DEMAND - Request a user initiated heartbeat
3505 * to be made immediately.
3506 *
3507 * SPP_PMTUD_ENABLE - This field will enable PMTU
3508 * discovery upon the specified address. Note that
3509 * if the address feild is empty then all addresses
3510 * on the association are effected.
3511 *
3512 * SPP_PMTUD_DISABLE - This field will disable PMTU
3513 * discovery upon the specified address. Note that
3514 * if the address feild is empty then all addresses
3515 * on the association are effected. Not also that
3516 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3517 * exclusive. Enabling both will have undetermined
3518 * results.
3519 *
3520 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3521 * on delayed sack. The time specified in spp_sackdelay
3522 * is used to specify the sack delay for this address. Note
3523 * that if spp_address is empty then all addresses will
3524 * enable delayed sack and take on the sack delay
3525 * value specified in spp_sackdelay.
3526 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3527 * off delayed sack. If the spp_address field is blank then
3528 * delayed sack is disabled for the entire association. Note
3529 * also that this field is mutually exclusive to
3530 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3531 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003532 */
3533static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08003534 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003535{
Frank Filz52ccb8e2005-12-22 11:36:46 -08003536 struct sctp_paddrparams params;
3537 struct sctp_transport *trans = NULL;
3538 struct sctp_association *asoc = NULL;
3539 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003540
3541 if (len != sizeof(struct sctp_paddrparams))
3542 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003543
Linus Torvalds1da177e2005-04-16 15:20:36 -07003544 if (copy_from_user(&params, optval, len))
3545 return -EFAULT;
3546
Frank Filz52ccb8e2005-12-22 11:36:46 -08003547 /* If an address other than INADDR_ANY is specified, and
3548 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003549 */
Frank Filz52ccb8e2005-12-22 11:36:46 -08003550 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3551 trans = sctp_addr_id2transport(sk, &params.spp_address,
3552 params.spp_assoc_id);
3553 if (!trans) {
3554 SCTP_DEBUG_PRINTK("Failed no transport\n");
3555 return -EINVAL;
3556 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003557 }
3558
Frank Filz52ccb8e2005-12-22 11:36:46 -08003559 /* Get association, if assoc_id != 0 and the socket is a one
3560 * to many style socket, and an association was not found, then
3561 * the id was invalid.
3562 */
3563 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3564 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3565 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003566 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003567 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003568
Frank Filz52ccb8e2005-12-22 11:36:46 -08003569 if (trans) {
3570 /* Fetch transport values. */
3571 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3572 params.spp_pathmtu = trans->pathmtu;
3573 params.spp_pathmaxrxt = trans->pathmaxrxt;
3574 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003575
Frank Filz52ccb8e2005-12-22 11:36:46 -08003576 /*draft-11 doesn't say what to return in spp_flags*/
3577 params.spp_flags = trans->param_flags;
3578 } else if (asoc) {
3579 /* Fetch association values. */
3580 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3581 params.spp_pathmtu = asoc->pathmtu;
3582 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3583 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003584
Frank Filz52ccb8e2005-12-22 11:36:46 -08003585 /*draft-11 doesn't say what to return in spp_flags*/
3586 params.spp_flags = asoc->param_flags;
3587 } else {
3588 /* Fetch socket values. */
3589 params.spp_hbinterval = sp->hbinterval;
3590 params.spp_pathmtu = sp->pathmtu;
3591 params.spp_sackdelay = sp->sackdelay;
3592 params.spp_pathmaxrxt = sp->pathmaxrxt;
3593
3594 /*draft-11 doesn't say what to return in spp_flags*/
3595 params.spp_flags = sp->param_flags;
3596 }
3597
Linus Torvalds1da177e2005-04-16 15:20:36 -07003598 if (copy_to_user(optval, &params, len))
3599 return -EFAULT;
3600
3601 if (put_user(len, optlen))
3602 return -EFAULT;
3603
3604 return 0;
3605}
3606
Frank Filz77086102005-12-22 11:37:30 -08003607/* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3608 *
3609 * This options will get or set the delayed ack timer. The time is set
3610 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3611 * endpoints default delayed ack timer value. If the assoc_id field is
3612 * non-zero, then the set or get effects the specified association.
3613 *
3614 * struct sctp_assoc_value {
3615 * sctp_assoc_t assoc_id;
3616 * uint32_t assoc_value;
3617 * };
3618 *
3619 * assoc_id - This parameter, indicates which association the
3620 * user is preforming an action upon. Note that if
3621 * this field's value is zero then the endpoints
3622 * default value is changed (effecting future
3623 * associations only).
3624 *
3625 * assoc_value - This parameter contains the number of milliseconds
3626 * that the user is requesting the delayed ACK timer
3627 * be set to. Note that this value is defined in
3628 * the standard to be between 200 and 500 milliseconds.
3629 *
3630 * Note: a value of zero will leave the value alone,
3631 * but disable SACK delay. A non-zero value will also
3632 * enable SACK delay.
3633 */
3634static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3635 char __user *optval,
3636 int __user *optlen)
3637{
3638 struct sctp_assoc_value params;
3639 struct sctp_association *asoc = NULL;
3640 struct sctp_sock *sp = sctp_sk(sk);
3641
3642 if (len != sizeof(struct sctp_assoc_value))
3643 return - EINVAL;
3644
3645 if (copy_from_user(&params, optval, len))
3646 return -EFAULT;
3647
3648 /* Get association, if assoc_id != 0 and the socket is a one
3649 * to many style socket, and an association was not found, then
3650 * the id was invalid.
3651 */
3652 asoc = sctp_id2assoc(sk, params.assoc_id);
3653 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3654 return -EINVAL;
3655
3656 if (asoc) {
3657 /* Fetch association values. */
3658 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3659 params.assoc_value = jiffies_to_msecs(
3660 asoc->sackdelay);
3661 else
3662 params.assoc_value = 0;
3663 } else {
3664 /* Fetch socket values. */
3665 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3666 params.assoc_value = sp->sackdelay;
3667 else
3668 params.assoc_value = 0;
3669 }
3670
3671 if (copy_to_user(optval, &params, len))
3672 return -EFAULT;
3673
3674 if (put_user(len, optlen))
3675 return -EFAULT;
3676
3677 return 0;
3678}
3679
Linus Torvalds1da177e2005-04-16 15:20:36 -07003680/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3681 *
3682 * Applications can specify protocol parameters for the default association
3683 * initialization. The option name argument to setsockopt() and getsockopt()
3684 * is SCTP_INITMSG.
3685 *
3686 * Setting initialization parameters is effective only on an unconnected
3687 * socket (for UDP-style sockets only future associations are effected
3688 * by the change). With TCP-style sockets, this option is inherited by
3689 * sockets derived from a listener socket.
