be2net: support multiple TX queues
[linux-2.6-block.git] / net / sctp / socket.c
CommitLineData
60c778b2 1/* SCTP kernel implementation
1da177e4
LT
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 *
60c778b2 9 * This file is part of the SCTP kernel implementation
1da177e4
LT
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 *
60c778b2 18 * This SCTP implementation is free software;
1da177e4
LT
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 *
60c778b2 24 * This SCTP implementation is distributed in the hope that it
1da177e4
LT
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
145ce502
JP
60#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
1da177e4
LT
62#include <linux/types.h>
63#include <linux/kernel.h>
64#include <linux/wait.h>
65#include <linux/time.h>
66#include <linux/ip.h>
4fc268d2 67#include <linux/capability.h>
1da177e4
LT
68#include <linux/fcntl.h>
69#include <linux/poll.h>
70#include <linux/init.h>
71#include <linux/crypto.h>
5a0e3ad6 72#include <linux/slab.h>
1da177e4
LT
73
74#include <net/ip.h>
75#include <net/icmp.h>
76#include <net/route.h>
77#include <net/ipv6.h>
78#include <net/inet_common.h>
79
80#include <linux/socket.h> /* for sa_family_t */
81#include <net/sock.h>
82#include <net/sctp/sctp.h>
83#include <net/sctp/sm.h>
84
85/* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
88 */
89
90/* Forward declarations for internal helper functions. */
91static int sctp_writeable(struct sock *sk);
92static void sctp_wfree(struct sk_buff *skb);
93static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97static int sctp_wait_for_accept(struct sock *sk, long timeo);
98static void sctp_wait_for_close(struct sock *sk, long timeo);
99static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108static int sctp_autobind(struct sock *sk);
109static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
112
4d93df0a 113extern struct kmem_cache *sctp_bucket_cachep;
8d987e5c 114extern long sysctl_sctp_mem[3];
4d93df0a
NH
115extern int sysctl_sctp_rmem[3];
116extern int sysctl_sctp_wmem[3];
117
b6fa1a4d 118static int sctp_memory_pressure;
8d987e5c 119static atomic_long_t sctp_memory_allocated;
1748376b 120struct percpu_counter sctp_sockets_allocated;
4d93df0a 121
5c52ba17 122static void sctp_enter_memory_pressure(struct sock *sk)
4d93df0a
NH
123{
124 sctp_memory_pressure = 1;
125}
126
127
1da177e4
LT
128/* Get the sndbuf space available at the time on the association. */
129static inline int sctp_wspace(struct sctp_association *asoc)
130{
4d93df0a 131 int amt;
1da177e4 132
4d93df0a
NH
133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
31e6d363 136 amt = sk_wmem_alloc_get(asoc->base.sk);
4d93df0a
NH
137
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
145 }
4eb701df 146 } else {
4d93df0a 147 amt = asoc->base.sk->sk_sndbuf - amt;
4eb701df 148 }
1da177e4
LT
149 return amt;
150}
151
152/* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
155 *
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
160 */
161static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
162{
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
165
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
168
4eb701df
NH
169 skb_set_owner_w(chunk->skb, sk);
170
1da177e4
LT
171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
174
4eb701df
NH
175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
178
4eb701df 179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
3ab224be
HA
180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
1da177e4
LT
182}
183
184/* Verify that this is a valid address. */
185static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
186 int len)
187{
188 struct sctp_af *af;
189
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 if (!af)
193 return -EINVAL;
194
195 /* Is this a valid SCTP address? */
5636bef7 196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
1da177e4
LT
197 return -EINVAL;
198
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
201
202 return 0;
203}
204
205/* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
207 */
208struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
209{
210 struct sctp_association *asoc = NULL;
211
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
217 */
218 if (!sctp_sstate(sk, ESTABLISHED))
219 return NULL;
220
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
225 return asoc;
226 }
227
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
230 return NULL;
231
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
235
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
237 return NULL;
238
239 return asoc;
240}
241
242/* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
244 * the same.
245 */
246static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
248 sctp_assoc_t id)
249{
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
253
1da177e4 254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
cd4ff034 255 laddr,
1da177e4 256 &transport);
1da177e4
LT
257
258 if (!addr_asoc)
259 return NULL;
260
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
264
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
267
268 return transport;
269}
270
271/* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
273 *
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
275 *
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
280 */
3f7a87d2 281SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
1da177e4
LT
282{
283 int retval = 0;
284
285 sctp_lock_sock(sk);
286
3f7a87d2
FF
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
1da177e4
LT
289
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
3f7a87d2 292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
1da177e4
LT
293 addr_len);
294 else
295 retval = -EINVAL;
296
297 sctp_release_sock(sk);
298
299 return retval;
300}
301
302static long sctp_get_port_local(struct sock *, union sctp_addr *);
303
304/* Verify this is a valid sockaddr. */
305static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
307{
308 struct sctp_af *af;
309
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
312 return NULL;
313
7dab83de
VY
314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
323 }
1da177e4
LT
324
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
327
328 if (len < af->sockaddr_len)
329 return NULL;
330
331 return af;
332}
333
334/* Bind a local address either to an endpoint or to an association. */
335SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
336{
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
340 struct sctp_af *af;
341 unsigned short snum;
342 int ret = 0;
343
1da177e4
LT
344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
3f7a87d2
FF
346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
1da177e4 349 return -EINVAL;
3f7a87d2
FF
350 }
351
352 snum = ntohs(addr->v4.sin_port);
353
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
356 sk,
357 addr,
358 bp->port, snum,
359 len);
1da177e4
LT
360
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
364
8b358056
VY
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
368 */
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
1da177e4
LT
374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
8b358056
VY
376 return -EINVAL;
377 }
1da177e4
LT
378 }
379
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
382
4e54064e
VY
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
385 */
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
388
1da177e4
LT
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
392 */
2772b495 393 addr->v4.sin_port = htons(snum);
1da177e4 394 if ((ret = sctp_get_port_local(sk, addr))) {
4e54064e 395 return -EADDRINUSE;
1da177e4
LT
396 }
397
398 /* Refresh ephemeral port. */
399 if (!bp->port)
c720c7e8 400 bp->port = inet_sk(sk)->inet_num;
1da177e4 401
559cf710
VY
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
404 */
f57d96b2 405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
1da177e4
LT
406
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
c720c7e8 409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
1da177e4
LT
410 af->to_sk_saddr(addr, sk);
411 }
412
413 return ret;
414}
415
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 *
d808ad9a 418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
1da177e4 419 * at any one time. If a sender, after sending an ASCONF chunk, decides
d808ad9a 420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
1da177e4 421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
d808ad9a
YH
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
1da177e4
LT
424 * from each endpoint).
425 */
426static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
428{
429 int retval = 0;
430
431 /* If there is an outstanding ASCONF chunk, queue it for later
432 * transmission.
d808ad9a 433 */
1da177e4 434 if (asoc->addip_last_asconf) {
79af02c2 435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
d808ad9a 436 goto out;
1da177e4
LT
437 }
438
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
446
447out:
448 return retval;
449}
450
451/* Add a list of addresses as bind addresses to local endpoint or
452 * association.
453 *
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
457 *
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
460 *
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
462 */
04675210 463static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
1da177e4
LT
464{
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
470
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
473
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
478 */
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
484 }
485
d808ad9a 486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
1da177e4
LT
487 af->sockaddr_len);
488
489 addr_buf += af->sockaddr_len;
490
491err_bindx_add:
492 if (retval < 0) {
493 /* Failed. Cleanup the ones that have been added */
494 if (cnt > 0)
495 sctp_bindx_rem(sk, addrs, cnt);
496 return retval;
497 }
498 }
499
500 return retval;
501}
502
503/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
506 *
d808ad9a 507 * If any of the addresses is already in the bind address list of the
1da177e4
LT
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
510 *
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 */
d808ad9a 513static int sctp_send_asconf_add_ip(struct sock *sk,
1da177e4
LT
514 struct sockaddr *addrs,
515 int addrcnt)
516{
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
dc022a98 524 union sctp_addr saveaddr;
1da177e4
LT
525 void *addr_buf;
526 struct sctp_af *af;
1da177e4
LT
527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!sctp_addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
0dc47877 538 __func__, sk, addrs, addrcnt);
1da177e4 539
9dbc15f0 540 list_for_each_entry(asoc, &ep->asocs, asocs) {
1da177e4
LT
541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
d808ad9a
YH
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
1da177e4
LT
554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
559cf710
VY
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
1da177e4 575 */
1da177e4
LT
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
5ae955cf 579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
1da177e4
LT
580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
dc022a98
SS
586 /* Add the new addresses to the bind address list with
587 * use_as_src set to 0.
1da177e4 588 */
dc022a98
SS
589 addr_buf = addrs;
590 for (i = 0; i < addrcnt; i++) {
591 addr = (union sctp_addr *)addr_buf;
592 af = sctp_get_af_specific(addr->v4.sin_family);
593 memcpy(&saveaddr, addr, af->sockaddr_len);
f57d96b2
VY
594 retval = sctp_add_bind_addr(bp, &saveaddr,
595 SCTP_ADDR_NEW, GFP_ATOMIC);
dc022a98
SS
596 addr_buf += af->sockaddr_len;
597 }
8a07eb0a
MH
598 if (asoc->src_out_of_asoc_ok) {
599 struct sctp_transport *trans;
600
601 list_for_each_entry(trans,
602 &asoc->peer.transport_addr_list, transports) {
603 /* Clear the source and route cache */
604 dst_release(trans->dst);
605 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
606 2*asoc->pathmtu, 4380));
607 trans->ssthresh = asoc->peer.i.a_rwnd;
608 trans->rto = asoc->rto_initial;
609 trans->rtt = trans->srtt = trans->rttvar = 0;
610 sctp_transport_route(trans, NULL,
611 sctp_sk(asoc->base.sk));
612 }
613 }
614 retval = sctp_send_asconf(asoc, chunk);
1da177e4
LT
615 }
616
617out:
618 return retval;
619}
620
621/* Remove a list of addresses from bind addresses list. Do not remove the
622 * last address.
623 *
624 * Basically run through each address specified in the addrs/addrcnt
625 * array/length pair, determine if it is IPv6 or IPv4 and call
626 * sctp_del_bind() on it.
627 *
628 * If any of them fails, then the operation will be reversed and the
629 * ones that were removed will be added back.
630 *
631 * At least one address has to be left; if only one address is
632 * available, the operation will return -EBUSY.
633 *
634 * Only sctp_setsockopt_bindx() is supposed to call this function.
635 */
04675210 636static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
1da177e4
LT
637{
638 struct sctp_sock *sp = sctp_sk(sk);
639 struct sctp_endpoint *ep = sp->ep;
640 int cnt;
641 struct sctp_bind_addr *bp = &ep->base.bind_addr;
642 int retval = 0;
1da177e4 643 void *addr_buf;
c9a08505 644 union sctp_addr *sa_addr;
1da177e4
LT
645 struct sctp_af *af;
646
647 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
648 sk, addrs, addrcnt);
649
650 addr_buf = addrs;
651 for (cnt = 0; cnt < addrcnt; cnt++) {
652 /* If the bind address list is empty or if there is only one
653 * bind address, there is nothing more to be removed (we need
654 * at least one address here).
655 */
656 if (list_empty(&bp->address_list) ||
657 (sctp_list_single_entry(&bp->address_list))) {
658 retval = -EBUSY;
659 goto err_bindx_rem;
660 }
661
c9a08505
AV
662 sa_addr = (union sctp_addr *)addr_buf;
663 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1da177e4
LT
664 if (!af) {
665 retval = -EINVAL;
666 goto err_bindx_rem;
667 }
0304ff8a
PG
668
669 if (!af->addr_valid(sa_addr, sp, NULL)) {
670 retval = -EADDRNOTAVAIL;
671 goto err_bindx_rem;
672 }
673
ee9cbaca
VY
674 if (sa_addr->v4.sin_port &&
675 sa_addr->v4.sin_port != htons(bp->port)) {
1da177e4
LT
676 retval = -EINVAL;
677 goto err_bindx_rem;
678 }
679
ee9cbaca
VY
680 if (!sa_addr->v4.sin_port)
681 sa_addr->v4.sin_port = htons(bp->port);
682
1da177e4
LT
683 /* FIXME - There is probably a need to check if sk->sk_saddr and
684 * sk->sk_rcv_addr are currently set to one of the addresses to
685 * be removed. This is something which needs to be looked into
686 * when we are fixing the outstanding issues with multi-homing
687 * socket routing and failover schemes. Refer to comments in
688 * sctp_do_bind(). -daisy
689 */
0ed90fb0 690 retval = sctp_del_bind_addr(bp, sa_addr);
1da177e4
LT
691
692 addr_buf += af->sockaddr_len;
693err_bindx_rem:
694 if (retval < 0) {
695 /* Failed. Add the ones that has been removed back */
696 if (cnt > 0)
697 sctp_bindx_add(sk, addrs, cnt);
698 return retval;
699 }
700 }
701
702 return retval;
703}
704
705/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
706 * the associations that are part of the endpoint indicating that a list of
707 * local addresses are removed from the endpoint.
708 *
d808ad9a 709 * If any of the addresses is already in the bind address list of the
1da177e4
LT
710 * association, we do not send the chunk for that association. But it will not
711 * affect other associations.
712 *
713 * Only sctp_setsockopt_bindx() is supposed to call this function.
714 */
715static int sctp_send_asconf_del_ip(struct sock *sk,
716 struct sockaddr *addrs,
717 int addrcnt)
718{
719 struct sctp_sock *sp;
720 struct sctp_endpoint *ep;
721 struct sctp_association *asoc;
dc022a98 722 struct sctp_transport *transport;
1da177e4
LT
723 struct sctp_bind_addr *bp;
724 struct sctp_chunk *chunk;
725 union sctp_addr *laddr;
726 void *addr_buf;
727 struct sctp_af *af;
dc022a98 728 struct sctp_sockaddr_entry *saddr;
1da177e4
LT
729 int i;
730 int retval = 0;
8a07eb0a 731 int stored = 0;
1da177e4 732
8a07eb0a 733 chunk = NULL;
1da177e4
LT
734 if (!sctp_addip_enable)
735 return retval;
736
737 sp = sctp_sk(sk);
738 ep = sp->ep;
739
740 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
0dc47877 741 __func__, sk, addrs, addrcnt);
1da177e4 742
9dbc15f0 743 list_for_each_entry(asoc, &ep->asocs, asocs) {
1da177e4
LT
744
745 if (!asoc->peer.asconf_capable)
746 continue;
747
748 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
749 continue;
750
751 if (!sctp_state(asoc, ESTABLISHED))
752 continue;
753
754 /* Check if any address in the packed array of addresses is
d808ad9a 755 * not present in the bind address list of the association.
1da177e4
LT
756 * If so, do not send the asconf chunk to its peer, but
757 * continue with other associations.
758 */
759 addr_buf = addrs;
760 for (i = 0; i < addrcnt; i++) {
761 laddr = (union sctp_addr *)addr_buf;
762 af = sctp_get_af_specific(laddr->v4.sin_family);
763 if (!af) {
764 retval = -EINVAL;
765 goto out;
766 }
767
768 if (!sctp_assoc_lookup_laddr(asoc, laddr))
769 break;
770
771 addr_buf += af->sockaddr_len;
772 }
773 if (i < addrcnt)
774 continue;
775
776 /* Find one address in the association's bind address list
777 * that is not in the packed array of addresses. This is to
778 * make sure that we do not delete all the addresses in the
779 * association.
780 */
1da177e4
LT
781 bp = &asoc->base.bind_addr;
782 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
783 addrcnt, sp);
8a07eb0a
MH
784 if ((laddr == NULL) && (addrcnt == 1)) {
785 if (asoc->asconf_addr_del_pending)
786 continue;
787 asoc->asconf_addr_del_pending =
788 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
6d65e5ee
MH
789 if (asoc->asconf_addr_del_pending == NULL) {
790 retval = -ENOMEM;
791 goto out;
792 }
8a07eb0a
MH
793 asoc->asconf_addr_del_pending->sa.sa_family =
794 addrs->sa_family;
795 asoc->asconf_addr_del_pending->v4.sin_port =
796 htons(bp->port);
797 if (addrs->sa_family == AF_INET) {
798 struct sockaddr_in *sin;
799
800 sin = (struct sockaddr_in *)addrs;
801 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
802 } else if (addrs->sa_family == AF_INET6) {
803 struct sockaddr_in6 *sin6;
804
805 sin6 = (struct sockaddr_in6 *)addrs;
806 ipv6_addr_copy(&asoc->asconf_addr_del_pending->v6.sin6_addr, &sin6->sin6_addr);
807 }
808 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
809 " at %p\n", asoc, asoc->asconf_addr_del_pending,
810 asoc->asconf_addr_del_pending);
811 asoc->src_out_of_asoc_ok = 1;
812 stored = 1;
813 goto skip_mkasconf;
814 }
1da177e4 815
559cf710
VY
816 /* We do not need RCU protection throughout this loop
817 * because this is done under a socket lock from the
818 * setsockopt call.
819 */
1da177e4
LT
820 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
821 SCTP_PARAM_DEL_IP);
822 if (!chunk) {
823 retval = -ENOMEM;
824 goto out;
825 }
826
8a07eb0a 827skip_mkasconf:
dc022a98
SS
828 /* Reset use_as_src flag for the addresses in the bind address
829 * list that are to be deleted.
