]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - net/sctp/socket.c
Merge branch 'bnxt_en-next'
[thirdparty/kernel/stable.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
4b2f13a2
JK
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
1da177e4
LT
33 *
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
91705c61 36 * lksctp developers <linux-sctp@vger.kernel.org>
1da177e4 37 *
1da177e4
LT
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
1da177e4
LT
51 */
52
145ce502
JP
53#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
54
5821c769 55#include <crypto/hash.h>
1da177e4
LT
56#include <linux/types.h>
57#include <linux/kernel.h>
58#include <linux/wait.h>
59#include <linux/time.h>
60#include <linux/ip.h>
4fc268d2 61#include <linux/capability.h>
1da177e4
LT
62#include <linux/fcntl.h>
63#include <linux/poll.h>
64#include <linux/init.h>
5a0e3ad6 65#include <linux/slab.h>
56b31d1c 66#include <linux/file.h>
ffd59393 67#include <linux/compat.h>
1da177e4
LT
68
69#include <net/ip.h>
70#include <net/icmp.h>
71#include <net/route.h>
72#include <net/ipv6.h>
73#include <net/inet_common.h>
8465a5fc 74#include <net/busy_poll.h>
1da177e4
LT
75
76#include <linux/socket.h> /* for sa_family_t */
bc3b2d7f 77#include <linux/export.h>
1da177e4
LT
78#include <net/sock.h>
79#include <net/sctp/sctp.h>
80#include <net/sctp/sm.h>
81
1da177e4
LT
82/* Forward declarations for internal helper functions. */
83static int sctp_writeable(struct sock *sk);
84static void sctp_wfree(struct sk_buff *skb);
85static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 size_t msg_len);
26ac8e5f 87static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
1da177e4
LT
88static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89static int sctp_wait_for_accept(struct sock *sk, long timeo);
90static void sctp_wait_for_close(struct sock *sk, long timeo);
0a2fbac1 91static void sctp_destruct_sock(struct sock *sk);
1da177e4
LT
92static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101static int sctp_autobind(struct sock *sk);
102static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
1da177e4 104
b6fa1a4d 105static int sctp_memory_pressure;
8d987e5c 106static atomic_long_t sctp_memory_allocated;
1748376b 107struct percpu_counter sctp_sockets_allocated;
4d93df0a 108
5c52ba17 109static void sctp_enter_memory_pressure(struct sock *sk)
4d93df0a
NH
110{
111 sctp_memory_pressure = 1;
112}
113
114
1da177e4
LT
115/* Get the sndbuf space available at the time on the association. */
116static inline int sctp_wspace(struct sctp_association *asoc)
117{
4d93df0a 118 int amt;
1da177e4 119
4d93df0a
NH
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
122 else
31e6d363 123 amt = sk_wmem_alloc_get(asoc->base.sk);
4d93df0a
NH
124
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
127 amt = 0;
128 else {
129 amt = sk_stream_wspace(asoc->base.sk);
130 if (amt < 0)
131 amt = 0;
132 }
4eb701df 133 } else {
4d93df0a 134 amt = asoc->base.sk->sk_sndbuf - amt;
4eb701df 135 }
1da177e4
LT
136 return amt;
137}
138
139/* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
142 *
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
146 * tracking.
147 */
148static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
149{
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
152
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
155
4eb701df
NH
156 skb_set_owner_w(chunk->skb, sk);
157
1da177e4
LT
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
f869c912 160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
1da177e4 161
4eb701df
NH
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
165
4eb701df 166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
3ab224be
HA
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
1da177e4
LT
169}
170
171/* Verify that this is a valid address. */
172static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
173 int len)
174{
175 struct sctp_af *af;
176
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
179 if (!af)
180 return -EINVAL;
181
182 /* Is this a valid SCTP address? */
5636bef7 183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
1da177e4
LT
184 return -EINVAL;
185
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
187 return -EINVAL;
188
189 return 0;
190}
191
192/* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
194 */
195struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
196{
197 struct sctp_association *asoc = NULL;
198
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
204 */
e5b13f34 205 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
1da177e4
LT
206 return NULL;
207
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
212 return asoc;
213 }
214
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
217 return NULL;
218
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
222
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
224 return NULL;
225
226 return asoc;
227}
228
229/* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
231 * the same.
232 */
233static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
235 sctp_assoc_t id)
236{
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_transport *transport;
239 union sctp_addr *laddr = (union sctp_addr *)addr;
240
1da177e4 241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
cd4ff034 242 laddr,
1da177e4 243 &transport);
1da177e4
LT
244
245 if (!addr_asoc)
246 return NULL;
247
248 id_asoc = sctp_id2assoc(sk, id);
249 if (id_asoc && (id_asoc != addr_asoc))
250 return NULL;
251
299ee123 252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
1da177e4
LT
253 (union sctp_addr *)addr);
254
255 return transport;
256}
257
258/* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
260 *
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
262 *
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
267 */
dda91928 268static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
1da177e4
LT
269{
270 int retval = 0;
271
048ed4b6 272 lock_sock(sk);
1da177e4 273
bb33381d
DB
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
275 addr, addr_len);
1da177e4
LT
276
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk)->ep->base.bind_addr.port)
3f7a87d2 279 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
1da177e4
LT
280 addr_len);
281 else
282 retval = -EINVAL;
283
048ed4b6 284 release_sock(sk);
1da177e4
LT
285
286 return retval;
287}
288
289static long sctp_get_port_local(struct sock *, union sctp_addr *);
290
291/* Verify this is a valid sockaddr. */
292static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293 union sctp_addr *addr, int len)
294{
295 struct sctp_af *af;
296
297 /* Check minimum size. */
298 if (len < sizeof (struct sockaddr))
299 return NULL;
300
7dab83de
VY
301 /* V4 mapped address are really of AF_INET family */
302 if (addr->sa.sa_family == AF_INET6 &&
303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304 if (!opt->pf->af_supported(AF_INET, opt))
305 return NULL;
306 } else {
307 /* Does this PF support this AF? */
308 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
309 return NULL;
310 }
1da177e4
LT
311
312 /* If we get this far, af is valid. */
313 af = sctp_get_af_specific(addr->sa.sa_family);
314
315 if (len < af->sockaddr_len)
316 return NULL;
317
318 return af;
319}
320
321/* Bind a local address either to an endpoint or to an association. */
dda91928 322static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
1da177e4 323{
3594698a 324 struct net *net = sock_net(sk);
1da177e4
LT
325 struct sctp_sock *sp = sctp_sk(sk);
326 struct sctp_endpoint *ep = sp->ep;
327 struct sctp_bind_addr *bp = &ep->base.bind_addr;
328 struct sctp_af *af;
329 unsigned short snum;
330 int ret = 0;
331
1da177e4
LT
332 /* Common sockaddr verification. */
333 af = sctp_sockaddr_af(sp, addr, len);
3f7a87d2 334 if (!af) {
bb33381d
DB
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__, sk, addr, len);
1da177e4 337 return -EINVAL;
3f7a87d2
FF
338 }
339
340 snum = ntohs(addr->v4.sin_port);
341
bb33381d
DB
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__, sk, &addr->sa, bp->port, snum, len);
1da177e4
LT
344
345 /* PF specific bind() address verification. */
346 if (!sp->pf->bind_verify(sp, addr))
347 return -EADDRNOTAVAIL;
348
8b358056
VY
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
352 */
353 if (bp->port) {
354 if (!snum)
355 snum = bp->port;
356 else if (snum != bp->port) {
bb33381d
DB
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__, snum, bp->port);
8b358056
VY
359 return -EINVAL;
360 }
1da177e4
LT
361 }
362
3594698a
EB
363 if (snum && snum < PROT_SOCK &&
364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1da177e4
LT
365 return -EACCES;
366
4e54064e
VY
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
369 */
370 if (sctp_bind_addr_match(bp, addr, sp))
371 return -EINVAL;
372
1da177e4
LT
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
375 * detection.
376 */
2772b495 377 addr->v4.sin_port = htons(snum);
1da177e4 378 if ((ret = sctp_get_port_local(sk, addr))) {
4e54064e 379 return -EADDRINUSE;
1da177e4
LT
380 }
381
382 /* Refresh ephemeral port. */
383 if (!bp->port)
c720c7e8 384 bp->port = inet_sk(sk)->inet_num;
1da177e4 385
559cf710
VY
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
388 */
133800d1
MRL
389 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
390 SCTP_ADDR_SRC, GFP_ATOMIC);
1da177e4
LT
391
392 /* Copy back into socket for getsockname() use. */
393 if (!ret) {
c720c7e8 394 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
299ee123 395 sp->pf->to_sk_saddr(addr, sk);
1da177e4
LT
396 }
397
398 return ret;
399}
400
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
402 *
d808ad9a 403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
1da177e4 404 * at any one time. If a sender, after sending an ASCONF chunk, decides
d808ad9a 405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
1da177e4 406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
d808ad9a
YH
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
1da177e4
LT
409 * from each endpoint).
410 */
411static int sctp_send_asconf(struct sctp_association *asoc,
412 struct sctp_chunk *chunk)
413{
55e26eb9 414 struct net *net = sock_net(asoc->base.sk);
1da177e4
LT
415 int retval = 0;
416
417 /* If there is an outstanding ASCONF chunk, queue it for later
418 * transmission.
d808ad9a 419 */
1da177e4 420 if (asoc->addip_last_asconf) {
79af02c2 421 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
d808ad9a 422 goto out;
1da177e4
LT
423 }
424
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk);
55e26eb9 427 retval = sctp_primitive_ASCONF(net, asoc, chunk);
1da177e4
LT
428 if (retval)
429 sctp_chunk_free(chunk);
430 else
431 asoc->addip_last_asconf = chunk;
432
433out:
434 return retval;
435}
436
437/* Add a list of addresses as bind addresses to local endpoint or
438 * association.
439 *
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
443 *
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
446 *
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
448 */
04675210 449static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
1da177e4
LT
450{
451 int cnt;
452 int retval = 0;
453 void *addr_buf;
454 struct sockaddr *sa_addr;
455 struct sctp_af *af;
456
bb33381d
DB
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
458 addrs, addrcnt);
1da177e4
LT
459
460 addr_buf = addrs;
461 for (cnt = 0; cnt < addrcnt; cnt++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
464 */
ea110733 465 sa_addr = addr_buf;
1da177e4
LT
466 af = sctp_get_af_specific(sa_addr->sa_family);
467 if (!af) {
468 retval = -EINVAL;
469 goto err_bindx_add;
470 }
471
d808ad9a 472 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
1da177e4
LT
473 af->sockaddr_len);
474
475 addr_buf += af->sockaddr_len;
476
477err_bindx_add:
478 if (retval < 0) {
479 /* Failed. Cleanup the ones that have been added */
480 if (cnt > 0)
481 sctp_bindx_rem(sk, addrs, cnt);
482 return retval;
483 }
484 }
485
486 return retval;
487}
488
489/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
492 *
d808ad9a 493 * If any of the addresses is already in the bind address list of the
1da177e4
LT
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
496 *
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 */
d808ad9a 499static int sctp_send_asconf_add_ip(struct sock *sk,
1da177e4
LT
500 struct sockaddr *addrs,
501 int addrcnt)
502{
e1fc3b14 503 struct net *net = sock_net(sk);
1da177e4
LT
504 struct sctp_sock *sp;
505 struct sctp_endpoint *ep;
506 struct sctp_association *asoc;
507 struct sctp_bind_addr *bp;
508 struct sctp_chunk *chunk;
509 struct sctp_sockaddr_entry *laddr;
510 union sctp_addr *addr;
dc022a98 511 union sctp_addr saveaddr;
1da177e4
LT
512 void *addr_buf;
513 struct sctp_af *af;
1da177e4
LT
514 struct list_head *p;
515 int i;
516 int retval = 0;
517
e1fc3b14 518 if (!net->sctp.addip_enable)
1da177e4
LT
519 return retval;
520
521 sp = sctp_sk(sk);
522 ep = sp->ep;
523
bb33381d
DB
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__, sk, addrs, addrcnt);
1da177e4 526
9dbc15f0 527 list_for_each_entry(asoc, &ep->asocs, asocs) {
1da177e4
LT
528 if (!asoc->peer.asconf_capable)
529 continue;
530
531 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
532 continue;
533
534 if (!sctp_state(asoc, ESTABLISHED))
535 continue;
536
537 /* Check if any address in the packed array of addresses is
d808ad9a
YH
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
1da177e4
LT
540 * other associations.
541 */
542 addr_buf = addrs;
543 for (i = 0; i < addrcnt; i++) {
ea110733 544 addr = addr_buf;
1da177e4
LT
545 af = sctp_get_af_specific(addr->v4.sin_family);
546 if (!af) {
547 retval = -EINVAL;
548 goto out;
549 }
550
551 if (sctp_assoc_lookup_laddr(asoc, addr))
552 break;
553
554 addr_buf += af->sockaddr_len;
555 }
556 if (i < addrcnt)
557 continue;
558
559cf710
VY
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
1da177e4 561 */
1da177e4
LT
562 bp = &asoc->base.bind_addr;
563 p = bp->address_list.next;
564 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
5ae955cf 565 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
1da177e4
LT
566 addrcnt, SCTP_PARAM_ADD_IP);
567 if (!chunk) {
568 retval = -ENOMEM;
569 goto out;
570 }
571
dc022a98
SS
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
1da177e4 574 */
dc022a98
SS
575 addr_buf = addrs;
576 for (i = 0; i < addrcnt; i++) {
ea110733 577 addr = addr_buf;
dc022a98
SS
578 af = sctp_get_af_specific(addr->v4.sin_family);
579 memcpy(&saveaddr, addr, af->sockaddr_len);
f57d96b2 580 retval = sctp_add_bind_addr(bp, &saveaddr,
133800d1 581 sizeof(saveaddr),
f57d96b2 582 SCTP_ADDR_NEW, GFP_ATOMIC);
dc022a98
SS
583 addr_buf += af->sockaddr_len;
584 }
8a07eb0a
MH
585 if (asoc->src_out_of_asoc_ok) {
586 struct sctp_transport *trans;
587
588 list_for_each_entry(trans,
589 &asoc->peer.transport_addr_list, transports) {
590 /* Clear the source and route cache */
591 dst_release(trans->dst);
592 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
593 2*asoc->pathmtu, 4380));
594 trans->ssthresh = asoc->peer.i.a_rwnd;
595 trans->rto = asoc->rto_initial;
196d6759 596 sctp_max_rto(asoc, trans);
8a07eb0a
MH
597 trans->rtt = trans->srtt = trans->rttvar = 0;
598 sctp_transport_route(trans, NULL,
599 sctp_sk(asoc->base.sk));
600 }
601 }
602 retval = sctp_send_asconf(asoc, chunk);
1da177e4
LT
603 }
604
605out:
606 return retval;
607}
608
609/* Remove a list of addresses from bind addresses list. Do not remove the
610 * last address.
611 *
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
615 *
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
618 *
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
621 *
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
623 */
04675210 624static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
1da177e4
LT
625{
626 struct sctp_sock *sp = sctp_sk(sk);
627 struct sctp_endpoint *ep = sp->ep;
628 int cnt;
629 struct sctp_bind_addr *bp = &ep->base.bind_addr;
630 int retval = 0;
1da177e4 631 void *addr_buf;
c9a08505 632 union sctp_addr *sa_addr;
1da177e4
LT
633 struct sctp_af *af;
634
bb33381d
DB
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__, sk, addrs, addrcnt);
1da177e4
LT
637
638 addr_buf = addrs;
639 for (cnt = 0; cnt < addrcnt; cnt++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
643 */
644 if (list_empty(&bp->address_list) ||
645 (sctp_list_single_entry(&bp->address_list))) {
646 retval = -EBUSY;
647 goto err_bindx_rem;
648 }
649
ea110733 650 sa_addr = addr_buf;
c9a08505 651 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1da177e4
LT
652 if (!af) {
653 retval = -EINVAL;
654 goto err_bindx_rem;
655 }
0304ff8a
PG
656
657 if (!af->addr_valid(sa_addr, sp, NULL)) {
658 retval = -EADDRNOTAVAIL;
659 goto err_bindx_rem;
660 }
661
ee9cbaca
VY
662 if (sa_addr->v4.sin_port &&
663 sa_addr->v4.sin_port != htons(bp->port)) {
1da177e4
LT
664 retval = -EINVAL;
665 goto err_bindx_rem;
666 }
667
ee9cbaca
VY
668 if (!sa_addr->v4.sin_port)
669 sa_addr->v4.sin_port = htons(bp->port);
670
1da177e4
LT
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
677 */
0ed90fb0 678 retval = sctp_del_bind_addr(bp, sa_addr);
1da177e4
LT
679
680 addr_buf += af->sockaddr_len;
681err_bindx_rem:
682 if (retval < 0) {
683 /* Failed. Add the ones that has been removed back */
684 if (cnt > 0)
685 sctp_bindx_add(sk, addrs, cnt);
686 return retval;
687 }
688 }
689
690 return retval;
691}
692
693/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
696 *
d808ad9a 697 * If any of the addresses is already in the bind address list of the
1da177e4
LT
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
700 *
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
702 */
703static int sctp_send_asconf_del_ip(struct sock *sk,
704 struct sockaddr *addrs,
705 int addrcnt)
706{
e1fc3b14 707 struct net *net = sock_net(sk);
1da177e4
LT
708 struct sctp_sock *sp;
709 struct sctp_endpoint *ep;
710 struct sctp_association *asoc;
dc022a98 711 struct sctp_transport *transport;
1da177e4
LT
712 struct sctp_bind_addr *bp;
713 struct sctp_chunk *chunk;
714 union sctp_addr *laddr;
715 void *addr_buf;
716 struct sctp_af *af;
dc022a98 717 struct sctp_sockaddr_entry *saddr;
1da177e4
LT
718 int i;
719 int retval = 0;
8a07eb0a 720 int stored = 0;
1da177e4 721
8a07eb0a 722 chunk = NULL;
e1fc3b14 723 if (!net->sctp.addip_enable)
1da177e4
LT
724 return retval;
725
726 sp = sctp_sk(sk);
727 ep = sp->ep;
728
bb33381d
DB
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__, sk, addrs, addrcnt);
1da177e4 731
9dbc15f0 732 list_for_each_entry(asoc, &ep->asocs, asocs) {
1da177e4
LT
733
734 if (!asoc->peer.asconf_capable)
735 continue;
736
737 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
738 continue;
739
740 if (!sctp_state(asoc, ESTABLISHED))
741 continue;
742
743 /* Check if any address in the packed array of addresses is
d808ad9a 744 * not present in the bind address list of the association.
1da177e4
LT
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
747 */
748 addr_buf = addrs;
749 for (i = 0; i < addrcnt; i++) {
ea110733 750 laddr = addr_buf;
1da177e4
LT
751 af = sctp_get_af_specific(laddr->v4.sin_family);
752 if (!af) {
753 retval = -EINVAL;
754 goto out;
755 }
756
757 if (!sctp_assoc_lookup_laddr(asoc, laddr))
758 break;
759
760 addr_buf += af->sockaddr_len;
761 }
762 if (i < addrcnt)
763 continue;
764
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
768 * association.
769 */
1da177e4
LT
770 bp = &asoc->base.bind_addr;
771 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
772 addrcnt, sp);
8a07eb0a
MH
773 if ((laddr == NULL) && (addrcnt == 1)) {
774 if (asoc->asconf_addr_del_pending)
775 continue;
776 asoc->asconf_addr_del_pending =
777 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
6d65e5ee
MH
778 if (asoc->asconf_addr_del_pending == NULL) {
779 retval = -ENOMEM;
780 goto out;
781 }
8a07eb0a
MH
782 asoc->asconf_addr_del_pending->sa.sa_family =
783 addrs->sa_family;
784 asoc->asconf_addr_del_pending->v4.sin_port =
785 htons(bp->port);
786 if (addrs->sa_family == AF_INET) {
787 struct sockaddr_in *sin;
788
789 sin = (struct sockaddr_in *)addrs;
790 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
791 } else if (addrs->sa_family == AF_INET6) {
792 struct sockaddr_in6 *sin6;
793
794 sin6 = (struct sockaddr_in6 *)addrs;
4e3fd7a0 795 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
8a07eb0a 796 }
bb33381d
DB
797
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
800 asoc->asconf_addr_del_pending);
801
8a07eb0a
MH
802 asoc->src_out_of_asoc_ok = 1;
803 stored = 1;
804 goto skip_mkasconf;
805 }
1da177e4 806
88362ad8
DB
807 if (laddr == NULL)
808 return -EINVAL;
809
559cf710
VY
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
812 * setsockopt call.