3690 */
3691static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3692{
3693 if (len != sizeof(struct sctp_initmsg))
3694 return -EINVAL;
3695 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3696 return -EFAULT;
3697 return 0;
3698}
3699
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003700static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3701 char __user *optval,
3702 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003703{
3704 sctp_assoc_t id;
3705 struct sctp_association *asoc;
3706 struct list_head *pos;
3707 int cnt = 0;
3708
3709 if (len != sizeof(sctp_assoc_t))
3710 return -EINVAL;
3711
3712 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3713 return -EFAULT;
3714
3715 /* For UDP-style sockets, id specifies the association to query. */
3716 asoc = sctp_id2assoc(sk, id);
3717 if (!asoc)
3718 return -EINVAL;
3719
3720 list_for_each(pos, &asoc->peer.transport_addr_list) {
3721 cnt ++;
3722 }
3723
3724 return cnt;
3725}
3726
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003727/*
3728 * Old API for getting list of peer addresses. Does not work for 32-bit
3729 * programs running on a 64-bit kernel
3730 */
3731static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3732 char __user *optval,
3733 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003734{
3735 struct sctp_association *asoc;
3736 struct list_head *pos;
3737 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003738 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003739 struct sctp_transport *from;
3740 void __user *to;
3741 union sctp_addr temp;
3742 struct sctp_sock *sp = sctp_sk(sk);
3743 int addrlen;
3744
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003745 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003746 return -EINVAL;
3747
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003748 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003749 return -EFAULT;
3750
3751 if (getaddrs.addr_num <= 0) return -EINVAL;
3752
3753 /* For UDP-style sockets, id specifies the association to query. */
3754 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3755 if (!asoc)
3756 return -EINVAL;
3757
3758 to = (void __user *)getaddrs.addrs;
3759 list_for_each(pos, &asoc->peer.transport_addr_list) {
3760 from = list_entry(pos, struct sctp_transport, transports);
3761 memcpy(&temp, &from->ipaddr, sizeof(temp));
3762 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3763 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3764 temp.v4.sin_port = htons(temp.v4.sin_port);
3765 if (copy_to_user(to, &temp, addrlen))
3766 return -EFAULT;
3767 to += addrlen ;
3768 cnt ++;
3769 if (cnt >= getaddrs.addr_num) break;
3770 }
3771 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003772 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003773 return -EFAULT;
3774
3775 return 0;
3776}
3777
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003778static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3779 char __user *optval, int __user *optlen)
3780{
3781 struct sctp_association *asoc;
3782 struct list_head *pos;
3783 int cnt = 0;
3784 struct sctp_getaddrs getaddrs;
3785 struct sctp_transport *from;
3786 void __user *to;
3787 union sctp_addr temp;
3788 struct sctp_sock *sp = sctp_sk(sk);
3789 int addrlen;
3790 size_t space_left;
3791 int bytes_copied;
3792
3793 if (len < sizeof(struct sctp_getaddrs))
3794 return -EINVAL;
3795
3796 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3797 return -EFAULT;
3798
3799 /* For UDP-style sockets, id specifies the association to query. */
3800 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3801 if (!asoc)
3802 return -EINVAL;
3803
3804 to = optval + offsetof(struct sctp_getaddrs,addrs);
3805 space_left = len - sizeof(struct sctp_getaddrs) -
3806 offsetof(struct sctp_getaddrs,addrs);
3807
3808 list_for_each(pos, &asoc->peer.transport_addr_list) {
3809 from = list_entry(pos, struct sctp_transport, transports);
3810 memcpy(&temp, &from->ipaddr, sizeof(temp));
3811 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3812 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3813 if(space_left < addrlen)
3814 return -ENOMEM;
3815 temp.v4.sin_port = htons(temp.v4.sin_port);
3816 if (copy_to_user(to, &temp, addrlen))
3817 return -EFAULT;
3818 to += addrlen;
3819 cnt++;
3820 space_left -= addrlen;
3821 }
3822
3823 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3824 return -EFAULT;
3825 bytes_copied = ((char __user *)to) - optval;
3826 if (put_user(bytes_copied, optlen))
3827 return -EFAULT;
3828
3829 return 0;
3830}
3831
3832static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3833 char __user *optval,
3834 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003835{
3836 sctp_assoc_t id;
3837 struct sctp_bind_addr *bp;
3838 struct sctp_association *asoc;
3839 struct list_head *pos;
3840 struct sctp_sockaddr_entry *addr;
3841 rwlock_t *addr_lock;
3842 unsigned long flags;
3843 int cnt = 0;
3844
3845 if (len != sizeof(sctp_assoc_t))
3846 return -EINVAL;
3847
3848 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3849 return -EFAULT;
3850
3851 /*
3852 * For UDP-style sockets, id specifies the association to query.
3853 * If the id field is set to the value '0' then the locally bound
3854 * addresses are returned without regard to any particular
3855 * association.
3856 */
3857 if (0 == id) {
3858 bp = &sctp_sk(sk)->ep->base.bind_addr;
3859 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3860 } else {
3861 asoc = sctp_id2assoc(sk, id);
3862 if (!asoc)
3863 return -EINVAL;
3864 bp = &asoc->base.bind_addr;
3865 addr_lock = &asoc->base.addr_lock;
3866 }
3867
3868 sctp_read_lock(addr_lock);
3869
3870 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3871 * addresses from the global local address list.
3872 */
3873 if (sctp_list_single_entry(&bp->address_list)) {
3874 addr = list_entry(bp->address_list.next,
3875 struct sctp_sockaddr_entry, list);
3876 if (sctp_is_any(&addr->a)) {
3877 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3878 list_for_each(pos, &sctp_local_addr_list) {
3879 addr = list_entry(pos,
3880 struct sctp_sockaddr_entry,
3881 list);
3882 if ((PF_INET == sk->sk_family) &&
3883 (AF_INET6 == addr->a.sa.sa_family))
3884 continue;
3885 cnt++;
3886 }
3887 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3888 flags);
3889 } else {
3890 cnt = 1;
3891 }
3892 goto done;
3893 }
3894
3895 list_for_each(pos, &bp->address_list) {
3896 cnt ++;
3897 }
3898
3899done:
3900 sctp_read_unlock(addr_lock);
3901 return cnt;
3902}
3903
3904/* Helper function that copies local addresses to user and returns the number
3905 * of addresses copied.
3906 */
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003907static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3908 void __user *to)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003909{
3910 struct list_head *pos;
3911 struct sctp_sockaddr_entry *addr;
3912 unsigned long flags;
3913 union sctp_addr temp;
3914 int cnt = 0;
3915 int addrlen;
3916
3917 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3918 list_for_each(pos, &sctp_local_addr_list) {
3919 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3920 if ((PF_INET == sk->sk_family) &&
3921 (AF_INET6 == addr->a.sa.sa_family))
3922 continue;
3923 memcpy(&temp, &addr->a, sizeof(temp));
3924 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3925 &temp);
3926 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3927 temp.v4.sin_port = htons(port);
3928 if (copy_to_user(to, &temp, addrlen)) {
3929 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3930 flags);
3931 return -EFAULT;
3932 }
3933 to += addrlen;
3934 cnt ++;
3935 if (cnt >= max_addrs) break;
3936 }
3937 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3938
3939 return cnt;
3940}
3941
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003942static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
Al Virod3a880e2005-12-15 09:18:30 +00003943 void __user **to, size_t space_left)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003944{
3945 struct list_head *pos;
3946 struct sctp_sockaddr_entry *addr;
3947 unsigned long flags;
3948 union sctp_addr temp;
3949 int cnt = 0;
3950 int addrlen;
3951
3952 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3953 list_for_each(pos, &sctp_local_addr_list) {
3954 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3955 if ((PF_INET == sk->sk_family) &&
3956 (AF_INET6 == addr->a.sa.sa_family))
3957 continue;
3958 memcpy(&temp, &addr->a, sizeof(temp));
3959 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3960 &temp);
3961 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3962 if(space_left<addrlen)
3963 return -ENOMEM;
3964 temp.v4.sin_port = htons(port);
3965 if (copy_to_user(*to, &temp, addrlen)) {
3966 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3967 flags);
3968 return -EFAULT;
3969 }
3970 *to += addrlen;
3971 cnt ++;
3972 space_left -= addrlen;
3973 }
3974 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3975
3976 return cnt;
3977}
3978
3979/* Old API for getting list of local addresses. Does not work for 32-bit
3980 * programs running on a 64-bit kernel
3981 */
3982static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3983 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003984{
3985 struct sctp_bind_addr *bp;
3986 struct sctp_association *asoc;
3987 struct list_head *pos;
3988 int cnt = 0;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003989 struct sctp_getaddrs_old getaddrs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003990 struct sctp_sockaddr_entry *addr;
3991 void __user *to;
3992 union sctp_addr temp;
3993 struct sctp_sock *sp = sctp_sk(sk);
3994 int addrlen;
3995 rwlock_t *addr_lock;
3996 int err = 0;
3997
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07003998 if (len != sizeof(struct sctp_getaddrs_old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003999 return -EINVAL;
4000
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004001 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004002 return -EFAULT;
4003
4004 if (getaddrs.addr_num <= 0) return -EINVAL;
4005 /*
4006 * For UDP-style sockets, id specifies the association to query.
4007 * If the id field is set to the value '0' then the locally bound
4008 * addresses are returned without regard to any particular
4009 * association.
4010 */
4011 if (0 == getaddrs.assoc_id) {
4012 bp = &sctp_sk(sk)->ep->base.bind_addr;
4013 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4014 } else {
4015 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4016 if (!asoc)
4017 return -EINVAL;
4018 bp = &asoc->base.bind_addr;
4019 addr_lock = &asoc->base.addr_lock;
4020 }
4021
4022 to = getaddrs.addrs;
4023
4024 sctp_read_lock(addr_lock);
4025
4026 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4027 * addresses from the global local address list.