830 */
dc022a98
SS
831 addr_buf = addrs;
832 for (i = 0; i < addrcnt; i++) {
833 laddr = (union sctp_addr *)addr_buf;
834 af = sctp_get_af_specific(laddr->v4.sin_family);
559cf710 835 list_for_each_entry(saddr, &bp->address_list, list) {
5f242a13 836 if (sctp_cmp_addr_exact(&saddr->a, laddr))
f57d96b2 837 saddr->state = SCTP_ADDR_DEL;
dc022a98
SS
838 }
839 addr_buf += af->sockaddr_len;
840 }
1da177e4 841
dc022a98
SS
842 /* Update the route and saddr entries for all the transports
843 * as some of the addresses in the bind address list are
844 * about to be deleted and cannot be used as source addresses.
1da177e4 845 */
9dbc15f0
RD
846 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
847 transports) {
dc022a98
SS
848 dst_release(transport->dst);
849 sctp_transport_route(transport, NULL,
850 sctp_sk(asoc->base.sk));
851 }
852
8a07eb0a
MH
853 if (stored)
854 /* We don't need to transmit ASCONF */
855 continue;
dc022a98 856 retval = sctp_send_asconf(asoc, chunk);
1da177e4
LT
857 }
858out:
859 return retval;
860}
861
9f7d653b
MH
862/* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
863int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
864{
865 struct sock *sk = sctp_opt2sk(sp);
866 union sctp_addr *addr;
867 struct sctp_af *af;
868
869 /* It is safe to write port space in caller. */
870 addr = &addrw->a;
871 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
872 af = sctp_get_af_specific(addr->sa.sa_family);
873 if (!af)
874 return -EINVAL;
875 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
876 return -EINVAL;
877
878 if (addrw->state == SCTP_ADDR_NEW)
879 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
880 else
881 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
882}
883
1da177e4
LT
884/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
885 *
886 * API 8.1
887 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
888 * int flags);
889 *
890 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
891 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
892 * or IPv6 addresses.
893 *
894 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
895 * Section 3.1.2 for this usage.
896 *
897 * addrs is a pointer to an array of one or more socket addresses. Each
898 * address is contained in its appropriate structure (i.e. struct
899 * sockaddr_in or struct sockaddr_in6) the family of the address type
23c435f7 900 * must be used to distinguish the address length (note that this
1da177e4
LT
901 * representation is termed a "packed array" of addresses). The caller
902 * specifies the number of addresses in the array with addrcnt.
903 *
904 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
905 * -1, and sets errno to the appropriate error code.
906 *
907 * For SCTP, the port given in each socket address must be the same, or
908 * sctp_bindx() will fail, setting errno to EINVAL.
909 *
910 * The flags parameter is formed from the bitwise OR of zero or more of
911 * the following currently defined flags:
912 *
913 * SCTP_BINDX_ADD_ADDR
914 *
915 * SCTP_BINDX_REM_ADDR
916 *
917 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
918 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
919 * addresses from the association. The two flags are mutually exclusive;
920 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
921 * not remove all addresses from an association; sctp_bindx() will
922 * reject such an attempt with EINVAL.
923 *
924 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
925 * additional addresses with an endpoint after calling bind(). Or use
926 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
927 * socket is associated with so that no new association accepted will be
928 * associated with those addresses. If the endpoint supports dynamic
929 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
930 * endpoint to send the appropriate message to the peer to change the
931 * peers address lists.
932 *
933 * Adding and removing addresses from a connected association is
934 * optional functionality. Implementations that do not support this
935 * functionality should return EOPNOTSUPP.
936 *
937 * Basically do nothing but copying the addresses from user to kernel
938 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
3f7a87d2
FF
939 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
940 * from userspace.
1da177e4
LT
941 *
942 * We don't use copy_from_user() for optimization: we first do the
943 * sanity checks (buffer size -fast- and access check-healthy
944 * pointer); if all of those succeed, then we can alloc the memory
945 * (expensive operation) needed to copy the data to kernel. Then we do
946 * the copying without checking the user space area
947 * (__copy_from_user()).
948 *
949 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
950 * it.
951 *
952 * sk The sk of the socket
953 * addrs The pointer to the addresses in user land
954 * addrssize Size of the addrs buffer
955 * op Operation to perform (add or remove, see the flags of
956 * sctp_bindx)
957 *
958 * Returns 0 if ok, <0 errno code on error.
959 */
960SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
961 struct sockaddr __user *addrs,
962 int addrs_size, int op)
963{
964 struct sockaddr *kaddrs;
965 int err;
966 int addrcnt = 0;
967 int walk_size = 0;
968 struct sockaddr *sa_addr;
969 void *addr_buf;
970 struct sctp_af *af;
971
972 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
973 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
974
975 if (unlikely(addrs_size <= 0))
976 return -EINVAL;
977
978 /* Check the user passed a healthy pointer. */
979 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
980 return -EFAULT;
981
982 /* Alloc space for the address array in kernel memory. */
8b3a7005 983 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1da177e4
LT
984 if (unlikely(!kaddrs))
985 return -ENOMEM;
986
987 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
988 kfree(kaddrs);
989 return -EFAULT;
990 }
991
d808ad9a 992 /* Walk through the addrs buffer and count the number of addresses. */
1da177e4
LT
993 addr_buf = kaddrs;
994 while (walk_size < addrs_size) {
d7e0d19a
DR
995 if (walk_size + sizeof(sa_family_t) > addrs_size) {
996 kfree(kaddrs);
997 return -EINVAL;
998 }
999
1da177e4
LT
1000 sa_addr = (struct sockaddr *)addr_buf;
1001 af = sctp_get_af_specific(sa_addr->sa_family);
1002
1003 /* If the address family is not supported or if this address
1004 * causes the address buffer to overflow return EINVAL.
d808ad9a 1005 */
1da177e4
LT
1006 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1007 kfree(kaddrs);
1008 return -EINVAL;
1009 }
1010 addrcnt++;
1011 addr_buf += af->sockaddr_len;
1012 walk_size += af->sockaddr_len;
1013 }
1014
1015 /* Do the work. */
1016 switch (op) {
1017 case SCTP_BINDX_ADD_ADDR:
1018 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1019 if (err)
1020 goto out;
1021 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1022 break;
1023
1024 case SCTP_BINDX_REM_ADDR:
1025 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1026 if (err)
1027 goto out;
1028 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1029 break;
1030
1031 default:
1032 err = -EINVAL;
1033 break;
3ff50b79 1034 }
1da177e4
LT
1035
1036out:
1037 kfree(kaddrs);
1038
1039 return err;
1040}
1041
3f7a87d2
FF
1042/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1043 *
1044 * Common routine for handling connect() and sctp_connectx().
1045 * Connect will come in with just a single address.
1046 */
1047static int __sctp_connect(struct sock* sk,
1048 struct sockaddr *kaddrs,
88a0a948
VY
1049 int addrs_size,
1050 sctp_assoc_t *assoc_id)
3f7a87d2
FF
1051{
1052 struct sctp_sock *sp;
1053 struct sctp_endpoint *ep;
1054 struct sctp_association *asoc = NULL;
1055 struct sctp_association *asoc2;
1056 struct sctp_transport *transport;
1057 union sctp_addr to;
1058 struct sctp_af *af;
1059 sctp_scope_t scope;
1060 long timeo;
1061 int err = 0;
1062 int addrcnt = 0;
1063 int walk_size = 0;
e4d1feab 1064 union sctp_addr *sa_addr = NULL;
3f7a87d2 1065 void *addr_buf;
16d00fb7 1066 unsigned short port;
f50f95ca 1067 unsigned int f_flags = 0;
3f7a87d2
FF
1068
1069 sp = sctp_sk(sk);
1070 ep = sp->ep;
1071
1072 /* connect() cannot be done on a socket that is already in ESTABLISHED
1073 * state - UDP-style peeled off socket or a TCP-style socket that
1074 * is already connected.
1075 * It cannot be done even on a TCP-style listening socket.
1076 */
1077 if (sctp_sstate(sk, ESTABLISHED) ||
1078 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1079 err = -EISCONN;
1080 goto out_free;
1081 }
1082
1083 /* Walk through the addrs buffer and count the number of addresses. */
1084 addr_buf = kaddrs;
1085 while (walk_size < addrs_size) {
d7e0d19a
DR
1086 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1087 err = -EINVAL;
1088 goto out_free;
1089 }
1090
4bdf4b5f
AV
1091 sa_addr = (union sctp_addr *)addr_buf;
1092 af = sctp_get_af_specific(sa_addr->sa.sa_family);
3f7a87d2
FF
1093
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1096 */
1097 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 err = -EINVAL;
1099 goto out_free;
1100 }
1101
d7e0d19a
DR
1102 port = ntohs(sa_addr->v4.sin_port);
1103
e4d1feab
VY
1104 /* Save current address so we can work with it */
1105 memcpy(&to, sa_addr, af->sockaddr_len);
1106
1107 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
3f7a87d2
FF
1108 if (err)
1109 goto out_free;
1110
16d00fb7
VY
1111 /* Make sure the destination port is correctly set
1112 * in all addresses.
1113 */
1114 if (asoc && asoc->peer.port && asoc->peer.port != port)
1115 goto out_free;
1116
3f7a87d2
FF
1117
1118 /* Check if there already is a matching association on the
1119 * endpoint (other than the one created here).
1120 */
e4d1feab 1121 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
3f7a87d2
FF
1122 if (asoc2 && asoc2 != asoc) {
1123 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1124 err = -EISCONN;
1125 else
1126 err = -EALREADY;
1127 goto out_free;
1128 }
1129
1130 /* If we could not find a matching association on the endpoint,
1131 * make sure that there is no peeled-off association matching
1132 * the peer address even on another socket.
1133 */
e4d1feab 1134 if (sctp_endpoint_is_peeled_off(ep, &to)) {
3f7a87d2
FF
1135 err = -EADDRNOTAVAIL;
1136 goto out_free;
1137 }
1138
1139 if (!asoc) {
1140 /* If a bind() or sctp_bindx() is not called prior to
1141 * an sctp_connectx() call, the system picks an
1142 * ephemeral port and will choose an address set
1143 * equivalent to binding with a wildcard address.
1144 */
1145 if (!ep->base.bind_addr.port) {
1146 if (sctp_autobind(sk)) {
1147 err = -EAGAIN;
1148 goto out_free;
1149 }
64a0c1c8
ISJ
1150 } else {
1151 /*
d808ad9a
YH
1152 * If an unprivileged user inherits a 1-many
1153 * style socket with open associations on a
1154 * privileged port, it MAY be permitted to
1155 * accept new associations, but it SHOULD NOT
64a0c1c8
ISJ
1156 * be permitted to open new associations.
1157 */
1158 if (ep->base.bind_addr.port < PROT_SOCK &&
1159 !capable(CAP_NET_BIND_SERVICE)) {
1160 err = -EACCES;
1161 goto out_free;
1162 }
3f7a87d2
FF
1163 }
1164
e4d1feab 1165 scope = sctp_scope(&to);
3f7a87d2
FF
1166 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1167 if (!asoc) {
1168 err = -ENOMEM;
1169 goto out_free;
1170 }
409b95af
VY
1171
1172 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1173 GFP_KERNEL);
1174 if (err < 0) {
1175 goto out_free;
1176 }
1177
3f7a87d2
FF
1178 }
1179
1180 /* Prime the peer's transport structures. */
e4d1feab 1181 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
3f7a87d2
FF
1182 SCTP_UNKNOWN);
1183 if (!transport) {
1184 err = -ENOMEM;
1185 goto out_free;
1186 }
1187
1188 addrcnt++;
1189 addr_buf += af->sockaddr_len;
1190 walk_size += af->sockaddr_len;
1191 }
1192
c6ba68a2
VY
1193 /* In case the user of sctp_connectx() wants an association
1194 * id back, assign one now.
1195 */
1196 if (assoc_id) {
1197 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1198 if (err < 0)
1199 goto out_free;
1200 }
1201
3f7a87d2
FF
1202 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1203 if (err < 0) {
1204 goto out_free;
1205 }
1206
1207 /* Initialize sk's dport and daddr for getpeername() */
c720c7e8 1208 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
e4d1feab
VY
1209 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1210 af->to_sk_daddr(sa_addr, sk);
8de8c873 1211 sk->sk_err = 0;
3f7a87d2 1212
f50f95ca
VY
1213 /* in-kernel sockets don't generally have a file allocated to them
1214 * if all they do is call sock_create_kern().
1215 */
1216 if (sk->sk_socket->file)
1217 f_flags = sk->sk_socket->file->f_flags;
1218
1219 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1220
3f7a87d2 1221 err = sctp_wait_for_connect(asoc, &timeo);
c6ba68a2 1222 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
88a0a948 1223 *assoc_id = asoc->assoc_id;
3f7a87d2
FF
1224
1225 /* Don't free association on exit. */
1226 asoc = NULL;
1227
1228out_free:
1229
1230 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
d808ad9a
YH
1231 " kaddrs: %p err: %d\n",
1232 asoc, kaddrs, err);
3f7a87d2
FF
1233 if (asoc)
1234 sctp_association_free(asoc);
1235 return err;
1236}
1237
1238/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1239 *
1240 * API 8.9
88a0a948
VY
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
3f7a87d2
FF
1243 *
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1247 *
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1250 *
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1257 *
88a0a948
VY
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
3f7a87d2
FF
1262 *
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1265 *
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
25985edc 1270 * the association is implementation dependent. This function only
3f7a87d2
FF
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1273 *
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1279 *
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1283 *
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1290 *
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1292 * it.
1293 *
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1297 *
88a0a948 1298 * Returns >=0 if ok, <0 errno code on error.
3f7a87d2 1299 */
88a0a948 1300SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
3f7a87d2 1301 struct sockaddr __user *addrs,
88a0a948
VY
1302 int addrs_size,
1303 sctp_assoc_t *assoc_id)
3f7a87d2
FF
1304{
1305 int err = 0;
1306 struct sockaddr *kaddrs;
1307
1308 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
0dc47877 1309 __func__, sk, addrs, addrs_size);
3f7a87d2
FF
1310
1311 if (unlikely(addrs_size <= 0))
1312 return -EINVAL;
1313
1314 /* Check the user passed a healthy pointer. */
1315 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1316 return -EFAULT;
1317
1318 /* Alloc space for the address array in kernel memory. */
8b3a7005 1319 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
3f7a87d2
FF
1320 if (unlikely(!kaddrs))
1321 return -ENOMEM;
1322
1323 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1324 err = -EFAULT;
1325 } else {
88a0a948 1326 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
3f7a87d2
FF
1327 }
1328
1329 kfree(kaddrs);
88a0a948 1330
3f7a87d2
FF
1331 return err;
1332}
1333
88a0a948
VY
1334/*
1335 * This is an older interface. It's kept for backward compatibility
1336 * to the option that doesn't provide association id.
1337 */
1338SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1339 struct sockaddr __user *addrs,
1340 int addrs_size)
1341{
1342 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1343}
1344
1345/*
1346 * New interface for the API. The since the API is done with a socket
1347 * option, to make it simple we feed back the association id is as a return
1348 * indication to the call. Error is always negative and association id is
1349 * always positive.
1350 */
1351SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1352 struct sockaddr __user *addrs,
1353 int addrs_size)
1354{
1355 sctp_assoc_t assoc_id = 0;
1356 int err = 0;
1357
1358 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1359
1360 if (err)
1361 return err;
1362 else
1363 return assoc_id;
1364}
1365
c6ba68a2 1366/*
f9c67811
VY
1367 * New (hopefully final) interface for the API.
1368 * We use the sctp_getaddrs_old structure so that use-space library
1369 * can avoid any unnecessary allocations. The only defferent part
1370 * is that we store the actual length of the address buffer into the
1371 * addrs_num structure member. That way we can re-use the existing
1372 * code.
c6ba68a2
VY
1373 */
1374SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1375 char __user *optval,
1376 int __user *optlen)
1377{
f9c67811 1378 struct sctp_getaddrs_old param;
c6ba68a2
VY
1379 sctp_assoc_t assoc_id = 0;
1380 int err = 0;
1381
f9c67811 1382 if (len < sizeof(param))
c6ba68a2
VY
1383 return -EINVAL;
1384
f9c67811
VY
1385 if (copy_from_user(&param, optval, sizeof(param)))
1386 return -EFAULT;
1387
c6ba68a2 1388 err = __sctp_setsockopt_connectx(sk,
f9c67811
VY
1389 (struct sockaddr __user *)param.addrs,
1390 param.addr_num, &assoc_id);
c6ba68a2
VY
1391
1392 if (err == 0 || err == -EINPROGRESS) {
1393 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1394 return -EFAULT;
1395 if (put_user(sizeof(assoc_id), optlen))
1396 return -EFAULT;
1397 }
1398
1399 return err;
1400}
1401
1da177e4
LT
1402/* API 3.1.4 close() - UDP Style Syntax
1403 * Applications use close() to perform graceful shutdown (as described in
1404 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1405 * by a UDP-style socket.
1406 *
1407 * The syntax is
1408 *
1409 * ret = close(int sd);
1410 *
1411 * sd - the socket descriptor of the associations to be closed.
1412 *
1413 * To gracefully shutdown a specific association represented by the
1414 * UDP-style socket, an application should use the sendmsg() call,
1415 * passing no user data, but including the appropriate flag in the
1416 * ancillary data (see Section xxxx).
1417 *
1418 * If sd in the close() call is a branched-off socket representing only
1419 * one association, the shutdown is performed on that association only.
1420 *
1421 * 4.1.6 close() - TCP Style Syntax
1422 *
1423 * Applications use close() to gracefully close down an association.
1424 *
1425 * The syntax is:
1426 *
1427 * int close(int sd);
1428 *
1429 * sd - the socket descriptor of the association to be closed.
1430 *
1431 * After an application calls close() on a socket descriptor, no further
1432 * socket operations will succeed on that descriptor.