813 */
1da177e4
LT
814 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
815 SCTP_PARAM_DEL_IP);
816 if (!chunk) {
817 retval = -ENOMEM;
818 goto out;
819 }
820
8a07eb0a 821skip_mkasconf:
dc022a98
SS
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
824 */
dc022a98
SS
825 addr_buf = addrs;
826 for (i = 0; i < addrcnt; i++) {
ea110733 827 laddr = addr_buf;
dc022a98 828 af = sctp_get_af_specific(laddr->v4.sin_family);
559cf710 829 list_for_each_entry(saddr, &bp->address_list, list) {
5f242a13 830 if (sctp_cmp_addr_exact(&saddr->a, laddr))
f57d96b2 831 saddr->state = SCTP_ADDR_DEL;
dc022a98
SS
832 }
833 addr_buf += af->sockaddr_len;
834 }
1da177e4 835
dc022a98
SS
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
1da177e4 839 */
9dbc15f0
RD
840 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
841 transports) {
dc022a98
SS
842 dst_release(transport->dst);
843 sctp_transport_route(transport, NULL,
844 sctp_sk(asoc->base.sk));
845 }
846
8a07eb0a
MH
847 if (stored)
848 /* We don't need to transmit ASCONF */
849 continue;
dc022a98 850 retval = sctp_send_asconf(asoc, chunk);
1da177e4
LT
851 }
852out:
853 return retval;
854}
855
9f7d653b
MH
856/* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
858{
859 struct sock *sk = sctp_opt2sk(sp);
860 union sctp_addr *addr;
861 struct sctp_af *af;
862
863 /* It is safe to write port space in caller. */
864 addr = &addrw->a;
865 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
866 af = sctp_get_af_specific(addr->sa.sa_family);
867 if (!af)
868 return -EINVAL;
869 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
870 return -EINVAL;
871
872 if (addrw->state == SCTP_ADDR_NEW)
873 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
874 else
875 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
876}
877
1da177e4
LT
878/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
879 *
880 * API 8.1
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
882 * int flags);
883 *
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
886 * or IPv6 addresses.
887 *
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
890 *
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
23c435f7 894 * must be used to distinguish the address length (note that this
1da177e4
LT
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
897 *
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
900 *
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
903 *
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
906 *
907 * SCTP_BINDX_ADD_ADDR
908 *
909 * SCTP_BINDX_REM_ADDR
910 *
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
917 *
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
926 *
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
930 *
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
3f7a87d2
FF
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
934 * from userspace.
1da177e4
LT
935 *
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
942 *
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
944 * it.
945 *
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
950 * sctp_bindx)
951 *
952 * Returns 0 if ok, <0 errno code on error.
953 */
26ac8e5f 954static int sctp_setsockopt_bindx(struct sock *sk,
dda91928
DB
955 struct sockaddr __user *addrs,
956 int addrs_size, int op)
1da177e4
LT
957{
958 struct sockaddr *kaddrs;
959 int err;
960 int addrcnt = 0;
961 int walk_size = 0;
962 struct sockaddr *sa_addr;
963 void *addr_buf;
964 struct sctp_af *af;
965
bb33381d
DB
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__, sk, addrs, addrs_size, op);
1da177e4
LT
968
969 if (unlikely(addrs_size <= 0))
970 return -EINVAL;
971
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
974 return -EFAULT;
975
976 /* Alloc space for the address array in kernel memory. */
cacc0621 977 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1da177e4
LT
978 if (unlikely(!kaddrs))
979 return -ENOMEM;
980
981 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
982 kfree(kaddrs);
983 return -EFAULT;
984 }
985
d808ad9a 986 /* Walk through the addrs buffer and count the number of addresses. */
1da177e4
LT
987 addr_buf = kaddrs;
988 while (walk_size < addrs_size) {
d7e0d19a
DR
989 if (walk_size + sizeof(sa_family_t) > addrs_size) {
990 kfree(kaddrs);
991 return -EINVAL;
992 }
993
ea110733 994 sa_addr = addr_buf;
1da177e4
LT
995 af = sctp_get_af_specific(sa_addr->sa_family);
996
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
d808ad9a 999 */
1da177e4
LT
1000 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1001 kfree(kaddrs);
1002 return -EINVAL;
1003 }
1004 addrcnt++;
1005 addr_buf += af->sockaddr_len;
1006 walk_size += af->sockaddr_len;
1007 }
1008
1009 /* Do the work. */
1010 switch (op) {
1011 case SCTP_BINDX_ADD_ADDR:
1012 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1013 if (err)
1014 goto out;
1015 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1016 break;
1017
1018 case SCTP_BINDX_REM_ADDR:
1019 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1020 if (err)
1021 goto out;
1022 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1023 break;
1024
1025 default:
1026 err = -EINVAL;
1027 break;
3ff50b79 1028 }
1da177e4
LT
1029
1030out:
1031 kfree(kaddrs);
1032
1033 return err;
1034}
1035
3f7a87d2
FF
1036/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1037 *
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1040 */
26ac8e5f 1041static int __sctp_connect(struct sock *sk,
3f7a87d2 1042 struct sockaddr *kaddrs,
88a0a948
VY
1043 int addrs_size,
1044 sctp_assoc_t *assoc_id)
3f7a87d2 1045{
55e26eb9 1046 struct net *net = sock_net(sk);
3f7a87d2
FF
1047 struct sctp_sock *sp;
1048 struct sctp_endpoint *ep;
1049 struct sctp_association *asoc = NULL;
1050 struct sctp_association *asoc2;
1051 struct sctp_transport *transport;
1052 union sctp_addr to;
3f7a87d2
FF
1053 sctp_scope_t scope;
1054 long timeo;
1055 int err = 0;
1056 int addrcnt = 0;
1057 int walk_size = 0;
e4d1feab 1058 union sctp_addr *sa_addr = NULL;
3f7a87d2 1059 void *addr_buf;
16d00fb7 1060 unsigned short port;
f50f95ca 1061 unsigned int f_flags = 0;
3f7a87d2
FF
1062
1063 sp = sctp_sk(sk);
1064 ep = sp->ep;
1065
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1070 */
e5b13f34 1071 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
3f7a87d2
FF
1072 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1073 err = -EISCONN;
1074 goto out_free;
1075 }
1076
1077 /* Walk through the addrs buffer and count the number of addresses. */
1078 addr_buf = kaddrs;
1079 while (walk_size < addrs_size) {
299ee123
JG
1080 struct sctp_af *af;
1081
d7e0d19a
DR
1082 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1083 err = -EINVAL;
1084 goto out_free;
1085 }
1086
ea110733 1087 sa_addr = addr_buf;
4bdf4b5f 1088 af = sctp_get_af_specific(sa_addr->sa.sa_family);
3f7a87d2
FF
1089
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1092 */
1093 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1094 err = -EINVAL;
1095 goto out_free;
1096 }
1097
d7e0d19a
DR
1098 port = ntohs(sa_addr->v4.sin_port);
1099
e4d1feab
VY
1100 /* Save current address so we can work with it */
1101 memcpy(&to, sa_addr, af->sockaddr_len);
1102
1103 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
3f7a87d2
FF
1104 if (err)
1105 goto out_free;
1106
16d00fb7
VY
1107 /* Make sure the destination port is correctly set
1108 * in all addresses.
1109 */
524fba6c
WY
1110 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1111 err = -EINVAL;
16d00fb7 1112 goto out_free;
524fba6c 1113 }
3f7a87d2
FF
1114
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1117 */
e4d1feab 1118 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
3f7a87d2
FF
1119 if (asoc2 && asoc2 != asoc) {
1120 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1121 err = -EISCONN;
1122 else
1123 err = -EALREADY;
1124 goto out_free;
1125 }
1126
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1130 */
e4d1feab 1131 if (sctp_endpoint_is_peeled_off(ep, &to)) {
3f7a87d2
FF
1132 err = -EADDRNOTAVAIL;
1133 goto out_free;
1134 }
1135
1136 if (!asoc) {
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1141 */
1142 if (!ep->base.bind_addr.port) {
1143 if (sctp_autobind(sk)) {
1144 err = -EAGAIN;
1145 goto out_free;
1146 }
64a0c1c8
ISJ
1147 } else {
1148 /*
d808ad9a
YH
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
64a0c1c8
ISJ
1153 * be permitted to open new associations.
1154 */
1155 if (ep->base.bind_addr.port < PROT_SOCK &&
3594698a 1156 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
64a0c1c8
ISJ
1157 err = -EACCES;
1158 goto out_free;
1159 }
3f7a87d2
FF
1160 }
1161
e4d1feab 1162 scope = sctp_scope(&to);
3f7a87d2
FF
1163 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1164 if (!asoc) {
1165 err = -ENOMEM;
1166 goto out_free;
1167 }
409b95af
VY
1168
1169 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1170 GFP_KERNEL);
1171 if (err < 0) {
1172 goto out_free;
1173 }
1174
3f7a87d2
FF
1175 }
1176
1177 /* Prime the peer's transport structures. */
e4d1feab 1178 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
3f7a87d2
FF
1179 SCTP_UNKNOWN);
1180 if (!transport) {
1181 err = -ENOMEM;
1182 goto out_free;
1183 }
1184
1185 addrcnt++;
1186 addr_buf += af->sockaddr_len;
1187 walk_size += af->sockaddr_len;
1188 }
1189
c6ba68a2
VY
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1192 */
1193 if (assoc_id) {
1194 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1195 if (err < 0)
1196 goto out_free;
1197 }
1198
55e26eb9 1199 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
3f7a87d2
FF
1200 if (err < 0) {
1201 goto out_free;
1202 }
1203
1204 /* Initialize sk's dport and daddr for getpeername() */
c720c7e8 1205 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
299ee123 1206 sp->pf->to_sk_daddr(sa_addr, sk);
8de8c873 1207 sk->sk_err = 0;
3f7a87d2 1208
f50f95ca
VY
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1211 */
1212 if (sk->sk_socket->file)
1213 f_flags = sk->sk_socket->file->f_flags;
1214
1215 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1216
7233bc84 1217 if (assoc_id)
88a0a948 1218 *assoc_id = asoc->assoc_id;
7233bc84
MRL
1219 err = sctp_wait_for_connect(asoc, &timeo);
1220 /* Note: the asoc may be freed after the return of
1221 * sctp_wait_for_connect.
1222 */
3f7a87d2
FF
1223
1224 /* Don't free association on exit. */
1225 asoc = NULL;
1226
1227out_free:
bb33381d
DB
1228 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1229 __func__, asoc, kaddrs, err);
3f7a87d2 1230
2eebc1e1
NH
1231 if (asoc) {
1232 /* sctp_primitive_ASSOCIATE may have added this association
1233 * To the hash table, try to unhash it, just in case, its a noop
1234 * if it wasn't hashed so we're safe
1235 */
3f7a87d2 1236 sctp_association_free(asoc);
2eebc1e1 1237 }
3f7a87d2
FF
1238 return err;
1239}
1240
1241/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1242 *
1243 * API 8.9
88a0a948
VY
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * sctp_assoc_t *asoc);
3f7a87d2
FF
1246 *
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1250 *
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1253 *
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1260 *
88a0a948
VY
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association. On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code. The assoc_id
1264 * is not touched by the kernel.
3f7a87d2
FF
1265 *
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1268 *
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached. The way the SCTP stack uses the list of addresses to set up
25985edc 1273 * the association is implementation dependent. This function only
3f7a87d2
FF
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1276 *
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association. It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association. If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1282 *
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 *
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1293 *
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1295 * it.
1296 *
1297 * sk The sk of the socket
1298 * addrs The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1300 *
88a0a948 1301 * Returns >=0 if ok, <0 errno code on error.
3f7a87d2 1302 */
26ac8e5f 1303static int __sctp_setsockopt_connectx(struct sock *sk,
3f7a87d2 1304 struct sockaddr __user *addrs,
88a0a948
VY
1305 int addrs_size,
1306 sctp_assoc_t *assoc_id)
3f7a87d2 1307{
3f7a87d2 1308 struct sockaddr *kaddrs;
9ba0b963
MRL
1309 gfp_t gfp = GFP_KERNEL;
1310 int err = 0;
3f7a87d2 1311
bb33381d
DB
1312 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1313 __func__, sk, addrs, addrs_size);
3f7a87d2
FF
1314
1315 if (unlikely(addrs_size <= 0))
1316 return -EINVAL;
1317
1318 /* Check the user passed a healthy pointer. */
1319 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1320 return -EFAULT;
1321
1322 /* Alloc space for the address array in kernel memory. */
9ba0b963
MRL
1323 if (sk->sk_socket->file)
1324 gfp = GFP_USER | __GFP_NOWARN;
1325 kaddrs = kmalloc(addrs_size, gfp);
3f7a87d2
FF
1326 if (unlikely(!kaddrs))
1327 return -ENOMEM;
1328
1329 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1330 err = -EFAULT;
1331 } else {
88a0a948 1332 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
3f7a87d2
FF
1333 }
1334
1335 kfree(kaddrs);
88a0a948 1336
3f7a87d2
FF
1337 return err;
1338}
1339
88a0a948
VY
1340/*
1341 * This is an older interface. It's kept for backward compatibility
1342 * to the option that doesn't provide association id.
1343 */
26ac8e5f 1344static int sctp_setsockopt_connectx_old(struct sock *sk,
dda91928
DB
1345 struct sockaddr __user *addrs,
1346 int addrs_size)
88a0a948
VY
1347{
1348 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1349}
1350
1351/*
1352 * New interface for the API. The since the API is done with a socket
1353 * option, to make it simple we feed back the association id is as a return
1354 * indication to the call. Error is always negative and association id is
1355 * always positive.
1356 */
26ac8e5f 1357static int sctp_setsockopt_connectx(struct sock *sk,
dda91928
DB
1358 struct sockaddr __user *addrs,
1359 int addrs_size)
88a0a948
VY
1360{
1361 sctp_assoc_t assoc_id = 0;
1362 int err = 0;
1363
1364 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1365
1366 if (err)
1367 return err;
1368 else
1369 return assoc_id;
1370}
1371
c6ba68a2 1372/*
f9c67811
VY
1373 * New (hopefully final) interface for the API.
1374 * We use the sctp_getaddrs_old structure so that use-space library
ffd59393 1375 * can avoid any unnecessary allocations. The only different part
f9c67811 1376 * is that we store the actual length of the address buffer into the
ffd59393 1377 * addrs_num structure member. That way we can re-use the existing
f9c67811 1378 * code.
c6ba68a2 1379 */
ffd59393
DB
1380#ifdef CONFIG_COMPAT
1381struct compat_sctp_getaddrs_old {
1382 sctp_assoc_t assoc_id;
1383 s32 addr_num;
1384 compat_uptr_t addrs; /* struct sockaddr * */
1385};
1386#endif
1387
26ac8e5f 1388static int sctp_getsockopt_connectx3(struct sock *sk, int len,
dda91928
DB
1389 char __user *optval,
1390 int __user *optlen)
c6ba68a2 1391{
f9c67811 1392 struct sctp_getaddrs_old param;
c6ba68a2
VY
1393 sctp_assoc_t assoc_id = 0;
1394 int err = 0;
1395
ffd59393 1396#ifdef CONFIG_COMPAT
96c0e0a9 1397 if (in_compat_syscall()) {
ffd59393 1398 struct compat_sctp_getaddrs_old param32;
c6ba68a2 1399
ffd59393
DB
1400 if (len < sizeof(param32))
1401 return -EINVAL;
1402 if (copy_from_user(&param32, optval, sizeof(param32)))
1403 return -EFAULT;
f9c67811 1404
ffd59393
DB
1405 param.assoc_id = param32.assoc_id;
1406 param.addr_num = param32.addr_num;
1407 param.addrs = compat_ptr(param32.addrs);
1408 } else
1409#endif
1410 {
1411 if (len < sizeof(param))
1412 return -EINVAL;
1413 if (copy_from_user(&param, optval, sizeof(param)))
1414 return -EFAULT;
1415 }
c6ba68a2 1416
ffd59393
DB
1417 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1418 param.addrs, param.addr_num,
1419 &assoc_id);
c6ba68a2
VY
1420 if (err == 0 || err == -EINPROGRESS) {
1421 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1422 return -EFAULT;
1423 if (put_user(sizeof(assoc_id), optlen))
1424 return -EFAULT;
1425 }
1426
1427 return err;
1428}
1429
1da177e4
LT
1430/* API 3.1.4 close() - UDP Style Syntax
1431 * Applications use close() to perform graceful shutdown (as described in
1432 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1433 * by a UDP-style socket.
1434 *
1435 * The syntax is
1436 *
1437 * ret = close(int sd);
1438 *
1439 * sd - the socket descriptor of the associations to be closed.
1440 *
1441 * To gracefully shutdown a specific association represented by the
1442 * UDP-style socket, an application should use the sendmsg() call,
1443 * passing no user data, but including the appropriate flag in the
1444 * ancillary data (see Section xxxx).
1445 *
1446 * If sd in the close() call is a branched-off socket representing only
1447 * one association, the shutdown is performed on that association only.
1448 *
1449 * 4.1.6 close() - TCP Style Syntax
1450 *
1451 * Applications use close() to gracefully close down an association.
1452 *
1453 * The syntax is:
1454 *
1455 * int close(int sd);
1456 *
1457 * sd - the socket descriptor of the association to be closed.
1458 *
1459 * After an application calls close() on a socket descriptor, no further
1460 * socket operations will succeed on that descriptor.
1461 *
1462 * API 7.1.4 SO_LINGER
1463 *
1464 * An application using the TCP-style socket can use this option to
1465 * perform the SCTP ABORT primitive. The linger option structure is:
1466 *
1467 * struct linger {
1468 * int l_onoff; // option on/off
1469 * int l_linger; // linger time
1470 * };
1471 *
1472 * To enable the option, set l_onoff to 1. If the l_linger value is set
1473 * to 0, calling close() is the same as the ABORT primitive. If the
1474 * value is set to a negative value, the setsockopt() call will return
1475 * an error. If the value is set to a positive value linger_time, the
1476 * close() can be blocked for at most linger_time ms. If the graceful
1477 * shutdown phase does not finish during this period, close() will
1478 * return but the graceful shutdown phase continues in the system.
1479 */
dda91928 1480static void sctp_close(struct sock *sk, long timeout)
1da177e4 1481{
55e26eb9 1482 struct net *net = sock_net(sk);
1da177e4
LT
1483 struct sctp_endpoint *ep;
1484 struct sctp_association *asoc;
1485 struct list_head *pos, *temp;
cd4fcc70 1486 unsigned int data_was_unread;
1da177e4 1487
bb33381d 1488 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1da177e4 1489
048ed4b6 1490 lock_sock(sk);
1da177e4 1491 sk->sk_shutdown = SHUTDOWN_MASK;
bec9640b 1492 sk->sk_state = SCTP_SS_CLOSING;
1da177e4
LT
1493
1494 ep = sctp_sk(sk)->ep;
1495
cd4fcc70
TG
1496 /* Clean up any skbs sitting on the receive queue. */
1497 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1498 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1499
61c9fed4 1500 /* Walk all associations on an endpoint. */
1da177e4
LT
1501 list_for_each_safe(pos, temp, &ep->asocs) {
1502 asoc = list_entry(pos, struct sctp_association, asocs);
1503
1504 if (sctp_style(sk, TCP)) {
1505 /* A closed association can still be in the list if
1506 * it belongs to a TCP-style listening socket that is
1507 * not yet accepted. If so, free it. If not, send an
1508 * ABORT or SHUTDOWN based on the linger options.
1509 */
1510 if (sctp_state(asoc, CLOSED)) {
1da177e4 1511 sctp_association_free(asoc);
b89498a1
VY
1512 continue;
1513 }
1514 }
1da177e4 1515
cd4fcc70
TG
1516 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1517 !skb_queue_empty(&asoc->ulpq.reasm) ||
1518 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
b9ac8672
SS
1519 struct sctp_chunk *chunk;
1520
1521 chunk = sctp_make_abort_user(asoc, NULL, 0);
068d8bd3 1522 sctp_primitive_ABORT(net, asoc, chunk);
b9ac8672 1523 } else
55e26eb9 1524 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1da177e4
LT
1525 }
1526
1da177e4
LT
1527 /* On a TCP-style socket, block for at most linger_time if set. */
1528 if (sctp_style(sk, TCP) && timeout)
1529 sctp_wait_for_close(sk, timeout);
1530
1531 /* This will run the backlog queue. */
048ed4b6 1532 release_sock(sk);
1da177e4
LT
1533
1534 /* Supposedly, no process has access to the socket, but
1535 * the net layers still may.
2d45a02d
MRL
1536 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1537 * held and that should be grabbed before socket lock.
1da177e4 1538 */
2d45a02d 1539 spin_lock_bh(&net->sctp.addr_wq_lock);
5bc1d1b4 1540 bh_lock_sock(sk);
1da177e4
LT
1541
1542 /* Hold the sock, since sk_common_release() will put sock_put()
1543 * and we have just a little more cleanup.