4028 */
4029 if (sctp_list_single_entry(&bp->address_list)) {
4030 addr = list_entry(bp->address_list.next,
4031 struct sctp_sockaddr_entry, list);
4032 if (sctp_is_any(&addr->a)) {
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004033 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4034 getaddrs.addr_num,
4035 to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004036 if (cnt < 0) {
4037 err = cnt;
4038 goto unlock;
4039 }
4040 goto copy_getaddrs;
4041 }
4042 }
4043
4044 list_for_each(pos, &bp->address_list) {
4045 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4046 memcpy(&temp, &addr->a, sizeof(temp));
4047 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4048 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4049 temp.v4.sin_port = htons(temp.v4.sin_port);
4050 if (copy_to_user(to, &temp, addrlen)) {
4051 err = -EFAULT;
4052 goto unlock;
4053 }
4054 to += addrlen;
4055 cnt ++;
4056 if (cnt >= getaddrs.addr_num) break;
4057 }
4058
4059copy_getaddrs:
4060 getaddrs.addr_num = cnt;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004061 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004062 err = -EFAULT;
4063
4064unlock:
4065 sctp_read_unlock(addr_lock);
4066 return err;
4067}
4068
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004069static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4070 char __user *optval, int __user *optlen)
4071{
4072 struct sctp_bind_addr *bp;
4073 struct sctp_association *asoc;
4074 struct list_head *pos;
4075 int cnt = 0;
4076 struct sctp_getaddrs getaddrs;
4077 struct sctp_sockaddr_entry *addr;
4078 void __user *to;
4079 union sctp_addr temp;
4080 struct sctp_sock *sp = sctp_sk(sk);
4081 int addrlen;
4082 rwlock_t *addr_lock;
4083 int err = 0;
4084 size_t space_left;
4085 int bytes_copied;
4086
4087 if (len <= sizeof(struct sctp_getaddrs))
4088 return -EINVAL;
4089
4090 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4091 return -EFAULT;
4092
4093 /*
4094 * For UDP-style sockets, id specifies the association to query.
4095 * If the id field is set to the value '0' then the locally bound
4096 * addresses are returned without regard to any particular
4097 * association.
4098 */
4099 if (0 == getaddrs.assoc_id) {
4100 bp = &sctp_sk(sk)->ep->base.bind_addr;
4101 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4102 } else {
4103 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4104 if (!asoc)
4105 return -EINVAL;
4106 bp = &asoc->base.bind_addr;
4107 addr_lock = &asoc->base.addr_lock;
4108 }
4109
4110 to = optval + offsetof(struct sctp_getaddrs,addrs);
4111 space_left = len - sizeof(struct sctp_getaddrs) -
4112 offsetof(struct sctp_getaddrs,addrs);
4113
4114 sctp_read_lock(addr_lock);
4115
4116 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4117 * addresses from the global local address list.
4118 */
4119 if (sctp_list_single_entry(&bp->address_list)) {
4120 addr = list_entry(bp->address_list.next,
4121 struct sctp_sockaddr_entry, list);
4122 if (sctp_is_any(&addr->a)) {
4123 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4124 &to, space_left);
4125 if (cnt < 0) {
4126 err = cnt;
4127 goto unlock;
4128 }
4129 goto copy_getaddrs;
4130 }
4131 }
4132
4133 list_for_each(pos, &bp->address_list) {
4134 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4135 memcpy(&temp, &addr->a, sizeof(temp));
4136 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4137 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4138 if(space_left < addrlen)
4139 return -ENOMEM; /*fixme: right error?*/
4140 temp.v4.sin_port = htons(temp.v4.sin_port);
4141 if (copy_to_user(to, &temp, addrlen)) {
4142 err = -EFAULT;
4143 goto unlock;
4144 }
4145 to += addrlen;
4146 cnt ++;
4147 space_left -= addrlen;
4148 }
4149
4150copy_getaddrs:
4151 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4152 return -EFAULT;
4153 bytes_copied = ((char __user *)to) - optval;
4154 if (put_user(bytes_copied, optlen))
4155 return -EFAULT;
4156
4157unlock:
4158 sctp_read_unlock(addr_lock);
4159 return err;
4160}
4161
Linus Torvalds1da177e2005-04-16 15:20:36 -07004162/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4163 *
4164 * Requests that the local SCTP stack use the enclosed peer address as
4165 * the association primary. The enclosed address must be one of the
4166 * association peer's addresses.
4167 */
4168static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4169 char __user *optval, int __user *optlen)
4170{
4171 struct sctp_prim prim;
4172 struct sctp_association *asoc;
4173 struct sctp_sock *sp = sctp_sk(sk);
4174
4175 if (len != sizeof(struct sctp_prim))
4176 return -EINVAL;
4177
4178 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4179 return -EFAULT;
4180
4181 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4182 if (!asoc)
4183 return -EINVAL;
4184
4185 if (!asoc->peer.primary_path)
4186 return -ENOTCONN;
4187
4188 asoc->peer.primary_path->ipaddr.v4.sin_port =
4189 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
4190 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4191 sizeof(union sctp_addr));
4192 asoc->peer.primary_path->ipaddr.v4.sin_port =
4193 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
4194
4195 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4196 (union sctp_addr *)&prim.ssp_addr);
4197
4198 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4199 return -EFAULT;
4200
4201 return 0;
4202}
4203
4204/*
4205 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4206 *
4207 * Requests that the local endpoint set the specified Adaption Layer
4208 * Indication parameter for all future INIT and INIT-ACK exchanges.
4209 */
4210static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4211 char __user *optval, int __user *optlen)
4212{
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004213 struct sctp_setadaption adaption;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004214
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004215 if (len != sizeof(struct sctp_setadaption))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004216 return -EINVAL;
4217
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004218 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4219 if (copy_to_user(optval, &adaption, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004220 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004221
Linus Torvalds1da177e2005-04-16 15:20:36 -07004222 return 0;
4223}
4224
4225/*
4226 *
4227 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4228 *
4229 * Applications that wish to use the sendto() system call may wish to
4230 * specify a default set of parameters that would normally be supplied
4231 * through the inclusion of ancillary data. This socket option allows
4232 * such an application to set the default sctp_sndrcvinfo structure.
4233
4234
4235 * The application that wishes to use this socket option simply passes
4236 * in to this call the sctp_sndrcvinfo structure defined in Section
4237 * 5.2.2) The input parameters accepted by this call include
4238 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4239 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4240 * to this call if the caller is using the UDP model.
4241 *
4242 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4243 */
4244static int sctp_getsockopt_default_send_param(struct sock *sk,
4245 int len, char __user *optval,
4246 int __user *optlen)
4247{
4248 struct sctp_sndrcvinfo info;
4249 struct sctp_association *asoc;
4250 struct sctp_sock *sp = sctp_sk(sk);
4251
4252 if (len != sizeof(struct sctp_sndrcvinfo))
4253 return -EINVAL;
4254 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4255 return -EFAULT;
4256
4257 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4258 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4259 return -EINVAL;
4260
4261 if (asoc) {
4262 info.sinfo_stream = asoc->default_stream;
4263 info.sinfo_flags = asoc->default_flags;
4264 info.sinfo_ppid = asoc->default_ppid;
4265 info.sinfo_context = asoc->default_context;
4266 info.sinfo_timetolive = asoc->default_timetolive;
4267 } else {
4268 info.sinfo_stream = sp->default_stream;
4269 info.sinfo_flags = sp->default_flags;
4270 info.sinfo_ppid = sp->default_ppid;
4271 info.sinfo_context = sp->default_context;
4272 info.sinfo_timetolive = sp->default_timetolive;
4273 }
4274
4275 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4276 return -EFAULT;
4277
4278 return 0;
4279}
4280
4281/*
4282 *
4283 * 7.1.5 SCTP_NODELAY
4284 *
4285 * Turn on/off any Nagle-like algorithm. This means that packets are
4286 * generally sent as soon as possible and no unnecessary delays are
4287 * introduced, at the cost of more packets in the network. Expects an
4288 * integer boolean flag.
4289 */
4290
4291static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4292 char __user *optval, int __user *optlen)
4293{
4294 int val;
4295
4296 if (len < sizeof(int))
4297 return -EINVAL;
4298
4299 len = sizeof(int);
4300 val = (sctp_sk(sk)->nodelay == 1);
4301 if (put_user(len, optlen))
4302 return -EFAULT;
4303 if (copy_to_user(optval, &val, len))
4304 return -EFAULT;
4305 return 0;
4306}
4307
4308/*
4309 *
4310 * 7.1.1 SCTP_RTOINFO
4311 *
4312 * The protocol parameters used to initialize and bound retransmission
4313 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4314 * and modify these parameters.
4315 * All parameters are time values, in milliseconds. A value of 0, when
4316 * modifying the parameters, indicates that the current value should not
4317 * be changed.
4318 *
4319 */
4320static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4321 char __user *optval,
4322 int __user *optlen) {
4323 struct sctp_rtoinfo rtoinfo;
4324 struct sctp_association *asoc;
4325
4326 if (len != sizeof (struct sctp_rtoinfo))
4327 return -EINVAL;
4328
4329 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4330 return -EFAULT;
4331
4332 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4333
4334 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4335 return -EINVAL;
4336
4337 /* Values corresponding to the specific association. */
4338 if (asoc) {
4339 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4340 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4341 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4342 } else {
4343 /* Values corresponding to the endpoint. */
4344 struct sctp_sock *sp = sctp_sk(sk);
4345
4346 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4347 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4348 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4349 }
4350
4351 if (put_user(len, optlen))
4352 return -EFAULT;
4353
4354 if (copy_to_user(optval, &rtoinfo, len))
4355 return -EFAULT;
4356
4357 return 0;
4358}
4359
4360/*
4361 *
4362 * 7.1.2 SCTP_ASSOCINFO
4363 *
4364 * This option is used to tune the the maximum retransmission attempts
4365 * of the association.