1433 *
1434 * API 7.1.4 SO_LINGER
1435 *
1436 * An application using the TCP-style socket can use this option to
1437 * perform the SCTP ABORT primitive. The linger option structure is:
1438 *
1439 * struct linger {
1440 * int l_onoff; // option on/off
1441 * int l_linger; // linger time
1442 * };
1443 *
1444 * To enable the option, set l_onoff to 1. If the l_linger value is set
1445 * to 0, calling close() is the same as the ABORT primitive. If the
1446 * value is set to a negative value, the setsockopt() call will return
1447 * an error. If the value is set to a positive value linger_time, the
1448 * close() can be blocked for at most linger_time ms. If the graceful
1449 * shutdown phase does not finish during this period, close() will
1450 * return but the graceful shutdown phase continues in the system.
1451 */
1452SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1453{
1454 struct sctp_endpoint *ep;
1455 struct sctp_association *asoc;
1456 struct list_head *pos, *temp;
1457
1458 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1459
1460 sctp_lock_sock(sk);
1461 sk->sk_shutdown = SHUTDOWN_MASK;
bec9640b 1462 sk->sk_state = SCTP_SS_CLOSING;
1da177e4
LT
1463
1464 ep = sctp_sk(sk)->ep;
1465
61c9fed4 1466 /* Walk all associations on an endpoint. */
1da177e4
LT
1467 list_for_each_safe(pos, temp, &ep->asocs) {
1468 asoc = list_entry(pos, struct sctp_association, asocs);
1469
1470 if (sctp_style(sk, TCP)) {
1471 /* A closed association can still be in the list if
1472 * it belongs to a TCP-style listening socket that is
1473 * not yet accepted. If so, free it. If not, send an
1474 * ABORT or SHUTDOWN based on the linger options.
1475 */
1476 if (sctp_state(asoc, CLOSED)) {
1477 sctp_unhash_established(asoc);
1478 sctp_association_free(asoc);
b89498a1
VY
1479 continue;
1480 }
1481 }
1da177e4 1482
b9ac8672
SS
1483 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1484 struct sctp_chunk *chunk;
1485
1486 chunk = sctp_make_abort_user(asoc, NULL, 0);
1487 if (chunk)
1488 sctp_primitive_ABORT(asoc, chunk);
1489 } else
1da177e4
LT
1490 sctp_primitive_SHUTDOWN(asoc, NULL);
1491 }
1492
1493 /* Clean up any skbs sitting on the receive queue. */
1494 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1496
1497 /* On a TCP-style socket, block for at most linger_time if set. */
1498 if (sctp_style(sk, TCP) && timeout)
1499 sctp_wait_for_close(sk, timeout);
1500
1501 /* This will run the backlog queue. */
1502 sctp_release_sock(sk);
1503
1504 /* Supposedly, no process has access to the socket, but
1505 * the net layers still may.
1506 */
1507 sctp_local_bh_disable();
1508 sctp_bh_lock_sock(sk);
1509
1510 /* Hold the sock, since sk_common_release() will put sock_put()
1511 * and we have just a little more cleanup.
1512 */
1513 sock_hold(sk);
1514 sk_common_release(sk);
1515
1516 sctp_bh_unlock_sock(sk);
1517 sctp_local_bh_enable();
1518
1519 sock_put(sk);
1520
1521 SCTP_DBG_OBJCNT_DEC(sock);
1522}
1523
1524/* Handle EPIPE error. */
1525static int sctp_error(struct sock *sk, int flags, int err)
1526{
1527 if (err == -EPIPE)
1528 err = sock_error(sk) ? : -EPIPE;
1529 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1530 send_sig(SIGPIPE, current, 0);
1531 return err;
1532}
1533
1534/* API 3.1.3 sendmsg() - UDP Style Syntax
1535 *
1536 * An application uses sendmsg() and recvmsg() calls to transmit data to
1537 * and receive data from its peer.
1538 *
1539 * ssize_t sendmsg(int socket, const struct msghdr *message,
1540 * int flags);
1541 *
1542 * socket - the socket descriptor of the endpoint.
1543 * message - pointer to the msghdr structure which contains a single
1544 * user message and possibly some ancillary data.
1545 *
1546 * See Section 5 for complete description of the data
1547 * structures.
1548 *
1549 * flags - flags sent or received with the user message, see Section
1550 * 5 for complete description of the flags.
1551 *
1552 * Note: This function could use a rewrite especially when explicit
1553 * connect support comes in.
1554 */
1555/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1556
1557SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1558
1559SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1560 struct msghdr *msg, size_t msg_len)
1561{
1562 struct sctp_sock *sp;
1563 struct sctp_endpoint *ep;
1564 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1565 struct sctp_transport *transport, *chunk_tp;
1566 struct sctp_chunk *chunk;
dce116ae 1567 union sctp_addr to;
1da177e4 1568 struct sockaddr *msg_name = NULL;
517aa0bc 1569 struct sctp_sndrcvinfo default_sinfo;
1da177e4
LT
1570 struct sctp_sndrcvinfo *sinfo;
1571 struct sctp_initmsg *sinit;
1572 sctp_assoc_t associd = 0;
1573 sctp_cmsgs_t cmsgs = { NULL };
1574 int err;
1575 sctp_scope_t scope;
1576 long timeo;
1577 __u16 sinfo_flags = 0;
1578 struct sctp_datamsg *datamsg;
1da177e4
LT
1579 int msg_flags = msg->msg_flags;
1580
1581 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1582 sk, msg, msg_len);
1583
1584 err = 0;
1585 sp = sctp_sk(sk);
1586 ep = sp->ep;
1587
3f7a87d2 1588 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1da177e4
LT
1589
1590 /* We cannot send a message over a TCP-style listening socket. */
1591 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1592 err = -EPIPE;
1593 goto out_nounlock;
1594 }
1595
1596 /* Parse out the SCTP CMSGs. */
1597 err = sctp_msghdr_parse(msg, &cmsgs);
1598
1599 if (err) {
1600 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1601 goto out_nounlock;
1602 }
1603
1604 /* Fetch the destination address for this packet. This
1605 * address only selects the association--it is not necessarily
1606 * the address we will send to.
1607 * For a peeled-off socket, msg_name is ignored.
1608 */
1609 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1610 int msg_namelen = msg->msg_namelen;
1611
1612 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1613 msg_namelen);
1614 if (err)
1615 return err;
1616
1617 if (msg_namelen > sizeof(to))
1618 msg_namelen = sizeof(to);
1619 memcpy(&to, msg->msg_name, msg_namelen);
1da177e4
LT
1620 msg_name = msg->msg_name;
1621 }
1622
1623 sinfo = cmsgs.info;
1624 sinit = cmsgs.init;
1625
1626 /* Did the user specify SNDRCVINFO? */
1627 if (sinfo) {
1628 sinfo_flags = sinfo->sinfo_flags;
1629 associd = sinfo->sinfo_assoc_id;
1630 }
1631
1632 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1633 msg_len, sinfo_flags);
1634
eaa5c54d
ISJ
1635 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1636 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1da177e4
LT
1637 err = -EINVAL;
1638 goto out_nounlock;
1639 }
1640
eaa5c54d
ISJ
1641 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1642 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1643 * If SCTP_ABORT is set, the message length could be non zero with
1da177e4 1644 * the msg_iov set to the user abort reason.
d808ad9a 1645 */
eaa5c54d
ISJ
1646 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1647 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1da177e4
LT
1648 err = -EINVAL;
1649 goto out_nounlock;
1650 }
1651
eaa5c54d 1652 /* If SCTP_ADDR_OVER is set, there must be an address
1da177e4
LT
1653 * specified in msg_name.
1654 */
eaa5c54d 1655 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1da177e4
LT
1656 err = -EINVAL;
1657 goto out_nounlock;
1658 }
1659
1660 transport = NULL;
1661
1662 SCTP_DEBUG_PRINTK("About to look up association.\n");
1663
1664 sctp_lock_sock(sk);
1665
1666 /* If a msg_name has been specified, assume this is to be used. */
1667 if (msg_name) {
1668 /* Look for a matching association on the endpoint. */
dce116ae 1669 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1da177e4
LT
1670 if (!asoc) {
1671 /* If we could not find a matching association on the
1672 * endpoint, make sure that it is not a TCP-style
1673 * socket that already has an association or there is
1674 * no peeled-off association on another socket.
1675 */
1676 if ((sctp_style(sk, TCP) &&
1677 sctp_sstate(sk, ESTABLISHED)) ||
dce116ae 1678 sctp_endpoint_is_peeled_off(ep, &to)) {
1da177e4
LT
1679 err = -EADDRNOTAVAIL;
1680 goto out_unlock;
1681 }
1682 }
1683 } else {
1684 asoc = sctp_id2assoc(sk, associd);
1685 if (!asoc) {
1686 err = -EPIPE;
1687 goto out_unlock;
1688 }
1689 }
1690
1691 if (asoc) {
1692 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1693
1694 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1695 * socket that has an association in CLOSED state. This can
1696 * happen when an accepted socket has an association that is
1697 * already CLOSED.
1698 */
1699 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1700 err = -EPIPE;
1701 goto out_unlock;
1702 }
1703
eaa5c54d 1704 if (sinfo_flags & SCTP_EOF) {
1da177e4
LT
1705 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1706 asoc);
1707 sctp_primitive_SHUTDOWN(asoc, NULL);
1708 err = 0;
1709 goto out_unlock;
1710 }
eaa5c54d 1711 if (sinfo_flags & SCTP_ABORT) {
c164a9ba
SS
1712
1713 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1714 if (!chunk) {
1715 err = -ENOMEM;
1716 goto out_unlock;
1717 }
1718
1da177e4 1719 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
c164a9ba 1720 sctp_primitive_ABORT(asoc, chunk);
1da177e4
LT
1721 err = 0;
1722 goto out_unlock;
1723 }
1724 }
1725
1726 /* Do we need to create the association? */
1727 if (!asoc) {
1728 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1729
eaa5c54d 1730 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1da177e4
LT
1731 err = -EINVAL;
1732 goto out_unlock;
1733 }
1734
1735 /* Check for invalid stream against the stream counts,
1736 * either the default or the user specified stream counts.
1737 */
1738 if (sinfo) {
1739 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1740 /* Check against the defaults. */
1741 if (sinfo->sinfo_stream >=
1742 sp->initmsg.sinit_num_ostreams) {
1743 err = -EINVAL;
1744 goto out_unlock;
1745 }
1746 } else {
1747 /* Check against the requested. */
1748 if (sinfo->sinfo_stream >=
1749 sinit->sinit_num_ostreams) {
1750 err = -EINVAL;
1751 goto out_unlock;
1752 }
1753 }
1754 }
1755
1756 /*
1757 * API 3.1.2 bind() - UDP Style Syntax
1758 * If a bind() or sctp_bindx() is not called prior to a
1759 * sendmsg() call that initiates a new association, the
1760 * system picks an ephemeral port and will choose an address
1761 * set equivalent to binding with a wildcard address.
1762 */
1763 if (!ep->base.bind_addr.port) {
1764 if (sctp_autobind(sk)) {
1765 err = -EAGAIN;
1766 goto out_unlock;
1767 }
64a0c1c8
ISJ
1768 } else {
1769 /*
1770 * If an unprivileged user inherits a one-to-many
1771 * style socket with open associations on a privileged
1772 * port, it MAY be permitted to accept new associations,
1773 * but it SHOULD NOT be permitted to open new
1774 * associations.
1775 */
1776 if (ep->base.bind_addr.port < PROT_SOCK &&
1777 !capable(CAP_NET_BIND_SERVICE)) {
1778 err = -EACCES;
1779 goto out_unlock;
1780 }
1da177e4
LT
1781 }
1782
1783 scope = sctp_scope(&to);
1784 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1785 if (!new_asoc) {
1786 err = -ENOMEM;
1787 goto out_unlock;
1788 }
1789 asoc = new_asoc;
409b95af
VY
1790 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1791 if (err < 0) {
1792 err = -ENOMEM;
1793 goto out_free;
1794 }
1da177e4
LT
1795
1796 /* If the SCTP_INIT ancillary data is specified, set all
1797 * the association init values accordingly.
1798 */
1799 if (sinit) {
1800 if (sinit->sinit_num_ostreams) {
1801 asoc->c.sinit_num_ostreams =
1802 sinit->sinit_num_ostreams;
1803 }
1804 if (sinit->sinit_max_instreams) {
1805 asoc->c.sinit_max_instreams =
1806 sinit->sinit_max_instreams;
1807 }
1808 if (sinit->sinit_max_attempts) {
1809 asoc->max_init_attempts
1810 = sinit->sinit_max_attempts;
1811 }
1812 if (sinit->sinit_max_init_timeo) {
d808ad9a 1813 asoc->max_init_timeo =
1da177e4
LT
1814 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1815 }
1816 }
1817
1818 /* Prime the peer's transport structures. */
dce116ae 1819 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1da177e4
LT
1820 if (!transport) {
1821 err = -ENOMEM;
1822 goto out_free;
1823 }
1da177e4
LT
1824 }
1825
1826 /* ASSERT: we have a valid association at this point. */
1827 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1828
1829 if (!sinfo) {
1830 /* If the user didn't specify SNDRCVINFO, make up one with
1831 * some defaults.
1832 */
517aa0bc 1833 memset(&default_sinfo, 0, sizeof(default_sinfo));
1da177e4
LT
1834 default_sinfo.sinfo_stream = asoc->default_stream;
1835 default_sinfo.sinfo_flags = asoc->default_flags;
1836 default_sinfo.sinfo_ppid = asoc->default_ppid;
1837 default_sinfo.sinfo_context = asoc->default_context;
1838 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1839 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1840 sinfo = &default_sinfo;
1841 }
1842
1843 /* API 7.1.7, the sndbuf size per association bounds the
1844 * maximum size of data that can be sent in a single send call.
1845 */
1846 if (msg_len > sk->sk_sndbuf) {
1847 err = -EMSGSIZE;
1848 goto out_free;
1849 }
1850
8a479491
VY
1851 if (asoc->pmtu_pending)
1852 sctp_assoc_pending_pmtu(asoc);
1853
1da177e4
LT
1854 /* If fragmentation is disabled and the message length exceeds the
1855 * association fragmentation point, return EMSGSIZE. The I-D
1856 * does not specify what this error is, but this looks like
1857 * a great fit.
1858 */
1859 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1860 err = -EMSGSIZE;
1861 goto out_free;
1862 }
1863
afd7614c
JP
1864 /* Check for invalid stream. */
1865 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1866 err = -EINVAL;
1867 goto out_free;
1da177e4
LT
1868 }
1869
1870 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1871 if (!sctp_wspace(asoc)) {
1872 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1873 if (err)
1874 goto out_free;
1875 }
1876
1877 /* If an address is passed with the sendto/sendmsg call, it is used
1878 * to override the primary destination address in the TCP model, or
eaa5c54d 1879 * when SCTP_ADDR_OVER flag is set in the UDP model.
1da177e4
LT
1880 */
1881 if ((sctp_style(sk, TCP) && msg_name) ||
eaa5c54d 1882 (sinfo_flags & SCTP_ADDR_OVER)) {
dce116ae 1883 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1da177e4
LT
1884 if (!chunk_tp) {
1885 err = -EINVAL;
1886 goto out_free;
1887 }
1888 } else
1889 chunk_tp = NULL;
1890
1891 /* Auto-connect, if we aren't connected already. */
1892 if (sctp_state(asoc, CLOSED)) {
1893 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1894 if (err < 0)
1895 goto out_free;
1896 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1897 }
1898
1899 /* Break the message into multiple chunks of maximum size. */
1900 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1901 if (!datamsg) {
1902 err = -ENOMEM;
1903 goto out_free;
1904 }
1905
1906 /* Now send the (possibly) fragmented message. */
9dbc15f0 1907 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
80445cfb 1908 sctp_chunk_hold(chunk);
1da177e4
LT
1909
1910 /* Do accounting for the write space. */
1911 sctp_set_owner_w(chunk);
1912
1913 chunk->transport = chunk_tp;
1da177e4
LT
1914 }
1915
9c5c62be
VY
1916 /* Send it to the lower layers. Note: all chunks
1917 * must either fail or succeed. The lower layer
1918 * works that way today. Keep it that way or this
1919 * breaks.
1920 */
1921 err = sctp_primitive_SEND(asoc, datamsg);
1922 /* Did the lower layer accept the chunk? */
1923 if (err)
1924 sctp_datamsg_free(datamsg);
1925 else
1926 sctp_datamsg_put(datamsg);
1927
1928 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1929
1da177e4
LT
1930 if (err)
1931 goto out_free;
1932 else
1933 err = msg_len;
1934
1935 /* If we are already past ASSOCIATE, the lower
1936 * layers are responsible for association cleanup.
1937 */
1938 goto out_unlock;
1939
1940out_free:
1941 if (new_asoc)
1942 sctp_association_free(asoc);
1943out_unlock:
1944 sctp_release_sock(sk);
1945
1946out_nounlock:
1947 return sctp_error(sk, msg_flags, err);
1948
1949#if 0
1950do_sock_err:
1951 if (msg_len)
1952 err = msg_len;
1953 else
1954 err = sock_error(sk);
1955 goto out;
1956
1957do_interrupted:
1958 if (msg_len)
1959 err = msg_len;
1960 goto out;
1961#endif /* 0 */
1962}
1963
1964/* This is an extended version of skb_pull() that removes the data from the
1965 * start of a skb even when data is spread across the list of skb's in the
1966 * frag_list. len specifies the total amount of data that needs to be removed.
1967 * when 'len' bytes could be removed from the skb, it returns 0.
1968 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1969 * could not be removed.