1544 */
1545 sock_hold(sk);
1546 sk_common_release(sk);
1547
5bc1d1b4 1548 bh_unlock_sock(sk);
2d45a02d 1549 spin_unlock_bh(&net->sctp.addr_wq_lock);
1da177e4
LT
1550
1551 sock_put(sk);
1552
1553 SCTP_DBG_OBJCNT_DEC(sock);
1554}
1555
1556/* Handle EPIPE error. */
1557static int sctp_error(struct sock *sk, int flags, int err)
1558{
1559 if (err == -EPIPE)
1560 err = sock_error(sk) ? : -EPIPE;
1561 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1562 send_sig(SIGPIPE, current, 0);
1563 return err;
1564}
1565
1566/* API 3.1.3 sendmsg() - UDP Style Syntax
1567 *
1568 * An application uses sendmsg() and recvmsg() calls to transmit data to
1569 * and receive data from its peer.
1570 *
1571 * ssize_t sendmsg(int socket, const struct msghdr *message,
1572 * int flags);
1573 *
1574 * socket - the socket descriptor of the endpoint.
1575 * message - pointer to the msghdr structure which contains a single
1576 * user message and possibly some ancillary data.
1577 *
1578 * See Section 5 for complete description of the data
1579 * structures.
1580 *
1581 * flags - flags sent or received with the user message, see Section
1582 * 5 for complete description of the flags.
1583 *
1584 * Note: This function could use a rewrite especially when explicit
1585 * connect support comes in.
1586 */
1587/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1588
dda91928 1589static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1da177e4 1590
1b784140 1591static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1da177e4 1592{
55e26eb9 1593 struct net *net = sock_net(sk);
1da177e4
LT
1594 struct sctp_sock *sp;
1595 struct sctp_endpoint *ep;
cb3f837b 1596 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1da177e4
LT
1597 struct sctp_transport *transport, *chunk_tp;
1598 struct sctp_chunk *chunk;
dce116ae 1599 union sctp_addr to;
1da177e4 1600 struct sockaddr *msg_name = NULL;
517aa0bc 1601 struct sctp_sndrcvinfo default_sinfo;
1da177e4
LT
1602 struct sctp_sndrcvinfo *sinfo;
1603 struct sctp_initmsg *sinit;
1604 sctp_assoc_t associd = 0;
1605 sctp_cmsgs_t cmsgs = { NULL };
1da177e4 1606 sctp_scope_t scope;
2061dcd6 1607 bool fill_sinfo_ttl = false, wait_connect = false;
1da177e4 1608 struct sctp_datamsg *datamsg;
1da177e4 1609 int msg_flags = msg->msg_flags;
63b94938
GOV
1610 __u16 sinfo_flags = 0;
1611 long timeo;
1612 int err;
1da177e4 1613
1da177e4
LT
1614 err = 0;
1615 sp = sctp_sk(sk);
1616 ep = sp->ep;
1617
bb33381d
DB
1618 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1619 msg, msg_len, ep);
1da177e4
LT
1620
1621 /* We cannot send a message over a TCP-style listening socket. */
1622 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1623 err = -EPIPE;
1624 goto out_nounlock;
1625 }
1626
1627 /* Parse out the SCTP CMSGs. */
1628 err = sctp_msghdr_parse(msg, &cmsgs);
1da177e4 1629 if (err) {
bb33381d 1630 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1da177e4
LT
1631 goto out_nounlock;
1632 }
1633
1634 /* Fetch the destination address for this packet. This
1635 * address only selects the association--it is not necessarily
1636 * the address we will send to.
1637 * For a peeled-off socket, msg_name is ignored.
1638 */
1639 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1640 int msg_namelen = msg->msg_namelen;
1641
1642 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1643 msg_namelen);
1644 if (err)
1645 return err;
1646
1647 if (msg_namelen > sizeof(to))
1648 msg_namelen = sizeof(to);
1649 memcpy(&to, msg->msg_name, msg_namelen);
1da177e4
LT
1650 msg_name = msg->msg_name;
1651 }
1652
1da177e4 1653 sinit = cmsgs.init;
63b94938
GOV
1654 if (cmsgs.sinfo != NULL) {
1655 memset(&default_sinfo, 0, sizeof(default_sinfo));
1656 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1657 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1658 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1659 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1660 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1da177e4 1661
63b94938
GOV
1662 sinfo = &default_sinfo;
1663 fill_sinfo_ttl = true;
1664 } else {
1665 sinfo = cmsgs.srinfo;
1666 }
1667 /* Did the user specify SNDINFO/SNDRCVINFO? */
1da177e4
LT
1668 if (sinfo) {
1669 sinfo_flags = sinfo->sinfo_flags;
1670 associd = sinfo->sinfo_assoc_id;
1671 }
1672
bb33381d
DB
1673 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1674 msg_len, sinfo_flags);
1da177e4 1675
eaa5c54d
ISJ
1676 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1677 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1da177e4
LT
1678 err = -EINVAL;
1679 goto out_nounlock;
1680 }
1681
eaa5c54d
ISJ
1682 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1683 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1684 * If SCTP_ABORT is set, the message length could be non zero with
1da177e4 1685 * the msg_iov set to the user abort reason.
d808ad9a 1686 */
eaa5c54d
ISJ
1687 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1688 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1da177e4
LT
1689 err = -EINVAL;
1690 goto out_nounlock;
1691 }
1692
eaa5c54d 1693 /* If SCTP_ADDR_OVER is set, there must be an address
1da177e4
LT
1694 * specified in msg_name.
1695 */
eaa5c54d 1696 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1da177e4
LT
1697 err = -EINVAL;
1698 goto out_nounlock;
1699 }
1700
1701 transport = NULL;
1702
bb33381d 1703 pr_debug("%s: about to look up association\n", __func__);
1da177e4 1704
048ed4b6 1705 lock_sock(sk);
1da177e4
LT
1706
1707 /* If a msg_name has been specified, assume this is to be used. */
1708 if (msg_name) {
1709 /* Look for a matching association on the endpoint. */
dce116ae 1710 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
e5b13f34
MRL
1711
1712 /* If we could not find a matching association on the
1713 * endpoint, make sure that it is not a TCP-style
1714 * socket that already has an association or there is
1715 * no peeled-off association on another socket.
1716 */
1717 if (!asoc &&
1718 ((sctp_style(sk, TCP) &&
1719 (sctp_sstate(sk, ESTABLISHED) ||
1720 sctp_sstate(sk, CLOSING))) ||
1721 sctp_endpoint_is_peeled_off(ep, &to))) {
1722 err = -EADDRNOTAVAIL;
1723 goto out_unlock;
1da177e4
LT
1724 }
1725 } else {
1726 asoc = sctp_id2assoc(sk, associd);
1727 if (!asoc) {
1728 err = -EPIPE;
1729 goto out_unlock;
1730 }
1731 }
1732
1733 if (asoc) {
bb33381d 1734 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1da177e4
LT
1735
1736 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1737 * socket that has an association in CLOSED state. This can
1738 * happen when an accepted socket has an association that is
1739 * already CLOSED.
1740 */
1741 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1742 err = -EPIPE;
1743 goto out_unlock;
1744 }
1745
eaa5c54d 1746 if (sinfo_flags & SCTP_EOF) {
bb33381d
DB
1747 pr_debug("%s: shutting down association:%p\n",
1748 __func__, asoc);
1749
55e26eb9 1750 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1da177e4
LT
1751 err = 0;
1752 goto out_unlock;
1753 }
eaa5c54d 1754 if (sinfo_flags & SCTP_ABORT) {
c164a9ba
SS
1755
1756 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1757 if (!chunk) {
1758 err = -ENOMEM;
1759 goto out_unlock;
1760 }
1761
bb33381d
DB
1762 pr_debug("%s: aborting association:%p\n",
1763 __func__, asoc);
1764
55e26eb9 1765 sctp_primitive_ABORT(net, asoc, chunk);
1da177e4
LT
1766 err = 0;
1767 goto out_unlock;
1768 }
1769 }
1770
1771 /* Do we need to create the association? */
1772 if (!asoc) {
bb33381d 1773 pr_debug("%s: there is no association yet\n", __func__);
1da177e4 1774
eaa5c54d 1775 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1da177e4
LT
1776 err = -EINVAL;
1777 goto out_unlock;
1778 }
1779
1780 /* Check for invalid stream against the stream counts,
1781 * either the default or the user specified stream counts.
1782 */
1783 if (sinfo) {
0e864b21 1784 if (!sinit || !sinit->sinit_num_ostreams) {
1da177e4
LT
1785 /* Check against the defaults. */
1786 if (sinfo->sinfo_stream >=
1787 sp->initmsg.sinit_num_ostreams) {
1788 err = -EINVAL;
1789 goto out_unlock;
1790 }
1791 } else {
1792 /* Check against the requested. */
1793 if (sinfo->sinfo_stream >=
1794 sinit->sinit_num_ostreams) {
1795 err = -EINVAL;
1796 goto out_unlock;
1797 }
1798 }
1799 }
1800
1801 /*
1802 * API 3.1.2 bind() - UDP Style Syntax
1803 * If a bind() or sctp_bindx() is not called prior to a
1804 * sendmsg() call that initiates a new association, the
1805 * system picks an ephemeral port and will choose an address
1806 * set equivalent to binding with a wildcard address.
1807 */
1808 if (!ep->base.bind_addr.port) {
1809 if (sctp_autobind(sk)) {
1810 err = -EAGAIN;
1811 goto out_unlock;
1812 }
64a0c1c8
ISJ
1813 } else {
1814 /*
1815 * If an unprivileged user inherits a one-to-many
1816 * style socket with open associations on a privileged
1817 * port, it MAY be permitted to accept new associations,
1818 * but it SHOULD NOT be permitted to open new
1819 * associations.
1820 */
1821 if (ep->base.bind_addr.port < PROT_SOCK &&
3594698a 1822 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
64a0c1c8
ISJ
1823 err = -EACCES;
1824 goto out_unlock;
1825 }
1da177e4
LT
1826 }
1827
1828 scope = sctp_scope(&to);
1829 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1830 if (!new_asoc) {
1831 err = -ENOMEM;
1832 goto out_unlock;
1833 }
1834 asoc = new_asoc;
409b95af
VY
1835 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1836 if (err < 0) {
1837 err = -ENOMEM;
1838 goto out_free;
1839 }
1da177e4
LT
1840
1841 /* If the SCTP_INIT ancillary data is specified, set all
1842 * the association init values accordingly.
1843 */
1844 if (sinit) {
1845 if (sinit->sinit_num_ostreams) {
1846 asoc->c.sinit_num_ostreams =
1847 sinit->sinit_num_ostreams;
1848 }
1849 if (sinit->sinit_max_instreams) {
1850 asoc->c.sinit_max_instreams =
1851 sinit->sinit_max_instreams;
1852 }
1853 if (sinit->sinit_max_attempts) {
1854 asoc->max_init_attempts
1855 = sinit->sinit_max_attempts;
1856 }
1857 if (sinit->sinit_max_init_timeo) {
d808ad9a 1858 asoc->max_init_timeo =
1da177e4
LT
1859 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1860 }
1861 }
1862
1863 /* Prime the peer's transport structures. */
dce116ae 1864 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1da177e4
LT
1865 if (!transport) {
1866 err = -ENOMEM;
1867 goto out_free;
1868 }
1da177e4
LT
1869 }
1870
1871 /* ASSERT: we have a valid association at this point. */
bb33381d 1872 pr_debug("%s: we have a valid association\n", __func__);
1da177e4
LT
1873
1874 if (!sinfo) {
63b94938
GOV
1875 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1876 * one with some defaults.
1da177e4 1877 */
517aa0bc 1878 memset(&default_sinfo, 0, sizeof(default_sinfo));
1da177e4
LT
1879 default_sinfo.sinfo_stream = asoc->default_stream;
1880 default_sinfo.sinfo_flags = asoc->default_flags;
1881 default_sinfo.sinfo_ppid = asoc->default_ppid;
1882 default_sinfo.sinfo_context = asoc->default_context;
1883 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1884 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
63b94938 1885
1da177e4 1886 sinfo = &default_sinfo;
63b94938
GOV
1887 } else if (fill_sinfo_ttl) {
1888 /* In case SNDINFO was specified, we still need to fill
1889 * it with a default ttl from the assoc here.
1890 */
1891 sinfo->sinfo_timetolive = asoc->default_timetolive;
1da177e4
LT
1892 }
1893
1894 /* API 7.1.7, the sndbuf size per association bounds the
1895 * maximum size of data that can be sent in a single send call.
1896 */
1897 if (msg_len > sk->sk_sndbuf) {
1898 err = -EMSGSIZE;
1899 goto out_free;
1900 }
1901
8a479491 1902 if (asoc->pmtu_pending)
02f3d4ce 1903 sctp_assoc_pending_pmtu(sk, asoc);
8a479491 1904
1da177e4
LT
1905 /* If fragmentation is disabled and the message length exceeds the
1906 * association fragmentation point, return EMSGSIZE. The I-D
1907 * does not specify what this error is, but this looks like
1908 * a great fit.
1909 */
1910 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1911 err = -EMSGSIZE;
1912 goto out_free;
1913 }
1914
afd7614c
JP
1915 /* Check for invalid stream. */
1916 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1917 err = -EINVAL;
1918 goto out_free;
1da177e4
LT
1919 }
1920
8dbdf1f5
XL
1921 if (sctp_wspace(asoc) < msg_len)
1922 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1923
1da177e4
LT
1924 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1925 if (!sctp_wspace(asoc)) {
1926 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1927 if (err)
1928 goto out_free;
1929 }
1930
1931 /* If an address is passed with the sendto/sendmsg call, it is used
1932 * to override the primary destination address in the TCP model, or
eaa5c54d 1933 * when SCTP_ADDR_OVER flag is set in the UDP model.
1da177e4
LT
1934 */
1935 if ((sctp_style(sk, TCP) && msg_name) ||
eaa5c54d 1936 (sinfo_flags & SCTP_ADDR_OVER)) {
dce116ae 1937 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1da177e4
LT
1938 if (!chunk_tp) {
1939 err = -EINVAL;
1940 goto out_free;
1941 }
1942 } else
1943 chunk_tp = NULL;
1944
1945 /* Auto-connect, if we aren't connected already. */
1946 if (sctp_state(asoc, CLOSED)) {
55e26eb9 1947 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1da177e4
LT
1948 if (err < 0)
1949 goto out_free;
bb33381d 1950
2061dcd6 1951 wait_connect = true;
bb33381d 1952 pr_debug("%s: we associated primitively\n", __func__);
1da177e4
LT
1953 }
1954
1955 /* Break the message into multiple chunks of maximum size. */
c0371da6 1956 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
6e51fe75
TR
1957 if (IS_ERR(datamsg)) {
1958 err = PTR_ERR(datamsg);
1da177e4
LT
1959 goto out_free;
1960 }
1961
1962 /* Now send the (possibly) fragmented message. */
9dbc15f0 1963 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
b61c654f
XL
1964 sctp_chunk_hold(chunk);
1965
1da177e4
LT
1966 /* Do accounting for the write space. */
1967 sctp_set_owner_w(chunk);
1968
1969 chunk->transport = chunk_tp;
1da177e4
LT
1970 }
1971
9c5c62be
VY
1972 /* Send it to the lower layers. Note: all chunks
1973 * must either fail or succeed. The lower layer
1974 * works that way today. Keep it that way or this
1975 * breaks.
1976 */
55e26eb9 1977 err = sctp_primitive_SEND(net, asoc, datamsg);
9c5c62be 1978 /* Did the lower layer accept the chunk? */
b61c654f
XL
1979 if (err) {
1980 sctp_datamsg_free(datamsg);
bb33381d 1981 goto out_free;
b61c654f 1982 }
9c5c62be 1983
bb33381d 1984 pr_debug("%s: we sent primitively\n", __func__);
9c5c62be 1985
b61c654f 1986 sctp_datamsg_put(datamsg);
bb33381d 1987 err = msg_len;
1da177e4 1988
2061dcd6
DB
1989 if (unlikely(wait_connect)) {
1990 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1991 sctp_wait_for_connect(asoc, &timeo);
1992 }
1993
1da177e4
LT
1994 /* If we are already past ASSOCIATE, the lower
1995 * layers are responsible for association cleanup.
1996 */
1997 goto out_unlock;
1998
1999out_free:
b5eff712 2000 if (new_asoc)
1da177e4
LT
2001 sctp_association_free(asoc);
2002out_unlock:
048ed4b6 2003 release_sock(sk);
1da177e4
LT
2004
2005out_nounlock:
2006 return sctp_error(sk, msg_flags, err);
2007
2008#if 0
2009do_sock_err:
2010 if (msg_len)
2011 err = msg_len;
2012 else
2013 err = sock_error(sk);
2014 goto out;
2015
2016do_interrupted:
2017 if (msg_len)
2018 err = msg_len;
2019 goto out;
2020#endif /* 0 */
2021}
2022
2023/* This is an extended version of skb_pull() that removes the data from the
2024 * start of a skb even when data is spread across the list of skb's in the
2025 * frag_list. len specifies the total amount of data that needs to be removed.
2026 * when 'len' bytes could be removed from the skb, it returns 0.
2027 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2028 * could not be removed.
2029 */
2030static int sctp_skb_pull(struct sk_buff *skb, int len)
2031{
2032 struct sk_buff *list;
2033 int skb_len = skb_headlen(skb);
2034 int rlen;
2035
2036 if (len <= skb_len) {
2037 __skb_pull(skb, len);
2038 return 0;
2039 }
2040 len -= skb_len;
2041 __skb_pull(skb, skb_len);
2042
1b003be3 2043 skb_walk_frags(skb, list) {
1da177e4
LT
2044 rlen = sctp_skb_pull(list, len);
2045 skb->len -= (len-rlen);
2046 skb->data_len -= (len-rlen);
2047
2048 if (!rlen)
2049 return 0;
2050
2051 len = rlen;
2052 }
2053
2054 return len;
2055}
2056
2057/* API 3.1.3 recvmsg() - UDP Style Syntax
2058 *
2059 * ssize_t recvmsg(int socket, struct msghdr *message,
2060 * int flags);
2061 *
2062 * socket - the socket descriptor of the endpoint.
2063 * message - pointer to the msghdr structure which contains a single
2064 * user message and possibly some ancillary data.
2065 *
2066 * See Section 5 for complete description of the data
2067 * structures.
2068 *
2069 * flags - flags sent or received with the user message, see Section
2070 * 5 for complete description of the flags.
2071 */
1b784140
YX
2072static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2073 int noblock, int flags, int *addr_len)
1da177e4
LT
2074{
2075 struct sctp_ulpevent *event = NULL;
2076 struct sctp_sock *sp = sctp_sk(sk);
1f45f78f 2077 struct sk_buff *skb, *head_skb;
1da177e4
LT
2078 int copied;
2079 int err = 0;
2080 int skb_len;
2081
bb33381d
DB
2082 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2083 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2084 addr_len);
1da177e4 2085
048ed4b6 2086 lock_sock(sk);
1da177e4 2087
e5b13f34 2088 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
e0878694 2089 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
1da177e4
LT
2090 err = -ENOTCONN;
2091 goto out;
2092 }
2093
2094 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2095 if (!skb)
2096 goto out;
2097
2098 /* Get the total length of the skb including any skb's in the
2099 * frag_list.
2100 */
2101 skb_len = skb->len;
2102
2103 copied = skb_len;
2104 if (copied > len)
2105 copied = len;
2106
51f3d02b 2107 err = skb_copy_datagram_msg(skb, 0, msg, copied);
1da177e4
LT
2108
2109 event = sctp_skb2event(skb);
2110
2111 if (err)
2112 goto out_free;
2113
1f45f78f
MRL
2114 if (event->chunk && event->chunk->head_skb)
2115 head_skb = event->chunk->head_skb;
2116 else
2117 head_skb = skb;
2118 sock_recv_ts_and_drops(msg, sk, head_skb);
1da177e4
LT
2119 if (sctp_ulpevent_is_notification(event)) {
2120 msg->msg_flags |= MSG_NOTIFICATION;
2121 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2122 } else {
1f45f78f 2123 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
1da177e4
LT
2124 }
2125
2347c80f
GOV
2126 /* Check if we allow SCTP_NXTINFO. */
2127 if (sp->recvnxtinfo)
2128 sctp_ulpevent_read_nxtinfo(event, msg, sk);
0d3a421d
GOV
2129 /* Check if we allow SCTP_RCVINFO. */
2130 if (sp->recvrcvinfo)
2131 sctp_ulpevent_read_rcvinfo(event, msg);
1da177e4
LT
2132 /* Check if we allow SCTP_SNDRCVINFO. */
2133 if (sp->subscribe.sctp_data_io_event)
2134 sctp_ulpevent_read_sndrcvinfo(event, msg);
0d3a421d 2135
1da177e4
LT
2136 err = copied;
2137
2138 /* If skb's length exceeds the user's buffer, update the skb and
2139 * push it back to the receive_queue so that the next call to
2140 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2141 */
2142 if (skb_len > copied) {
2143 msg->msg_flags &= ~MSG_EOR;
2144 if (flags & MSG_PEEK)
2145 goto out_free;
2146 sctp_skb_pull(skb, copied);
2147 skb_queue_head(&sk->sk_receive_queue, skb);
2148
362d5204
DB
2149 /* When only partial message is copied to the user, increase
2150 * rwnd by that amount. If all the data in the skb is read,
2151 * rwnd is updated when the event is freed.