4366 * Returns an error if the new association retransmission value is
4367 * greater than the sum of the retransmission value of the peer.
4368 * See [SCTP] for more information.
4369 *
4370 */
4371static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4372 char __user *optval,
4373 int __user *optlen)
4374{
4375
4376 struct sctp_assocparams assocparams;
4377 struct sctp_association *asoc;
4378 struct list_head *pos;
4379 int cnt = 0;
4380
4381 if (len != sizeof (struct sctp_assocparams))
4382 return -EINVAL;
4383
4384 if (copy_from_user(&assocparams, optval,
4385 sizeof (struct sctp_assocparams)))
4386 return -EFAULT;
4387
4388 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4389
4390 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4391 return -EINVAL;
4392
4393 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07004394 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004395 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4396 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4397 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4398 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4399 * 1000) +
4400 (asoc->cookie_life.tv_usec
4401 / 1000);
4402
4403 list_for_each(pos, &asoc->peer.transport_addr_list) {
4404 cnt ++;
4405 }
4406
4407 assocparams.sasoc_number_peer_destinations = cnt;
4408 } else {
4409 /* Values corresponding to the endpoint */
4410 struct sctp_sock *sp = sctp_sk(sk);
4411
4412 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4413 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4414 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4415 assocparams.sasoc_cookie_life =
4416 sp->assocparams.sasoc_cookie_life;
4417 assocparams.sasoc_number_peer_destinations =
4418 sp->assocparams.
4419 sasoc_number_peer_destinations;
4420 }
4421
4422 if (put_user(len, optlen))
4423 return -EFAULT;
4424
4425 if (copy_to_user(optval, &assocparams, len))
4426 return -EFAULT;
4427
4428 return 0;
4429}
4430
4431/*
4432 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4433 *
4434 * This socket option is a boolean flag which turns on or off mapped V4
4435 * addresses. If this option is turned on and the socket is type
4436 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4437 * If this option is turned off, then no mapping will be done of V4
4438 * addresses and a user will receive both PF_INET6 and PF_INET type
4439 * addresses on the socket.
4440 */
4441static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4442 char __user *optval, int __user *optlen)
4443{
4444 int val;
4445 struct sctp_sock *sp = sctp_sk(sk);
4446
4447 if (len < sizeof(int))
4448 return -EINVAL;
4449
4450 len = sizeof(int);
4451 val = sp->v4mapped;
4452 if (put_user(len, optlen))
4453 return -EFAULT;
4454 if (copy_to_user(optval, &val, len))
4455 return -EFAULT;
4456
4457 return 0;
4458}
4459
4460/*
4461 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4462 *
4463 * This socket option specifies the maximum size to put in any outgoing
4464 * SCTP chunk. If a message is larger than this size it will be
4465 * fragmented by SCTP into the specified size. Note that the underlying
4466 * SCTP implementation may fragment into smaller sized chunks when the
4467 * PMTU of the underlying association is smaller than the value set by
4468 * the user.
4469 */
4470static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4471 char __user *optval, int __user *optlen)
4472{
4473 int val;
4474
4475 if (len < sizeof(int))
4476 return -EINVAL;
4477
4478 len = sizeof(int);
4479
4480 val = sctp_sk(sk)->user_frag;
4481 if (put_user(len, optlen))
4482 return -EFAULT;
4483 if (copy_to_user(optval, &val, len))
4484 return -EFAULT;
4485
4486 return 0;
4487}
4488
4489SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4490 char __user *optval, int __user *optlen)
4491{
4492 int retval = 0;
4493 int len;
4494
Frank Filz3f7a87d2005-06-20 13:14:57 -07004495 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4496 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004497
4498 /* I can hardly begin to describe how wrong this is. This is
4499 * so broken as to be worse than useless. The API draft
4500 * REALLY is NOT helpful here... I am not convinced that the
4501 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4502 * are at all well-founded.
4503 */
4504 if (level != SOL_SCTP) {
4505 struct sctp_af *af = sctp_sk(sk)->pf->af;
4506
4507 retval = af->getsockopt(sk, level, optname, optval, optlen);
4508 return retval;
4509 }
4510
4511 if (get_user(len, optlen))
4512 return -EFAULT;
4513
4514 sctp_lock_sock(sk);
4515
4516 switch (optname) {
4517 case SCTP_STATUS:
4518 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4519 break;
4520 case SCTP_DISABLE_FRAGMENTS:
4521 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4522 optlen);
4523 break;
4524 case SCTP_EVENTS:
4525 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4526 break;
4527 case SCTP_AUTOCLOSE:
4528 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4529 break;
4530 case SCTP_SOCKOPT_PEELOFF:
4531 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4532 break;
4533 case SCTP_PEER_ADDR_PARAMS:
4534 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4535 optlen);
4536 break;
Frank Filz77086102005-12-22 11:37:30 -08004537 case SCTP_DELAYED_ACK_TIME:
4538 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4539 optlen);
4540 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004541 case SCTP_INITMSG:
4542 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4543 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004544 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4545 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4546 optlen);
4547 break;
4548 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4549 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4550 optlen);
4551 break;
4552 case SCTP_GET_PEER_ADDRS_OLD:
4553 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004554 optlen);
4555 break;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004556 case SCTP_GET_LOCAL_ADDRS_OLD:
4557 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004558 optlen);
4559 break;
4560 case SCTP_GET_PEER_ADDRS:
4561 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4562 optlen);
4563 break;
4564 case SCTP_GET_LOCAL_ADDRS:
4565 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4566 optlen);
4567 break;
4568 case SCTP_DEFAULT_SEND_PARAM:
4569 retval = sctp_getsockopt_default_send_param(sk, len,
4570 optval, optlen);
4571 break;
4572 case SCTP_PRIMARY_ADDR:
4573 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4574 break;
4575 case SCTP_NODELAY:
4576 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4577 break;
4578 case SCTP_RTOINFO:
4579 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4580 break;
4581 case SCTP_ASSOCINFO:
4582 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4583 break;
4584 case SCTP_I_WANT_MAPPED_V4_ADDR:
4585 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4586 break;
4587 case SCTP_MAXSEG:
4588 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4589 break;
4590 case SCTP_GET_PEER_ADDR_INFO:
4591 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4592 optlen);
4593 break;
4594 case SCTP_ADAPTION_LAYER:
4595 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4596 optlen);
4597 break;
4598 default:
4599 retval = -ENOPROTOOPT;
4600 break;
4601 };
4602
4603 sctp_release_sock(sk);
4604 return retval;
4605}
4606
4607static void sctp_hash(struct sock *sk)
4608{
4609 /* STUB */
4610}
4611
4612static void sctp_unhash(struct sock *sk)
4613{
4614 /* STUB */
4615}
4616
4617/* Check if port is acceptable. Possibly find first available port.
4618 *
4619 * The port hash table (contained in the 'global' SCTP protocol storage
4620 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4621 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4622 * list (the list number is the port number hashed out, so as you
4623 * would expect from a hash function, all the ports in a given list have
4624 * such a number that hashes out to the same list number; you were
4625 * expecting that, right?); so each list has a set of ports, with a
4626 * link to the socket (struct sock) that uses it, the port number and
4627 * a fastreuse flag (FIXME: NPI ipg).
4628 */
4629static struct sctp_bind_bucket *sctp_bucket_create(
4630 struct sctp_bind_hashbucket *head, unsigned short snum);
4631
4632static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4633{
4634 struct sctp_bind_hashbucket *head; /* hash list */
4635 struct sctp_bind_bucket *pp; /* hash list port iterator */
4636 unsigned short snum;
4637 int ret;
4638
4639 /* NOTE: Remember to put this back to net order. */
4640 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4641 snum = addr->v4.sin_port;
4642
4643 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4644 sctp_local_bh_disable();
4645
4646 if (snum == 0) {
4647 /* Search for an available port.
4648 *
4649 * 'sctp_port_rover' was the last port assigned, so
4650 * we start to search from 'sctp_port_rover +
4651 * 1'. What we do is first check if port 'rover' is
4652 * already in the hash table; if not, we use that; if
4653 * it is, we try next.