1970 */
1971static int sctp_skb_pull(struct sk_buff *skb, int len)
1972{
1973 struct sk_buff *list;
1974 int skb_len = skb_headlen(skb);
1975 int rlen;
1976
1977 if (len <= skb_len) {
1978 __skb_pull(skb, len);
1979 return 0;
1980 }
1981 len -= skb_len;
1982 __skb_pull(skb, skb_len);
1983
1b003be3 1984 skb_walk_frags(skb, list) {
1da177e4
LT
1985 rlen = sctp_skb_pull(list, len);
1986 skb->len -= (len-rlen);
1987 skb->data_len -= (len-rlen);
1988
1989 if (!rlen)
1990 return 0;
1991
1992 len = rlen;
1993 }
1994
1995 return len;
1996}
1997
1998/* API 3.1.3 recvmsg() - UDP Style Syntax
1999 *
2000 * ssize_t recvmsg(int socket, struct msghdr *message,
2001 * int flags);
2002 *
2003 * socket - the socket descriptor of the endpoint.
2004 * message - pointer to the msghdr structure which contains a single
2005 * user message and possibly some ancillary data.
2006 *
2007 * See Section 5 for complete description of the data
2008 * structures.
2009 *
2010 * flags - flags sent or received with the user message, see Section
2011 * 5 for complete description of the flags.
2012 */
2013static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2014
2015SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2016 struct msghdr *msg, size_t len, int noblock,
2017 int flags, int *addr_len)
2018{
2019 struct sctp_ulpevent *event = NULL;
2020 struct sctp_sock *sp = sctp_sk(sk);
2021 struct sk_buff *skb;
2022 int copied;
2023 int err = 0;
2024 int skb_len;
2025
2026 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2027 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2028 "len", len, "knoblauch", noblock,
2029 "flags", flags, "addr_len", addr_len);
2030
2031 sctp_lock_sock(sk);
2032
2033 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2034 err = -ENOTCONN;
2035 goto out;
2036 }
2037
2038 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2039 if (!skb)
2040 goto out;
2041
2042 /* Get the total length of the skb including any skb's in the
2043 * frag_list.
2044 */
2045 skb_len = skb->len;
2046
2047 copied = skb_len;
2048 if (copied > len)
2049 copied = len;
2050
2051 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2052
2053 event = sctp_skb2event(skb);
2054
2055 if (err)
2056 goto out_free;
2057
3b885787 2058 sock_recv_ts_and_drops(msg, sk, skb);
1da177e4
LT
2059 if (sctp_ulpevent_is_notification(event)) {
2060 msg->msg_flags |= MSG_NOTIFICATION;
2061 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2062 } else {
2063 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2064 }
2065
2066 /* Check if we allow SCTP_SNDRCVINFO. */
2067 if (sp->subscribe.sctp_data_io_event)
2068 sctp_ulpevent_read_sndrcvinfo(event, msg);
2069#if 0
2070 /* FIXME: we should be calling IP/IPv6 layers. */
2071 if (sk->sk_protinfo.af_inet.cmsg_flags)
2072 ip_cmsg_recv(msg, skb);
2073#endif
2074
2075 err = copied;
2076
2077 /* If skb's length exceeds the user's buffer, update the skb and
2078 * push it back to the receive_queue so that the next call to
2079 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2080 */
2081 if (skb_len > copied) {
2082 msg->msg_flags &= ~MSG_EOR;
2083 if (flags & MSG_PEEK)
2084 goto out_free;
2085 sctp_skb_pull(skb, copied);
2086 skb_queue_head(&sk->sk_receive_queue, skb);
2087
2088 /* When only partial message is copied to the user, increase
2089 * rwnd by that amount. If all the data in the skb is read,
2090 * rwnd is updated when the event is freed.
2091 */
0eca8fee
VY
2092 if (!sctp_ulpevent_is_notification(event))
2093 sctp_assoc_rwnd_increase(event->asoc, copied);
1da177e4
LT
2094 goto out;
2095 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2096 (event->msg_flags & MSG_EOR))
2097 msg->msg_flags |= MSG_EOR;
2098 else
2099 msg->msg_flags &= ~MSG_EOR;
2100
2101out_free:
2102 if (flags & MSG_PEEK) {
2103 /* Release the skb reference acquired after peeking the skb in
2104 * sctp_skb_recv_datagram().
2105 */
2106 kfree_skb(skb);
2107 } else {
2108 /* Free the event which includes releasing the reference to
2109 * the owner of the skb, freeing the skb and updating the
2110 * rwnd.
2111 */
2112 sctp_ulpevent_free(event);
2113 }
2114out:
2115 sctp_release_sock(sk);
2116 return err;
2117}
2118
2119/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2120 *
2121 * This option is a on/off flag. If enabled no SCTP message
2122 * fragmentation will be performed. Instead if a message being sent
2123 * exceeds the current PMTU size, the message will NOT be sent and
2124 * instead a error will be indicated to the user.
2125 */
2126static int sctp_setsockopt_disable_fragments(struct sock *sk,
b7058842
DM
2127 char __user *optval,
2128 unsigned int optlen)
1da177e4
LT
2129{
2130 int val;
2131
2132 if (optlen < sizeof(int))
2133 return -EINVAL;
2134
2135 if (get_user(val, (int __user *)optval))
2136 return -EFAULT;
2137
2138 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2139
2140 return 0;
2141}
2142
2143static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
b7058842 2144 unsigned int optlen)
1da177e4 2145{
7e8616d8 2146 if (optlen > sizeof(struct sctp_event_subscribe))
1da177e4
LT
2147 return -EINVAL;
2148 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2149 return -EFAULT;
2150 return 0;
2151}
2152
2153/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2154 *
2155 * This socket option is applicable to the UDP-style socket only. When
2156 * set it will cause associations that are idle for more than the
2157 * specified number of seconds to automatically close. An association
2158 * being idle is defined an association that has NOT sent or received
2159 * user data. The special value of '0' indicates that no automatic
2160 * close of any associations should be performed. The option expects an
2161 * integer defining the number of seconds of idle time before an
2162 * association is closed.
2163 */
2164static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
b7058842 2165 unsigned int optlen)
1da177e4
LT
2166{
2167 struct sctp_sock *sp = sctp_sk(sk);
2168
2169 /* Applicable to UDP-style socket only */
2170 if (sctp_style(sk, TCP))
2171 return -EOPNOTSUPP;
2172 if (optlen != sizeof(int))
2173 return -EINVAL;
2174 if (copy_from_user(&sp->autoclose, optval, optlen))
2175 return -EFAULT;
f6778aab 2176 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
8ffd3208 2177 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
1da177e4 2178
1da177e4
LT
2179 return 0;
2180}
2181
2182/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2183 *
2184 * Applications can enable or disable heartbeats for any peer address of
2185 * an association, modify an address's heartbeat interval, force a
2186 * heartbeat to be sent immediately, and adjust the address's maximum
2187 * number of retransmissions sent before an address is considered
2188 * unreachable. The following structure is used to access and modify an
2189 * address's parameters:
2190 *
2191 * struct sctp_paddrparams {
52ccb8e9
FF
2192 * sctp_assoc_t spp_assoc_id;
2193 * struct sockaddr_storage spp_address;
2194 * uint32_t spp_hbinterval;
2195 * uint16_t spp_pathmaxrxt;
2196 * uint32_t spp_pathmtu;
2197 * uint32_t spp_sackdelay;
2198 * uint32_t spp_flags;
2199 * };
2200 *
2201 * spp_assoc_id - (one-to-many style socket) This is filled in the
2202 * application, and identifies the association for
2203 * this query.
1da177e4
LT
2204 * spp_address - This specifies which address is of interest.
2205 * spp_hbinterval - This contains the value of the heartbeat interval,
52ccb8e9
FF
2206 * in milliseconds. If a value of zero
2207 * is present in this field then no changes are to
2208 * be made to this parameter.
1da177e4
LT
2209 * spp_pathmaxrxt - This contains the maximum number of
2210 * retransmissions before this address shall be
52ccb8e9
FF
2211 * considered unreachable. If a value of zero
2212 * is present in this field then no changes are to
2213 * be made to this parameter.
2214 * spp_pathmtu - When Path MTU discovery is disabled the value
2215 * specified here will be the "fixed" path mtu.
2216 * Note that if the spp_address field is empty
2217 * then all associations on this address will
2218 * have this fixed path mtu set upon them.
2219 *
2220 * spp_sackdelay - When delayed sack is enabled, this value specifies
2221 * the number of milliseconds that sacks will be delayed
2222 * for. This value will apply to all addresses of an
2223 * association if the spp_address field is empty. Note
2224 * also, that if delayed sack is enabled and this
2225 * value is set to 0, no change is made to the last
2226 * recorded delayed sack timer value.
2227 *
2228 * spp_flags - These flags are used to control various features
2229 * on an association. The flag field may contain
2230 * zero or more of the following options.
2231 *
2232 * SPP_HB_ENABLE - Enable heartbeats on the
2233 * specified address. Note that if the address
2234 * field is empty all addresses for the association
2235 * have heartbeats enabled upon them.
2236 *
2237 * SPP_HB_DISABLE - Disable heartbeats on the
2238 * speicifed address. Note that if the address
2239 * field is empty all addresses for the association
2240 * will have their heartbeats disabled. Note also
2241 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2242 * mutually exclusive, only one of these two should
2243 * be specified. Enabling both fields will have
2244 * undetermined results.
2245 *
2246 * SPP_HB_DEMAND - Request a user initiated heartbeat
2247 * to be made immediately.
2248 *
bdf3092a
VY
2249 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2250 * heartbeat delayis to be set to the value of 0
2251 * milliseconds.
2252 *
52ccb8e9
FF
2253 * SPP_PMTUD_ENABLE - This field will enable PMTU
2254 * discovery upon the specified address. Note that
2255 * if the address feild is empty then all addresses
2256 * on the association are effected.
2257 *
2258 * SPP_PMTUD_DISABLE - This field will disable PMTU
2259 * discovery upon the specified address. Note that
2260 * if the address feild is empty then all addresses
2261 * on the association are effected. Not also that
2262 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2263 * exclusive. Enabling both will have undetermined
2264 * results.
2265 *
2266 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2267 * on delayed sack. The time specified in spp_sackdelay
2268 * is used to specify the sack delay for this address. Note
2269 * that if spp_address is empty then all addresses will
2270 * enable delayed sack and take on the sack delay
2271 * value specified in spp_sackdelay.
2272 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2273 * off delayed sack. If the spp_address field is blank then
2274 * delayed sack is disabled for the entire association. Note
2275 * also that this field is mutually exclusive to
2276 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2277 * results.
1da177e4 2278 */
16164366
AB
2279static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2280 struct sctp_transport *trans,
2281 struct sctp_association *asoc,
2282 struct sctp_sock *sp,
2283 int hb_change,
2284 int pmtud_change,
2285 int sackdelay_change)
52ccb8e9
FF
2286{
2287 int error;
2288
2289 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2290 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2291 if (error)
2292 return error;
2293 }
2294
bdf3092a
VY
2295 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2296 * this field is ignored. Note also that a value of zero indicates
2297 * the current setting should be left unchanged.
2298 */
2299 if (params->spp_flags & SPP_HB_ENABLE) {
2300
2301 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2302 * set. This lets us use 0 value when this flag
2303 * is set.
2304 */
2305 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2306 params->spp_hbinterval = 0;
2307
2308 if (params->spp_hbinterval ||
2309 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2310 if (trans) {
2311 trans->hbinterval =
2312 msecs_to_jiffies(params->spp_hbinterval);
2313 } else if (asoc) {
2314 asoc->hbinterval =
2315 msecs_to_jiffies(params->spp_hbinterval);
2316 } else {
2317 sp->hbinterval = params->spp_hbinterval;
2318 }
52ccb8e9
FF
2319 }
2320 }
2321
2322 if (hb_change) {
2323 if (trans) {
2324 trans->param_flags =
2325 (trans->param_flags & ~SPP_HB) | hb_change;
2326 } else if (asoc) {
2327 asoc->param_flags =
2328 (asoc->param_flags & ~SPP_HB) | hb_change;
2329 } else {
2330 sp->param_flags =
2331 (sp->param_flags & ~SPP_HB) | hb_change;
2332 }
2333 }
2334
bdf3092a
VY
2335 /* When Path MTU discovery is disabled the value specified here will
2336 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2337 * include the flag SPP_PMTUD_DISABLE for this field to have any
2338 * effect).
2339 */
2340 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
52ccb8e9
FF
2341 if (trans) {
2342 trans->pathmtu = params->spp_pathmtu;
2343 sctp_assoc_sync_pmtu(asoc);
2344 } else if (asoc) {
2345 asoc->pathmtu = params->spp_pathmtu;
f68b2e05 2346 sctp_frag_point(asoc, params->spp_pathmtu);
52ccb8e9
FF
2347 } else {
2348 sp->pathmtu = params->spp_pathmtu;
2349 }
2350 }
2351
2352 if (pmtud_change) {
2353 if (trans) {
2354 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2355 (params->spp_flags & SPP_PMTUD_ENABLE);
2356 trans->param_flags =
2357 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2358 if (update) {
9914ae3c 2359 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
52ccb8e9
FF
2360 sctp_assoc_sync_pmtu(asoc);
2361 }
2362 } else if (asoc) {
2363 asoc->param_flags =
2364 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2365 } else {
2366 sp->param_flags =
2367 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2368 }
2369 }
2370
bdf3092a
VY
2371 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2372 * value of this field is ignored. Note also that a value of zero
2373 * indicates the current setting should be left unchanged.
2374 */
2375 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
52ccb8e9
FF
2376 if (trans) {
2377 trans->sackdelay =
2378 msecs_to_jiffies(params->spp_sackdelay);
2379 } else if (asoc) {
2380 asoc->sackdelay =
2381 msecs_to_jiffies(params->spp_sackdelay);
2382 } else {
2383 sp->sackdelay = params->spp_sackdelay;
2384 }
2385 }
2386
2387 if (sackdelay_change) {
2388 if (trans) {
2389 trans->param_flags =
2390 (trans->param_flags & ~SPP_SACKDELAY) |
2391 sackdelay_change;
2392 } else if (asoc) {
2393 asoc->param_flags =
2394 (asoc->param_flags & ~SPP_SACKDELAY) |
2395 sackdelay_change;
2396 } else {
2397 sp->param_flags =
2398 (sp->param_flags & ~SPP_SACKDELAY) |
2399 sackdelay_change;
2400 }
2401 }
2402
37051f73
APO
2403 /* Note that a value of zero indicates the current setting should be
2404 left unchanged.
bdf3092a 2405 */
37051f73 2406 if (params->spp_pathmaxrxt) {
52ccb8e9
FF
2407 if (trans) {
2408 trans->pathmaxrxt = params->spp_pathmaxrxt;
2409 } else if (asoc) {
2410 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2411 } else {
2412 sp->pathmaxrxt = params->spp_pathmaxrxt;
2413 }
2414 }
2415
2416 return 0;
2417}
2418
1da177e4 2419static int sctp_setsockopt_peer_addr_params(struct sock *sk,
b7058842
DM
2420 char __user *optval,
2421 unsigned int optlen)
1da177e4 2422{
52ccb8e9
FF
2423 struct sctp_paddrparams params;
2424 struct sctp_transport *trans = NULL;
2425 struct sctp_association *asoc = NULL;
2426 struct sctp_sock *sp = sctp_sk(sk);
1da177e4 2427 int error;
52ccb8e9 2428 int hb_change, pmtud_change, sackdelay_change;
1da177e4
LT
2429
2430 if (optlen != sizeof(struct sctp_paddrparams))
52ccb8e9
FF
2431 return - EINVAL;
2432
1da177e4
LT
2433 if (copy_from_user(&params, optval, optlen))
2434 return -EFAULT;
2435
52ccb8e9
FF
2436 /* Validate flags and value parameters. */
2437 hb_change = params.spp_flags & SPP_HB;
2438 pmtud_change = params.spp_flags & SPP_PMTUD;
2439 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2440
2441 if (hb_change == SPP_HB ||
2442 pmtud_change == SPP_PMTUD ||
2443 sackdelay_change == SPP_SACKDELAY ||
2444 params.spp_sackdelay > 500 ||
f64f9e71
JP
2445 (params.spp_pathmtu &&
2446 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
52ccb8e9 2447 return -EINVAL;
1da177e4 2448
52ccb8e9
FF
2449 /* If an address other than INADDR_ANY is specified, and
2450 * no transport is found, then the request is invalid.
2451 */
52cae8f0 2452 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
52ccb8e9
FF
2453 trans = sctp_addr_id2transport(sk, &params.spp_address,
2454 params.spp_assoc_id);
2455 if (!trans)
1da177e4 2456 return -EINVAL;
1da177e4
LT
2457 }
2458
52ccb8e9
FF
2459 /* Get association, if assoc_id != 0 and the socket is a one
2460 * to many style socket, and an association was not found, then
2461 * the id was invalid.
2462 */
2463 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2464 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
1da177e4
LT
2465 return -EINVAL;
2466
52ccb8e9
FF
2467 /* Heartbeat demand can only be sent on a transport or
2468 * association, but not a socket.
1da177e4 2469 */
52ccb8e9
FF
2470 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2471 return -EINVAL;
2472
2473 /* Process parameters. */
2474 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2475 hb_change, pmtud_change,
2476 sackdelay_change);
1da177e4 2477
52ccb8e9
FF
2478 if (error)
2479 return error;
2480
2481 /* If changes are for association, also apply parameters to each
2482 * transport.
1da177e4 2483 */
52ccb8e9 2484 if (!trans && asoc) {
9dbc15f0
RD
2485 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2486 transports) {
52ccb8e9
FF
2487 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2488 hb_change, pmtud_change,
2489 sackdelay_change);
2490 }
2491 }
1da177e4
LT
2492
2493 return 0;
2494}
2495
d364d927
WY
2496/*
2497 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2498 *
2499 * This option will effect the way delayed acks are performed. This
2500 * option allows you to get or set the delayed ack time, in
2501 * milliseconds. It also allows changing the delayed ack frequency.