2152 */
2153 if (!sctp_ulpevent_is_notification(event))
2154 sctp_assoc_rwnd_increase(event->asoc, copied);
1da177e4
LT
2155 goto out;
2156 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2157 (event->msg_flags & MSG_EOR))
2158 msg->msg_flags |= MSG_EOR;
2159 else
2160 msg->msg_flags &= ~MSG_EOR;
2161
2162out_free:
2163 if (flags & MSG_PEEK) {
2164 /* Release the skb reference acquired after peeking the skb in
2165 * sctp_skb_recv_datagram().
2166 */
2167 kfree_skb(skb);
2168 } else {
2169 /* Free the event which includes releasing the reference to
2170 * the owner of the skb, freeing the skb and updating the
2171 * rwnd.
2172 */
2173 sctp_ulpevent_free(event);
2174 }
2175out:
048ed4b6 2176 release_sock(sk);
1da177e4
LT
2177 return err;
2178}
2179
2180/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2181 *
2182 * This option is a on/off flag. If enabled no SCTP message
2183 * fragmentation will be performed. Instead if a message being sent
2184 * exceeds the current PMTU size, the message will NOT be sent and
2185 * instead a error will be indicated to the user.
2186 */
2187static int sctp_setsockopt_disable_fragments(struct sock *sk,
b7058842
DM
2188 char __user *optval,
2189 unsigned int optlen)
1da177e4
LT
2190{
2191 int val;
2192
2193 if (optlen < sizeof(int))
2194 return -EINVAL;
2195
2196 if (get_user(val, (int __user *)optval))
2197 return -EFAULT;
2198
2199 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2200
2201 return 0;
2202}
2203
2204static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
b7058842 2205 unsigned int optlen)
1da177e4 2206{
94912301
WY
2207 struct sctp_association *asoc;
2208 struct sctp_ulpevent *event;
2209
7e8616d8 2210 if (optlen > sizeof(struct sctp_event_subscribe))
1da177e4
LT
2211 return -EINVAL;
2212 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2213 return -EFAULT;
94912301 2214
bbbea41d 2215 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
94912301
WY
2216 * if there is no data to be sent or retransmit, the stack will
2217 * immediately send up this notification.
2218 */
2219 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2220 &sctp_sk(sk)->subscribe)) {
2221 asoc = sctp_id2assoc(sk, 0);
2222
2223 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2224 event = sctp_ulpevent_make_sender_dry_event(asoc,
2225 GFP_ATOMIC);
2226 if (!event)
2227 return -ENOMEM;
2228
2229 sctp_ulpq_tail_event(&asoc->ulpq, event);
2230 }
2231 }
2232
1da177e4
LT
2233 return 0;
2234}
2235
2236/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2237 *
2238 * This socket option is applicable to the UDP-style socket only. When
2239 * set it will cause associations that are idle for more than the
2240 * specified number of seconds to automatically close. An association
2241 * being idle is defined an association that has NOT sent or received
2242 * user data. The special value of '0' indicates that no automatic
2243 * close of any associations should be performed. The option expects an
2244 * integer defining the number of seconds of idle time before an
2245 * association is closed.
2246 */
2247static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
b7058842 2248 unsigned int optlen)
1da177e4
LT
2249{
2250 struct sctp_sock *sp = sctp_sk(sk);
9f70f46b 2251 struct net *net = sock_net(sk);
1da177e4
LT
2252
2253 /* Applicable to UDP-style socket only */
2254 if (sctp_style(sk, TCP))
2255 return -EOPNOTSUPP;
2256 if (optlen != sizeof(int))
2257 return -EINVAL;
2258 if (copy_from_user(&sp->autoclose, optval, optlen))
2259 return -EFAULT;
2260
9f70f46b
NH
2261 if (sp->autoclose > net->sctp.max_autoclose)
2262 sp->autoclose = net->sctp.max_autoclose;
2263
1da177e4
LT
2264 return 0;
2265}
2266
2267/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2268 *
2269 * Applications can enable or disable heartbeats for any peer address of
2270 * an association, modify an address's heartbeat interval, force a
2271 * heartbeat to be sent immediately, and adjust the address's maximum
2272 * number of retransmissions sent before an address is considered
2273 * unreachable. The following structure is used to access and modify an
2274 * address's parameters:
2275 *
2276 * struct sctp_paddrparams {
52ccb8e9
FF
2277 * sctp_assoc_t spp_assoc_id;
2278 * struct sockaddr_storage spp_address;
2279 * uint32_t spp_hbinterval;
2280 * uint16_t spp_pathmaxrxt;
2281 * uint32_t spp_pathmtu;
2282 * uint32_t spp_sackdelay;
2283 * uint32_t spp_flags;
2284 * };
2285 *
2286 * spp_assoc_id - (one-to-many style socket) This is filled in the
2287 * application, and identifies the association for
2288 * this query.
1da177e4
LT
2289 * spp_address - This specifies which address is of interest.
2290 * spp_hbinterval - This contains the value of the heartbeat interval,
52ccb8e9
FF
2291 * in milliseconds. If a value of zero
2292 * is present in this field then no changes are to
2293 * be made to this parameter.
1da177e4
LT
2294 * spp_pathmaxrxt - This contains the maximum number of
2295 * retransmissions before this address shall be
52ccb8e9
FF
2296 * considered unreachable. If a value of zero
2297 * is present in this field then no changes are to
2298 * be made to this parameter.
2299 * spp_pathmtu - When Path MTU discovery is disabled the value
2300 * specified here will be the "fixed" path mtu.
2301 * Note that if the spp_address field is empty
2302 * then all associations on this address will
2303 * have this fixed path mtu set upon them.
2304 *
2305 * spp_sackdelay - When delayed sack is enabled, this value specifies
2306 * the number of milliseconds that sacks will be delayed
2307 * for. This value will apply to all addresses of an
2308 * association if the spp_address field is empty. Note
2309 * also, that if delayed sack is enabled and this
2310 * value is set to 0, no change is made to the last
2311 * recorded delayed sack timer value.
2312 *
2313 * spp_flags - These flags are used to control various features
2314 * on an association. The flag field may contain
2315 * zero or more of the following options.
2316 *
2317 * SPP_HB_ENABLE - Enable heartbeats on the
2318 * specified address. Note that if the address
2319 * field is empty all addresses for the association
2320 * have heartbeats enabled upon them.
2321 *
2322 * SPP_HB_DISABLE - Disable heartbeats on the
2323 * speicifed address. Note that if the address
2324 * field is empty all addresses for the association
2325 * will have their heartbeats disabled. Note also
2326 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2327 * mutually exclusive, only one of these two should
2328 * be specified. Enabling both fields will have
2329 * undetermined results.
2330 *
2331 * SPP_HB_DEMAND - Request a user initiated heartbeat
2332 * to be made immediately.
2333 *
bdf3092a
VY
2334 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2335 * heartbeat delayis to be set to the value of 0
2336 * milliseconds.
2337 *
52ccb8e9
FF
2338 * SPP_PMTUD_ENABLE - This field will enable PMTU
2339 * discovery upon the specified address. Note that
2340 * if the address feild is empty then all addresses
2341 * on the association are effected.
2342 *
2343 * SPP_PMTUD_DISABLE - This field will disable PMTU
2344 * discovery upon the specified address. Note that
2345 * if the address feild is empty then all addresses
2346 * on the association are effected. Not also that
2347 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2348 * exclusive. Enabling both will have undetermined
2349 * results.
2350 *
2351 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2352 * on delayed sack. The time specified in spp_sackdelay
2353 * is used to specify the sack delay for this address. Note
2354 * that if spp_address is empty then all addresses will
2355 * enable delayed sack and take on the sack delay
2356 * value specified in spp_sackdelay.
2357 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2358 * off delayed sack. If the spp_address field is blank then
2359 * delayed sack is disabled for the entire association. Note
2360 * also that this field is mutually exclusive to
2361 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2362 * results.
1da177e4 2363 */
16164366
AB
2364static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2365 struct sctp_transport *trans,
2366 struct sctp_association *asoc,
2367 struct sctp_sock *sp,
2368 int hb_change,
2369 int pmtud_change,
2370 int sackdelay_change)
52ccb8e9
FF
2371{
2372 int error;
2373
2374 if (params->spp_flags & SPP_HB_DEMAND && trans) {
55e26eb9
EB
2375 struct net *net = sock_net(trans->asoc->base.sk);
2376
2377 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
52ccb8e9
FF
2378 if (error)
2379 return error;
2380 }
2381
bdf3092a
VY
2382 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2383 * this field is ignored. Note also that a value of zero indicates
2384 * the current setting should be left unchanged.
2385 */
2386 if (params->spp_flags & SPP_HB_ENABLE) {
2387
2388 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2389 * set. This lets us use 0 value when this flag
2390 * is set.
2391 */
2392 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2393 params->spp_hbinterval = 0;
2394
2395 if (params->spp_hbinterval ||
2396 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2397 if (trans) {
2398 trans->hbinterval =
2399 msecs_to_jiffies(params->spp_hbinterval);
2400 } else if (asoc) {
2401 asoc->hbinterval =
2402 msecs_to_jiffies(params->spp_hbinterval);
2403 } else {
2404 sp->hbinterval = params->spp_hbinterval;
2405 }
52ccb8e9
FF
2406 }
2407 }
2408
2409 if (hb_change) {
2410 if (trans) {
2411 trans->param_flags =
2412 (trans->param_flags & ~SPP_HB) | hb_change;
2413 } else if (asoc) {
2414 asoc->param_flags =
2415 (asoc->param_flags & ~SPP_HB) | hb_change;
2416 } else {
2417 sp->param_flags =
2418 (sp->param_flags & ~SPP_HB) | hb_change;
2419 }
2420 }
2421
bdf3092a
VY
2422 /* When Path MTU discovery is disabled the value specified here will
2423 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2424 * include the flag SPP_PMTUD_DISABLE for this field to have any
2425 * effect).
2426 */
2427 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
52ccb8e9
FF
2428 if (trans) {
2429 trans->pathmtu = params->spp_pathmtu;
02f3d4ce 2430 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
52ccb8e9
FF
2431 } else if (asoc) {
2432 asoc->pathmtu = params->spp_pathmtu;
f68b2e05 2433 sctp_frag_point(asoc, params->spp_pathmtu);
52ccb8e9
FF
2434 } else {
2435 sp->pathmtu = params->spp_pathmtu;
2436 }
2437 }
2438
2439 if (pmtud_change) {
2440 if (trans) {
2441 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2442 (params->spp_flags & SPP_PMTUD_ENABLE);
2443 trans->param_flags =
2444 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2445 if (update) {
9914ae3c 2446 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
02f3d4ce 2447 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
52ccb8e9
FF
2448 }
2449 } else if (asoc) {
2450 asoc->param_flags =
2451 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2452 } else {
2453 sp->param_flags =
2454 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2455 }
2456 }
2457
bdf3092a
VY
2458 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2459 * value of this field is ignored. Note also that a value of zero
2460 * indicates the current setting should be left unchanged.
2461 */
2462 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
52ccb8e9
FF
2463 if (trans) {
2464 trans->sackdelay =
2465 msecs_to_jiffies(params->spp_sackdelay);
2466 } else if (asoc) {
2467 asoc->sackdelay =
2468 msecs_to_jiffies(params->spp_sackdelay);
2469 } else {
2470 sp->sackdelay = params->spp_sackdelay;
2471 }
2472 }
2473
2474 if (sackdelay_change) {
2475 if (trans) {
2476 trans->param_flags =
2477 (trans->param_flags & ~SPP_SACKDELAY) |
2478 sackdelay_change;
2479 } else if (asoc) {
2480 asoc->param_flags =
2481 (asoc->param_flags & ~SPP_SACKDELAY) |
2482 sackdelay_change;
2483 } else {
2484 sp->param_flags =
2485 (sp->param_flags & ~SPP_SACKDELAY) |
2486 sackdelay_change;
2487 }
2488 }
2489
37051f73
APO
2490 /* Note that a value of zero indicates the current setting should be
2491 left unchanged.
bdf3092a 2492 */
37051f73 2493 if (params->spp_pathmaxrxt) {
52ccb8e9
FF
2494 if (trans) {
2495 trans->pathmaxrxt = params->spp_pathmaxrxt;
2496 } else if (asoc) {
2497 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2498 } else {
2499 sp->pathmaxrxt = params->spp_pathmaxrxt;
2500 }
2501 }
2502
2503 return 0;
2504}
2505
1da177e4 2506static int sctp_setsockopt_peer_addr_params(struct sock *sk,
b7058842
DM
2507 char __user *optval,
2508 unsigned int optlen)
1da177e4 2509{
52ccb8e9
FF
2510 struct sctp_paddrparams params;
2511 struct sctp_transport *trans = NULL;
2512 struct sctp_association *asoc = NULL;
2513 struct sctp_sock *sp = sctp_sk(sk);
1da177e4 2514 int error;
52ccb8e9 2515 int hb_change, pmtud_change, sackdelay_change;
1da177e4
LT
2516
2517 if (optlen != sizeof(struct sctp_paddrparams))
cb3f837b 2518 return -EINVAL;
52ccb8e9 2519
1da177e4
LT
2520 if (copy_from_user(&params, optval, optlen))
2521 return -EFAULT;
2522
52ccb8e9
FF
2523 /* Validate flags and value parameters. */
2524 hb_change = params.spp_flags & SPP_HB;
2525 pmtud_change = params.spp_flags & SPP_PMTUD;
2526 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2527
2528 if (hb_change == SPP_HB ||
2529 pmtud_change == SPP_PMTUD ||
2530 sackdelay_change == SPP_SACKDELAY ||
2531 params.spp_sackdelay > 500 ||
f64f9e71
JP
2532 (params.spp_pathmtu &&
2533 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
52ccb8e9 2534 return -EINVAL;
1da177e4 2535
52ccb8e9
FF
2536 /* If an address other than INADDR_ANY is specified, and
2537 * no transport is found, then the request is invalid.
2538 */
cb3f837b 2539 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
52ccb8e9
FF
2540 trans = sctp_addr_id2transport(sk, &params.spp_address,
2541 params.spp_assoc_id);
2542 if (!trans)
1da177e4 2543 return -EINVAL;
1da177e4
LT
2544 }
2545
52ccb8e9
FF
2546 /* Get association, if assoc_id != 0 and the socket is a one
2547 * to many style socket, and an association was not found, then
2548 * the id was invalid.
2549 */
2550 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2551 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
1da177e4
LT
2552 return -EINVAL;
2553
52ccb8e9
FF
2554 /* Heartbeat demand can only be sent on a transport or
2555 * association, but not a socket.
1da177e4 2556 */
52ccb8e9
FF
2557 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2558 return -EINVAL;
2559
2560 /* Process parameters. */
2561 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2562 hb_change, pmtud_change,
2563 sackdelay_change);
1da177e4 2564
52ccb8e9
FF
2565 if (error)
2566 return error;
2567
2568 /* If changes are for association, also apply parameters to each
2569 * transport.
1da177e4 2570 */
52ccb8e9 2571 if (!trans && asoc) {
9dbc15f0
RD
2572 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2573 transports) {
52ccb8e9
FF
2574 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2575 hb_change, pmtud_change,
2576 sackdelay_change);
2577 }
2578 }
1da177e4
LT
2579
2580 return 0;
2581}
2582
0ea5e4df 2583static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2584{
2585 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2586}
2587
2588static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2589{
2590 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2591}
2592
d364d927
WY
2593/*
2594 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2595 *
2596 * This option will effect the way delayed acks are performed. This
2597 * option allows you to get or set the delayed ack time, in
2598 * milliseconds. It also allows changing the delayed ack frequency.
2599 * Changing the frequency to 1 disables the delayed sack algorithm. If
2600 * the assoc_id is 0, then this sets or gets the endpoints default
2601 * values. If the assoc_id field is non-zero, then the set or get
2602 * effects the specified association for the one to many model (the
2603 * assoc_id field is ignored by the one to one model). Note that if
2604 * sack_delay or sack_freq are 0 when setting this option, then the
2605 * current values will remain unchanged.
2606 *
2607 * struct sctp_sack_info {
2608 * sctp_assoc_t sack_assoc_id;
2609 * uint32_t sack_delay;
2610 * uint32_t sack_freq;
2611 * };
2612 *
2613 * sack_assoc_id - This parameter, indicates which association the user
2614 * is performing an action upon. Note that if this field's value is
2615 * zero then the endpoints default value is changed (effecting future
2616 * associations only).
2617 *
2618 * sack_delay - This parameter contains the number of milliseconds that
2619 * the user is requesting the delayed ACK timer be set to. Note that
2620 * this value is defined in the standard to be between 200 and 500
2621 * milliseconds.
2622 *
2623 * sack_freq - This parameter contains the number of packets that must
2624 * be received before a sack is sent without waiting for the delay
2625 * timer to expire. The default value for this is 2, setting this
2626 * value to 1 will disable the delayed sack algorithm.
7708610b
FF
2627 */
2628
d364d927 2629static int sctp_setsockopt_delayed_ack(struct sock *sk,
b7058842 2630 char __user *optval, unsigned int optlen)
7708610b 2631{
d364d927 2632 struct sctp_sack_info params;
7708610b
FF
2633 struct sctp_transport *trans = NULL;
2634 struct sctp_association *asoc = NULL;
2635 struct sctp_sock *sp = sctp_sk(sk);
2636
d364d927
WY
2637 if (optlen == sizeof(struct sctp_sack_info)) {
2638 if (copy_from_user(&params, optval, optlen))
2639 return -EFAULT;
7708610b 2640
d364d927
WY
2641 if (params.sack_delay == 0 && params.sack_freq == 0)
2642 return 0;
2643 } else if (optlen == sizeof(struct sctp_assoc_value)) {
94f65193 2644 pr_warn_ratelimited(DEPRECATED
f916ec96 2645 "%s (pid %d) "
94f65193 2646 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
f916ec96
NH
2647 "Use struct sctp_sack_info instead\n",
2648 current->comm, task_pid_nr(current));
d364d927
WY
2649 if (copy_from_user(&params, optval, optlen))
2650 return -EFAULT;
2651
2652 if (params.sack_delay == 0)
2653 params.sack_freq = 1;
2654 else
2655 params.sack_freq = 0;
2656 } else
cb3f837b 2657 return -EINVAL;
7708610b
FF
2658
2659 /* Validate value parameter. */
d364d927 2660 if (params.sack_delay > 500)
7708610b
FF
2661 return -EINVAL;
2662
d364d927 2663 /* Get association, if sack_assoc_id != 0 and the socket is a one
7708610b
FF
2664 * to many style socket, and an association was not found, then
2665 * the id was invalid.
d808ad9a 2666 */
d364d927
WY
2667 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2668 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
7708610b
FF
2669 return -EINVAL;
2670
d364d927 2671 if (params.sack_delay) {
7708610b
FF
2672 if (asoc) {
2673 asoc->sackdelay =
d364d927 2674 msecs_to_jiffies(params.sack_delay);
d808ad9a 2675 asoc->param_flags =
0ea5e4df 2676 sctp_spp_sackdelay_enable(asoc->param_flags);
7708610b 2677 } else {
d364d927 2678 sp->sackdelay = params.sack_delay;
d808ad9a 2679 sp->param_flags =
0ea5e4df 2680 sctp_spp_sackdelay_enable(sp->param_flags);
7708610b 2681 }
d364d927
WY
2682 }
2683
2684 if (params.sack_freq == 1) {
7708610b 2685 if (asoc) {
d808ad9a 2686 asoc->param_flags =
0ea5e4df 2687 sctp_spp_sackdelay_disable(asoc->param_flags);
7708610b 2688 } else {
d808ad9a 2689 sp->param_flags =
0ea5e4df 2690 sctp_spp_sackdelay_disable(sp->param_flags);
7708610b 2691 }
d364d927
WY
2692 } else if (params.sack_freq > 1) {
2693 if (asoc) {
2694 asoc->sackfreq = params.sack_freq;
2695 asoc->param_flags =
0ea5e4df 2696 sctp_spp_sackdelay_enable(asoc->param_flags);
d364d927
WY
2697 } else {
2698 sp->sackfreq = params.sack_freq;
2699 sp->param_flags =
0ea5e4df 2700 sctp_spp_sackdelay_enable(sp->param_flags);
d364d927 2701 }
7708610b
FF
2702 }
2703
2704 /* If change is for association, also apply to each transport. */
2705 if (asoc) {
9dbc15f0
RD
2706 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2707 transports) {
d364d927 2708 if (params.sack_delay) {
7708610b 2709 trans->sackdelay =
d364d927 2710 msecs_to_jiffies(params.sack_delay);
d808ad9a 2711 trans->param_flags =
0ea5e4df 2712 sctp_spp_sackdelay_enable(trans->param_flags);
d364d927 2713 }
7bfe8bdb 2714 if (params.sack_freq == 1) {
d808ad9a 2715 trans->param_flags =
0ea5e4df 2716 sctp_spp_sackdelay_disable(trans->param_flags);
d364d927
WY
2717 } else if (params.sack_freq > 1) {
2718 trans->sackfreq = params.sack_freq;
2719 trans->param_flags =
0ea5e4df 2720 sctp_spp_sackdelay_enable(trans->param_flags);
7708610b
FF
2721 }
2722 }
2723 }
d808ad9a 2724
7708610b
FF
2725 return 0;
2726}
2727
1da177e4
LT
2728/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2729 *
2730 * Applications can specify protocol parameters for the default association
2731 * initialization. The option name argument to setsockopt() and getsockopt()
2732 * is SCTP_INITMSG.