4654 */
4655 int low = sysctl_local_port_range[0];
4656 int high = sysctl_local_port_range[1];
4657 int remaining = (high - low) + 1;
4658 int rover;
4659 int index;
4660
4661 sctp_spin_lock(&sctp_port_alloc_lock);
4662 rover = sctp_port_rover;
4663 do {
4664 rover++;
4665 if ((rover < low) || (rover > high))
4666 rover = low;
4667 index = sctp_phashfn(rover);
4668 head = &sctp_port_hashtable[index];
4669 sctp_spin_lock(&head->lock);
4670 for (pp = head->chain; pp; pp = pp->next)
4671 if (pp->port == rover)
4672 goto next;
4673 break;
4674 next:
4675 sctp_spin_unlock(&head->lock);
4676 } while (--remaining > 0);
4677 sctp_port_rover = rover;
4678 sctp_spin_unlock(&sctp_port_alloc_lock);
4679
4680 /* Exhausted local port range during search? */
4681 ret = 1;
4682 if (remaining <= 0)
4683 goto fail;
4684
4685 /* OK, here is the one we will use. HEAD (the port
4686 * hash table list entry) is non-NULL and we hold it's
4687 * mutex.
4688 */
4689 snum = rover;
4690 } else {
4691 /* We are given an specific port number; we verify
4692 * that it is not being used. If it is used, we will
4693 * exahust the search in the hash list corresponding
4694 * to the port number (snum) - we detect that with the
4695 * port iterator, pp being NULL.
4696 */
4697 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4698 sctp_spin_lock(&head->lock);
4699 for (pp = head->chain; pp; pp = pp->next) {
4700 if (pp->port == snum)
4701 goto pp_found;
4702 }
4703 }
4704 pp = NULL;
4705 goto pp_not_found;
4706pp_found:
4707 if (!hlist_empty(&pp->owner)) {
4708 /* We had a port hash table hit - there is an
4709 * available port (pp != NULL) and it is being
4710 * used by other socket (pp->owner not empty); that other
4711 * socket is going to be sk2.
4712 */
4713 int reuse = sk->sk_reuse;
4714 struct sock *sk2;
4715 struct hlist_node *node;
4716
4717 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4718 if (pp->fastreuse && sk->sk_reuse)
4719 goto success;
4720
4721 /* Run through the list of sockets bound to the port
4722 * (pp->port) [via the pointers bind_next and
4723 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4724 * we get the endpoint they describe and run through
4725 * the endpoint's list of IP (v4 or v6) addresses,
4726 * comparing each of the addresses with the address of
4727 * the socket sk. If we find a match, then that means
4728 * that this port/socket (sk) combination are already
4729 * in an endpoint.
4730 */
4731 sk_for_each_bound(sk2, node, &pp->owner) {
4732 struct sctp_endpoint *ep2;
4733 ep2 = sctp_sk(sk2)->ep;
4734
4735 if (reuse && sk2->sk_reuse)
4736 continue;
4737
4738 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4739 sctp_sk(sk))) {
4740 ret = (long)sk2;
4741 goto fail_unlock;
4742 }
4743 }
4744 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4745 }
4746pp_not_found:
4747 /* If there was a hash table miss, create a new port. */
4748 ret = 1;
4749 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4750 goto fail_unlock;
4751
4752 /* In either case (hit or miss), make sure fastreuse is 1 only
4753 * if sk->sk_reuse is too (that is, if the caller requested
4754 * SO_REUSEADDR on this socket -sk-).
4755 */
4756 if (hlist_empty(&pp->owner))
4757 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4758 else if (pp->fastreuse && !sk->sk_reuse)
4759 pp->fastreuse = 0;
4760
4761 /* We are set, so fill up all the data in the hash table
4762 * entry, tie the socket list information with the rest of the
4763 * sockets FIXME: Blurry, NPI (ipg).
4764 */
4765success:
4766 inet_sk(sk)->num = snum;
4767 if (!sctp_sk(sk)->bind_hash) {
4768 sk_add_bind_node(sk, &pp->owner);
4769 sctp_sk(sk)->bind_hash = pp;
4770 }
4771 ret = 0;
4772
4773fail_unlock:
4774 sctp_spin_unlock(&head->lock);
4775
4776fail:
4777 sctp_local_bh_enable();
4778 addr->v4.sin_port = htons(addr->v4.sin_port);
4779 return ret;
4780}
4781
4782/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4783 * port is requested.
4784 */
4785static int sctp_get_port(struct sock *sk, unsigned short snum)
4786{
4787 long ret;
4788 union sctp_addr addr;
4789 struct sctp_af *af = sctp_sk(sk)->pf->af;
4790
4791 /* Set up a dummy address struct from the sk. */
4792 af->from_sk(&addr, sk);
4793 addr.v4.sin_port = htons(snum);
4794
4795 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4796 ret = sctp_get_port_local(sk, &addr);
4797
4798 return (ret ? 1 : 0);
4799}
4800
4801/*
4802 * 3.1.3 listen() - UDP Style Syntax
4803 *
4804 * By default, new associations are not accepted for UDP style sockets.
4805 * An application uses listen() to mark a socket as being able to
4806 * accept new associations.
4807 */
4808SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4809{
4810 struct sctp_sock *sp = sctp_sk(sk);
4811 struct sctp_endpoint *ep = sp->ep;
4812
4813 /* Only UDP style sockets that are not peeled off are allowed to
4814 * listen().
4815 */
4816 if (!sctp_style(sk, UDP))
4817 return -EINVAL;
4818
4819 /* If backlog is zero, disable listening. */
4820 if (!backlog) {
4821 if (sctp_sstate(sk, CLOSED))
4822 return 0;
4823
4824 sctp_unhash_endpoint(ep);
4825 sk->sk_state = SCTP_SS_CLOSED;
4826 }
4827
4828 /* Return if we are already listening. */
4829 if (sctp_sstate(sk, LISTENING))
4830 return 0;
4831
4832 /*
4833 * If a bind() or sctp_bindx() is not called prior to a listen()
4834 * call that allows new associations to be accepted, the system
4835 * picks an ephemeral port and will choose an address set equivalent
4836 * to binding with a wildcard address.
4837 *
4838 * This is not currently spelled out in the SCTP sockets
4839 * extensions draft, but follows the practice as seen in TCP
4840 * sockets.
4841 */
4842 if (!ep->base.bind_addr.port) {
4843 if (sctp_autobind(sk))
4844 return -EAGAIN;
4845 }
4846 sk->sk_state = SCTP_SS_LISTENING;
4847 sctp_hash_endpoint(ep);
4848 return 0;
4849}
4850
4851/*
4852 * 4.1.3 listen() - TCP Style Syntax
4853 *
4854 * Applications uses listen() to ready the SCTP endpoint for accepting
4855 * inbound associations.
4856 */
4857SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4858{
4859 struct sctp_sock *sp = sctp_sk(sk);
4860 struct sctp_endpoint *ep = sp->ep;
4861
4862 /* If backlog is zero, disable listening. */
4863 if (!backlog) {
4864 if (sctp_sstate(sk, CLOSED))
4865 return 0;
4866
4867 sctp_unhash_endpoint(ep);
4868 sk->sk_state = SCTP_SS_CLOSED;
4869 }
4870
4871 if (sctp_sstate(sk, LISTENING))
4872 return 0;
4873
4874 /*
4875 * If a bind() or sctp_bindx() is not called prior to a listen()
4876 * call that allows new associations to be accepted, the system
4877 * picks an ephemeral port and will choose an address set equivalent
4878 * to binding with a wildcard address.
4879 *
4880 * This is not currently spelled out in the SCTP sockets
4881 * extensions draft, but follows the practice as seen in TCP
4882 * sockets.
4883 */
4884 if (!ep->base.bind_addr.port) {
4885 if (sctp_autobind(sk))
4886 return -EAGAIN;
4887 }
4888 sk->sk_state = SCTP_SS_LISTENING;
4889 sk->sk_max_ack_backlog = backlog;
4890 sctp_hash_endpoint(ep);
4891 return 0;
4892}
4893
4894/*
4895 * Move a socket to LISTENING state.
4896 */
4897int sctp_inet_listen(struct socket *sock, int backlog)
4898{
4899 struct sock *sk = sock->sk;
Herbert Xu1b489e12006-08-20 15:07:14 +10004900 struct crypto_hash *tfm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004901 int err = -EINVAL;
4902
4903 if (unlikely(backlog < 0))
4904 goto out;
4905
4906 sctp_lock_sock(sk);
4907
4908 if (sock->state != SS_UNCONNECTED)
4909 goto out;
4910
4911 /* Allocate HMAC for generating cookie. */
4912 if (sctp_hmac_alg) {
Herbert Xu1b489e12006-08-20 15:07:14 +10004913 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004914 if (!tfm) {
4915 err = -ENOSYS;
4916 goto out;
4917 }
4918 }
4919
4920 switch (sock->type) {
4921 case SOCK_SEQPACKET:
4922 err = sctp_seqpacket_listen(sk, backlog);
4923 break;
4924 case SOCK_STREAM:
4925 err = sctp_stream_listen(sk, backlog);
4926 break;
4927 default:
4928 break;
4929 };
4930 if (err)
4931 goto cleanup;
4932
4933 /* Store away the transform reference. */
4934 sctp_sk(sk)->hmac = tfm;
4935out:
4936 sctp_release_sock(sk);
4937 return err;
4938cleanup:
Herbert Xu1b489e12006-08-20 15:07:14 +10004939 crypto_free_hash(tfm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004940 goto out;
4941}
4942
4943/*
4944 * This function is done by modeling the current datagram_poll() and the
4945 * tcp_poll(). Note that, based on these implementations, we don't
4946 * lock the socket in this function, even though it seems that,
4947 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08004948 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07004949 * in this place. We assume that we don't need locks either until proven
4950 * otherwise.