2502 * Changing the frequency to 1 disables the delayed sack algorithm. If
2503 * the assoc_id is 0, then this sets or gets the endpoints default
2504 * values. If the assoc_id field is non-zero, then the set or get
2505 * effects the specified association for the one to many model (the
2506 * assoc_id field is ignored by the one to one model). Note that if
2507 * sack_delay or sack_freq are 0 when setting this option, then the
2508 * current values will remain unchanged.
2509 *
2510 * struct sctp_sack_info {
2511 * sctp_assoc_t sack_assoc_id;
2512 * uint32_t sack_delay;
2513 * uint32_t sack_freq;
2514 * };
2515 *
2516 * sack_assoc_id - This parameter, indicates which association the user
2517 * is performing an action upon. Note that if this field's value is
2518 * zero then the endpoints default value is changed (effecting future
2519 * associations only).
2520 *
2521 * sack_delay - This parameter contains the number of milliseconds that
2522 * the user is requesting the delayed ACK timer be set to. Note that
2523 * this value is defined in the standard to be between 200 and 500
2524 * milliseconds.
2525 *
2526 * sack_freq - This parameter contains the number of packets that must
2527 * be received before a sack is sent without waiting for the delay
2528 * timer to expire. The default value for this is 2, setting this
2529 * value to 1 will disable the delayed sack algorithm.
7708610b
FF
2530 */
2531
d364d927 2532static int sctp_setsockopt_delayed_ack(struct sock *sk,
b7058842 2533 char __user *optval, unsigned int optlen)
7708610b 2534{
d364d927 2535 struct sctp_sack_info params;
7708610b
FF
2536 struct sctp_transport *trans = NULL;
2537 struct sctp_association *asoc = NULL;
2538 struct sctp_sock *sp = sctp_sk(sk);
2539
d364d927
WY
2540 if (optlen == sizeof(struct sctp_sack_info)) {
2541 if (copy_from_user(&params, optval, optlen))
2542 return -EFAULT;
7708610b 2543
d364d927
WY
2544 if (params.sack_delay == 0 && params.sack_freq == 0)
2545 return 0;
2546 } else if (optlen == sizeof(struct sctp_assoc_value)) {
145ce502
JP
2547 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2548 pr_warn("Use struct sctp_sack_info instead\n");
d364d927
WY
2549 if (copy_from_user(&params, optval, optlen))
2550 return -EFAULT;
2551
2552 if (params.sack_delay == 0)
2553 params.sack_freq = 1;
2554 else
2555 params.sack_freq = 0;
2556 } else
2557 return - EINVAL;
7708610b
FF
2558
2559 /* Validate value parameter. */
d364d927 2560 if (params.sack_delay > 500)
7708610b
FF
2561 return -EINVAL;
2562
d364d927 2563 /* Get association, if sack_assoc_id != 0 and the socket is a one
7708610b
FF
2564 * to many style socket, and an association was not found, then
2565 * the id was invalid.
d808ad9a 2566 */
d364d927
WY
2567 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2568 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
7708610b
FF
2569 return -EINVAL;
2570
d364d927 2571 if (params.sack_delay) {
7708610b
FF
2572 if (asoc) {
2573 asoc->sackdelay =
d364d927 2574 msecs_to_jiffies(params.sack_delay);
d808ad9a 2575 asoc->param_flags =
7708610b
FF
2576 (asoc->param_flags & ~SPP_SACKDELAY) |
2577 SPP_SACKDELAY_ENABLE;
2578 } else {
d364d927 2579 sp->sackdelay = params.sack_delay;
d808ad9a 2580 sp->param_flags =
7708610b
FF
2581 (sp->param_flags & ~SPP_SACKDELAY) |
2582 SPP_SACKDELAY_ENABLE;
2583 }
d364d927
WY
2584 }
2585
2586 if (params.sack_freq == 1) {
7708610b 2587 if (asoc) {
d808ad9a 2588 asoc->param_flags =
7708610b
FF
2589 (asoc->param_flags & ~SPP_SACKDELAY) |
2590 SPP_SACKDELAY_DISABLE;
2591 } else {
d808ad9a 2592 sp->param_flags =
7708610b
FF
2593 (sp->param_flags & ~SPP_SACKDELAY) |
2594 SPP_SACKDELAY_DISABLE;
2595 }
d364d927
WY
2596 } else if (params.sack_freq > 1) {
2597 if (asoc) {
2598 asoc->sackfreq = params.sack_freq;
2599 asoc->param_flags =
2600 (asoc->param_flags & ~SPP_SACKDELAY) |
2601 SPP_SACKDELAY_ENABLE;
2602 } else {
2603 sp->sackfreq = params.sack_freq;
2604 sp->param_flags =
2605 (sp->param_flags & ~SPP_SACKDELAY) |
2606 SPP_SACKDELAY_ENABLE;
2607 }
7708610b
FF
2608 }
2609
2610 /* If change is for association, also apply to each transport. */
2611 if (asoc) {
9dbc15f0
RD
2612 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2613 transports) {
d364d927 2614 if (params.sack_delay) {
7708610b 2615 trans->sackdelay =
d364d927 2616 msecs_to_jiffies(params.sack_delay);
d808ad9a 2617 trans->param_flags =
7708610b
FF
2618 (trans->param_flags & ~SPP_SACKDELAY) |
2619 SPP_SACKDELAY_ENABLE;
d364d927 2620 }
7bfe8bdb 2621 if (params.sack_freq == 1) {
d808ad9a 2622 trans->param_flags =
7708610b
FF
2623 (trans->param_flags & ~SPP_SACKDELAY) |
2624 SPP_SACKDELAY_DISABLE;
d364d927
WY
2625 } else if (params.sack_freq > 1) {
2626 trans->sackfreq = params.sack_freq;
2627 trans->param_flags =
2628 (trans->param_flags & ~SPP_SACKDELAY) |
2629 SPP_SACKDELAY_ENABLE;
7708610b
FF
2630 }
2631 }
2632 }
d808ad9a 2633
7708610b
FF
2634 return 0;
2635}
2636
1da177e4
LT
2637/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2638 *
2639 * Applications can specify protocol parameters for the default association
2640 * initialization. The option name argument to setsockopt() and getsockopt()
2641 * is SCTP_INITMSG.
2642 *
2643 * Setting initialization parameters is effective only on an unconnected
2644 * socket (for UDP-style sockets only future associations are effected
2645 * by the change). With TCP-style sockets, this option is inherited by
2646 * sockets derived from a listener socket.
2647 */
b7058842 2648static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
2649{
2650 struct sctp_initmsg sinit;
2651 struct sctp_sock *sp = sctp_sk(sk);
2652
2653 if (optlen != sizeof(struct sctp_initmsg))
2654 return -EINVAL;
2655 if (copy_from_user(&sinit, optval, optlen))
2656 return -EFAULT;
2657
2658 if (sinit.sinit_num_ostreams)
d808ad9a 2659 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
1da177e4 2660 if (sinit.sinit_max_instreams)
d808ad9a 2661 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
1da177e4 2662 if (sinit.sinit_max_attempts)
d808ad9a 2663 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
1da177e4 2664 if (sinit.sinit_max_init_timeo)
d808ad9a 2665 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
1da177e4
LT
2666
2667 return 0;
2668}
2669
2670/*
2671 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2672 *
2673 * Applications that wish to use the sendto() system call may wish to
2674 * specify a default set of parameters that would normally be supplied
2675 * through the inclusion of ancillary data. This socket option allows
2676 * such an application to set the default sctp_sndrcvinfo structure.
2677 * The application that wishes to use this socket option simply passes
2678 * in to this call the sctp_sndrcvinfo structure defined in Section
2679 * 5.2.2) The input parameters accepted by this call include
2680 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2681 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2682 * to this call if the caller is using the UDP model.
2683 */
2684static int sctp_setsockopt_default_send_param(struct sock *sk,
b7058842
DM
2685 char __user *optval,
2686 unsigned int optlen)
1da177e4
LT
2687{
2688 struct sctp_sndrcvinfo info;
2689 struct sctp_association *asoc;
2690 struct sctp_sock *sp = sctp_sk(sk);
2691
2692 if (optlen != sizeof(struct sctp_sndrcvinfo))
2693 return -EINVAL;
2694 if (copy_from_user(&info, optval, optlen))
2695 return -EFAULT;
2696
2697 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2698 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2699 return -EINVAL;
2700
2701 if (asoc) {
2702 asoc->default_stream = info.sinfo_stream;
2703 asoc->default_flags = info.sinfo_flags;
2704 asoc->default_ppid = info.sinfo_ppid;
2705 asoc->default_context = info.sinfo_context;
2706 asoc->default_timetolive = info.sinfo_timetolive;
2707 } else {
2708 sp->default_stream = info.sinfo_stream;
2709 sp->default_flags = info.sinfo_flags;
2710 sp->default_ppid = info.sinfo_ppid;
2711 sp->default_context = info.sinfo_context;
2712 sp->default_timetolive = info.sinfo_timetolive;
2713 }
2714
2715 return 0;
2716}
2717
2718/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2719 *
2720 * Requests that the local SCTP stack use the enclosed peer address as
2721 * the association primary. The enclosed address must be one of the
2722 * association peer's addresses.
2723 */
2724static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
b7058842 2725 unsigned int optlen)
1da177e4
LT
2726{
2727 struct sctp_prim prim;
2728 struct sctp_transport *trans;
2729
2730 if (optlen != sizeof(struct sctp_prim))
2731 return -EINVAL;
2732
2733 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2734 return -EFAULT;
2735
2736 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2737 if (!trans)
2738 return -EINVAL;
2739
2740 sctp_assoc_set_primary(trans->asoc, trans);
2741
2742 return 0;
2743}
2744
2745/*
2746 * 7.1.5 SCTP_NODELAY
2747 *
2748 * Turn on/off any Nagle-like algorithm. This means that packets are
2749 * generally sent as soon as possible and no unnecessary delays are
2750 * introduced, at the cost of more packets in the network. Expects an
2751 * integer boolean flag.
2752 */
2753static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
b7058842 2754 unsigned int optlen)
1da177e4
LT
2755{
2756 int val;
2757
2758 if (optlen < sizeof(int))
2759 return -EINVAL;
2760 if (get_user(val, (int __user *)optval))
2761 return -EFAULT;
2762
2763 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2764 return 0;
2765}
2766
2767/*
2768 *
2769 * 7.1.1 SCTP_RTOINFO
2770 *
2771 * The protocol parameters used to initialize and bound retransmission
2772 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2773 * and modify these parameters.
2774 * All parameters are time values, in milliseconds. A value of 0, when
2775 * modifying the parameters, indicates that the current value should not
2776 * be changed.
2777 *
2778 */
b7058842
DM
2779static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2780{
1da177e4
LT
2781 struct sctp_rtoinfo rtoinfo;
2782 struct sctp_association *asoc;
2783
2784 if (optlen != sizeof (struct sctp_rtoinfo))
2785 return -EINVAL;
2786
2787 if (copy_from_user(&rtoinfo, optval, optlen))
2788 return -EFAULT;
2789
2790 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2791
2792 /* Set the values to the specific association */
2793 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2794 return -EINVAL;
2795
2796 if (asoc) {
2797 if (rtoinfo.srto_initial != 0)
d808ad9a 2798 asoc->rto_initial =
1da177e4
LT
2799 msecs_to_jiffies(rtoinfo.srto_initial);
2800 if (rtoinfo.srto_max != 0)
2801 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2802 if (rtoinfo.srto_min != 0)
2803 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2804 } else {
2805 /* If there is no association or the association-id = 0
2806 * set the values to the endpoint.
2807 */
2808 struct sctp_sock *sp = sctp_sk(sk);
2809
2810 if (rtoinfo.srto_initial != 0)
2811 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2812 if (rtoinfo.srto_max != 0)
2813 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2814 if (rtoinfo.srto_min != 0)
2815 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2816 }
2817
2818 return 0;
2819}
2820
2821/*
2822 *
2823 * 7.1.2 SCTP_ASSOCINFO
2824 *
59c51591 2825 * This option is used to tune the maximum retransmission attempts
1da177e4
LT
2826 * of the association.
2827 * Returns an error if the new association retransmission value is
2828 * greater than the sum of the retransmission value of the peer.
2829 * See [SCTP] for more information.
2830 *
2831 */
b7058842 2832static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
2833{
2834
2835 struct sctp_assocparams assocparams;
2836 struct sctp_association *asoc;
2837
2838 if (optlen != sizeof(struct sctp_assocparams))
2839 return -EINVAL;
2840 if (copy_from_user(&assocparams, optval, optlen))
2841 return -EFAULT;
2842
2843 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2844
2845 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2846 return -EINVAL;
2847
2848 /* Set the values to the specific association */
2849 if (asoc) {
402d68c4
VY
2850 if (assocparams.sasoc_asocmaxrxt != 0) {
2851 __u32 path_sum = 0;
2852 int paths = 0;
402d68c4
VY
2853 struct sctp_transport *peer_addr;
2854
9dbc15f0
RD
2855 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2856 transports) {
402d68c4
VY
2857 path_sum += peer_addr->pathmaxrxt;
2858 paths++;
2859 }
2860
025dfdaf 2861 /* Only validate asocmaxrxt if we have more than
402d68c4
VY
2862 * one path/transport. We do this because path
2863 * retransmissions are only counted when we have more
2864 * then one path.
2865 */
2866 if (paths > 1 &&
2867 assocparams.sasoc_asocmaxrxt > path_sum)
2868 return -EINVAL;
2869
1da177e4 2870 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
402d68c4
VY
2871 }
2872
1da177e4
LT
2873 if (assocparams.sasoc_cookie_life != 0) {
2874 asoc->cookie_life.tv_sec =
2875 assocparams.sasoc_cookie_life / 1000;
2876 asoc->cookie_life.tv_usec =
2877 (assocparams.sasoc_cookie_life % 1000)
2878 * 1000;
2879 }
2880 } else {
2881 /* Set the values to the endpoint */
2882 struct sctp_sock *sp = sctp_sk(sk);
2883
2884 if (assocparams.sasoc_asocmaxrxt != 0)
2885 sp->assocparams.sasoc_asocmaxrxt =
2886 assocparams.sasoc_asocmaxrxt;
2887 if (assocparams.sasoc_cookie_life != 0)
2888 sp->assocparams.sasoc_cookie_life =
2889 assocparams.sasoc_cookie_life;
2890 }
2891 return 0;
2892}
2893
2894/*
2895 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2896 *
2897 * This socket option is a boolean flag which turns on or off mapped V4
2898 * addresses. If this option is turned on and the socket is type
2899 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2900 * If this option is turned off, then no mapping will be done of V4
2901 * addresses and a user will receive both PF_INET6 and PF_INET type
2902 * addresses on the socket.
2903 */
b7058842 2904static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
2905{
2906 int val;
2907 struct sctp_sock *sp = sctp_sk(sk);
2908
2909 if (optlen < sizeof(int))
2910 return -EINVAL;
2911 if (get_user(val, (int __user *)optval))
2912 return -EFAULT;
2913 if (val)
2914 sp->v4mapped = 1;
2915 else
2916 sp->v4mapped = 0;
2917
2918 return 0;
2919}
2920
2921/*
e89c2095
WY
2922 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2923 * This option will get or set the maximum size to put in any outgoing
2924 * SCTP DATA chunk. If a message is larger than this size it will be
1da177e4
LT
2925 * fragmented by SCTP into the specified size. Note that the underlying
2926 * SCTP implementation may fragment into smaller sized chunks when the
2927 * PMTU of the underlying association is smaller than the value set by
e89c2095
WY
2928 * the user. The default value for this option is '0' which indicates
2929 * the user is NOT limiting fragmentation and only the PMTU will effect
2930 * SCTP's choice of DATA chunk size. Note also that values set larger
2931 * than the maximum size of an IP datagram will effectively let SCTP
2932 * control fragmentation (i.e. the same as setting this option to 0).
2933 *
2934 * The following structure is used to access and modify this parameter:
2935 *
2936 * struct sctp_assoc_value {
2937 * sctp_assoc_t assoc_id;
2938 * uint32_t assoc_value;
2939 * };
2940 *
2941 * assoc_id: This parameter is ignored for one-to-one style sockets.
2942 * For one-to-many style sockets this parameter indicates which
2943 * association the user is performing an action upon. Note that if
2944 * this field's value is zero then the endpoints default value is
2945 * changed (effecting future associations only).
2946 * assoc_value: This parameter specifies the maximum size in bytes.