2733 *
2734 * Setting initialization parameters is effective only on an unconnected
2735 * socket (for UDP-style sockets only future associations are effected
2736 * by the change). With TCP-style sockets, this option is inherited by
2737 * sockets derived from a listener socket.
2738 */
b7058842 2739static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
2740{
2741 struct sctp_initmsg sinit;
2742 struct sctp_sock *sp = sctp_sk(sk);
2743
2744 if (optlen != sizeof(struct sctp_initmsg))
2745 return -EINVAL;
2746 if (copy_from_user(&sinit, optval, optlen))
2747 return -EFAULT;
2748
2749 if (sinit.sinit_num_ostreams)
d808ad9a 2750 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
1da177e4 2751 if (sinit.sinit_max_instreams)
d808ad9a 2752 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
1da177e4 2753 if (sinit.sinit_max_attempts)
d808ad9a 2754 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
1da177e4 2755 if (sinit.sinit_max_init_timeo)
d808ad9a 2756 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
1da177e4
LT
2757
2758 return 0;
2759}
2760
2761/*
2762 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2763 *
2764 * Applications that wish to use the sendto() system call may wish to
2765 * specify a default set of parameters that would normally be supplied
2766 * through the inclusion of ancillary data. This socket option allows
2767 * such an application to set the default sctp_sndrcvinfo structure.
2768 * The application that wishes to use this socket option simply passes
2769 * in to this call the sctp_sndrcvinfo structure defined in Section
2770 * 5.2.2) The input parameters accepted by this call include
2771 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2772 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2773 * to this call if the caller is using the UDP model.
2774 */
2775static int sctp_setsockopt_default_send_param(struct sock *sk,
b7058842
DM
2776 char __user *optval,
2777 unsigned int optlen)
1da177e4 2778{
1da177e4 2779 struct sctp_sock *sp = sctp_sk(sk);
6b3fd5f3
GOV
2780 struct sctp_association *asoc;
2781 struct sctp_sndrcvinfo info;
1da177e4 2782
6b3fd5f3 2783 if (optlen != sizeof(info))
1da177e4
LT
2784 return -EINVAL;
2785 if (copy_from_user(&info, optval, optlen))
2786 return -EFAULT;
6b3fd5f3
GOV
2787 if (info.sinfo_flags &
2788 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2789 SCTP_ABORT | SCTP_EOF))
2790 return -EINVAL;
1da177e4
LT
2791
2792 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2793 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2794 return -EINVAL;
1da177e4
LT
2795 if (asoc) {
2796 asoc->default_stream = info.sinfo_stream;
2797 asoc->default_flags = info.sinfo_flags;
2798 asoc->default_ppid = info.sinfo_ppid;
2799 asoc->default_context = info.sinfo_context;
2800 asoc->default_timetolive = info.sinfo_timetolive;
2801 } else {
2802 sp->default_stream = info.sinfo_stream;
2803 sp->default_flags = info.sinfo_flags;
2804 sp->default_ppid = info.sinfo_ppid;
2805 sp->default_context = info.sinfo_context;
2806 sp->default_timetolive = info.sinfo_timetolive;
2807 }
2808
2809 return 0;
2810}
2811
6b3fd5f3
GOV
2812/* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2813 * (SCTP_DEFAULT_SNDINFO)
2814 */
2815static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2816 char __user *optval,
2817 unsigned int optlen)
2818{
2819 struct sctp_sock *sp = sctp_sk(sk);
2820 struct sctp_association *asoc;
2821 struct sctp_sndinfo info;
2822
2823 if (optlen != sizeof(info))
2824 return -EINVAL;
2825 if (copy_from_user(&info, optval, optlen))
2826 return -EFAULT;
2827 if (info.snd_flags &
2828 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2829 SCTP_ABORT | SCTP_EOF))
2830 return -EINVAL;
2831
2832 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2833 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2834 return -EINVAL;
2835 if (asoc) {
2836 asoc->default_stream = info.snd_sid;
2837 asoc->default_flags = info.snd_flags;
2838 asoc->default_ppid = info.snd_ppid;
2839 asoc->default_context = info.snd_context;
2840 } else {
2841 sp->default_stream = info.snd_sid;
2842 sp->default_flags = info.snd_flags;
2843 sp->default_ppid = info.snd_ppid;
2844 sp->default_context = info.snd_context;
2845 }
2846
2847 return 0;
2848}
2849
1da177e4
LT
2850/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2851 *
2852 * Requests that the local SCTP stack use the enclosed peer address as
2853 * the association primary. The enclosed address must be one of the
2854 * association peer's addresses.
2855 */
2856static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
b7058842 2857 unsigned int optlen)
1da177e4
LT
2858{
2859 struct sctp_prim prim;
2860 struct sctp_transport *trans;
2861
2862 if (optlen != sizeof(struct sctp_prim))
2863 return -EINVAL;
2864
2865 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2866 return -EFAULT;
2867
2868 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2869 if (!trans)
2870 return -EINVAL;
2871
2872 sctp_assoc_set_primary(trans->asoc, trans);
2873
2874 return 0;
2875}
2876
2877/*
2878 * 7.1.5 SCTP_NODELAY
2879 *
2880 * Turn on/off any Nagle-like algorithm. This means that packets are
2881 * generally sent as soon as possible and no unnecessary delays are
2882 * introduced, at the cost of more packets in the network. Expects an
2883 * integer boolean flag.
2884 */
2885static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
b7058842 2886 unsigned int optlen)
1da177e4
LT
2887{
2888 int val;
2889
2890 if (optlen < sizeof(int))
2891 return -EINVAL;
2892 if (get_user(val, (int __user *)optval))
2893 return -EFAULT;
2894
2895 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2896 return 0;
2897}
2898
2899/*
2900 *
2901 * 7.1.1 SCTP_RTOINFO
2902 *
2903 * The protocol parameters used to initialize and bound retransmission
2904 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2905 * and modify these parameters.
2906 * All parameters are time values, in milliseconds. A value of 0, when
2907 * modifying the parameters, indicates that the current value should not
2908 * be changed.
2909 *
2910 */
b7058842
DM
2911static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2912{
1da177e4
LT
2913 struct sctp_rtoinfo rtoinfo;
2914 struct sctp_association *asoc;
85f935d4 2915 unsigned long rto_min, rto_max;
2916 struct sctp_sock *sp = sctp_sk(sk);
1da177e4
LT
2917
2918 if (optlen != sizeof (struct sctp_rtoinfo))
2919 return -EINVAL;
2920
2921 if (copy_from_user(&rtoinfo, optval, optlen))
2922 return -EFAULT;
2923
2924 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2925
2926 /* Set the values to the specific association */
2927 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2928 return -EINVAL;
2929
85f935d4 2930 rto_max = rtoinfo.srto_max;
2931 rto_min = rtoinfo.srto_min;
2932
2933 if (rto_max)
2934 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2935 else
2936 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2937
2938 if (rto_min)
2939 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2940 else
2941 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2942
2943 if (rto_min > rto_max)
2944 return -EINVAL;
2945
1da177e4
LT
2946 if (asoc) {
2947 if (rtoinfo.srto_initial != 0)
d808ad9a 2948 asoc->rto_initial =
1da177e4 2949 msecs_to_jiffies(rtoinfo.srto_initial);
85f935d4 2950 asoc->rto_max = rto_max;
2951 asoc->rto_min = rto_min;
1da177e4
LT
2952 } else {
2953 /* If there is no association or the association-id = 0
2954 * set the values to the endpoint.
2955 */
1da177e4
LT
2956 if (rtoinfo.srto_initial != 0)
2957 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
85f935d4 2958 sp->rtoinfo.srto_max = rto_max;
2959 sp->rtoinfo.srto_min = rto_min;
1da177e4
LT
2960 }
2961
2962 return 0;
2963}
2964
2965/*
2966 *
2967 * 7.1.2 SCTP_ASSOCINFO
2968 *
59c51591 2969 * This option is used to tune the maximum retransmission attempts
1da177e4
LT
2970 * of the association.
2971 * Returns an error if the new association retransmission value is
2972 * greater than the sum of the retransmission value of the peer.
2973 * See [SCTP] for more information.
2974 *
2975 */
b7058842 2976static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
2977{
2978
2979 struct sctp_assocparams assocparams;
2980 struct sctp_association *asoc;
2981
2982 if (optlen != sizeof(struct sctp_assocparams))
2983 return -EINVAL;
2984 if (copy_from_user(&assocparams, optval, optlen))
2985 return -EFAULT;
2986
2987 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2988
2989 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2990 return -EINVAL;
2991
2992 /* Set the values to the specific association */
2993 if (asoc) {
402d68c4
VY
2994 if (assocparams.sasoc_asocmaxrxt != 0) {
2995 __u32 path_sum = 0;
2996 int paths = 0;
402d68c4
VY
2997 struct sctp_transport *peer_addr;
2998
9dbc15f0
RD
2999 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3000 transports) {
402d68c4
VY
3001 path_sum += peer_addr->pathmaxrxt;
3002 paths++;
3003 }
3004
025dfdaf 3005 /* Only validate asocmaxrxt if we have more than
402d68c4
VY
3006 * one path/transport. We do this because path
3007 * retransmissions are only counted when we have more
3008 * then one path.
3009 */
3010 if (paths > 1 &&
3011 assocparams.sasoc_asocmaxrxt > path_sum)
3012 return -EINVAL;
3013
1da177e4 3014 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
402d68c4
VY
3015 }
3016
52db882f
DB
3017 if (assocparams.sasoc_cookie_life != 0)
3018 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
1da177e4
LT
3019 } else {
3020 /* Set the values to the endpoint */
3021 struct sctp_sock *sp = sctp_sk(sk);
3022
3023 if (assocparams.sasoc_asocmaxrxt != 0)
3024 sp->assocparams.sasoc_asocmaxrxt =
3025 assocparams.sasoc_asocmaxrxt;
3026 if (assocparams.sasoc_cookie_life != 0)
3027 sp->assocparams.sasoc_cookie_life =
3028 assocparams.sasoc_cookie_life;
3029 }
3030 return 0;
3031}
3032
3033/*
3034 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3035 *
3036 * This socket option is a boolean flag which turns on or off mapped V4
3037 * addresses. If this option is turned on and the socket is type
3038 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3039 * If this option is turned off, then no mapping will be done of V4
3040 * addresses and a user will receive both PF_INET6 and PF_INET type
3041 * addresses on the socket.
3042 */
b7058842 3043static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4
LT
3044{
3045 int val;
3046 struct sctp_sock *sp = sctp_sk(sk);
3047
3048 if (optlen < sizeof(int))
3049 return -EINVAL;
3050 if (get_user(val, (int __user *)optval))
3051 return -EFAULT;
3052 if (val)
3053 sp->v4mapped = 1;
3054 else
3055 sp->v4mapped = 0;
3056
3057 return 0;
3058}
3059
3060/*
e89c2095
WY
3061 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3062 * This option will get or set the maximum size to put in any outgoing
3063 * SCTP DATA chunk. If a message is larger than this size it will be
1da177e4
LT
3064 * fragmented by SCTP into the specified size. Note that the underlying
3065 * SCTP implementation may fragment into smaller sized chunks when the
3066 * PMTU of the underlying association is smaller than the value set by
e89c2095
WY
3067 * the user. The default value for this option is '0' which indicates
3068 * the user is NOT limiting fragmentation and only the PMTU will effect
3069 * SCTP's choice of DATA chunk size. Note also that values set larger
3070 * than the maximum size of an IP datagram will effectively let SCTP
3071 * control fragmentation (i.e. the same as setting this option to 0).
3072 *
3073 * The following structure is used to access and modify this parameter:
3074 *
3075 * struct sctp_assoc_value {
3076 * sctp_assoc_t assoc_id;
3077 * uint32_t assoc_value;
3078 * };
3079 *
3080 * assoc_id: This parameter is ignored for one-to-one style sockets.
3081 * For one-to-many style sockets this parameter indicates which
3082 * association the user is performing an action upon. Note that if
3083 * this field's value is zero then the endpoints default value is
3084 * changed (effecting future associations only).
3085 * assoc_value: This parameter specifies the maximum size in bytes.
1da177e4 3086 */
b7058842 3087static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
1da177e4 3088{
e89c2095 3089 struct sctp_assoc_value params;
1da177e4 3090 struct sctp_association *asoc;
1da177e4
LT
3091 struct sctp_sock *sp = sctp_sk(sk);
3092 int val;
3093
e89c2095 3094 if (optlen == sizeof(int)) {
94f65193 3095 pr_warn_ratelimited(DEPRECATED
f916ec96 3096 "%s (pid %d) "
94f65193 3097 "Use of int in maxseg socket option.\n"
f916ec96
NH
3098 "Use struct sctp_assoc_value instead\n",
3099 current->comm, task_pid_nr(current));
e89c2095
WY
3100 if (copy_from_user(&val, optval, optlen))
3101 return -EFAULT;
3102 params.assoc_id = 0;
3103 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3104 if (copy_from_user(&params, optval, optlen))
3105 return -EFAULT;
3106 val = params.assoc_value;
3107 } else
1da177e4 3108 return -EINVAL;
e89c2095 3109
96a33998 3110 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
1da177e4 3111 return -EINVAL;
1da177e4 3112
e89c2095
WY
3113 asoc = sctp_id2assoc(sk, params.assoc_id);
3114 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3115 return -EINVAL;
3116
3117 if (asoc) {
3118 if (val == 0) {
3119 val = asoc->pathmtu;
3120 val -= sp->pf->af->net_header_len;
3121 val -= sizeof(struct sctphdr) +
3122 sizeof(struct sctp_data_chunk);
3123 }
f68b2e05
VY
3124 asoc->user_frag = val;
3125 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
e89c2095
WY
3126 } else {
3127 sp->user_frag = val;
1da177e4
LT
3128 }
3129
3130 return 0;
3131}
3132
3133
3134/*
3135 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3136 *
3137 * Requests that the peer mark the enclosed address as the association
3138 * primary. The enclosed address must be one of the association's
3139 * locally bound addresses. The following structure is used to make a
3140 * set primary request:
3141 */
3142static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
b7058842 3143 unsigned int optlen)
1da177e4 3144{
e1fc3b14 3145 struct net *net = sock_net(sk);
1da177e4 3146 struct sctp_sock *sp;
1da177e4
LT
3147 struct sctp_association *asoc = NULL;
3148 struct sctp_setpeerprim prim;
3149 struct sctp_chunk *chunk;
40a01039 3150 struct sctp_af *af;
1da177e4
LT
3151 int err;
3152
3153 sp = sctp_sk(sk);
1da177e4 3154
e1fc3b14 3155 if (!net->sctp.addip_enable)
1da177e4
LT
3156 return -EPERM;
3157
3158 if (optlen != sizeof(struct sctp_setpeerprim))
3159 return -EINVAL;
3160
3161 if (copy_from_user(&prim, optval, optlen))
3162 return -EFAULT;
3163
3164 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
d808ad9a 3165 if (!asoc)
1da177e4
LT
3166 return -EINVAL;
3167
3168 if (!asoc->peer.asconf_capable)
3169 return -EPERM;
3170
3171 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3172 return -EPERM;
3173
3174 if (!sctp_state(asoc, ESTABLISHED))
3175 return -ENOTCONN;
3176
40a01039
WY
3177 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3178 if (!af)
3179 return -EINVAL;
3180
3181 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3182 return -EADDRNOTAVAIL;
3183
1da177e4
LT
3184 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3185 return -EADDRNOTAVAIL;
3186
3187 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3188 chunk = sctp_make_asconf_set_prim(asoc,
3189 (union sctp_addr *)&prim.sspp_addr);
3190 if (!chunk)
3191 return -ENOMEM;
3192
3193 err = sctp_send_asconf(asoc, chunk);
3194
bb33381d 3195 pr_debug("%s: we set peer primary addr primitively\n", __func__);
1da177e4
LT
3196
3197 return err;
3198}
3199
0f3fffd8 3200static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
b7058842 3201 unsigned int optlen)
1da177e4 3202{
0f3fffd8 3203 struct sctp_setadaptation adaptation;
1da177e4 3204
0f3fffd8 3205 if (optlen != sizeof(struct sctp_setadaptation))
1da177e4 3206 return -EINVAL;
0f3fffd8 3207 if (copy_from_user(&adaptation, optval, optlen))
1da177e4
LT
3208 return -EFAULT;
3209
0f3fffd8 3210 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
1da177e4
LT
3211
3212 return 0;
3213}
3214
6ab792f5
ISJ
3215/*
3216 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3217 *
3218 * The context field in the sctp_sndrcvinfo structure is normally only
3219 * used when a failed message is retrieved holding the value that was
3220 * sent down on the actual send call. This option allows the setting of
3221 * a default context on an association basis that will be received on
3222 * reading messages from the peer. This is especially helpful in the
3223 * one-2-many model for an application to keep some reference to an
3224 * internal state machine that is processing messages on the
3225 * association. Note that the setting of this value only effects
3226 * received messages from the peer and does not effect the value that is
3227 * saved with outbound messages.
3228 */
3229static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
b7058842 3230 unsigned int optlen)
6ab792f5
ISJ
3231{
3232 struct sctp_assoc_value params;
3233 struct sctp_sock *sp;
3234 struct sctp_association *asoc;
3235
3236 if (optlen != sizeof(struct sctp_assoc_value))
3237 return -EINVAL;
3238 if (copy_from_user(&params, optval, optlen))
3239 return -EFAULT;
3240
3241 sp = sctp_sk(sk);
3242
3243 if (params.assoc_id != 0) {
3244 asoc = sctp_id2assoc(sk, params.assoc_id);
3245 if (!asoc)
3246 return -EINVAL;
3247 asoc->default_rcv_context = params.assoc_value;
3248 } else {
3249 sp->default_rcv_context = params.assoc_value;
3250 }
3251
3252 return 0;
3253}
3254
b6e1331f
VY
3255/*
3256 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3257 *
3258 * This options will at a minimum specify if the implementation is doing
3259 * fragmented interleave. Fragmented interleave, for a one to many
3260 * socket, is when subsequent calls to receive a message may return
3261 * parts of messages from different associations. Some implementations
3262 * may allow you to turn this value on or off. If so, when turned off,
3263 * no fragment interleave will occur (which will cause a head of line
3264 * blocking amongst multiple associations sharing the same one to many
3265 * socket). When this option is turned on, then each receive call may
3266 * come from a different association (thus the user must receive data
3267 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3268 * association each receive belongs to.
3269 *
3270 * This option takes a boolean value. A non-zero value indicates that
3271 * fragmented interleave is on. A value of zero indicates that
3272 * fragmented interleave is off.
3273 *
3274 * Note that it is important that an implementation that allows this
3275 * option to be turned on, have it off by default. Otherwise an unaware
3276 * application using the one to many model may become confused and act
3277 * incorrectly.
3278 */
3279static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3280 char __user *optval,
b7058842 3281 unsigned int optlen)
b6e1331f
VY
3282{
3283 int val;
3284
3285 if (optlen != sizeof(int))
3286 return -EINVAL;
3287 if (get_user(val, (int __user *)optval))
3288 return -EFAULT;
3289
3290 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3291
3292 return 0;
3293}
3294
d49d91d7 3295/*
8510b937 3296 * 8.1.21. Set or Get the SCTP Partial Delivery Point
d49d91d7 3297 * (SCTP_PARTIAL_DELIVERY_POINT)
8510b937 3298 *
d49d91d7
VY
3299 * This option will set or get the SCTP partial delivery point. This
3300 * point is the size of a message where the partial delivery API will be
3301 * invoked to help free up rwnd space for the peer. Setting this to a
8510b937 3302 * lower value will cause partial deliveries to happen more often. The
d49d91d7 3303 * calls argument is an integer that sets or gets the partial delivery
8510b937
WY
3304 * point. Note also that the call will fail if the user attempts to set
3305 * this value larger than the socket receive buffer size.