4951 *
4952 * Another thing to note is that we include the Async I/O support
4953 * here, again, by modeling the current TCP/UDP code. We don't have
4954 * a good way to test with it yet.
4955 */
4956unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4957{
4958 struct sock *sk = sock->sk;
4959 struct sctp_sock *sp = sctp_sk(sk);
4960 unsigned int mask;
4961
4962 poll_wait(file, sk->sk_sleep, wait);
4963
4964 /* A TCP-style listening socket becomes readable when the accept queue
4965 * is not empty.
4966 */
4967 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4968 return (!list_empty(&sp->ep->asocs)) ?
4969 (POLLIN | POLLRDNORM) : 0;
4970
4971 mask = 0;
4972
4973 /* Is there any exceptional events? */
4974 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4975 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08004976 if (sk->sk_shutdown & RCV_SHUTDOWN)
4977 mask |= POLLRDHUP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004978 if (sk->sk_shutdown == SHUTDOWN_MASK)
4979 mask |= POLLHUP;
4980
4981 /* Is it readable? Reconsider this code with TCP-style support. */
4982 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4983 (sk->sk_shutdown & RCV_SHUTDOWN))
4984 mask |= POLLIN | POLLRDNORM;
4985
4986 /* The association is either gone or not ready. */
4987 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4988 return mask;
4989
4990 /* Is it writable? */
4991 if (sctp_writeable(sk)) {
4992 mask |= POLLOUT | POLLWRNORM;
4993 } else {
4994 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4995 /*
4996 * Since the socket is not locked, the buffer
4997 * might be made available after the writeable check and
4998 * before the bit is set. This could cause a lost I/O
4999 * signal. tcp_poll() has a race breaker for this race
5000 * condition. Based on their implementation, we put
5001 * in the following code to cover it as well.
5002 */
5003 if (sctp_writeable(sk))
5004 mask |= POLLOUT | POLLWRNORM;
5005 }
5006 return mask;
5007}
5008
5009/********************************************************************
5010 * 2nd Level Abstractions
5011 ********************************************************************/
5012
5013static struct sctp_bind_bucket *sctp_bucket_create(
5014 struct sctp_bind_hashbucket *head, unsigned short snum)
5015{
5016 struct sctp_bind_bucket *pp;
5017
5018 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
5019 SCTP_DBG_OBJCNT_INC(bind_bucket);
5020 if (pp) {
5021 pp->port = snum;
5022 pp->fastreuse = 0;
5023 INIT_HLIST_HEAD(&pp->owner);
5024 if ((pp->next = head->chain) != NULL)
5025 pp->next->pprev = &pp->next;
5026 head->chain = pp;
5027 pp->pprev = &head->chain;
5028 }
5029 return pp;
5030}
5031
5032/* Caller must hold hashbucket lock for this tb with local BH disabled */
5033static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5034{
Sridhar Samudrala37fa6872006-07-21 14:45:47 -07005035 if (pp && hlist_empty(&pp->owner)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005036 if (pp->next)
5037 pp->next->pprev = pp->pprev;
5038 *(pp->pprev) = pp->next;
5039 kmem_cache_free(sctp_bucket_cachep, pp);
5040 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5041 }
5042}
5043
5044/* Release this socket's reference to a local port. */
5045static inline void __sctp_put_port(struct sock *sk)
5046{
5047 struct sctp_bind_hashbucket *head =
5048 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5049 struct sctp_bind_bucket *pp;
5050
5051 sctp_spin_lock(&head->lock);
5052 pp = sctp_sk(sk)->bind_hash;
5053 __sk_del_bind_node(sk);
5054 sctp_sk(sk)->bind_hash = NULL;
5055 inet_sk(sk)->num = 0;
5056 sctp_bucket_destroy(pp);
5057 sctp_spin_unlock(&head->lock);
5058}
5059
5060void sctp_put_port(struct sock *sk)
5061{
5062 sctp_local_bh_disable();
5063 __sctp_put_port(sk);
5064 sctp_local_bh_enable();
5065}
5066
5067/*
5068 * The system picks an ephemeral port and choose an address set equivalent
5069 * to binding with a wildcard address.
5070 * One of those addresses will be the primary address for the association.
5071 * This automatically enables the multihoming capability of SCTP.
5072 */
5073static int sctp_autobind(struct sock *sk)
5074{
5075 union sctp_addr autoaddr;
5076 struct sctp_af *af;
5077 unsigned short port;
5078
5079 /* Initialize a local sockaddr structure to INADDR_ANY. */
5080 af = sctp_sk(sk)->pf->af;
5081
5082 port = htons(inet_sk(sk)->num);
5083 af->inaddr_any(&autoaddr, port);
5084
5085 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5086}
5087
5088/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5089 *
5090 * From RFC 2292
5091 * 4.2 The cmsghdr Structure *
5092 *
5093 * When ancillary data is sent or received, any number of ancillary data
5094 * objects can be specified by the msg_control and msg_controllen members of
5095 * the msghdr structure, because each object is preceded by
5096 * a cmsghdr structure defining the object's length (the cmsg_len member).
5097 * Historically Berkeley-derived implementations have passed only one object
5098 * at a time, but this API allows multiple objects to be
5099 * passed in a single call to sendmsg() or recvmsg(). The following example
5100 * shows two ancillary data objects in a control buffer.
5101 *
5102 * |<--------------------------- msg_controllen -------------------------->|
5103 * | |
5104 *
5105 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5106 *
5107 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5108 * | | |
5109 *
5110 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5111 *
5112 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5113 * | | | | |
5114 *
5115 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5116 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5117 *
5118 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5119 *
5120 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5121 * ^
5122 * |
5123 *
5124 * msg_control
5125 * points here
5126 */
5127SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5128 sctp_cmsgs_t *cmsgs)
5129{
5130 struct cmsghdr *cmsg;
5131
5132 for (cmsg = CMSG_FIRSTHDR(msg);
5133 cmsg != NULL;
5134 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5135 if (!CMSG_OK(msg, cmsg))
5136 return -EINVAL;
5137
5138 /* Should we parse this header or ignore? */
5139 if (cmsg->cmsg_level != IPPROTO_SCTP)
5140 continue;
5141
5142 /* Strictly check lengths following example in SCM code. */
5143 switch (cmsg->cmsg_type) {
5144 case SCTP_INIT:
5145 /* SCTP Socket API Extension
5146 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5147 *
5148 * This cmsghdr structure provides information for
5149 * initializing new SCTP associations with sendmsg().
5150 * The SCTP_INITMSG socket option uses this same data
5151 * structure. This structure is not used for
5152 * recvmsg().
5153 *
5154 * cmsg_level cmsg_type cmsg_data[]
5155 * ------------ ------------ ----------------------
5156 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5157 */
5158 if (cmsg->cmsg_len !=
5159 CMSG_LEN(sizeof(struct sctp_initmsg)))
5160 return -EINVAL;
5161 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5162 break;
5163
5164 case SCTP_SNDRCV:
5165 /* SCTP Socket API Extension
5166 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5167 *
5168 * This cmsghdr structure specifies SCTP options for
5169 * sendmsg() and describes SCTP header information
5170 * about a received message through recvmsg().
5171 *
5172 * cmsg_level cmsg_type cmsg_data[]
5173 * ------------ ------------ ----------------------
5174 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5175 */
5176 if (cmsg->cmsg_len !=
5177 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5178 return -EINVAL;
5179
5180 cmsgs->info =
5181 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5182
5183 /* Minimally, validate the sinfo_flags. */
5184 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07005185 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5186 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005187 return -EINVAL;
5188 break;
5189
5190 default:
5191 return -EINVAL;
5192 };
5193 }
5194 return 0;
5195}
5196
5197/*
5198 * Wait for a packet..
5199 * Note: This function is the same function as in core/datagram.c
5200 * with a few modifications to make lksctp work.