1da177e4 2947 */
b7058842 2948static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4 2949{
e89c2095 2950 struct sctp_assoc_value params;
1da177e4 2951 struct sctp_association *asoc;
1da177e4
LT
2952 struct sctp_sock *sp = sctp_sk(sk);
2953 int val;
2954
e89c2095 2955 if (optlen == sizeof(int)) {
145ce502
JP
2956 pr_warn("Use of int in maxseg socket option deprecated\n");
2957 pr_warn("Use struct sctp_assoc_value instead\n");
e89c2095
WY
2958 if (copy_from_user(&val, optval, optlen))
2959 return -EFAULT;
2960 params.assoc_id = 0;
2961 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2962 if (copy_from_user(&params, optval, optlen))
2963 return -EFAULT;
2964 val = params.assoc_value;
2965 } else
1da177e4 2966 return -EINVAL;
e89c2095 2967
96a33998 2968 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
1da177e4 2969 return -EINVAL;
1da177e4 2970
e89c2095
WY
2971 asoc = sctp_id2assoc(sk, params.assoc_id);
2972 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2973 return -EINVAL;
2974
2975 if (asoc) {
2976 if (val == 0) {
2977 val = asoc->pathmtu;
2978 val -= sp->pf->af->net_header_len;
2979 val -= sizeof(struct sctphdr) +
2980 sizeof(struct sctp_data_chunk);
2981 }
f68b2e05
VY
2982 asoc->user_frag = val;
2983 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
e89c2095
WY
2984 } else {
2985 sp->user_frag = val;
1da177e4
LT
2986 }
2987
2988 return 0;
2989}
2990
2991
2992/*
2993 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2994 *
2995 * Requests that the peer mark the enclosed address as the association
2996 * primary. The enclosed address must be one of the association's
2997 * locally bound addresses. The following structure is used to make a
2998 * set primary request:
2999 */
3000static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
b7058842 3001 unsigned int optlen)
1da177e4
LT
3002{
3003 struct sctp_sock *sp;
1da177e4
LT
3004 struct sctp_association *asoc = NULL;
3005 struct sctp_setpeerprim prim;
3006 struct sctp_chunk *chunk;
40a01039 3007 struct sctp_af *af;
1da177e4
LT
3008 int err;
3009
3010 sp = sctp_sk(sk);
1da177e4
LT
3011
3012 if (!sctp_addip_enable)
3013 return -EPERM;
3014
3015 if (optlen != sizeof(struct sctp_setpeerprim))
3016 return -EINVAL;
3017
3018 if (copy_from_user(&prim, optval, optlen))
3019 return -EFAULT;
3020
3021 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
d808ad9a 3022 if (!asoc)
1da177e4
LT
3023 return -EINVAL;
3024
3025 if (!asoc->peer.asconf_capable)
3026 return -EPERM;
3027
3028 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3029 return -EPERM;
3030
3031 if (!sctp_state(asoc, ESTABLISHED))
3032 return -ENOTCONN;
3033
40a01039
WY
3034 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3035 if (!af)
3036 return -EINVAL;
3037
3038 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3039 return -EADDRNOTAVAIL;
3040
1da177e4
LT
3041 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3042 return -EADDRNOTAVAIL;
3043
3044 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3045 chunk = sctp_make_asconf_set_prim(asoc,
3046 (union sctp_addr *)&prim.sspp_addr);
3047 if (!chunk)
3048 return -ENOMEM;
3049
3050 err = sctp_send_asconf(asoc, chunk);
3051
3052 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3053
3054 return err;
3055}
3056
0f3fffd8 3057static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
b7058842 3058 unsigned int optlen)
1da177e4 3059{
0f3fffd8 3060 struct sctp_setadaptation adaptation;
1da177e4 3061
0f3fffd8 3062 if (optlen != sizeof(struct sctp_setadaptation))
1da177e4 3063 return -EINVAL;
0f3fffd8 3064 if (copy_from_user(&adaptation, optval, optlen))
1da177e4
LT
3065 return -EFAULT;
3066
0f3fffd8 3067 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
1da177e4
LT
3068
3069 return 0;
3070}
3071
6ab792f5
ISJ
3072/*
3073 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3074 *
3075 * The context field in the sctp_sndrcvinfo structure is normally only
3076 * used when a failed message is retrieved holding the value that was
3077 * sent down on the actual send call. This option allows the setting of
3078 * a default context on an association basis that will be received on
3079 * reading messages from the peer. This is especially helpful in the
3080 * one-2-many model for an application to keep some reference to an
3081 * internal state machine that is processing messages on the
3082 * association. Note that the setting of this value only effects
3083 * received messages from the peer and does not effect the value that is
3084 * saved with outbound messages.
3085 */
3086static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
b7058842 3087 unsigned int optlen)
6ab792f5
ISJ
3088{
3089 struct sctp_assoc_value params;
3090 struct sctp_sock *sp;
3091 struct sctp_association *asoc;
3092
3093 if (optlen != sizeof(struct sctp_assoc_value))
3094 return -EINVAL;
3095 if (copy_from_user(&params, optval, optlen))
3096 return -EFAULT;
3097
3098 sp = sctp_sk(sk);
3099
3100 if (params.assoc_id != 0) {
3101 asoc = sctp_id2assoc(sk, params.assoc_id);
3102 if (!asoc)
3103 return -EINVAL;
3104 asoc->default_rcv_context = params.assoc_value;
3105 } else {
3106 sp->default_rcv_context = params.assoc_value;
3107 }
3108
3109 return 0;
3110}
3111
b6e1331f
VY
3112/*
3113 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3114 *
3115 * This options will at a minimum specify if the implementation is doing
3116 * fragmented interleave. Fragmented interleave, for a one to many
3117 * socket, is when subsequent calls to receive a message may return
3118 * parts of messages from different associations. Some implementations
3119 * may allow you to turn this value on or off. If so, when turned off,
3120 * no fragment interleave will occur (which will cause a head of line
3121 * blocking amongst multiple associations sharing the same one to many
3122 * socket). When this option is turned on, then each receive call may
3123 * come from a different association (thus the user must receive data
3124 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3125 * association each receive belongs to.
3126 *
3127 * This option takes a boolean value. A non-zero value indicates that
3128 * fragmented interleave is on. A value of zero indicates that
3129 * fragmented interleave is off.
3130 *
3131 * Note that it is important that an implementation that allows this
3132 * option to be turned on, have it off by default. Otherwise an unaware
3133 * application using the one to many model may become confused and act
3134 * incorrectly.
3135 */
3136static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3137 char __user *optval,
b7058842 3138 unsigned int optlen)
b6e1331f
VY
3139{
3140 int val;
3141
3142 if (optlen != sizeof(int))
3143 return -EINVAL;
3144 if (get_user(val, (int __user *)optval))
3145 return -EFAULT;
3146
3147 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3148
3149 return 0;
3150}
3151
d49d91d7 3152/*
8510b937 3153 * 8.1.21. Set or Get the SCTP Partial Delivery Point
d49d91d7 3154 * (SCTP_PARTIAL_DELIVERY_POINT)
8510b937 3155 *
d49d91d7
VY
3156 * This option will set or get the SCTP partial delivery point. This
3157 * point is the size of a message where the partial delivery API will be
3158 * invoked to help free up rwnd space for the peer. Setting this to a
8510b937 3159 * lower value will cause partial deliveries to happen more often. The
d49d91d7 3160 * calls argument is an integer that sets or gets the partial delivery
8510b937
WY
3161 * point. Note also that the call will fail if the user attempts to set
3162 * this value larger than the socket receive buffer size.
3163 *
3164 * Note that any single message having a length smaller than or equal to
3165 * the SCTP partial delivery point will be delivered in one single read
3166 * call as long as the user provided buffer is large enough to hold the
3167 * message.
d49d91d7
VY
3168 */
3169static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3170 char __user *optval,
b7058842 3171 unsigned int optlen)
d49d91d7
VY
3172{
3173 u32 val;
3174
3175 if (optlen != sizeof(u32))
3176 return -EINVAL;
3177 if (get_user(val, (int __user *)optval))
3178 return -EFAULT;
3179
8510b937
WY
3180 /* Note: We double the receive buffer from what the user sets
3181 * it to be, also initial rwnd is based on rcvbuf/2.
3182 */
3183 if (val > (sk->sk_rcvbuf >> 1))
3184 return -EINVAL;
3185
d49d91d7
VY
3186 sctp_sk(sk)->pd_point = val;
3187
3188 return 0; /* is this the right error code? */
3189}
3190
70331571
VY
3191/*
3192 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3193 *
3194 * This option will allow a user to change the maximum burst of packets
3195 * that can be emitted by this association. Note that the default value
3196 * is 4, and some implementations may restrict this setting so that it
3197 * can only be lowered.
3198 *
3199 * NOTE: This text doesn't seem right. Do this on a socket basis with
3200 * future associations inheriting the socket value.
3201 */
3202static int sctp_setsockopt_maxburst(struct sock *sk,
3203 char __user *optval,
b7058842 3204 unsigned int optlen)
70331571 3205{
219b99a9
NH
3206 struct sctp_assoc_value params;
3207 struct sctp_sock *sp;
3208 struct sctp_association *asoc;
70331571 3209 int val;
219b99a9 3210 int assoc_id = 0;
70331571 3211
219b99a9 3212 if (optlen == sizeof(int)) {
145ce502
JP
3213 pr_warn("Use of int in max_burst socket option deprecated\n");
3214 pr_warn("Use struct sctp_assoc_value instead\n");
219b99a9
NH
3215 if (copy_from_user(&val, optval, optlen))
3216 return -EFAULT;
3217 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3218 if (copy_from_user(&params, optval, optlen))
3219 return -EFAULT;
3220 val = params.assoc_value;
3221 assoc_id = params.assoc_id;
3222 } else
70331571
VY
3223 return -EINVAL;
3224
219b99a9
NH
3225 sp = sctp_sk(sk);
3226
3227 if (assoc_id != 0) {
3228 asoc = sctp_id2assoc(sk, assoc_id);
3229 if (!asoc)
3230 return -EINVAL;
3231 asoc->max_burst = val;
3232 } else
3233 sp->max_burst = val;
70331571
VY
3234
3235 return 0;
3236}
3237
65b07e5d
VY
3238/*
3239 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3240 *
3241 * This set option adds a chunk type that the user is requesting to be
3242 * received only in an authenticated way. Changes to the list of chunks
3243 * will only effect future associations on the socket.
3244 */
3245static int sctp_setsockopt_auth_chunk(struct sock *sk,
b7058842
DM
3246 char __user *optval,
3247 unsigned int optlen)
65b07e5d
VY
3248{
3249 struct sctp_authchunk val;
3250
5e739d17
VY
3251 if (!sctp_auth_enable)
3252 return -EACCES;
3253
65b07e5d
VY
3254 if (optlen != sizeof(struct sctp_authchunk))
3255 return -EINVAL;
3256 if (copy_from_user(&val, optval, optlen))
3257 return -EFAULT;
3258
3259 switch (val.sauth_chunk) {
3260 case SCTP_CID_INIT:
3261 case SCTP_CID_INIT_ACK:
3262 case SCTP_CID_SHUTDOWN_COMPLETE:
3263 case SCTP_CID_AUTH:
3264 return -EINVAL;
3265 }
3266
3267 /* add this chunk id to the endpoint */
3268 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3269}
3270
3271/*
3272 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3273 *
3274 * This option gets or sets the list of HMAC algorithms that the local
3275 * endpoint requires the peer to use.
3276 */
3277static int sctp_setsockopt_hmac_ident(struct sock *sk,
b7058842
DM
3278 char __user *optval,
3279 unsigned int optlen)
65b07e5d
VY
3280{
3281 struct sctp_hmacalgo *hmacs;
d9724055 3282 u32 idents;
65b07e5d
VY
3283 int err;
3284
5e739d17
VY
3285 if (!sctp_auth_enable)
3286 return -EACCES;
3287
65b07e5d
VY
3288 if (optlen < sizeof(struct sctp_hmacalgo))
3289 return -EINVAL;
3290
934253a7
SW
3291 hmacs= memdup_user(optval, optlen);
3292 if (IS_ERR(hmacs))
3293 return PTR_ERR(hmacs);
65b07e5d 3294
d9724055
VY
3295 idents = hmacs->shmac_num_idents;
3296 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3297 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
65b07e5d
VY
3298 err = -EINVAL;
3299 goto out;
3300 }
3301
3302 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3303out:
3304 kfree(hmacs);
3305 return err;
3306}
3307
3308/*
3309 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3310 *
3311 * This option will set a shared secret key which is used to build an
3312 * association shared key.
3313 */
3314static int sctp_setsockopt_auth_key(struct sock *sk,
3315 char __user *optval,
b7058842 3316 unsigned int optlen)
65b07e5d
VY
3317{
3318 struct sctp_authkey *authkey;
3319 struct sctp_association *asoc;
3320 int ret;
3321
5e739d17
VY
3322 if (!sctp_auth_enable)
3323 return -EACCES;
3324
65b07e5d
VY
3325 if (optlen <= sizeof(struct sctp_authkey))
3326 return -EINVAL;
3327
934253a7
SW
3328 authkey= memdup_user(optval, optlen);
3329 if (IS_ERR(authkey))
3330 return PTR_ERR(authkey);
65b07e5d 3331
328fc47e 3332 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
30c2235c
VY
3333 ret = -EINVAL;
3334 goto out;
3335 }
3336
65b07e5d
VY
3337 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3338 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3339 ret = -EINVAL;
3340 goto out;
3341 }
3342
3343 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3344out:
3345 kfree(authkey);
3346 return ret;
3347}
3348
3349/*
3350 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3351 *
3352 * This option will get or set the active shared key to be used to build
3353 * the association shared key.
3354 */
3355static int sctp_setsockopt_active_key(struct sock *sk,
b7058842
DM
3356 char __user *optval,
3357 unsigned int optlen)
65b07e5d
VY
3358{
3359 struct sctp_authkeyid val;
3360 struct sctp_association *asoc;
3361
5e739d17
VY
3362 if (!sctp_auth_enable)
3363 return -EACCES;
3364
65b07e5d
VY
3365 if (optlen != sizeof(struct sctp_authkeyid))
3366 return -EINVAL;
3367 if (copy_from_user(&val, optval, optlen))
3368 return -EFAULT;
3369
3370 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3371 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3372 return -EINVAL;
3373
3374 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3375 val.scact_keynumber);
3376}
3377
3378/*
3379 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3380 *
3381 * This set option will delete a shared secret key from use.
3382 */
3383static int sctp_setsockopt_del_key(struct sock *sk,
b7058842
DM
3384 char __user *optval,
3385 unsigned int optlen)
65b07e5d
VY
3386{
3387 struct sctp_authkeyid val;
3388 struct sctp_association *asoc;
3389
5e739d17
VY
3390 if (!sctp_auth_enable)
3391 return -EACCES;
3392
65b07e5d
VY
3393 if (optlen != sizeof(struct sctp_authkeyid))
3394 return -EINVAL;
3395 if (copy_from_user(&val, optval, optlen))
3396 return -EFAULT;
3397
3398 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3399 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3400 return -EINVAL;
3401
3402 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3403 val.scact_keynumber);
3404
3405}
3406
7dc04d71
MH
3407/*
3408 * 8.1.23 SCTP_AUTO_ASCONF
3409 *
3410 * This option will enable or disable the use of the automatic generation of
3411 * ASCONF chunks to add and delete addresses to an existing association. Note
3412 * that this option has two caveats namely: a) it only affects sockets that
3413 * are bound to all addresses available to the SCTP stack, and b) the system
3414 * administrator may have an overriding control that turns the ASCONF feature
3415 * off no matter what setting the socket option may have.
3416 * This option expects an integer boolean flag, where a non-zero value turns on
3417 * the option, and a zero value turns off the option.
3418 * Note. In this implementation, socket operation overrides default parameter
3419 * being set by sysctl as well as FreeBSD implementation
3420 */
3421static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3422 unsigned int optlen)
3423{
3424 int val;
3425 struct sctp_sock *sp = sctp_sk(sk);
3426
3427 if (optlen < sizeof(int))
3428 return -EINVAL;
3429 if (get_user(val, (int __user *)optval))
3430 return -EFAULT;
3431 if (!sctp_is_ep_boundall(sk) && val)
3432 return -EINVAL;
3433 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3434 return 0;
3435
3436 if (val == 0 && sp->do_auto_asconf) {
3437 list_del(&sp->auto_asconf_list);
3438 sp->do_auto_asconf = 0;
3439 } else if (val && !sp->do_auto_asconf) {
3440 list_add_tail(&sp->auto_asconf_list,
3441 &sctp_auto_asconf_splist);
3442 sp->do_auto_asconf = 1;
3443 }
3444 return 0;
3445}
3446
65b07e5d 3447
1da177e4
LT
3448/* API 6.2 setsockopt(), getsockopt()
3449 *
3450 * Applications use setsockopt() and getsockopt() to set or retrieve
3451 * socket options. Socket options are used to change the default
3452 * behavior of sockets calls. They are described in Section 7.
3453 *
3454 * The syntax is:
3455 *
3456 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3457 * int __user *optlen);
3458 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3459 * int optlen);
3460 *
3461 * sd - the socket descript.
3462 * level - set to IPPROTO_SCTP for all SCTP options.
3463 * optname - the option name.
3464 * optval - the buffer to store the value of the option.
3465 * optlen - the size of the buffer.
3466 */
3467SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
b7058842 3468 char __user *optval, unsigned int optlen)
1da177e4
LT
3469{
3470 int retval = 0;
3471
3472 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3473 sk, optname);
3474
3475 /* I can hardly begin to describe how wrong this is. This is
3476 * so broken as to be worse than useless. The API draft
3477 * REALLY is NOT helpful here... I am not convinced that the
3478 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3479 * are at all well-founded.
3480 */
3481 if (level != SOL_SCTP) {
3482 struct sctp_af *af = sctp_sk(sk)->pf->af;
3483 retval = af->setsockopt(sk, level, optname, optval, optlen);
3484 goto out_nounlock;
3485 }
3486
3487 sctp_lock_sock(sk);
3488
3489 switch (optname) {
3490 case SCTP_SOCKOPT_BINDX_ADD:
3491 /* 'optlen' is the size of the addresses buffer. */
3492 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3493 optlen, SCTP_BINDX_ADD_ADDR);
3494 break;
3495
3496 case SCTP_SOCKOPT_BINDX_REM:
3497 /* 'optlen' is the size of the addresses buffer. */
3498 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3499 optlen, SCTP_BINDX_REM_ADDR);
3500 break;
3501
88a0a948
VY
3502 case SCTP_SOCKOPT_CONNECTX_OLD:
3503 /* 'optlen' is the size of the addresses buffer. */
3504 retval = sctp_setsockopt_connectx_old(sk,
3505 (struct sockaddr __user *)optval,
3506 optlen);
3507 break;
3508
3f7a87d2
FF
3509 case SCTP_SOCKOPT_CONNECTX:
3510 /* 'optlen' is the size of the addresses buffer. */
88a0a948
VY
3511 retval = sctp_setsockopt_connectx(sk,
3512 (struct sockaddr __user *)optval,
3513 optlen);
3f7a87d2
FF
3514 break;
3515
1da177e4
LT
3516 case SCTP_DISABLE_FRAGMENTS:
3517 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3518 break;
3519
3520 case SCTP_EVENTS:
3521 retval = sctp_setsockopt_events(sk, optval, optlen);
3522 break;
3523
3524 case SCTP_AUTOCLOSE:
3525 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3526 break;
3527
3528 case SCTP_PEER_ADDR_PARAMS:
3529 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3530 break;
3531
4580ccc0 3532 case SCTP_DELAYED_SACK:
d364d927 3533 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
7708610b 3534 break;
d49d91d7
VY
3535 case SCTP_PARTIAL_DELIVERY_POINT:
3536 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3537 break;
7708610b 3538
1da177e4
LT
3539 case SCTP_INITMSG:
3540 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3541 break;
3542 case SCTP_DEFAULT_SEND_PARAM:
3543 retval = sctp_setsockopt_default_send_param(sk, optval,
3544 optlen);
3545 break;
3546 case SCTP_PRIMARY_ADDR:
3547 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3548 break;
3549 case SCTP_SET_PEER_PRIMARY_ADDR:
3550 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3551 break;
3552 case SCTP_NODELAY:
3553 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3554 break;
3555 case SCTP_RTOINFO:
3556 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3557 break;
3558 case SCTP_ASSOCINFO:
3559 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3560 break;
3561 case SCTP_I_WANT_MAPPED_V4_ADDR:
3562 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3563 break;
3564 case SCTP_MAXSEG:
3565 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3566 break;
0f3fffd8
ISJ
3567 case SCTP_ADAPTATION_LAYER:
3568 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
1da177e4 3569 break;
6ab792f5
ISJ
3570 case SCTP_CONTEXT:
3571 retval = sctp_setsockopt_context(sk, optval, optlen);
3572 break;
b6e1331f
VY
3573 case SCTP_FRAGMENT_INTERLEAVE:
3574 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3575 break;
70331571
VY
3576 case SCTP_MAX_BURST:
3577 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3578 break;
65b07e5d
VY
3579 case SCTP_AUTH_CHUNK:
3580 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3581 break;
3582 case SCTP_HMAC_IDENT:
3583 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3584 break;
3585 case SCTP_AUTH_KEY:
3586 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3587 break;
3588 case SCTP_AUTH_ACTIVE_KEY:
3589 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3590 break;
3591 case SCTP_AUTH_DELETE_KEY:
3592 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3593 break;
7dc04d71
MH
3594 case SCTP_AUTO_ASCONF:
3595 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3596 break;
1da177e4
LT
3597 default:
3598 retval = -ENOPROTOOPT;
3599 break;
3ff50b79 3600 }
1da177e4
LT
3601
3602 sctp_release_sock(sk);
3603
3604out_nounlock:
3605 return retval;
3606}
3607
3608/* API 3.1.6 connect() - UDP Style Syntax
3609 *
3610 * An application may use the connect() call in the UDP model to initiate an
3611 * association without sending data.
3612 *
3613 * The syntax is:
3614 *
3615 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3616 *
3617 * sd: the socket descriptor to have a new association added to.
3618 *
3619 * nam: the address structure (either struct sockaddr_in or struct
3620 * sockaddr_in6 defined in RFC2553 [7]).
3621 *
3622 * len: the size of the address.
3623 */
3f7a87d2 3624SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
1da177e4
LT
3625 int addr_len)
3626{
1da177e4 3627 int err = 0;
3f7a87d2 3628 struct sctp_af *af;
1da177e4
LT
3629
3630 sctp_lock_sock(sk);
3631
3f7a87d2 3632 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
0dc47877 3633 __func__, sk, addr, addr_len);
1da177e4 3634
3f7a87d2
FF
3635 /* Validate addr_len before calling common connect/connectx routine. */
3636 af = sctp_get_af_specific(addr->sa_family);
3637 if (!af || addr_len < af->sockaddr_len) {
3638 err = -EINVAL;
3639 } else {
3640 /* Pass correct addr len to common routine (so it knows there
3641 * is only one address being passed.
3642 */
88a0a948 3643 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
1da177e4
LT
3644 }
3645
1da177e4 3646 sctp_release_sock(sk);
1da177e4
LT
3647 return err;
3648}
3649
3650/* FIXME: Write comments. */
3651SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3652{
3653 return -EOPNOTSUPP; /* STUB */
3654}
3655
3656/* 4.1.4 accept() - TCP Style Syntax
3657 *
3658 * Applications use accept() call to remove an established SCTP
3659 * association from the accept queue of the endpoint. A new socket
3660 * descriptor will be returned from accept() to represent the newly
3661 * formed association.
3662 */
3663SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3664{
3665 struct sctp_sock *sp;
3666 struct sctp_endpoint *ep;
3667 struct sock *newsk = NULL;
3668 struct sctp_association *asoc;
3669 long timeo;
3670 int error = 0;
3671
3672 sctp_lock_sock(sk);
3673
3674 sp = sctp_sk(sk);
3675 ep = sp->ep;
3676
3677 if (!sctp_style(sk, TCP)) {
3678 error = -EOPNOTSUPP;
3679 goto out;
3680 }
3681
3682 if (!sctp_sstate(sk, LISTENING)) {
3683 error = -EINVAL;
3684 goto out;
3685 }
3686
8abfedd8 3687 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1da177e4
LT
3688
3689 error = sctp_wait_for_accept(sk, timeo);
3690 if (error)
3691 goto out;
3692
3693 /* We treat the list of associations on the endpoint as the accept
3694 * queue and pick the first association on the list.
3695 */
3696 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3697
3698 newsk = sp->pf->create_accept_sk(sk, asoc);
3699 if (!newsk) {
3700 error = -ENOMEM;
3701 goto out;
3702 }
3703
3704 /* Populate the fields of the newsk from the oldsk and migrate the
3705 * asoc to the newsk.
3706 */
3707 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3708
3709out:
3710 sctp_release_sock(sk);
d808ad9a 3711 *err = error;
1da177e4
LT
3712 return newsk;
3713}
3714
3715/* The SCTP ioctl handler. */
3716SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3717{
65040c33
DEFP
3718 int rc = -ENOTCONN;
3719
3720 sctp_lock_sock(sk);
3721
3722 /*
3723 * SEQPACKET-style sockets in LISTENING state are valid, for
3724 * SCTP, so only discard TCP-style sockets in LISTENING state.
3725 */
3726 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3727 goto out;
3728
3729 switch (cmd) {
3730 case SIOCINQ: {
3731 struct sk_buff *skb;
3732 unsigned int amount = 0;
3733
3734 skb = skb_peek(&sk->sk_receive_queue);
3735 if (skb != NULL) {
3736 /*
3737 * We will only return the amount of this packet since
3738 * that is all that will be read.
3739 */
3740 amount = skb->len;
3741 }
3742 rc = put_user(amount, (int __user *)arg);
65040c33 3743 break;
9a7241c2 3744 }
65040c33
DEFP
3745 default:
3746 rc = -ENOIOCTLCMD;
3747 break;
3748 }
3749out:
3750 sctp_release_sock(sk);
3751 return rc;
1da177e4
LT
3752}
3753
3754/* This is the function which gets called during socket creation to
3755 * initialized the SCTP-specific portion of the sock.
3756 * The sock structure should already be zero-filled memory.
3757 */
3758SCTP_STATIC int sctp_init_sock(struct sock *sk)
3759{
3760 struct sctp_endpoint *ep;
3761 struct sctp_sock *sp;
3762
3763 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3764
3765 sp = sctp_sk(sk);
3766
3767 /* Initialize the SCTP per socket area. */
3768 switch (sk->sk_type) {
3769 case SOCK_SEQPACKET:
3770 sp->type = SCTP_SOCKET_UDP;
3771 break;
3772 case SOCK_STREAM:
3773 sp->type = SCTP_SOCKET_TCP;
3774 break;
3775 default:
3776 return -ESOCKTNOSUPPORT;
3777 }
3778
3779 /* Initialize default send parameters. These parameters can be
3780 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3781 */
3782 sp->default_stream = 0;
3783 sp->default_ppid = 0;
3784 sp->default_flags = 0;
3785 sp->default_context = 0;
3786 sp->default_timetolive = 0;
3787
6ab792f5 3788 sp->default_rcv_context = 0;
70331571 3789 sp->max_burst = sctp_max_burst;
6ab792f5 3790
1da177e4
LT
3791 /* Initialize default setup parameters. These parameters
3792 * can be modified with the SCTP_INITMSG socket option or
3793 * overridden by the SCTP_INIT CMSG.
3794 */
3795 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3796 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3797 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3fd091e7 3798 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
1da177e4
LT
3799
3800 /* Initialize default RTO related parameters. These parameters can
3801 * be modified for with the SCTP_RTOINFO socket option.
3802 */
3fd091e7
VY
3803 sp->rtoinfo.srto_initial = sctp_rto_initial;
3804 sp->rtoinfo.srto_max = sctp_rto_max;
3805 sp->rtoinfo.srto_min = sctp_rto_min;
1da177e4
LT
3806
3807 /* Initialize default association related parameters. These parameters
3808 * can be modified with the SCTP_ASSOCINFO socket option.
3809 */
3810 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3811 sp->assocparams.sasoc_number_peer_destinations = 0;
3812 sp->assocparams.sasoc_peer_rwnd = 0;
3813 sp->assocparams.sasoc_local_rwnd = 0;
3fd091e7 3814 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
1da177e4
LT
3815
3816 /* Initialize default event subscriptions. By default, all the
d808ad9a 3817 * options are off.
1da177e4
LT
3818 */
3819 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3820
3821 /* Default Peer Address Parameters. These defaults can
3822 * be modified via SCTP_PEER_ADDR_PARAMS
3823 */
3fd091e7 3824 sp->hbinterval = sctp_hb_interval;
52ccb8e9
FF
3825 sp->pathmaxrxt = sctp_max_retrans_path;
3826 sp->pathmtu = 0; // allow default discovery
3fd091e7 3827 sp->sackdelay = sctp_sack_timeout;
7bfe8bdb 3828 sp->sackfreq = 2;
52ccb8e9 3829 sp->param_flags = SPP_HB_ENABLE |
d808ad9a
YH
3830 SPP_PMTUD_ENABLE |
3831 SPP_SACKDELAY_ENABLE;
1da177e4
LT
3832
3833 /* If enabled no SCTP message fragmentation will be performed.
3834 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3835 */
3836 sp->disable_fragments = 0;
3837
208edef6
SS
3838 /* Enable Nagle algorithm by default. */
3839 sp->nodelay = 0;
1da177e4
LT
3840
3841 /* Enable by default. */
3842 sp->v4mapped = 1;
3843
3844 /* Auto-close idle associations after the configured
3845 * number of seconds. A value of 0 disables this
3846 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3847 * for UDP-style sockets only.
3848 */
3849 sp->autoclose = 0;
3850
3851 /* User specified fragmentation limit. */
3852 sp->user_frag = 0;
3853
0f3fffd8 3854 sp->adaptation_ind = 0;
1da177e4
LT
3855
3856 sp->pf = sctp_get_pf_specific(sk->sk_family);
3857
3858 /* Control variables for partial data delivery. */
b6e1331f 3859 atomic_set(&sp->pd_mode, 0);
1da177e4 3860 skb_queue_head_init(&sp->pd_lobby);
b6e1331f 3861 sp->frag_interleave = 0;
1da177e4
LT
3862
3863 /* Create a per socket endpoint structure. Even if we
3864 * change the data structure relationships, this may still
3865 * be useful for storing pre-connect address information.
3866 */
3867 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3868 if (!ep)
3869 return -ENOMEM;
3870
3871 sp->ep = ep;
3872 sp->hmac = NULL;
3873
3874 SCTP_DBG_OBJCNT_INC(sock);
6f756a8c
DM
3875
3876 local_bh_disable();
81419d86 3877 percpu_counter_inc(&sctp_sockets_allocated);
9a57f7fa 3878 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
9f7d653b
MH
3879 if (sctp_default_auto_asconf) {
3880 list_add_tail(&sp->auto_asconf_list,
3881 &sctp_auto_asconf_splist);
3882 sp->do_auto_asconf = 1;
3883 } else
3884 sp->do_auto_asconf = 0;
6f756a8c
DM
3885 local_bh_enable();
3886
1da177e4
LT
3887 return 0;
3888}
3889
3890/* Cleanup any SCTP per socket resources. */
7d06b2e0 3891SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
1da177e4 3892{
9f7d653b 3893 struct sctp_sock *sp;
1da177e4
LT
3894
3895 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3896
3897 /* Release our hold on the endpoint. */
9f7d653b
MH
3898 sp = sctp_sk(sk);
3899 if (sp->do_auto_asconf) {
3900 sp->do_auto_asconf = 0;
3901 list_del(&sp->auto_asconf_list);
3902 }
3903 sctp_endpoint_free(sp->ep);
5bc0b3bf 3904 local_bh_disable();
81419d86 3905 percpu_counter_dec(&sctp_sockets_allocated);
9a57f7fa 3906 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
5bc0b3bf 3907 local_bh_enable();
1da177e4
LT
3908}
3909
3910/* API 4.1.7 shutdown() - TCP Style Syntax
3911 * int shutdown(int socket, int how);
3912 *
3913 * sd - the socket descriptor of the association to be closed.
3914 * how - Specifies the type of shutdown. The values are
3915 * as follows:
3916 * SHUT_RD
3917 * Disables further receive operations. No SCTP
3918 * protocol action is taken.
3919 * SHUT_WR
3920 * Disables further send operations, and initiates
3921 * the SCTP shutdown sequence.
3922 * SHUT_RDWR
3923 * Disables further send and receive operations
3924 * and initiates the SCTP shutdown sequence.
3925 */
3926SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3927{
3928 struct sctp_endpoint *ep;
3929 struct sctp_association *asoc;
3930
3931 if (!sctp_style(sk, TCP))
3932 return;
3933
3934 if (how & SEND_SHUTDOWN) {
3935 ep = sctp_sk(sk)->ep;
3936 if (!list_empty(&ep->asocs)) {
3937 asoc = list_entry(ep->asocs.next,
3938 struct sctp_association, asocs);
3939 sctp_primitive_SHUTDOWN(asoc, NULL);
3940 }
3941 }
3942}
3943
3944/* 7.2.1 Association Status (SCTP_STATUS)
3945
3946 * Applications can retrieve current status information about an
3947 * association, including association state, peer receiver window size,
3948 * number of unacked data chunks, and number of data chunks pending
3949 * receipt. This information is read-only.
3950 */
3951static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3952 char __user *optval,
3953 int __user *optlen)
3954{
3955 struct sctp_status status;
3956 struct sctp_association *asoc = NULL;
3957 struct sctp_transport *transport;
3958 sctp_assoc_t associd;
3959 int retval = 0;
3960
408f22e8 3961 if (len < sizeof(status)) {
1da177e4
LT
3962 retval = -EINVAL;
3963 goto out;
3964 }
3965
408f22e8
NH
3966 len = sizeof(status);
3967 if (copy_from_user(&status, optval, len)) {
1da177e4
LT
3968 retval = -EFAULT;
3969 goto out;
3970 }
3971
3972 associd = status.sstat_assoc_id;
3973 asoc = sctp_id2assoc(sk, associd);
3974 if (!asoc) {
3975 retval = -EINVAL;
3976 goto out;
3977 }
3978
3979 transport = asoc->peer.primary_path;
3980
3981 status.sstat_assoc_id = sctp_assoc2id(asoc);
3982 status.sstat_state = asoc->state;
3983 status.sstat_rwnd = asoc->peer.rwnd;
3984 status.sstat_unackdata = asoc->unack_data;
3985
3986 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3987 status.sstat_instrms = asoc->c.sinit_max_instreams;
3988 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3989 status.sstat_fragmentation_point = asoc->frag_point;
3990 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
8cec6b80
AV
3991 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3992 transport->af_specific->sockaddr_len);
1da177e4
LT
3993 /* Map ipv4 address into v4-mapped-on-v6 address. */
3994 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3995 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3f7a87d2 3996 status.sstat_primary.spinfo_state = transport->state;
1da177e4
LT
3997 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3998 status.sstat_primary.spinfo_srtt = transport->srtt;
3999 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
52ccb8e9 4000 status.sstat_primary.spinfo_mtu = transport->pathmtu;
1da177e4 4001
3f7a87d2
FF
4002 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4003 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4004
1da177e4
LT
4005 if (put_user(len, optlen)) {
4006 retval = -EFAULT;
4007 goto out;
4008 }
4009
4010 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4011 len, status.sstat_state, status.sstat_rwnd,
4012 status.sstat_assoc_id);
4013
4014 if (copy_to_user(optval, &status, len)) {
4015 retval = -EFAULT;
4016 goto out;
4017 }
4018
4019out:
a02cec21 4020 return retval;
1da177e4
LT
4021}
4022
4023
4024/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4025 *
4026 * Applications can retrieve information about a specific peer address
4027 * of an association, including its reachability state, congestion
4028 * window, and retransmission timer values. This information is
4029 * read-only.
4030 */
4031static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4032 char __user *optval,
4033 int __user *optlen)
4034{
4035 struct sctp_paddrinfo pinfo;
4036 struct sctp_transport *transport;
4037 int retval = 0;
4038
408f22e8 4039 if (len < sizeof(pinfo)) {
1da177e4
LT
4040 retval = -EINVAL;
4041 goto out;
4042 }
4043
408f22e8
NH
4044 len = sizeof(pinfo);
4045 if (copy_from_user(&pinfo, optval, len)) {
1da177e4
LT
4046 retval = -EFAULT;
4047 goto out;
4048 }
4049
4050 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4051 pinfo.spinfo_assoc_id);
4052 if (!transport)
4053 return -EINVAL;
4054
4055 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3f7a87d2 4056 pinfo.spinfo_state = transport->state;
1da177e4
LT
4057 pinfo.spinfo_cwnd = transport->cwnd;
4058 pinfo.spinfo_srtt = transport->srtt;
4059 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
52ccb8e9 4060 pinfo.spinfo_mtu = transport->pathmtu;
1da177e4 4061
3f7a87d2
FF
4062 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4063 pinfo.spinfo_state = SCTP_ACTIVE;
4064
1da177e4
LT
4065 if (put_user(len, optlen)) {
4066 retval = -EFAULT;
4067 goto out;
4068 }
4069
4070 if (copy_to_user(optval, &pinfo, len)) {
4071 retval = -EFAULT;
4072 goto out;
4073 }
4074
4075out:
a02cec21 4076 return retval;
1da177e4
LT
4077}
4078
4079/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4080 *
4081 * This option is a on/off flag. If enabled no SCTP message
4082 * fragmentation will be performed. Instead if a message being sent
4083 * exceeds the current PMTU size, the message will NOT be sent and
4084 * instead a error will be indicated to the user.
4085 */
4086static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4087 char __user *optval, int __user *optlen)
4088{
4089 int val;
4090
4091 if (len < sizeof(int))
4092 return -EINVAL;
4093
4094 len = sizeof(int);
4095 val = (sctp_sk(sk)->disable_fragments == 1);
4096 if (put_user(len, optlen))
4097 return -EFAULT;
4098 if (copy_to_user(optval, &val, len))
4099 return -EFAULT;
4100 return 0;
4101}
4102
4103/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4104 *
4105 * This socket option is used to specify various notifications and
4106 * ancillary data the user wishes to receive.
4107 */
4108static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4109 int __user *optlen)
4110{
408f22e8 4111 if (len < sizeof(struct sctp_event_subscribe))
1da177e4 4112 return -EINVAL;
408f22e8
NH
4113 len = sizeof(struct sctp_event_subscribe);
4114 if (put_user(len, optlen))
4115 return -EFAULT;
1da177e4
LT
4116 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4117 return -EFAULT;
4118 return 0;
4119}
4120
4121/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4122 *
4123 * This socket option is applicable to the UDP-style socket only. When
4124 * set it will cause associations that are idle for more than the
4125 * specified number of seconds to automatically close. An association
4126 * being idle is defined an association that has NOT sent or received
4127 * user data. The special value of '0' indicates that no automatic
4128 * close of any associations should be performed. The option expects an
4129 * integer defining the number of seconds of idle time before an
4130 * association is closed.
4131 */
4132static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4133{
4134 /* Applicable to UDP-style socket only */
4135 if (sctp_style(sk, TCP))
4136 return -EOPNOTSUPP;
408f22e8 4137 if (len < sizeof(int))
1da177e4 4138 return -EINVAL;
408f22e8
NH
4139 len = sizeof(int);
4140 if (put_user(len, optlen))
4141 return -EFAULT;
4142 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
1da177e4
LT
4143 return -EFAULT;
4144 return 0;
4145}
4146
4147/* Helper routine to branch off an association to a new socket. */
4148SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4149 struct socket **sockp)
4150{
4151 struct sock *sk = asoc->base.sk;
4152 struct socket *sock;
d570ee49 4153 struct sctp_af *af;
1da177e4
LT
4154 int err = 0;
4155
4156 /* An association cannot be branched off from an already peeled-off
4157 * socket, nor is this supported for tcp style sockets.
4158 */
4159 if (!sctp_style(sk, UDP))
4160 return -EINVAL;
4161
4162 /* Create a new socket. */
4163 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4164 if (err < 0)
4165 return err;
4166
914e1c8b 4167 sctp_copy_sock(sock->sk, sk, asoc);
4f444308
VY
4168
4169 /* Make peeled-off sockets more like 1-1 accepted sockets.
4170 * Set the daddr and initialize id to something more random
4171 */
d570ee49
VY
4172 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4173 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
914e1c8b
VY
4174
4175 /* Populate the fields of the newsk from the oldsk and migrate the
4176 * asoc to the newsk.
4177 */
4178 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4f444308 4179
1da177e4
LT
4180 *sockp = sock;
4181
4182 return err;
4183}
4184
4185static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4186{
4187 sctp_peeloff_arg_t peeloff;
4188 struct socket *newsock;
4189 int retval = 0;
4190 struct sctp_association *asoc;
4191
408f22e8 4192 if (len < sizeof(sctp_peeloff_arg_t))
1da177e4 4193 return -EINVAL;
408f22e8 4194 len = sizeof(sctp_peeloff_arg_t);
1da177e4
LT
4195 if (copy_from_user(&peeloff, optval, len))
4196 return -EFAULT;
4197
4198 asoc = sctp_id2assoc(sk, peeloff.associd);
4199 if (!asoc) {
4200 retval = -EINVAL;
4201 goto out;
4202 }
4203
0dc47877 4204 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
1da177e4
LT
4205
4206 retval = sctp_do_peeloff(asoc, &newsock);
4207 if (retval < 0)
4208 goto out;
4209
4210 /* Map the socket to an unused fd that can be returned to the user. */
a677a039 4211 retval = sock_map_fd(newsock, 0);
1da177e4
LT
4212 if (retval < 0) {
4213 sock_release(newsock);
4214 goto out;
4215 }
4216
4217 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
0dc47877 4218 __func__, sk, asoc, newsock->sk, retval);
1da177e4
LT
4219
4220 /* Return the fd mapped to the new socket. */
4221 peeloff.sd = retval;
408f22e8
NH
4222 if (put_user(len, optlen))
4223 return -EFAULT;
1da177e4
LT
4224 if (copy_to_user(optval, &peeloff, len))
4225 retval = -EFAULT;
4226
4227out:
4228 return retval;
4229}
4230
4231/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4232 *
4233 * Applications can enable or disable heartbeats for any peer address of
4234 * an association, modify an address's heartbeat interval, force a
4235 * heartbeat to be sent immediately, and adjust the address's maximum
4236 * number of retransmissions sent before an address is considered
4237 * unreachable. The following structure is used to access and modify an
4238 * address's parameters:
4239 *
4240 * struct sctp_paddrparams {
52ccb8e9
FF
4241 * sctp_assoc_t spp_assoc_id;
4242 * struct sockaddr_storage spp_address;
4243 * uint32_t spp_hbinterval;
4244 * uint16_t spp_pathmaxrxt;
4245 * uint32_t spp_pathmtu;
4246 * uint32_t spp_sackdelay;
4247 * uint32_t spp_flags;
4248 * };
4249 *
4250 * spp_assoc_id - (one-to-many style socket) This is filled in the
4251 * application, and identifies the association for
4252 * this query.
1da177e4
LT
4253 * spp_address - This specifies which address is of interest.
4254 * spp_hbinterval - This contains the value of the heartbeat interval,
52ccb8e9
FF
4255 * in milliseconds. If a value of zero
4256 * is present in this field then no changes are to
4257 * be made to this parameter.
1da177e4
LT
4258 * spp_pathmaxrxt - This contains the maximum number of
4259 * retransmissions before this address shall be
52ccb8e9
FF
4260 * considered unreachable. If a value of zero
4261 * is present in this field then no changes are to
4262 * be made to this parameter.
4263 * spp_pathmtu - When Path MTU discovery is disabled the value
4264 * specified here will be the "fixed" path mtu.
4265 * Note that if the spp_address field is empty
4266 * then all associations on this address will
4267 * have this fixed path mtu set upon them.
4268 *
4269 * spp_sackdelay - When delayed sack is enabled, this value specifies
4270 * the number of milliseconds that sacks will be delayed
4271 * for. This value will apply to all addresses of an
4272 * association if the spp_address field is empty. Note
4273 * also, that if delayed sack is enabled and this
4274 * value is set to 0, no change is made to the last
4275 * recorded delayed sack timer value.
4276 *
4277 * spp_flags - These flags are used to control various features
4278 * on an association. The flag field may contain
4279 * zero or more of the following options.
4280 *
4281 * SPP_HB_ENABLE - Enable heartbeats on the
4282 * specified address. Note that if the address
4283 * field is empty all addresses for the association
4284 * have heartbeats enabled upon them.
4285 *
4286 * SPP_HB_DISABLE - Disable heartbeats on the
4287 * speicifed address. Note that if the address
4288 * field is empty all addresses for the association
4289 * will have their heartbeats disabled. Note also
4290 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4291 * mutually exclusive, only one of these two should
4292 * be specified. Enabling both fields will have
4293 * undetermined results.
4294 *
4295 * SPP_HB_DEMAND - Request a user initiated heartbeat
4296 * to be made immediately.
4297 *
4298 * SPP_PMTUD_ENABLE - This field will enable PMTU
4299 * discovery upon the specified address. Note that
4300 * if the address feild is empty then all addresses
4301 * on the association are effected.
4302 *
4303 * SPP_PMTUD_DISABLE - This field will disable PMTU
4304 * discovery upon the specified address. Note that
4305 * if the address feild is empty then all addresses
4306 * on the association are effected. Not also that
4307 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4308 * exclusive. Enabling both will have undetermined
4309 * results.
4310 *
4311 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4312 * on delayed sack. The time specified in spp_sackdelay
4313 * is used to specify the sack delay for this address. Note
4314 * that if spp_address is empty then all addresses will
4315 * enable delayed sack and take on the sack delay
4316 * value specified in spp_sackdelay.
4317 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4318 * off delayed sack. If the spp_address field is blank then
4319 * delayed sack is disabled for the entire association. Note
4320 * also that this field is mutually exclusive to
4321 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4322 * results.
1da177e4
LT
4323 */
4324static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
52ccb8e9 4325 char __user *optval, int __user *optlen)
1da177e4 4326{
52ccb8e9
FF
4327 struct sctp_paddrparams params;
4328 struct sctp_transport *trans = NULL;
4329 struct sctp_association *asoc = NULL;
4330 struct sctp_sock *sp = sctp_sk(sk);
1da177e4 4331
408f22e8 4332 if (len < sizeof(struct sctp_paddrparams))
1da177e4 4333 return -EINVAL;
408f22e8 4334 len = sizeof(struct sctp_paddrparams);
1da177e4
LT
4335 if (copy_from_user(&params, optval, len))
4336 return -EFAULT;
4337
52ccb8e9
FF
4338 /* If an address other than INADDR_ANY is specified, and
4339 * no transport is found, then the request is invalid.
1da177e4 4340 */
52cae8f0 4341 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
52ccb8e9
FF
4342 trans = sctp_addr_id2transport(sk, &params.spp_address,
4343 params.spp_assoc_id);
4344 if (!trans) {
4345 SCTP_DEBUG_PRINTK("Failed no transport\n");
4346 return -EINVAL;
4347 }
1da177e4
LT
4348 }
4349
52ccb8e9
FF
4350 /* Get association, if assoc_id != 0 and the socket is a one
4351 * to many style socket, and an association was not found, then
4352 * the id was invalid.
4353 */
4354 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4355 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4356 SCTP_DEBUG_PRINTK("Failed no association\n");
1da177e4 4357 return -EINVAL;
52ccb8e9 4358 }
1da177e4 4359
52ccb8e9
FF
4360 if (trans) {
4361 /* Fetch transport values. */
4362 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4363 params.spp_pathmtu = trans->pathmtu;
4364 params.spp_pathmaxrxt = trans->pathmaxrxt;
4365 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4366
4367 /*draft-11 doesn't say what to return in spp_flags*/
4368 params.spp_flags = trans->param_flags;
4369 } else if (asoc) {
4370 /* Fetch association values. */
4371 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4372 params.spp_pathmtu = asoc->pathmtu;
4373 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4374 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4375
4376 /*draft-11 doesn't say what to return in spp_flags*/
4377 params.spp_flags = asoc->param_flags;
4378 } else {
4379 /* Fetch socket values. */
4380 params.spp_hbinterval = sp->hbinterval;
4381 params.spp_pathmtu = sp->pathmtu;
4382 params.spp_sackdelay = sp->sackdelay;
4383 params.spp_pathmaxrxt = sp->pathmaxrxt;
1da177e4 4384
52ccb8e9
FF
4385 /*draft-11 doesn't say what to return in spp_flags*/
4386 params.spp_flags = sp->param_flags;
4387 }
1da177e4 4388
1da177e4
LT
4389 if (copy_to_user(optval, &params, len))
4390 return -EFAULT;
4391
4392 if (put_user(len, optlen))
4393 return -EFAULT;
4394
4395 return 0;
4396}
4397
d364d927
WY
4398/*
4399 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4400 *
4401 * This option will effect the way delayed acks are performed. This
4402 * option allows you to get or set the delayed ack time, in
4403 * milliseconds. It also allows changing the delayed ack frequency.
4404 * Changing the frequency to 1 disables the delayed sack algorithm. If
4405 * the assoc_id is 0, then this sets or gets the endpoints default
4406 * values. If the assoc_id field is non-zero, then the set or get
4407 * effects the specified association for the one to many model (the
4408 * assoc_id field is ignored by the one to one model). Note that if
4409 * sack_delay or sack_freq are 0 when setting this option, then the
4410 * current values will remain unchanged.
4411 *
4412 * struct sctp_sack_info {
4413 * sctp_assoc_t sack_assoc_id;
4414 * uint32_t sack_delay;
4415 * uint32_t sack_freq;
4416 * };
7708610b 4417 *
d364d927
WY
4418 * sack_assoc_id - This parameter, indicates which association the user
4419 * is performing an action upon. Note that if this field's value is
4420 * zero then the endpoints default value is changed (effecting future
4421 * associations only).
7708610b 4422 *
d364d927
WY
4423 * sack_delay - This parameter contains the number of milliseconds that
4424 * the user is requesting the delayed ACK timer be set to. Note that
4425 * this value is defined in the standard to be between 200 and 500
4426 * milliseconds.
7708610b 4427 *
d364d927
WY
4428 * sack_freq - This parameter contains the number of packets that must
4429 * be received before a sack is sent without waiting for the delay
4430 * timer to expire. The default value for this is 2, setting this
4431 * value to 1 will disable the delayed sack algorithm.
7708610b 4432 */
d364d927 4433static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
7708610b
FF
4434 char __user *optval,
4435 int __user *optlen)
4436{
d364d927 4437 struct sctp_sack_info params;
7708610b
FF
4438 struct sctp_association *asoc = NULL;
4439 struct sctp_sock *sp = sctp_sk(sk);
4440
d364d927
WY
4441 if (len >= sizeof(struct sctp_sack_info)) {
4442 len = sizeof(struct sctp_sack_info);
7708610b 4443
d364d927
WY
4444 if (copy_from_user(&params, optval, len))
4445 return -EFAULT;
4446 } else if (len == sizeof(struct sctp_assoc_value)) {
145ce502
JP
4447 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4448 pr_warn("Use struct sctp_sack_info instead\n");
d364d927
WY
4449 if (copy_from_user(&params, optval, len))
4450 return -EFAULT;
4451 } else
4452 return - EINVAL;
7708610b 4453
d364d927 4454 /* Get association, if sack_assoc_id != 0 and the socket is a one
7708610b
FF
4455 * to many style socket, and an association was not found, then
4456 * the id was invalid.
d808ad9a 4457 */
d364d927
WY
4458 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4459 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
7708610b
FF
4460 return -EINVAL;
4461
4462 if (asoc) {
4463 /* Fetch association values. */
d364d927
WY
4464 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4465 params.sack_delay = jiffies_to_msecs(
7708610b 4466 asoc->sackdelay);
d364d927
WY
4467 params.sack_freq = asoc->sackfreq;
4468
4469 } else {
4470 params.sack_delay = 0;
4471 params.sack_freq = 1;
4472 }
7708610b
FF
4473 } else {
4474 /* Fetch socket values. */
d364d927
WY
4475 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4476 params.sack_delay = sp->sackdelay;
4477 params.sack_freq = sp->sackfreq;
4478 } else {
4479 params.sack_delay = 0;
4480 params.sack_freq = 1;
4481 }
7708610b
FF
4482 }
4483
4484 if (copy_to_user(optval, &params, len))
4485 return -EFAULT;
4486
4487 if (put_user(len, optlen))
4488 return -EFAULT;
4489
4490 return 0;
4491}
4492
1da177e4
LT
4493/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4494 *
4495 * Applications can specify protocol parameters for the default association
4496 * initialization. The option name argument to setsockopt() and getsockopt()
4497 * is SCTP_INITMSG.
4498 *
4499 * Setting initialization parameters is effective only on an unconnected
4500 * socket (for UDP-style sockets only future associations are effected
4501 * by the change). With TCP-style sockets, this option is inherited by
4502 * sockets derived from a listener socket.
4503 */
4504static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4505{
408f22e8 4506 if (len < sizeof(struct sctp_initmsg))
1da177e4 4507 return -EINVAL;
408f22e8
NH
4508 len = sizeof(struct sctp_initmsg);
4509 if (put_user(len, optlen))
4510 return -EFAULT;
1da177e4
LT
4511 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4512 return -EFAULT;
4513 return 0;
4514}
4515