3306 *
3307 * Note that any single message having a length smaller than or equal to
3308 * the SCTP partial delivery point will be delivered in one single read
3309 * call as long as the user provided buffer is large enough to hold the
3310 * message.
d49d91d7
VY
3311 */
3312static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3313 char __user *optval,
b7058842 3314 unsigned int optlen)
d49d91d7
VY
3315{
3316 u32 val;
3317
3318 if (optlen != sizeof(u32))
3319 return -EINVAL;
3320 if (get_user(val, (int __user *)optval))
3321 return -EFAULT;
3322
8510b937
WY
3323 /* Note: We double the receive buffer from what the user sets
3324 * it to be, also initial rwnd is based on rcvbuf/2.
3325 */
3326 if (val > (sk->sk_rcvbuf >> 1))
3327 return -EINVAL;
3328
d49d91d7
VY
3329 sctp_sk(sk)->pd_point = val;
3330
3331 return 0; /* is this the right error code? */
3332}
3333
70331571
VY
3334/*
3335 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3336 *
3337 * This option will allow a user to change the maximum burst of packets
3338 * that can be emitted by this association. Note that the default value
3339 * is 4, and some implementations may restrict this setting so that it
3340 * can only be lowered.
3341 *
3342 * NOTE: This text doesn't seem right. Do this on a socket basis with
3343 * future associations inheriting the socket value.
3344 */
3345static int sctp_setsockopt_maxburst(struct sock *sk,
3346 char __user *optval,
b7058842 3347 unsigned int optlen)
70331571 3348{
219b99a9
NH
3349 struct sctp_assoc_value params;
3350 struct sctp_sock *sp;
3351 struct sctp_association *asoc;
70331571 3352 int val;
219b99a9 3353 int assoc_id = 0;
70331571 3354
219b99a9 3355 if (optlen == sizeof(int)) {
94f65193 3356 pr_warn_ratelimited(DEPRECATED
f916ec96 3357 "%s (pid %d) "
94f65193 3358 "Use of int in max_burst socket option deprecated.\n"
f916ec96
NH
3359 "Use struct sctp_assoc_value instead\n",
3360 current->comm, task_pid_nr(current));
219b99a9
NH
3361 if (copy_from_user(&val, optval, optlen))
3362 return -EFAULT;
3363 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3364 if (copy_from_user(&params, optval, optlen))
3365 return -EFAULT;
3366 val = params.assoc_value;
3367 assoc_id = params.assoc_id;
3368 } else
70331571
VY
3369 return -EINVAL;
3370
219b99a9
NH
3371 sp = sctp_sk(sk);
3372
3373 if (assoc_id != 0) {
3374 asoc = sctp_id2assoc(sk, assoc_id);
3375 if (!asoc)
3376 return -EINVAL;
3377 asoc->max_burst = val;
3378 } else
3379 sp->max_burst = val;
70331571
VY
3380
3381 return 0;
3382}
3383
65b07e5d
VY
3384/*
3385 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3386 *
3387 * This set option adds a chunk type that the user is requesting to be
3388 * received only in an authenticated way. Changes to the list of chunks
3389 * will only effect future associations on the socket.
3390 */
3391static int sctp_setsockopt_auth_chunk(struct sock *sk,
b7058842
DM
3392 char __user *optval,
3393 unsigned int optlen)
65b07e5d 3394{
b14878cc 3395 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
65b07e5d
VY
3396 struct sctp_authchunk val;
3397
b14878cc 3398 if (!ep->auth_enable)
5e739d17
VY
3399 return -EACCES;
3400
65b07e5d
VY
3401 if (optlen != sizeof(struct sctp_authchunk))
3402 return -EINVAL;
3403 if (copy_from_user(&val, optval, optlen))
3404 return -EFAULT;
3405
3406 switch (val.sauth_chunk) {
7fd71b1e
JP
3407 case SCTP_CID_INIT:
3408 case SCTP_CID_INIT_ACK:
3409 case SCTP_CID_SHUTDOWN_COMPLETE:
3410 case SCTP_CID_AUTH:
3411 return -EINVAL;
65b07e5d
VY
3412 }
3413
3414 /* add this chunk id to the endpoint */
b14878cc 3415 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
65b07e5d
VY
3416}
3417
3418/*
3419 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3420 *
3421 * This option gets or sets the list of HMAC algorithms that the local
3422 * endpoint requires the peer to use.
3423 */
3424static int sctp_setsockopt_hmac_ident(struct sock *sk,
b7058842
DM
3425 char __user *optval,
3426 unsigned int optlen)
65b07e5d 3427{
b14878cc 3428 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
65b07e5d 3429 struct sctp_hmacalgo *hmacs;
d9724055 3430 u32 idents;
65b07e5d
VY
3431 int err;
3432
b14878cc 3433 if (!ep->auth_enable)
5e739d17
VY
3434 return -EACCES;
3435
65b07e5d
VY
3436 if (optlen < sizeof(struct sctp_hmacalgo))
3437 return -EINVAL;
3438
cb3f837b 3439 hmacs = memdup_user(optval, optlen);
934253a7
SW
3440 if (IS_ERR(hmacs))
3441 return PTR_ERR(hmacs);
65b07e5d 3442
d9724055
VY
3443 idents = hmacs->shmac_num_idents;
3444 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3445 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
65b07e5d
VY
3446 err = -EINVAL;
3447 goto out;
3448 }
3449
b14878cc 3450 err = sctp_auth_ep_set_hmacs(ep, hmacs);
65b07e5d
VY
3451out:
3452 kfree(hmacs);
3453 return err;
3454}
3455
3456/*
3457 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3458 *
3459 * This option will set a shared secret key which is used to build an
3460 * association shared key.
3461 */
3462static int sctp_setsockopt_auth_key(struct sock *sk,
3463 char __user *optval,
b7058842 3464 unsigned int optlen)
65b07e5d 3465{
b14878cc 3466 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
65b07e5d
VY
3467 struct sctp_authkey *authkey;
3468 struct sctp_association *asoc;
3469 int ret;
3470
b14878cc 3471 if (!ep->auth_enable)
5e739d17
VY
3472 return -EACCES;
3473
65b07e5d
VY
3474 if (optlen <= sizeof(struct sctp_authkey))
3475 return -EINVAL;
3476
cb3f837b 3477 authkey = memdup_user(optval, optlen);
934253a7
SW
3478 if (IS_ERR(authkey))
3479 return PTR_ERR(authkey);
65b07e5d 3480
328fc47e 3481 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
30c2235c
VY
3482 ret = -EINVAL;
3483 goto out;
3484 }
3485
65b07e5d
VY
3486 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3487 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3488 ret = -EINVAL;
3489 goto out;
3490 }
3491
b14878cc 3492 ret = sctp_auth_set_key(ep, asoc, authkey);
65b07e5d 3493out:
6ba542a2 3494 kzfree(authkey);
65b07e5d
VY
3495 return ret;
3496}
3497
3498/*
3499 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3500 *
3501 * This option will get or set the active shared key to be used to build
3502 * the association shared key.
3503 */
3504static int sctp_setsockopt_active_key(struct sock *sk,
b7058842
DM
3505 char __user *optval,
3506 unsigned int optlen)
65b07e5d 3507{
b14878cc 3508 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
65b07e5d
VY
3509 struct sctp_authkeyid val;
3510 struct sctp_association *asoc;
3511
b14878cc 3512 if (!ep->auth_enable)
5e739d17
VY
3513 return -EACCES;
3514
65b07e5d
VY
3515 if (optlen != sizeof(struct sctp_authkeyid))
3516 return -EINVAL;
3517 if (copy_from_user(&val, optval, optlen))
3518 return -EFAULT;
3519
3520 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3521 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3522 return -EINVAL;
3523
b14878cc 3524 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
65b07e5d
VY
3525}
3526
3527/*
3528 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3529 *
3530 * This set option will delete a shared secret key from use.
3531 */
3532static int sctp_setsockopt_del_key(struct sock *sk,
b7058842
DM
3533 char __user *optval,
3534 unsigned int optlen)
65b07e5d 3535{
b14878cc 3536 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
65b07e5d
VY
3537 struct sctp_authkeyid val;
3538 struct sctp_association *asoc;
3539
b14878cc 3540 if (!ep->auth_enable)
5e739d17
VY
3541 return -EACCES;
3542
65b07e5d
VY
3543 if (optlen != sizeof(struct sctp_authkeyid))
3544 return -EINVAL;
3545 if (copy_from_user(&val, optval, optlen))
3546 return -EFAULT;
3547
3548 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3549 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3550 return -EINVAL;
3551
b14878cc 3552 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
65b07e5d
VY
3553
3554}
3555
7dc04d71
MH
3556/*
3557 * 8.1.23 SCTP_AUTO_ASCONF
3558 *
3559 * This option will enable or disable the use of the automatic generation of
3560 * ASCONF chunks to add and delete addresses to an existing association. Note
3561 * that this option has two caveats namely: a) it only affects sockets that
3562 * are bound to all addresses available to the SCTP stack, and b) the system
3563 * administrator may have an overriding control that turns the ASCONF feature
3564 * off no matter what setting the socket option may have.
3565 * This option expects an integer boolean flag, where a non-zero value turns on
3566 * the option, and a zero value turns off the option.
3567 * Note. In this implementation, socket operation overrides default parameter
3568 * being set by sysctl as well as FreeBSD implementation
3569 */
3570static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3571 unsigned int optlen)
3572{
3573 int val;
3574 struct sctp_sock *sp = sctp_sk(sk);
3575
3576 if (optlen < sizeof(int))
3577 return -EINVAL;
3578 if (get_user(val, (int __user *)optval))
3579 return -EFAULT;
3580 if (!sctp_is_ep_boundall(sk) && val)
3581 return -EINVAL;
3582 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3583 return 0;
3584
2d45a02d 3585 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
7dc04d71
MH
3586 if (val == 0 && sp->do_auto_asconf) {
3587 list_del(&sp->auto_asconf_list);
3588 sp->do_auto_asconf = 0;
3589 } else if (val && !sp->do_auto_asconf) {
3590 list_add_tail(&sp->auto_asconf_list,
4db67e80 3591 &sock_net(sk)->sctp.auto_asconf_splist);
7dc04d71
MH
3592 sp->do_auto_asconf = 1;
3593 }
2d45a02d 3594 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
7dc04d71
MH
3595 return 0;
3596}
3597
5aa93bcf
NH
3598/*
3599 * SCTP_PEER_ADDR_THLDS
3600 *
3601 * This option allows us to alter the partially failed threshold for one or all
3602 * transports in an association. See Section 6.1 of:
3603 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3604 */
3605static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3606 char __user *optval,
3607 unsigned int optlen)
3608{
3609 struct sctp_paddrthlds val;
3610 struct sctp_transport *trans;
3611 struct sctp_association *asoc;
3612
3613 if (optlen < sizeof(struct sctp_paddrthlds))
3614 return -EINVAL;
3615 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3616 sizeof(struct sctp_paddrthlds)))
3617 return -EFAULT;
3618
3619
3620 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3621 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3622 if (!asoc)
3623 return -ENOENT;
3624 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3625 transports) {
3626 if (val.spt_pathmaxrxt)
3627 trans->pathmaxrxt = val.spt_pathmaxrxt;
3628 trans->pf_retrans = val.spt_pathpfthld;
3629 }
3630
3631 if (val.spt_pathmaxrxt)
3632 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3633 asoc->pf_retrans = val.spt_pathpfthld;
3634 } else {
3635 trans = sctp_addr_id2transport(sk, &val.spt_address,
3636 val.spt_assoc_id);
3637 if (!trans)
3638 return -ENOENT;
3639
3640 if (val.spt_pathmaxrxt)
3641 trans->pathmaxrxt = val.spt_pathmaxrxt;
3642 trans->pf_retrans = val.spt_pathpfthld;
3643 }
3644
3645 return 0;
3646}
3647
0d3a421d
GOV
3648static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3649 char __user *optval,
3650 unsigned int optlen)
3651{
3652 int val;
3653
3654 if (optlen < sizeof(int))
3655 return -EINVAL;
3656 if (get_user(val, (int __user *) optval))
3657 return -EFAULT;
3658
3659 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3660
3661 return 0;
3662}
3663
2347c80f
GOV
3664static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3665 char __user *optval,
3666 unsigned int optlen)
3667{
3668 int val;
3669
3670 if (optlen < sizeof(int))
3671 return -EINVAL;
3672 if (get_user(val, (int __user *) optval))
3673 return -EFAULT;
3674
3675 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3676
3677 return 0;
3678}
3679
28aa4c26
XL
3680static int sctp_setsockopt_pr_supported(struct sock *sk,
3681 char __user *optval,
3682 unsigned int optlen)
3683{
3684 struct sctp_assoc_value params;
3685 struct sctp_association *asoc;
3686 int retval = -EINVAL;
3687
3688 if (optlen != sizeof(params))
3689 goto out;
3690
3691 if (copy_from_user(&params, optval, optlen)) {
3692 retval = -EFAULT;
3693 goto out;
3694 }
3695
3696 asoc = sctp_id2assoc(sk, params.assoc_id);
3697 if (asoc) {
3698 asoc->prsctp_enable = !!params.assoc_value;
3699 } else if (!params.assoc_id) {
3700 struct sctp_sock *sp = sctp_sk(sk);
3701
3702 sp->ep->prsctp_enable = !!params.assoc_value;
3703 } else {
3704 goto out;
3705 }
3706
3707 retval = 0;
3708
3709out:
3710 return retval;
3711}
3712
f959fb44
XL
3713static int sctp_setsockopt_default_prinfo(struct sock *sk,
3714 char __user *optval,
3715 unsigned int optlen)
3716{
3717 struct sctp_default_prinfo info;
3718 struct sctp_association *asoc;
3719 int retval = -EINVAL;
3720
3721 if (optlen != sizeof(info))
3722 goto out;
3723
3724 if (copy_from_user(&info, optval, sizeof(info))) {
3725 retval = -EFAULT;
3726 goto out;
3727 }
3728
3729 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3730 goto out;
3731
3732 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3733 info.pr_value = 0;
3734
3735 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3736 if (asoc) {
3737 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3738 asoc->default_timetolive = info.pr_value;
3739 } else if (!info.pr_assoc_id) {
3740 struct sctp_sock *sp = sctp_sk(sk);
3741
3742 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3743 sp->default_timetolive = info.pr_value;
3744 } else {
3745 goto out;
3746 }
3747
3748 retval = 0;
3749
3750out:
3751 return retval;
3752}
3753
1da177e4
LT
3754/* API 6.2 setsockopt(), getsockopt()
3755 *
3756 * Applications use setsockopt() and getsockopt() to set or retrieve
3757 * socket options. Socket options are used to change the default
3758 * behavior of sockets calls. They are described in Section 7.
3759 *
3760 * The syntax is:
3761 *
3762 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3763 * int __user *optlen);
3764 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3765 * int optlen);
3766 *
3767 * sd - the socket descript.
3768 * level - set to IPPROTO_SCTP for all SCTP options.
3769 * optname - the option name.
3770 * optval - the buffer to store the value of the option.
3771 * optlen - the size of the buffer.
3772 */
dda91928
DB
3773static int sctp_setsockopt(struct sock *sk, int level, int optname,
3774 char __user *optval, unsigned int optlen)
1da177e4
LT
3775{
3776 int retval = 0;
3777
bb33381d 3778 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
1da177e4
LT
3779
3780 /* I can hardly begin to describe how wrong this is. This is
3781 * so broken as to be worse than useless. The API draft
3782 * REALLY is NOT helpful here... I am not convinced that the
3783 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3784 * are at all well-founded.
3785 */
3786 if (level != SOL_SCTP) {
3787 struct sctp_af *af = sctp_sk(sk)->pf->af;
3788 retval = af->setsockopt(sk, level, optname, optval, optlen);
3789 goto out_nounlock;
3790 }
3791
048ed4b6 3792 lock_sock(sk);
1da177e4
LT
3793
3794 switch (optname) {
3795 case SCTP_SOCKOPT_BINDX_ADD:
3796 /* 'optlen' is the size of the addresses buffer. */
3797 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3798 optlen, SCTP_BINDX_ADD_ADDR);
3799 break;
3800
3801 case SCTP_SOCKOPT_BINDX_REM:
3802 /* 'optlen' is the size of the addresses buffer. */
3803 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3804 optlen, SCTP_BINDX_REM_ADDR);
3805 break;
3806
88a0a948
VY
3807 case SCTP_SOCKOPT_CONNECTX_OLD:
3808 /* 'optlen' is the size of the addresses buffer. */
3809 retval = sctp_setsockopt_connectx_old(sk,
3810 (struct sockaddr __user *)optval,
3811 optlen);
3812 break;
3813
3f7a87d2
FF
3814 case SCTP_SOCKOPT_CONNECTX:
3815 /* 'optlen' is the size of the addresses buffer. */
88a0a948
VY
3816 retval = sctp_setsockopt_connectx(sk,
3817 (struct sockaddr __user *)optval,
3818 optlen);
3f7a87d2
FF
3819 break;
3820
1da177e4
LT
3821 case SCTP_DISABLE_FRAGMENTS:
3822 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3823 break;
3824
3825 case SCTP_EVENTS:
3826 retval = sctp_setsockopt_events(sk, optval, optlen);
3827 break;
3828
3829 case SCTP_AUTOCLOSE:
3830 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3831 break;
3832
3833 case SCTP_PEER_ADDR_PARAMS:
3834 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3835 break;
3836
4580ccc0 3837 case SCTP_DELAYED_SACK:
d364d927 3838 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
7708610b 3839 break;
d49d91d7
VY
3840 case SCTP_PARTIAL_DELIVERY_POINT:
3841 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3842 break;
7708610b 3843
1da177e4
LT
3844 case SCTP_INITMSG:
3845 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3846 break;
3847 case SCTP_DEFAULT_SEND_PARAM:
3848 retval = sctp_setsockopt_default_send_param(sk, optval,
3849 optlen);
3850 break;
6b3fd5f3
GOV
3851 case SCTP_DEFAULT_SNDINFO:
3852 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3853 break;
1da177e4
LT
3854 case SCTP_PRIMARY_ADDR:
3855 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3856 break;
3857 case SCTP_SET_PEER_PRIMARY_ADDR:
3858 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3859 break;
3860 case SCTP_NODELAY:
3861 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3862 break;
3863 case SCTP_RTOINFO:
3864 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3865 break;
3866 case SCTP_ASSOCINFO:
3867 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3868 break;
3869 case SCTP_I_WANT_MAPPED_V4_ADDR:
3870 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3871 break;
3872 case SCTP_MAXSEG:
3873 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3874 break;
0f3fffd8
ISJ
3875 case SCTP_ADAPTATION_LAYER:
3876 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
1da177e4 3877 break;
6ab792f5
ISJ
3878 case SCTP_CONTEXT:
3879 retval = sctp_setsockopt_context(sk, optval, optlen);
3880 break;
b6e1331f
VY
3881 case SCTP_FRAGMENT_INTERLEAVE:
3882 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3883 break;
70331571
VY
3884 case SCTP_MAX_BURST:
3885 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3886 break;
65b07e5d
VY
3887 case SCTP_AUTH_CHUNK:
3888 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3889 break;
3890 case SCTP_HMAC_IDENT:
3891 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3892 break;
3893 case SCTP_AUTH_KEY:
3894 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3895 break;
3896 case SCTP_AUTH_ACTIVE_KEY:
3897 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3898 break;
3899 case SCTP_AUTH_DELETE_KEY:
3900 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3901 break;
7dc04d71
MH
3902 case SCTP_AUTO_ASCONF:
3903 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3904 break;
5aa93bcf
NH
3905 case SCTP_PEER_ADDR_THLDS:
3906 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3907 break;
0d3a421d
GOV
3908 case SCTP_RECVRCVINFO:
3909 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3910 break;
2347c80f
GOV
3911 case SCTP_RECVNXTINFO:
3912 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3913 break;
28aa4c26
XL
3914 case SCTP_PR_SUPPORTED:
3915 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3916 break;
f959fb44
XL
3917 case SCTP_DEFAULT_PRINFO:
3918 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3919 break;
1da177e4
LT
3920 default:
3921 retval = -ENOPROTOOPT;
3922 break;
3ff50b79 3923 }
1da177e4 3924
048ed4b6 3925 release_sock(sk);
1da177e4
LT
3926
3927out_nounlock:
3928 return retval;
3929}
3930
3931/* API 3.1.6 connect() - UDP Style Syntax
3932 *
3933 * An application may use the connect() call in the UDP model to initiate an
3934 * association without sending data.
3935 *
3936 * The syntax is:
3937 *
3938 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3939 *
3940 * sd: the socket descriptor to have a new association added to.
3941 *
3942 * nam: the address structure (either struct sockaddr_in or struct
3943 * sockaddr_in6 defined in RFC2553 [7]).
3944 *
3945 * len: the size of the address.
3946 */
dda91928
DB
3947static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3948 int addr_len)
1da177e4 3949{
1da177e4 3950 int err = 0;
3f7a87d2 3951 struct sctp_af *af;
1da177e4 3952
048ed4b6 3953 lock_sock(sk);
1da177e4 3954
bb33381d
DB
3955 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3956 addr, addr_len);
1da177e4 3957
3f7a87d2
FF
3958 /* Validate addr_len before calling common connect/connectx routine. */
3959 af = sctp_get_af_specific(addr->sa_family);
3960 if (!af || addr_len < af->sockaddr_len) {
3961 err = -EINVAL;
3962 } else {
3963 /* Pass correct addr len to common routine (so it knows there
3964 * is only one address being passed.
3965 */
88a0a948 3966 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
1da177e4
LT
3967 }
3968
048ed4b6 3969 release_sock(sk);
1da177e4
LT
3970 return err;
3971}
3972
3973/* FIXME: Write comments. */
dda91928 3974static int sctp_disconnect(struct sock *sk, int flags)
1da177e4
LT
3975{
3976 return -EOPNOTSUPP; /* STUB */
3977}
3978
3979/* 4.1.4 accept() - TCP Style Syntax
3980 *
3981 * Applications use accept() call to remove an established SCTP
3982 * association from the accept queue of the endpoint. A new socket
3983 * descriptor will be returned from accept() to represent the newly
3984 * formed association.
3985 */
dda91928 3986static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
1da177e4
LT
3987{
3988 struct sctp_sock *sp;
3989 struct sctp_endpoint *ep;
3990 struct sock *newsk = NULL;
3991 struct sctp_association *asoc;
3992 long timeo;
3993 int error = 0;
3994
048ed4b6 3995 lock_sock(sk);
1da177e4
LT
3996
3997 sp = sctp_sk(sk);
3998 ep = sp->ep;
3999
4000 if (!sctp_style(sk, TCP)) {
4001 error = -EOPNOTSUPP;
4002 goto out;
4003 }
4004
4005 if (!sctp_sstate(sk, LISTENING)) {
4006 error = -EINVAL;
4007 goto out;
4008 }
4009
8abfedd8 4010 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1da177e4
LT
4011
4012 error = sctp_wait_for_accept(sk, timeo);
4013 if (error)
4014 goto out;
4015
4016 /* We treat the list of associations on the endpoint as the accept
4017 * queue and pick the first association on the list.
4018 */
4019 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4020
4021 newsk = sp->pf->create_accept_sk(sk, asoc);
4022 if (!newsk) {
4023 error = -ENOMEM;
4024 goto out;
4025 }
4026
4027 /* Populate the fields of the newsk from the oldsk and migrate the
4028 * asoc to the newsk.
4029 */
4030 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4031
4032out:
048ed4b6 4033 release_sock(sk);
d808ad9a 4034 *err = error;
1da177e4
LT
4035 return newsk;
4036}
4037
4038/* The SCTP ioctl handler. */
dda91928 4039static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1da177e4 4040{
65040c33
DEFP
4041 int rc = -ENOTCONN;
4042
048ed4b6 4043 lock_sock(sk);
65040c33
DEFP
4044
4045 /*
4046 * SEQPACKET-style sockets in LISTENING state are valid, for
4047 * SCTP, so only discard TCP-style sockets in LISTENING state.
4048 */
4049 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4050 goto out;
4051
4052 switch (cmd) {
4053 case SIOCINQ: {
4054 struct sk_buff *skb;
4055 unsigned int amount = 0;
4056
4057 skb = skb_peek(&sk->sk_receive_queue);
4058 if (skb != NULL) {
4059 /*
4060 * We will only return the amount of this packet since
4061 * that is all that will be read.
4062 */
4063 amount = skb->len;
4064 }
4065 rc = put_user(amount, (int __user *)arg);
65040c33 4066 break;
9a7241c2 4067 }
65040c33
DEFP
4068 default:
4069 rc = -ENOIOCTLCMD;
4070 break;
4071 }
4072out:
048ed4b6 4073 release_sock(sk);
65040c33 4074 return rc;
1da177e4
LT
4075}
4076
4077/* This is the function which gets called during socket creation to
4078 * initialized the SCTP-specific portion of the sock.
4079 * The sock structure should already be zero-filled memory.
4080 */
dda91928 4081static int sctp_init_sock(struct sock *sk)
1da177e4 4082{
e1fc3b14 4083 struct net *net = sock_net(sk);
1da177e4
LT
4084 struct sctp_sock *sp;
4085
bb33381d 4086 pr_debug("%s: sk:%p\n", __func__, sk);
1da177e4
LT
4087
4088 sp = sctp_sk(sk);
4089
4090 /* Initialize the SCTP per socket area. */
4091 switch (sk->sk_type) {
4092 case SOCK_SEQPACKET:
4093 sp->type = SCTP_SOCKET_UDP;
4094 break;
4095 case SOCK_STREAM:
4096 sp->type = SCTP_SOCKET_TCP;
4097 break;
4098 default:
4099 return -ESOCKTNOSUPPORT;
4100 }
4101
90017acc
MRL
4102 sk->sk_gso_type = SKB_GSO_SCTP;
4103
1da177e4
LT
4104 /* Initialize default send parameters. These parameters can be
4105 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4106 */
4107 sp->default_stream = 0;
4108 sp->default_ppid = 0;
4109 sp->default_flags = 0;
4110 sp->default_context = 0;
4111 sp->default_timetolive = 0;
4112
6ab792f5 4113 sp->default_rcv_context = 0;
e1fc3b14 4114 sp->max_burst = net->sctp.max_burst;
6ab792f5 4115
3c68198e
NH
4116 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4117
1da177e4
LT
4118 /* Initialize default setup parameters. These parameters
4119 * can be modified with the SCTP_INITMSG socket option or
4120 * overridden by the SCTP_INIT CMSG.
4121 */
4122 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4123 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
e1fc3b14
EB
4124 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4125 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
1da177e4
LT
4126
4127 /* Initialize default RTO related parameters. These parameters can
4128 * be modified for with the SCTP_RTOINFO socket option.
4129 */
e1fc3b14
EB
4130 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4131 sp->rtoinfo.srto_max = net->sctp.rto_max;
4132 sp->rtoinfo.srto_min = net->sctp.rto_min;
1da177e4
LT
4133
4134 /* Initialize default association related parameters. These parameters
4135 * can be modified with the SCTP_ASSOCINFO socket option.
4136 */
e1fc3b14 4137 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
1da177e4
LT
4138 sp->assocparams.sasoc_number_peer_destinations = 0;
4139 sp->assocparams.sasoc_peer_rwnd = 0;
4140 sp->assocparams.sasoc_local_rwnd = 0;
e1fc3b14 4141 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
1da177e4
LT
4142
4143 /* Initialize default event subscriptions. By default, all the
d808ad9a 4144 * options are off.
1da177e4
LT
4145 */
4146 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4147
4148 /* Default Peer Address Parameters. These defaults can
4149 * be modified via SCTP_PEER_ADDR_PARAMS
4150 */
e1fc3b14
EB
4151 sp->hbinterval = net->sctp.hb_interval;
4152 sp->pathmaxrxt = net->sctp.max_retrans_path;
4e2d52bf 4153 sp->pathmtu = 0; /* allow default discovery */
e1fc3b14 4154 sp->sackdelay = net->sctp.sack_timeout;
7bfe8bdb 4155 sp->sackfreq = 2;
52ccb8e9 4156 sp->param_flags = SPP_HB_ENABLE |
d808ad9a
YH
4157 SPP_PMTUD_ENABLE |
4158 SPP_SACKDELAY_ENABLE;
1da177e4
LT
4159
4160 /* If enabled no SCTP message fragmentation will be performed.
4161 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4162 */
4163 sp->disable_fragments = 0;
4164
208edef6
SS
4165 /* Enable Nagle algorithm by default. */
4166 sp->nodelay = 0;
1da177e4 4167
0d3a421d 4168 sp->recvrcvinfo = 0;
2347c80f 4169 sp->recvnxtinfo = 0;
0d3a421d 4170
1da177e4
LT
4171 /* Enable by default. */
4172 sp->v4mapped = 1;
4173
4174 /* Auto-close idle associations after the configured
4175 * number of seconds. A value of 0 disables this
4176 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4177 * for UDP-style sockets only.
4178 */
4179 sp->autoclose = 0;
4180
4181 /* User specified fragmentation limit. */
4182 sp->user_frag = 0;
4183
0f3fffd8 4184 sp->adaptation_ind = 0;
1da177e4
LT
4185
4186 sp->pf = sctp_get_pf_specific(sk->sk_family);
4187
4188 /* Control variables for partial data delivery. */
b6e1331f 4189 atomic_set(&sp->pd_mode, 0);
1da177e4 4190 skb_queue_head_init(&sp->pd_lobby);
b6e1331f 4191 sp->frag_interleave = 0;
1da177e4
LT
4192
4193 /* Create a per socket endpoint structure. Even if we
4194 * change the data structure relationships, this may still
4195 * be useful for storing pre-connect address information.
4196 */
c164b838
DB
4197 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4198 if (!sp->ep)
1da177e4
LT
4199 return -ENOMEM;
4200
1da177e4
LT
4201 sp->hmac = NULL;
4202
0a2fbac1
DB
4203 sk->sk_destruct = sctp_destruct_sock;
4204
1da177e4 4205 SCTP_DBG_OBJCNT_INC(sock);
6f756a8c
DM
4206
4207 local_bh_disable();
81419d86 4208 percpu_counter_inc(&sctp_sockets_allocated);
e1fc3b14 4209 sock_prot_inuse_add(net, sk->sk_prot, 1);
2d45a02d
MRL
4210
4211 /* Nothing can fail after this block, otherwise
4212 * sctp_destroy_sock() will be called without addr_wq_lock held
4213 */
e1fc3b14 4214 if (net->sctp.default_auto_asconf) {
2d45a02d 4215 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
9f7d653b 4216 list_add_tail(&sp->auto_asconf_list,
e1fc3b14 4217 &net->sctp.auto_asconf_splist);
9f7d653b 4218 sp->do_auto_asconf = 1;
2d45a02d
MRL
4219 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4220 } else {
9f7d653b 4221 sp->do_auto_asconf = 0;
2d45a02d
MRL
4222 }
4223
6f756a8c
DM
4224 local_bh_enable();
4225
1da177e4
LT
4226 return 0;
4227}
4228
2d45a02d
MRL
4229/* Cleanup any SCTP per socket resources. Must be called with
4230 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4231 */
dda91928 4232static void sctp_destroy_sock(struct sock *sk)
1da177e4 4233{
9f7d653b 4234 struct sctp_sock *sp;
1da177e4 4235
bb33381d 4236 pr_debug("%s: sk:%p\n", __func__, sk);
1da177e4
LT
4237
4238 /* Release our hold on the endpoint. */
9f7d653b 4239 sp = sctp_sk(sk);
1abd165e
DB
4240 /* This could happen during socket init, thus we bail out
4241 * early, since the rest of the below is not setup either.
4242 */
4243 if (sp->ep == NULL)
4244 return;
4245
9f7d653b
MH
4246 if (sp->do_auto_asconf) {
4247 sp->do_auto_asconf = 0;
4248 list_del(&sp->auto_asconf_list);
4249 }
4250 sctp_endpoint_free(sp->ep);
5bc0b3bf 4251 local_bh_disable();
81419d86 4252 percpu_counter_dec(&sctp_sockets_allocated);
9a57f7fa 4253 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
5bc0b3bf 4254 local_bh_enable();
1da177e4
LT
4255}
4256
0a2fbac1
DB
4257/* Triggered when there are no references on the socket anymore */
4258static void sctp_destruct_sock(struct sock *sk)
4259{
4260 struct sctp_sock *sp = sctp_sk(sk);
4261
4262 /* Free up the HMAC transform. */
5821c769 4263 crypto_free_shash(sp->hmac);
0a2fbac1
DB
4264
4265 inet_sock_destruct(sk);
4266}
4267
1da177e4
LT
4268/* API 4.1.7 shutdown() - TCP Style Syntax
4269 * int shutdown(int socket, int how);
4270 *
4271 * sd - the socket descriptor of the association to be closed.
4272 * how - Specifies the type of shutdown. The values are
4273 * as follows:
4274 * SHUT_RD
4275 * Disables further receive operations. No SCTP
4276 * protocol action is taken.
4277 * SHUT_WR
4278 * Disables further send operations, and initiates
4279 * the SCTP shutdown sequence.
4280 * SHUT_RDWR
4281 * Disables further send and receive operations
4282 * and initiates the SCTP shutdown sequence.
4283 */
dda91928 4284static void sctp_shutdown(struct sock *sk, int how)
1da177e4 4285{
55e26eb9 4286 struct net *net = sock_net(sk);
1da177e4 4287 struct sctp_endpoint *ep;
1da177e4
LT
4288
4289 if (!sctp_style(sk, TCP))
4290 return;
4291
5bf35ddf
XL
4292 ep = sctp_sk(sk)->ep;
4293 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4294 struct sctp_association *asoc;
4295
d46e416c 4296 sk->sk_state = SCTP_SS_CLOSING;
5bf35ddf
XL
4297 asoc = list_entry(ep->asocs.next,
4298 struct sctp_association, asocs);
4299 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1da177e4
LT
4300 }
4301}
4302
52c52a61
XL
4303int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4304 struct sctp_info *info)
4305{
4306 struct sctp_transport *prim;
4307 struct list_head *pos;
4308 int mask;
4309
4310 memset(info, 0, sizeof(*info));
4311 if (!asoc) {
4312 struct sctp_sock *sp = sctp_sk(sk);
4313
4314 info->sctpi_s_autoclose = sp->autoclose;
4315 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4316 info->sctpi_s_pd_point = sp->pd_point;
4317 info->sctpi_s_nodelay = sp->nodelay;
4318 info->sctpi_s_disable_fragments = sp->disable_fragments;
4319 info->sctpi_s_v4mapped = sp->v4mapped;
4320 info->sctpi_s_frag_interleave = sp->frag_interleave;
40eb90e9 4321 info->sctpi_s_type = sp->type;
52c52a61
XL
4322
4323 return 0;
4324 }
4325
4326 info->sctpi_tag = asoc->c.my_vtag;
4327 info->sctpi_state = asoc->state;
4328 info->sctpi_rwnd = asoc->a_rwnd;
4329 info->sctpi_unackdata = asoc->unack_data;
4330 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4331 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4332 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4333 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4334 info->sctpi_inqueue++;
4335 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4336 info->sctpi_outqueue++;
4337 info->sctpi_overall_error = asoc->overall_error_count;
4338 info->sctpi_max_burst = asoc->max_burst;
4339 info->sctpi_maxseg = asoc->frag_point;
4340 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4341 info->sctpi_peer_tag = asoc->c.peer_vtag;
4342
4343 mask = asoc->peer.ecn_capable << 1;
4344 mask = (mask | asoc->peer.ipv4_address) << 1;
4345 mask = (mask | asoc->peer.ipv6_address) << 1;
4346 mask = (mask | asoc->peer.hostname_address) << 1;
4347 mask = (mask | asoc->peer.asconf_capable) << 1;
4348 mask = (mask | asoc->peer.prsctp_capable) << 1;
4349 mask = (mask | asoc->peer.auth_capable);
4350 info->sctpi_peer_capable = mask;
4351 mask = asoc->peer.sack_needed << 1;
4352 mask = (mask | asoc->peer.sack_generation) << 1;
4353 mask = (mask | asoc->peer.zero_window_announced);
4354 info->sctpi_peer_sack = mask;
4355
4356 info->sctpi_isacks = asoc->stats.isacks;
4357 info->sctpi_osacks = asoc->stats.osacks;
4358 info->sctpi_opackets = asoc->stats.opackets;
4359 info->sctpi_ipackets = asoc->stats.ipackets;
4360 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4361 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4362 info->sctpi_idupchunks = asoc->stats.idupchunks;
4363 info->sctpi_gapcnt = asoc->stats.gapcnt;
4364 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4365 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4366 info->sctpi_oodchunks = asoc->stats.oodchunks;
4367 info->sctpi_iodchunks = asoc->stats.iodchunks;
4368 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4369 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4370
4371 prim = asoc->peer.primary_path;
4372 memcpy(&info->sctpi_p_address, &prim->ipaddr,
4373 sizeof(struct sockaddr_storage));
4374 info->sctpi_p_state = prim->state;
4375 info->sctpi_p_cwnd = prim->cwnd;
4376 info->sctpi_p_srtt = prim->srtt;
4377 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4378 info->sctpi_p_hbinterval = prim->hbinterval;
4379 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4380 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4381 info->sctpi_p_ssthresh = prim->ssthresh;
4382 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4383 info->sctpi_p_flight_size = prim->flight_size;
4384 info->sctpi_p_error = prim->error_count;
4385
4386 return 0;
4387}
4388EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4389
626d16f5
XL
4390/* use callback to avoid exporting the core structure */
4391int sctp_transport_walk_start(struct rhashtable_iter *iter)
4392{
4393 int err;
4394
4395 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4396 GFP_KERNEL);
4397 if (err)
4398 return err;
4399
4400 err = rhashtable_walk_start(iter);
53fa1036 4401 if (err && err != -EAGAIN) {
5fc382d8 4402 rhashtable_walk_stop(iter);
53fa1036
XL
4403 rhashtable_walk_exit(iter);
4404 return err;
4405 }
626d16f5 4406
53fa1036 4407 return 0;
626d16f5
XL
4408}
4409
4410void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4411{
4412 rhashtable_walk_stop(iter);
4413 rhashtable_walk_exit(iter);
4414}
4415
4416struct sctp_transport *sctp_transport_get_next(struct net *net,
4417 struct rhashtable_iter *iter)
4418{
4419 struct sctp_transport *t;
4420
4421 t = rhashtable_walk_next(iter);
4422 for (; t; t = rhashtable_walk_next(iter)) {
4423 if (IS_ERR(t)) {
4424 if (PTR_ERR(t) == -EAGAIN)
4425 continue;
4426 break;
4427 }
4428
4429 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4430 t->asoc->peer.primary_path == t)
4431 break;
4432 }
4433
4434 return t;
4435}
4436
4437struct sctp_transport *sctp_transport_get_idx(struct net *net,
4438 struct rhashtable_iter *iter,
4439 int pos)
4440{
4441 void *obj = SEQ_START_TOKEN;
4442
4443 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4444 !IS_ERR(obj))
4445 pos--;
4446
4447 return obj;
4448}
4449
4450int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4451 void *p) {
4452 int err = 0;
4453 int hash = 0;
4454 struct sctp_ep_common *epb;
4455 struct sctp_hashbucket *head;
4456
4457 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4458 hash++, head++) {
4459 read_lock(&head->lock);
4460 sctp_for_each_hentry(epb, &head->chain) {
4461 err = cb(sctp_ep(epb), p);
4462 if (err)
4463 break;
4464 }
4465 read_unlock(&head->lock);
4466 }
4467
4468 return err;
4469}
4470EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4471
4472int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4473 struct net *net,
4474 const union sctp_addr *laddr,
4475 const union sctp_addr *paddr, void *p)
4476{
4477 struct sctp_transport *transport;
1cceda78 4478 int err = -ENOENT;
626d16f5
XL
4479
4480 rcu_read_lock();
4481 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4482 if (!transport || !sctp_transport_hold(transport))
4483 goto out;
1cceda78 4484
626d16f5 4485 rcu_read_unlock();
1cceda78 4486 err = cb(transport, p);
cd26da4f 4487 sctp_transport_put(transport);
1cceda78
XL
4488
4489out:
626d16f5
XL
4490 return err;
4491}
4492EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4493
4494int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4495 struct net *net, int pos, void *p) {
4496 struct rhashtable_iter hti;
626d16f5 4497 void *obj;
53fa1036 4498 int err;
626d16f5 4499
53fa1036
XL
4500 err = sctp_transport_walk_start(&hti);
4501 if (err)
4502 return err;
626d16f5
XL
4503
4504 sctp_transport_get_idx(net, &hti, pos);
4505 obj = sctp_transport_get_next(net, &hti);
4506 for (; obj && !IS_ERR(obj); obj = sctp_transport_get_next(net, &hti)) {
4507 struct sctp_transport *transport = obj;
4508
4509 if (!sctp_transport_hold(transport))
4510 continue;
4511 err = cb(transport, p);
4512 sctp_transport_put(transport);
4513 if (err)
4514 break;
4515 }
626d16f5 4516 sctp_transport_walk_stop(&hti);
53fa1036 4517
626d16f5
XL
4518 return err;
4519}
4520EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4521
1da177e4
LT
4522/* 7.2.1 Association Status (SCTP_STATUS)
4523
4524 * Applications can retrieve current status information about an
4525 * association, including association state, peer receiver window size,
4526 * number of unacked data chunks, and number of data chunks pending
4527 * receipt. This information is read-only.
4528 */
4529static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4530 char __user *optval,
4531 int __user *optlen)
4532{
4533 struct sctp_status status;
4534 struct sctp_association *asoc = NULL;
4535 struct sctp_transport *transport;
4536 sctp_assoc_t associd;
4537 int retval = 0;
4538
408f22e8 4539 if (len < sizeof(status)) {
1da177e4
LT
4540 retval = -EINVAL;
4541 goto out;
4542 }
4543
408f22e8
NH
4544 len = sizeof(status);
4545 if (copy_from_user(&status, optval, len)) {
1da177e4
LT
4546 retval = -EFAULT;
4547 goto out;
4548 }
4549
4550 associd = status.sstat_assoc_id;
4551 asoc = sctp_id2assoc(sk, associd);
4552 if (!asoc) {
4553 retval = -EINVAL;
4554 goto out;
4555 }
4556
4557 transport = asoc->peer.primary_path;
4558
4559 status.sstat_assoc_id = sctp_assoc2id(asoc);
38ab1fa9 4560 status.sstat_state = sctp_assoc_to_state(asoc);
1da177e4
LT
4561 status.sstat_rwnd = asoc->peer.rwnd;
4562 status.sstat_unackdata = asoc->unack_data;
4563
4564 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4565 status.sstat_instrms = asoc->c.sinit_max_instreams;
4566 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4567 status.sstat_fragmentation_point = asoc->frag_point;
4568 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
8cec6b80
AV
4569 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4570 transport->af_specific->sockaddr_len);
1da177e4 4571 /* Map ipv4 address into v4-mapped-on-v6 address. */
299ee123 4572 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
1da177e4 4573 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3f7a87d2 4574 status.sstat_primary.spinfo_state = transport->state;
1da177e4
LT
4575 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4576 status.sstat_primary.spinfo_srtt = transport->srtt;
4577 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
52ccb8e9 4578 status.sstat_primary.spinfo_mtu = transport->pathmtu;
1da177e4 4579
3f7a87d2
FF
4580 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4581 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4582
1da177e4
LT
4583 if (put_user(len, optlen)) {
4584 retval = -EFAULT;
4585 goto out;
4586 }
4587
bb33381d
DB
4588 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4589 __func__, len, status.sstat_state, status.sstat_rwnd,
4590 status.sstat_assoc_id);
1da177e4
LT
4591
4592 if (copy_to_user(optval, &status, len)) {
4593 retval = -EFAULT;
4594 goto out;
4595 }
4596
4597out:
a02cec21 4598 return retval;
1da177e4
LT
4599}
4600
4601
4602/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4603 *
4604 * Applications can retrieve information about a specific peer address
4605 * of an association, including its reachability state, congestion
4606 * window, and retransmission timer values. This information is
4607 * read-only.
4608 */
4609static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4610 char __user *optval,
4611 int __user *optlen)
4612{
4613 struct sctp_paddrinfo pinfo;
4614 struct sctp_transport *transport;
4615 int retval = 0;
4616
408f22e8 4617 if (len < sizeof(pinfo)) {
1da177e4
LT
4618 retval = -EINVAL;
4619 goto out;
4620 }
4621
408f22e8
NH
4622 len = sizeof(pinfo);
4623 if (copy_from_user(&pinfo, optval, len)) {
1da177e4
LT
4624 retval = -EFAULT;
4625 goto out;
4626 }
4627
4628 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4629 pinfo.spinfo_assoc_id);
4630 if (!transport)
4631 return -EINVAL;
4632
4633 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3f7a87d2 4634 pinfo.spinfo_state = transport->state;
1da177e4
LT
4635 pinfo.spinfo_cwnd = transport->cwnd;
4636 pinfo.spinfo_srtt = transport->srtt;
4637 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
52ccb8e9 4638 pinfo.spinfo_mtu = transport->pathmtu;
1da177e4 4639
3f7a87d2
FF
4640 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4641 pinfo.spinfo_state = SCTP_ACTIVE;
4642
1da177e4
LT
4643 if (put_user(len, optlen)) {
4644 retval = -EFAULT;
4645 goto out;
4646 }
4647
4648 if (copy_to_user(optval, &pinfo, len)) {
4649 retval = -EFAULT;
4650 goto out;
4651 }
4652
4653out:
a02cec21 4654 return retval;
1da177e4
LT
4655}
4656
4657/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4658 *
4659 * This option is a on/off flag. If enabled no SCTP message
4660 * fragmentation will be performed. Instead if a message being sent
4661 * exceeds the current PMTU size, the message will NOT be sent and
4662 * instead a error will be indicated to the user.
4663 */
4664static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4665 char __user *optval, int __user *optlen)
4666{
4667 int val;
4668
4669 if (len < sizeof(int))
4670 return -EINVAL;
4671
4672 len = sizeof(int);
4673 val = (sctp_sk(sk)->disable_fragments == 1);
4674 if (put_user(len, optlen))
4675 return -EFAULT;
4676 if (copy_to_user(optval, &val, len))
4677 return -EFAULT;
4678 return 0;
4679}
4680
4681/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4682 *
4683 * This socket option is used to specify various notifications and
4684 * ancillary data the user wishes to receive.
4685 */
4686static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4687 int __user *optlen)
4688{
a4b8e71b 4689 if (len == 0)
1da177e4 4690 return -EINVAL;
acdd5985
TG
4691 if (len > sizeof(struct sctp_event_subscribe))
4692 len = sizeof(struct sctp_event_subscribe);
408f22e8
NH
4693 if (put_user(len, optlen))
4694 return -EFAULT;
1da177e4
LT
4695 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4696 return -EFAULT;
4697 return 0;
4698}
4699
4700/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4701 *
4702 * This socket option is applicable to the UDP-style socket only. When
4703 * set it will cause associations that are idle for more than the
4704 * specified number of seconds to automatically close. An association
4705 * being idle is defined an association that has NOT sent or received
4706 * user data. The special value of '0' indicates that no automatic
4707 * close of any associations should be performed. The option expects an
4708 * integer defining the number of seconds of idle time before an
4709 * association is closed.
4710 */
4711static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4712{
4713 /* Applicable to UDP-style socket only */
4714 if (sctp_style(sk, TCP))
4715 return -EOPNOTSUPP;
408f22e8 4716 if (len < sizeof(int))
1da177e4 4717 return -EINVAL;
408f22e8
NH
4718 len = sizeof(int);
4719 if (put_user(len, optlen))
4720 return -EFAULT;
4721 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
1da177e4
LT
4722 return -EFAULT;
4723 return 0;
4724}
4725
4726/* Helper routine to branch off an association to a new socket. */
0343c554 4727int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
1da177e4 4728{
0343c554 4729 struct sctp_association *asoc = sctp_id2assoc(sk, id);
299ee123 4730 struct sctp_sock *sp = sctp_sk(sk);
1da177e4
LT
4731 struct socket *sock;
4732 int err = 0;
4733
0343c554
BP
4734 if (!asoc)
4735 return -EINVAL;
4736
1da177e4
LT
4737 /* An association cannot be branched off from an already peeled-off
4738 * socket, nor is this supported for tcp style sockets.
4739 */
4740 if (!sctp_style(sk, UDP))
4741 return -EINVAL;
4742
4743 /* Create a new socket. */
4744 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4745 if (err < 0)
4746 return err;
4747
914e1c8b 4748 sctp_copy_sock(sock->sk, sk, asoc);
4f444308
VY
4749
4750 /* Make peeled-off sockets more like 1-1 accepted sockets.
4751 * Set the daddr and initialize id to something more random
4752 */
299ee123 4753 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
914e1c8b
VY
4754
4755 /* Populate the fields of the newsk from the oldsk and migrate the
4756 * asoc to the newsk.
4757 */
4758 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4f444308 4759
1da177e4
LT
4760 *sockp = sock;
4761
4762 return err;
4763}
0343c554 4764EXPORT_SYMBOL(sctp_do_peeloff);
1da177e4
LT
4765
4766static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4767{
4768 sctp_peeloff_arg_t peeloff;
4769 struct socket *newsock;
56b31d1c 4770 struct file *newfile;
1da177e4 4771 int retval = 0;
1da177e4 4772
408f22e8 4773 if (len < sizeof(sctp_peeloff_arg_t))
1da177e4 4774 return -EINVAL;
408f22e8 4775 len = sizeof(sctp_peeloff_arg_t);
1da177e4
LT
4776 if (copy_from_user(&peeloff, optval, len))
4777 return -EFAULT;
4778
0343c554 4779 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
1da177e4
LT
4780 if (retval < 0)
4781 goto out;
4782
4783 /* Map the socket to an unused fd that can be returned to the user. */
8a59bd3e 4784 retval = get_unused_fd_flags(0);
1da177e4
LT
4785 if (retval < 0) {
4786 sock_release(newsock);
4787 goto out;
4788 }
4789
aab174f0 4790 newfile = sock_alloc_file(newsock, 0, NULL);
b5ffe634 4791 if (IS_ERR(newfile)) {
56b31d1c
AV
4792 put_unused_fd(retval);
4793 sock_release(newsock);
4794 return PTR_ERR(newfile);
4795 }
4796
bb33381d
DB
4797 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4798 retval);
1da177e4
LT
4799
4800 /* Return the fd mapped to the new socket. */
56b31d1c
AV
4801 if (put_user(len, optlen)) {
4802 fput(newfile);
4803 put_unused_fd(retval);
4804 return -EFAULT;
4805 }
1da177e4 4806 peeloff.sd = retval;
56b31d1c
AV
4807 if (copy_to_user(optval, &peeloff, len)) {
4808 fput(newfile);
4809 put_unused_fd(retval);
408f22e8 4810 return -EFAULT;
56b31d1c
AV
4811 }
4812 fd_install(retval, newfile);
1da177e4
LT
4813out:
4814 return retval;
4815}
4816
4817/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4818 *
4819 * Applications can enable or disable heartbeats for any peer address of
4820 * an association, modify an address's heartbeat interval, force a
4821 * heartbeat to be sent immediately, and adjust the address's maximum
4822 * number of retransmissions sent before an address is considered
4823 * unreachable. The following structure is used to access and modify an
4824 * address's parameters:
4825 *
4826 * struct sctp_paddrparams {
52ccb8e9
FF
4827 * sctp_assoc_t spp_assoc_id;
4828 * struct sockaddr_storage spp_address;
4829 * uint32_t spp_hbinterval;
4830 * uint16_t spp_pathmaxrxt;
4831 * uint32_t spp_pathmtu;
4832 * uint32_t spp_sackdelay;
4833 * uint32_t spp_flags;
4834 * };
4835 *
4836 * spp_assoc_id - (one-to-many style socket) This is filled in the
4837 * application, and identifies the association for
4838 * this query.
1da177e4
LT
4839 * spp_address - This specifies which address is of interest.
4840 * spp_hbinterval - This contains the value of the heartbeat interval,
52ccb8e9
FF
4841 * in milliseconds. If a value of zero
4842 * is present in this field then no changes are to
4843 * be made to this parameter.
1da177e4
LT
4844 * spp_pathmaxrxt - This contains the maximum number of
4845 * retransmissions before this address shall be
52ccb8e9
FF
4846 * considered unreachable. If a value of zero
4847 * is present in this field then no changes are to
4848 * be made to this parameter.
4849 * spp_pathmtu - When Path MTU discovery is disabled the value
4850 * specified here will be the "fixed" path mtu.
4851 * Note that if the spp_address field is empty
4852 * then all associations on this address will
4853 * have this fixed path mtu set upon them.
4854 *
4855 * spp_sackdelay - When delayed sack is enabled, this value specifies
4856 * the number of milliseconds that sacks will be delayed
4857 * for. This value will apply to all addresses of an
4858 * association if the spp_address field is empty. Note
4859 * also, that if delayed sack is enabled and this
4860 * value is set to 0, no change is made to the last
4861 * recorded delayed sack timer value.
4862 *
4863 * spp_flags - These flags are used to control various features
4864 * on an association. The flag field may contain
4865 * zero or more of the following options.
4866 *
4867 * SPP_HB_ENABLE - Enable heartbeats on the
4868 * specified address. Note that if the address
4869 * field is empty all addresses for the association
4870 * have heartbeats enabled upon them.
4871 *
4872 * SPP_HB_DISABLE - Disable heartbeats on the
4873 * speicifed address. Note that if the address
4874 * field is empty all addresses for the association
4875 * will have their heartbeats disabled. Note also
4876 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4877 * mutually exclusive, only one of these two should
4878 * be specified. Enabling both fields will have
4879 * undetermined results.
4880 *
4881 * SPP_HB_DEMAND - Request a user initiated heartbeat
4882 * to be made immediately.
4883 *
4884 * SPP_PMTUD_ENABLE - This field will enable PMTU
4885 * discovery upon the specified address. Note that
4886 * if the address feild is empty then all addresses
4887 * on the association are effected.
4888 *
4889 * SPP_PMTUD_DISABLE - This field will disable PMTU
4890 * discovery upon the specified address. Note that
4891 * if the address feild is empty then all addresses
4892 * on the association are effected. Not also that
4893 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4894 * exclusive. Enabling both will have undetermined
4895 * results.
4896 *
4897 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4898 * on delayed sack. The time specified in spp_sackdelay
4899 * is used to specify the sack delay for this address. Note
4900 * that if spp_address is empty then all addresses will
4901 * enable delayed sack and take on the sack delay
4902 * value specified in spp_sackdelay.
4903 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4904 * off delayed sack. If the spp_address field is blank then
4905 * delayed sack is disabled for the entire association. Note
4906 * also that this field is mutually exclusive to
4907 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4908 * results.
1da177e4
LT
4909 */
4910static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
52ccb8e9 4911 char __user *optval, int __user *optlen)
1da177e4 4912{
52ccb8e9
FF
4913 struct sctp_paddrparams params;
4914 struct sctp_transport *trans = NULL;
4915 struct sctp_association *asoc = NULL;
4916 struct sctp_sock *sp = sctp_sk(sk);
1da177e4 4917
408f22e8 4918 if (len < sizeof(struct sctp_paddrparams))
1da177e4 4919 return -EINVAL;
408f22e8 4920 len = sizeof(struct sctp_paddrparams);
1da177e4
LT
4921 if (copy_from_user(&params, optval, len))
4922 return -EFAULT;
4923
52ccb8e9
FF
4924 /* If an address other than INADDR_ANY is specified, and
4925 * no transport is found, then the request is invalid.
1da177e4 4926 */
cb3f837b 4927 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
52ccb8e9
FF
4928 trans = sctp_addr_id2transport(sk, &params.spp_address,
4929 params.spp_assoc_id);
4930 if (!trans) {
bb33381d 4931 pr_debug("%s: failed no transport\n", __func__);
52ccb8e9
FF
4932 return -EINVAL;
4933 }
1da177e4
LT
4934 }
4935
52ccb8e9
FF
4936 /* Get association, if assoc_id != 0 and the socket is a one
4937 * to many style socket, and an association was not found, then
4938 * the id was invalid.
4939 */
4940 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4941 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
bb33381d 4942 pr_debug("%s: failed no association\n", __func__);
1da177e4 4943 return -EINVAL;
52ccb8e9 4944 }
1da177e4 4945
52ccb8e9
FF
4946 if (trans) {
4947 /* Fetch transport values. */
4948 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4949 params.spp_pathmtu = trans->pathmtu;
4950 params.spp_pathmaxrxt = trans->pathmaxrxt;
4951 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4952
4953 /*draft-11 doesn't say what to return in spp_flags*/
4954 params.spp_flags = trans->param_flags;
4955 } else if (asoc) {
4956 /* Fetch association values. */
4957 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4958 params.spp_pathmtu = asoc->pathmtu;
4959 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4960 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4961
4962 /*draft-11 doesn't say what to return in spp_flags*/
4963 params.spp_flags = asoc->param_flags;
4964 } else {
4965 /* Fetch socket values. */
4966 params.spp_hbinterval = sp->hbinterval;
4967 params.spp_pathmtu = sp->pathmtu;
4968 params.spp_sackdelay = sp->sackdelay;
4969 params.spp_pathmaxrxt = sp->pathmaxrxt;
1da177e4 4970
52ccb8e9
FF
4971 /*draft-11 doesn't say what to return in spp_flags*/
4972 params.spp_flags = sp->param_flags;
4973 }
1da177e4 4974
1da177e4
LT
4975 if (copy_to_user(optval, &params, len))
4976 return -EFAULT;
4977
4978 if (put_user(len, optlen))
4979 return -EFAULT;
4980
4981 return 0;
4982}
4983
d364d927
WY
4984/*
4985 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4986 *
4987 * This option will effect the way delayed acks are performed. This
4988 * option allows you to get or set the delayed ack time, in
4989 * milliseconds. It also allows changing the delayed ack frequency.
4990 * Changing the frequency to 1 disables the delayed sack algorithm. If
4991 * the assoc_id is 0, then this sets or gets the endpoints default
4992 * values. If the assoc_id field is non-zero, then the set or get
4993 * effects the specified association for the one to many model (the
4994 * assoc_id field is ignored by the one to one model). Note that if
4995 * sack_delay or sack_freq are 0 when setting this option, then the
4996 * current values will remain unchanged.
4997 *
4998 * struct sctp_sack_info {
4999 * sctp_assoc_t sack_assoc_id;
5000 * uint32_t sack_delay;
5001 * uint32_t sack_freq;
5002 * };
7708610b 5003 *
d364d927
WY
5004 * sack_assoc_id - This parameter, indicates which association the user
5005 * is performing an action upon. Note that if this field's value is
5006 * zero then the endpoints default value is changed (effecting future
5007 * associations only).
7708610b 5008 *
d364d927
WY
5009 * sack_delay - This parameter contains the number of milliseconds that
5010 * the user is requesting the delayed ACK timer be set to. Note that
5011 * this value is defined in the standard to be between 200 and 500
5012 * milliseconds.
7708610b 5013 *
d364d927
WY
5014 * sack_freq - This parameter contains the number of packets that must
5015 * be received before a sack is sent without waiting for the delay
5016 * timer to expire. The default value for this is 2, setting this
5017 * value to 1 will disable the delayed sack algorithm.
7708610b 5018 */
d364d927 5019static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
7708610b
FF
5020 char __user *optval,
5021 int __user *optlen)
5022{
d364d927 5023 struct sctp_sack_info params;
7708610b
FF
5024 struct sctp_association *asoc = NULL;
5025 struct sctp_sock *sp = sctp_sk(sk);
5026
d364d927
WY
5027 if (len >= sizeof(struct sctp_sack_info)) {
5028 len = sizeof(struct sctp_sack_info);
7708610b 5029
d364d927
WY
5030 if (copy_from_user(&params, optval, len))
5031 return -EFAULT;
5032 } else if (len == sizeof(struct sctp_assoc_value)) {
94f65193 5033 pr_warn_ratelimited(DEPRECATED
f916ec96 5034 "%s (pid %d) "
94f65193 5035 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
f916ec96
NH
5036 "Use struct sctp_sack_info instead\n",
5037 current->comm, task_pid_nr(current));
d364d927
WY
5038 if (copy_from_user(&params, optval, len))
5039 return -EFAULT;
5040 } else
cb3f837b 5041 return -EINVAL;
7708610b 5042
d364d927 5043 /* Get association, if sack_assoc_id != 0 and the socket is a one
7708610b
FF
5044 * to many style socket, and an association was not found, then
5045 * the id was invalid.
d808ad9a 5046 */
d364d927
WY
5047 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5048 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
7708610b
FF
5049 return -EINVAL;
5050
5051 if (asoc) {
5052 /* Fetch association values. */
d364d927
WY
5053 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5054 params.sack_delay = jiffies_to_msecs(
7708610b 5055 asoc->sackdelay);
d364d927
WY
5056 params.sack_freq = asoc->sackfreq;
5057
5058 } else {
5059 params.sack_delay = 0;
5060 params.sack_freq = 1;
5061 }
7708610b
FF
5062 } else {
5063 /* Fetch socket values. */
d364d927
WY
5064 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5065 params.sack_delay = sp->sackdelay;
5066 params.sack_freq = sp->sackfreq;
5067 } else {
5068 params.sack_delay = 0;
5069 params.sack_freq = 1;
5070 }
7708610b
FF
5071 }
5072
5073 if (copy_to_user(optval, &params, len))
5074 return -EFAULT;
5075
5076 if (put_user(len, optlen))
5077 return -EFAULT;
5078
5079 return 0;
5080}
5081
1da177e4
LT
5082/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5083 *
5084 * Applications can specify protocol parameters for the default association
5085 * initialization. The option name argument to setsockopt() and getsockopt()
5086 * is SCTP_INITMSG.
5087 *
5088 * Setting initialization parameters is effective only on an unconnected
5089 * socket (for UDP-style sockets only future associations are effected
5090 * by the change). With TCP-style sockets, this option is inherited by
5091 * sockets derived from a listener socket.
5092 */
5093static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5094{
408f22e8 5095 if (len < sizeof(struct sctp_initmsg))
1da177e4 5096 return -EINVAL;
408f22e8
NH
5097 len = sizeof(struct sctp_initmsg);
5098 if (put_user(len, optlen))
5099 return -EFAULT;
1da177e4
LT
5100 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5101 return -EFAULT;
5102 return 0;
5103}
5104