5201 */
5202static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5203{
5204 int error;
5205 DEFINE_WAIT(wait);
5206
5207 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5208
5209 /* Socket errors? */
5210 error = sock_error(sk);
5211 if (error)
5212 goto out;
5213
5214 if (!skb_queue_empty(&sk->sk_receive_queue))
5215 goto ready;
5216
5217 /* Socket shut down? */
5218 if (sk->sk_shutdown & RCV_SHUTDOWN)
5219 goto out;
5220
5221 /* Sequenced packets can come disconnected. If so we report the
5222 * problem.
5223 */
5224 error = -ENOTCONN;
5225
5226 /* Is there a good reason to think that we may receive some data? */
5227 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5228 goto out;
5229
5230 /* Handle signals. */
5231 if (signal_pending(current))
5232 goto interrupted;
5233
5234 /* Let another process have a go. Since we are going to sleep
5235 * anyway. Note: This may cause odd behaviors if the message
5236 * does not fit in the user's buffer, but this seems to be the
5237 * only way to honor MSG_DONTWAIT realistically.
5238 */
5239 sctp_release_sock(sk);
5240 *timeo_p = schedule_timeout(*timeo_p);
5241 sctp_lock_sock(sk);
5242
5243ready:
5244 finish_wait(sk->sk_sleep, &wait);
5245 return 0;
5246
5247interrupted:
5248 error = sock_intr_errno(*timeo_p);
5249
5250out:
5251 finish_wait(sk->sk_sleep, &wait);
5252 *err = error;
5253 return error;
5254}
5255
5256/* Receive a datagram.
5257 * Note: This is pretty much the same routine as in core/datagram.c
5258 * with a few changes to make lksctp work.
5259 */
5260static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5261 int noblock, int *err)
5262{
5263 int error;
5264 struct sk_buff *skb;
5265 long timeo;
5266
Linus Torvalds1da177e2005-04-16 15:20:36 -07005267 timeo = sock_rcvtimeo(sk, noblock);
5268
5269 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5270 timeo, MAX_SCHEDULE_TIMEOUT);
5271
5272 do {
5273 /* Again only user level code calls this function,
5274 * so nothing interrupt level
5275 * will suddenly eat the receive_queue.
5276 *
5277 * Look at current nfs client by the way...
5278 * However, this function was corrent in any case. 8)
5279 */
5280 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07005281 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005282 skb = skb_peek(&sk->sk_receive_queue);
5283 if (skb)
5284 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07005285 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005286 } else {
5287 skb = skb_dequeue(&sk->sk_receive_queue);
5288 }
5289
5290 if (skb)
5291 return skb;
5292
Neil Horman6736dc32005-12-02 20:30:06 -08005293 /* Caller is allowed not to check sk->sk_err before calling. */
5294 error = sock_error(sk);
5295 if (error)
5296 goto no_packet;
5297
Linus Torvalds1da177e2005-04-16 15:20:36 -07005298 if (sk->sk_shutdown & RCV_SHUTDOWN)
5299 break;
5300
5301 /* User doesn't want to wait. */
5302 error = -EAGAIN;
5303 if (!timeo)
5304 goto no_packet;
5305 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5306
5307 return NULL;
5308
5309no_packet:
5310 *err = error;
5311 return NULL;
5312}
5313
5314/* If sndbuf has changed, wake up per association sndbuf waiters. */
5315static void __sctp_write_space(struct sctp_association *asoc)
5316{
5317 struct sock *sk = asoc->base.sk;
5318 struct socket *sock = sk->sk_socket;
5319
5320 if ((sctp_wspace(asoc) > 0) && sock) {
5321 if (waitqueue_active(&asoc->wait))
5322 wake_up_interruptible(&asoc->wait);
5323
5324 if (sctp_writeable(sk)) {
5325 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5326 wake_up_interruptible(sk->sk_sleep);
5327
5328 /* Note that we try to include the Async I/O support
5329 * here by modeling from the current TCP/UDP code.
5330 * We have not tested with it yet.
5331 */
5332 if (sock->fasync_list &&
5333 !(sk->sk_shutdown & SEND_SHUTDOWN))
5334 sock_wake_async(sock, 2, POLL_OUT);
5335 }
5336 }
5337}
5338
5339/* Do accounting for the sndbuf space.
5340 * Decrement the used sndbuf space of the corresponding association by the
5341 * data size which was just transmitted(freed).
5342 */
5343static void sctp_wfree(struct sk_buff *skb)
5344{
5345 struct sctp_association *asoc;
5346 struct sctp_chunk *chunk;
5347 struct sock *sk;
5348
5349 /* Get the saved chunk pointer. */
5350 chunk = *((struct sctp_chunk **)(skb->cb));
5351 asoc = chunk->asoc;
5352 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07005353 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5354 sizeof(struct sk_buff) +
5355 sizeof(struct sctp_chunk);
5356
Neil Horman4eb701d2005-04-28 12:02:04 -07005357 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5358
5359 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005360 __sctp_write_space(asoc);
5361
5362 sctp_association_put(asoc);
5363}
5364
5365/* Helper function to wait for space in the sndbuf. */
5366static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5367 size_t msg_len)
5368{
5369 struct sock *sk = asoc->base.sk;
5370 int err = 0;
5371 long current_timeo = *timeo_p;
5372 DEFINE_WAIT(wait);
5373
5374 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5375 asoc, (long)(*timeo_p), msg_len);
5376
5377 /* Increment the association's refcnt. */
5378 sctp_association_hold(asoc);
5379
5380 /* Wait on the association specific sndbuf space. */
5381 for (;;) {
5382 prepare_to_wait_exclusive(&asoc->wait, &wait,
5383 TASK_INTERRUPTIBLE);
5384 if (!*timeo_p)
5385 goto do_nonblock;
5386 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5387 asoc->base.dead)
5388 goto do_error;
5389 if (signal_pending(current))
5390 goto do_interrupted;
5391 if (msg_len <= sctp_wspace(asoc))
5392 break;
5393
5394 /* Let another process have a go. Since we are going
5395 * to sleep anyway.
5396 */
5397 sctp_release_sock(sk);
5398 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005399 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005400 sctp_lock_sock(sk);
5401
5402 *timeo_p = current_timeo;
5403 }
5404
5405out:
5406 finish_wait(&asoc->wait, &wait);
5407
5408 /* Release the association's refcnt. */
5409 sctp_association_put(asoc);
5410
5411 return err;
5412
5413do_error:
5414 err = -EPIPE;
5415 goto out;
5416
5417do_interrupted:
5418 err = sock_intr_errno(*timeo_p);
5419 goto out;
5420
5421do_nonblock:
5422 err = -EAGAIN;
5423 goto out;
5424}
5425
5426/* If socket sndbuf has changed, wake up all per association waiters. */
5427void sctp_write_space(struct sock *sk)
5428{
5429 struct sctp_association *asoc;
5430 struct list_head *pos;
5431
5432 /* Wake up the tasks in each wait queue. */
5433 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5434 asoc = list_entry(pos, struct sctp_association, asocs);
5435 __sctp_write_space(asoc);
5436 }
5437}
5438
5439/* Is there any sndbuf space available on the socket?
5440 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08005441 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07005442 * associations on the same socket. For a UDP-style socket with
5443 * multiple associations, it is possible for it to be "unwriteable"
5444 * prematurely. I assume that this is acceptable because
5445 * a premature "unwriteable" is better than an accidental "writeable" which
5446 * would cause an unwanted block under certain circumstances. For the 1-1
5447 * UDP-style sockets or TCP-style sockets, this code should work.
5448 * - Daisy
5449 */
5450static int sctp_writeable(struct sock *sk)
5451{
5452 int amt = 0;
5453
Neil Horman9bffc4a2005-12-19 14:24:40 -08005454 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005455 if (amt < 0)
5456 amt = 0;
5457 return amt;
5458}
5459
5460/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5461 * returns immediately with EINPROGRESS.
5462 */
5463static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5464{
5465 struct sock *sk = asoc->base.sk;
5466 int err = 0;
5467 long current_timeo = *timeo_p;
5468 DEFINE_WAIT(wait);
5469
5470 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5471 (long)(*timeo_p));
5472
5473 /* Increment the association's refcnt. */
5474 sctp_association_hold(asoc);
5475
5476 for (;;) {
5477 prepare_to_wait_exclusive(&asoc->wait, &wait,
5478 TASK_INTERRUPTIBLE);
5479 if (!*timeo_p)
5480 goto do_nonblock;
5481 if (sk->sk_shutdown & RCV_SHUTDOWN)
5482 break;
5483 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5484 asoc->base.dead)
5485 goto do_error;
5486 if (signal_pending(current))
5487 goto do_interrupted;
5488
5489 if (sctp_state(asoc, ESTABLISHED))
5490 break;
5491
5492 /* Let another process have a go. Since we are going
5493 * to sleep anyway.
5494 */
5495 sctp_release_sock(sk);
5496 current_timeo = schedule_timeout(current_timeo);
5497 sctp_lock_sock(sk);
5498
5499 *timeo_p = current_timeo;
5500 }
5501
5502out:
5503 finish_wait(&asoc->wait, &wait);
5504
5505 /* Release the association's refcnt. */
5506 sctp_association_put(asoc);
5507
5508 return err;
5509
5510do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08005511 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005512 err = -ETIMEDOUT;
5513 else
5514 err = -ECONNREFUSED;
5515 goto out;
5516
5517do_interrupted:
5518 err = sock_intr_errno(*timeo_p);
5519 goto out;
5520
5521do_nonblock:
5522 err = -EINPROGRESS;
5523 goto out;
5524}
5525
5526static int sctp_wait_for_accept(struct sock *sk, long timeo)
5527{
5528 struct sctp_endpoint *ep;
5529 int err = 0;
5530 DEFINE_WAIT(wait);
5531
5532 ep = sctp_sk(sk)->ep;
5533
5534
5535 for (;;) {
5536 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5537 TASK_INTERRUPTIBLE);
5538
5539 if (list_empty(&ep->asocs)) {
5540 sctp_release_sock(sk);
5541 timeo = schedule_timeout(timeo);
5542 sctp_lock_sock(sk);
5543 }
5544
5545 err = -EINVAL;
5546 if (!sctp_sstate(sk, LISTENING))
5547 break;
5548
5549 err = 0;
5550 if (!list_empty(&ep->asocs))
5551 break;
5552
5553 err = sock_intr_errno(timeo);
5554 if (signal_pending(current))
5555 break;
5556
5557 err = -EAGAIN;
5558 if (!timeo)
5559 break;
5560 }
5561
5562 finish_wait(sk->sk_sleep, &wait);
5563
5564 return err;
5565}
5566
5567void sctp_wait_for_close(struct sock *sk, long timeout)
5568{
5569 DEFINE_WAIT(wait);
5570
5571 do {
5572 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5573 if (list_empty(&sctp_sk(sk)->ep->asocs))
5574 break;
5575 sctp_release_sock(sk);
5576 timeout = schedule_timeout(timeout);
5577 sctp_lock_sock(sk);
5578 } while (!signal_pending(current) && timeout);
5579
5580 finish_wait(sk->sk_sleep, &wait);
5581}
5582
5583/* Populate the fields of the newsk from the oldsk and migrate the assoc
5584 * and its messages to the newsk.
5585 */
5586static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5587 struct sctp_association *assoc,
5588 sctp_socket_type_t type)
5589{
5590 struct sctp_sock *oldsp = sctp_sk(oldsk);
5591 struct sctp_sock *newsp = sctp_sk(newsk);
5592 struct sctp_bind_bucket *pp; /* hash list port iterator */
5593 struct sctp_endpoint *newep = newsp->ep;
5594 struct sk_buff *skb, *tmp;
5595 struct sctp_ulpevent *event;
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005596 int flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005597
5598 /* Migrate socket buffer sizes and all the socket level options to the
5599 * new socket.
5600 */
5601 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5602 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5603 /* Brute force copy old sctp opt. */
5604 inet_sk_copy_descendant(newsk, oldsk);
5605
5606 /* Restore the ep value that was overwritten with the above structure
5607 * copy.
5608 */
5609 newsp->ep = newep;
5610 newsp->hmac = NULL;
5611
5612 /* Hook this new socket in to the bind_hash list. */
5613 pp = sctp_sk(oldsk)->bind_hash;
5614 sk_add_bind_node(newsk, &pp->owner);
5615 sctp_sk(newsk)->bind_hash = pp;
5616 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5617
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005618 /* Copy the bind_addr list from the original endpoint to the new
5619 * endpoint so that we can handle restarts properly
5620 */
Vladislav Yasevicheb5fa392006-08-22 00:23:13 -07005621 if (PF_INET6 == assoc->base.sk->sk_family)
5622 flags = SCTP_ADDR6_ALLOWED;
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07005623 if (assoc->peer.ipv4_address)
5624 flags |= SCTP_ADDR4_PEERSUPP;
5625 if (assoc->peer.ipv6_address)
5626 flags |= SCTP_ADDR6_PEERSUPP;
5627 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5628 &oldsp->ep->base.bind_addr,
5629 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5630
Linus Torvalds1da177e2005-04-16 15:20:36 -07005631 /* Move any messages in the old socket's receive queue that are for the
5632 * peeled off association to the new socket's receive queue.
5633 */
5634 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5635 event = sctp_skb2event(skb);
5636 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005637 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005638 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005639 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005640 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005641 }
5642 }
5643
5644 /* Clean up any messages pending delivery due to partial
5645 * delivery. Three cases:
5646 * 1) No partial deliver; no work.
5647 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5648 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5649 */
5650 skb_queue_head_init(&newsp->pd_lobby);
5651 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5652
5653 if (sctp_sk(oldsk)->pd_mode) {
5654 struct sk_buff_head *queue;
5655
5656 /* Decide which queue to move pd_lobby skbs to. */
5657 if (assoc->ulpq.pd_mode) {
5658 queue = &newsp->pd_lobby;
5659 } else
5660 queue = &newsk->sk_receive_queue;
5661
5662 /* Walk through the pd_lobby, looking for skbs that
5663 * need moved to the new socket.
5664 */
5665 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5666 event = sctp_skb2event(skb);
5667 if (event->asoc == assoc) {
Neil Horman049b3ff2005-11-11 16:08:24 -08005668 sock_rfree(skb);
David S. Miller8728b832005-08-09 19:25:21 -07005669 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005670 __skb_queue_tail(queue, skb);
Neil Horman049b3ff2005-11-11 16:08:24 -08005671 skb_set_owner_r(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005672 }
5673 }
5674
5675 /* Clear up any skbs waiting for the partial
5676 * delivery to finish.
5677 */
5678 if (assoc->ulpq.pd_mode)
5679 sctp_clear_pd(oldsk);
5680
5681 }
5682
5683 /* Set the type of socket to indicate that it is peeled off from the
5684 * original UDP-style socket or created with the accept() call on a
5685 * TCP-style socket..
5686 */
5687 newsp->type = type;
5688
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005689 /* Mark the new socket "in-use" by the user so that any packets
5690 * that may arrive on the association after we've moved it are
5691 * queued to the backlog. This prevents a potential race between
5692 * backlog processing on the old socket and new-packet processing
5693 * on the new socket.
5694 */
5695 sctp_lock_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005696 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005697
5698 /* If the association on the newsk is already closed before accept()
5699 * is called, set RCV_SHUTDOWN flag.
5700 */
5701 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5702 newsk->sk_shutdown |= RCV_SHUTDOWN;
5703
5704 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07005705 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005706}
5707
5708/* This proto struct describes the ULP interface for SCTP. */
5709struct proto sctp_prot = {
5710 .name = "SCTP",
5711 .owner = THIS_MODULE,
5712 .close = sctp_close,
5713 .connect = sctp_connect,
5714 .disconnect = sctp_disconnect,
5715 .accept = sctp_accept,
5716 .ioctl = sctp_ioctl,
5717 .init = sctp_init_sock,
5718 .destroy = sctp_destroy_sock,
5719 .shutdown = sctp_shutdown,
5720 .setsockopt = sctp_setsockopt,
5721 .getsockopt = sctp_getsockopt,
5722 .sendmsg = sctp_sendmsg,
5723 .recvmsg = sctp_recvmsg,
5724 .bind = sctp_bind,
5725 .backlog_rcv = sctp_backlog_rcv,
5726 .hash = sctp_hash,
5727 .unhash = sctp_unhash,
5728 .get_port = sctp_get_port,
5729 .obj_size = sizeof(struct sctp_sock),
5730};
5731
5732#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5733struct proto sctpv6_prot = {
5734 .name = "SCTPv6",
5735 .owner = THIS_MODULE,
5736 .close = sctp_close,
5737 .connect = sctp_connect,
5738 .disconnect = sctp_disconnect,
5739 .accept = sctp_accept,
5740 .ioctl = sctp_ioctl,
5741 .init = sctp_init_sock,
5742 .destroy = sctp_destroy_sock,
5743 .shutdown = sctp_shutdown,
5744 .setsockopt = sctp_setsockopt,
5745 .getsockopt = sctp_getsockopt,
5746 .sendmsg = sctp_sendmsg,
5747 .recvmsg = sctp_recvmsg,
5748 .bind = sctp_bind,
5749 .backlog_rcv = sctp_backlog_rcv,
5750 .hash = sctp_hash,
5751 .unhash = sctp_unhash,
5752 .get_port = sctp_get_port,
5753 .obj_size = sizeof(struct sctp6_sock),
5754};
5755#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */