blob: f694ee1167469d6affae0dde164fd933fd205afd [file] [log] [blame]
Vlad Yasevich60c778b2008-01-11 09:57:09 -05001/* SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * (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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -05009 * This file is part of the SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 *
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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050018 * This SCTP implementation is free software;
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * 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 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050024 * This SCTP implementation is distributed in the hope that it
Linus Torvalds1da177e2005-04-16 15:20:36 -070025 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
Joe Perches145ce502010-08-24 13:21:08 +000060#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/types.h>
63#include <linux/kernel.h>
64#include <linux/wait.h>
65#include <linux/time.h>
66#include <linux/ip.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080067#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/fcntl.h>
69#include <linux/poll.h>
70#include <linux/init.h>
71#include <linux/crypto.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090072#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74#include <net/ip.h>
75#include <net/icmp.h>
76#include <net/route.h>
77#include <net/ipv6.h>
78#include <net/inet_common.h>
79
80#include <linux/socket.h> /* for sa_family_t */
81#include <net/sock.h>
82#include <net/sctp/sctp.h>
83#include <net/sctp/sm.h>
84
85/* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
88 */
89
90/* Forward declarations for internal helper functions. */
91static int sctp_writeable(struct sock *sk);
92static void sctp_wfree(struct sk_buff *skb);
93static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97static int sctp_wait_for_accept(struct sock *sk, long timeo);
98static void sctp_wait_for_close(struct sock *sk, long timeo);
99static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108static int sctp_autobind(struct sock *sk);
109static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
112
Neil Horman4d93df02007-08-15 16:07:44 -0700113extern struct kmem_cache *sctp_bucket_cachep;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000114extern long sysctl_sctp_mem[3];
Neil Horman4d93df02007-08-15 16:07:44 -0700115extern int sysctl_sctp_rmem[3];
116extern int sysctl_sctp_wmem[3];
117
Adrian Bunkb6fa1a42007-09-12 15:18:00 +0200118static int sctp_memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000119static atomic_long_t sctp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800120struct percpu_counter sctp_sockets_allocated;
Neil Horman4d93df02007-08-15 16:07:44 -0700121
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700122static void sctp_enter_memory_pressure(struct sock *sk)
Neil Horman4d93df02007-08-15 16:07:44 -0700123{
124 sctp_memory_pressure = 1;
125}
126
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128/* Get the sndbuf space available at the time on the association. */
129static inline int sctp_wspace(struct sctp_association *asoc)
130{
Neil Horman4d93df02007-08-15 16:07:44 -0700131 int amt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Neil Horman4d93df02007-08-15 16:07:44 -0700133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
Eric Dumazet31e6d362009-06-17 19:05:41 -0700136 amt = sk_wmem_alloc_get(asoc->base.sk);
Neil Horman4d93df02007-08-15 16:07:44 -0700137
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
145 }
Neil Horman4eb701d2005-04-28 12:02:04 -0700146 } else {
Neil Horman4d93df02007-08-15 16:07:44 -0700147 amt = asoc->base.sk->sk_sndbuf - amt;
Neil Horman4eb701d2005-04-28 12:02:04 -0700148 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 return amt;
150}
151
152/* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
155 *
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
160 */
161static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
162{
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
165
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
168
Neil Horman4eb701d2005-04-28 12:02:04 -0700169 skb_set_owner_w(chunk->skb, sk);
170
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
174
Neil Horman4eb701d2005-04-28 12:02:04 -0700175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
178
Neil Horman4eb701d2005-04-28 12:02:04 -0700179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182}
183
184/* Verify that this is a valid address. */
185static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
186 int len)
187{
188 struct sctp_af *af;
189
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 if (!af)
193 return -EINVAL;
194
195 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return -EINVAL;
198
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
201
202 return 0;
203}
204
205/* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
207 */
208struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
209{
210 struct sctp_association *asoc = NULL;
211
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
217 */
218 if (!sctp_sstate(sk, ESTABLISHED))
219 return NULL;
220
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
225 return asoc;
226 }
227
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
230 return NULL;
231
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
235
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
237 return NULL;
238
239 return asoc;
240}
241
242/* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
244 * the same.
245 */
246static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
248 sctp_assoc_t id)
249{
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
253
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
Al Virocd4ff032006-11-20 17:11:33 -0800255 laddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 if (!addr_asoc)
259 return NULL;
260
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
264
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
267
268 return transport;
269}
270
271/* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
273 *
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
275 *
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
280 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700281SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282{
283 int retval = 0;
284
285 sctp_lock_sock(sk);
286
Frank Filz3f7a87d2005-06-20 13:14:57 -0700287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 addr_len);
294 else
295 retval = -EINVAL;
296
297 sctp_release_sock(sk);
298
299 return retval;
300}
301
302static long sctp_get_port_local(struct sock *, union sctp_addr *);
303
304/* Verify this is a valid sockaddr. */
305static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
307{
308 struct sctp_af *af;
309
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
312 return NULL;
313
Vlad Yasevich7dab83d2008-07-18 23:05:40 -0700314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
323 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
327
328 if (len < af->sockaddr_len)
329 return NULL;
330
331 return af;
332}
333
334/* Bind a local address either to an endpoint or to an association. */
335SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
336{
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
340 struct sctp_af *af;
341 unsigned short snum;
342 int ret = 0;
343
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700350 }
351
352 snum = ntohs(addr->v4.sin_port);
353
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
356 sk,
357 addr,
358 bp->port, snum,
359 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
364
Vlad Yasevich8b358052007-05-15 17:14:58 -0400365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
368 */
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
Vlad Yasevich8b358052007-05-15 17:14:58 -0400376 return -EINVAL;
377 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 }
379
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
382
Vlad Yasevich4e540642008-07-18 23:06:32 -0700383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
385 */
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
388
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
392 */
Vlad Yasevich2772b492007-08-21 14:24:30 +0900393 addr->v4.sin_port = htons(snum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 if ((ret = sctp_get_port_local(sk, addr))) {
Vlad Yasevich4e540642008-07-18 23:06:32 -0700395 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 }
397
398 /* Refresh ephemeral port. */
399 if (!bp->port)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000400 bp->port = inet_sk(sk)->inet_num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
Vlad Yasevich559cf712007-09-16 16:03:28 -0700402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
404 */
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 af->to_sk_saddr(addr, sk);
411 }
412
413 return ret;
414}
415
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * at any one time. If a sender, after sending an ASCONF chunk, decides
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * from each endpoint).
425 */
426static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
428{
429 int retval = 0;
430
431 /* If there is an outstanding ASCONF chunk, queue it for later
432 * transmission.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900433 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900436 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 }
438
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
446
447out:
448 return retval;
449}
450
451/* Add a list of addresses as bind addresses to local endpoint or
452 * association.
453 *
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
457 *
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
460 *
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
462 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200463static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464{
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
470
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
473
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
478 */
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
484 }
485
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 af->sockaddr_len);
488
489 addr_buf += af->sockaddr_len;
490
491err_bindx_add:
492 if (retval < 0) {
493 /* Failed. Cleanup the ones that have been added */
494 if (cnt > 0)
495 sctp_bindx_rem(sk, addrs, cnt);
496 return retval;
497 }
498 }
499
500 return retval;
501}
502
503/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
506 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900507 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
510 *
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 */
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900513static int sctp_send_asconf_add_ip(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 struct sockaddr *addrs,
515 int addrcnt)
516{
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700524 union sctp_addr saveaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 void *addr_buf;
526 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!sctp_addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800538 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700540 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
Vlad Yasevich559cf712007-09-16 16:03:28 -0700573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
Al Viro5ae955c2006-11-20 17:22:08 -0800579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
586 retval = sctp_send_asconf(asoc, chunk);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700587 if (retval)
588 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700593 addr_buf = addrs;
594 for (i = 0; i < addrcnt; i++) {
595 addr = (union sctp_addr *)addr_buf;
596 af = sctp_get_af_specific(addr->v4.sin_family);
597 memcpy(&saveaddr, addr, af->sockaddr_len);
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800598 retval = sctp_add_bind_addr(bp, &saveaddr,
599 SCTP_ADDR_NEW, GFP_ATOMIC);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700600 addr_buf += af->sockaddr_len;
601 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 }
603
604out:
605 return retval;
606}
607
608/* Remove a list of addresses from bind addresses list. Do not remove the
609 * last address.
610 *
611 * Basically run through each address specified in the addrs/addrcnt
612 * array/length pair, determine if it is IPv6 or IPv4 and call
613 * sctp_del_bind() on it.
614 *
615 * If any of them fails, then the operation will be reversed and the
616 * ones that were removed will be added back.
617 *
618 * At least one address has to be left; if only one address is
619 * available, the operation will return -EBUSY.
620 *
621 * Only sctp_setsockopt_bindx() is supposed to call this function.
622 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200623static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 struct sctp_sock *sp = sctp_sk(sk);
626 struct sctp_endpoint *ep = sp->ep;
627 int cnt;
628 struct sctp_bind_addr *bp = &ep->base.bind_addr;
629 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 void *addr_buf;
Al Viroc9a08502006-11-20 17:07:48 -0800631 union sctp_addr *sa_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 struct sctp_af *af;
633
634 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
635 sk, addrs, addrcnt);
636
637 addr_buf = addrs;
638 for (cnt = 0; cnt < addrcnt; cnt++) {
639 /* If the bind address list is empty or if there is only one
640 * bind address, there is nothing more to be removed (we need
641 * at least one address here).
642 */
643 if (list_empty(&bp->address_list) ||
644 (sctp_list_single_entry(&bp->address_list))) {
645 retval = -EBUSY;
646 goto err_bindx_rem;
647 }
648
Al Viroc9a08502006-11-20 17:07:48 -0800649 sa_addr = (union sctp_addr *)addr_buf;
650 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 if (!af) {
652 retval = -EINVAL;
653 goto err_bindx_rem;
654 }
Paolo Galtieri0304ff8a2007-04-17 12:52:36 -0700655
656 if (!af->addr_valid(sa_addr, sp, NULL)) {
657 retval = -EADDRNOTAVAIL;
658 goto err_bindx_rem;
659 }
660
Vlad Yasevichee9cbac2011-04-18 19:14:47 +0000661 if (sa_addr->v4.sin_port &&
662 sa_addr->v4.sin_port != htons(bp->port)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 retval = -EINVAL;
664 goto err_bindx_rem;
665 }
666
Vlad Yasevichee9cbac2011-04-18 19:14:47 +0000667 if (!sa_addr->v4.sin_port)
668 sa_addr->v4.sin_port = htons(bp->port);
669
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 /* FIXME - There is probably a need to check if sk->sk_saddr and
671 * sk->sk_rcv_addr are currently set to one of the addresses to
672 * be removed. This is something which needs to be looked into
673 * when we are fixing the outstanding issues with multi-homing
674 * socket routing and failover schemes. Refer to comments in
675 * sctp_do_bind(). -daisy
676 */
Vlad Yasevich0ed90fb2007-10-24 16:10:00 -0400677 retval = sctp_del_bind_addr(bp, sa_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
679 addr_buf += af->sockaddr_len;
680err_bindx_rem:
681 if (retval < 0) {
682 /* Failed. Add the ones that has been removed back */
683 if (cnt > 0)
684 sctp_bindx_add(sk, addrs, cnt);
685 return retval;
686 }
687 }
688
689 return retval;
690}
691
692/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
693 * the associations that are part of the endpoint indicating that a list of
694 * local addresses are removed from the endpoint.
695 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900696 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 * association, we do not send the chunk for that association. But it will not
698 * affect other associations.
699 *
700 * Only sctp_setsockopt_bindx() is supposed to call this function.
701 */
702static int sctp_send_asconf_del_ip(struct sock *sk,
703 struct sockaddr *addrs,
704 int addrcnt)
705{
706 struct sctp_sock *sp;
707 struct sctp_endpoint *ep;
708 struct sctp_association *asoc;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700709 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 struct sctp_bind_addr *bp;
711 struct sctp_chunk *chunk;
712 union sctp_addr *laddr;
713 void *addr_buf;
714 struct sctp_af *af;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700715 struct sctp_sockaddr_entry *saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 int i;
717 int retval = 0;
718
719 if (!sctp_addip_enable)
720 return retval;
721
722 sp = sctp_sk(sk);
723 ep = sp->ep;
724
725 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800726 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700728 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729
730 if (!asoc->peer.asconf_capable)
731 continue;
732
733 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
734 continue;
735
736 if (!sctp_state(asoc, ESTABLISHED))
737 continue;
738
739 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900740 * not present in the bind address list of the association.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 * If so, do not send the asconf chunk to its peer, but
742 * continue with other associations.
743 */
744 addr_buf = addrs;
745 for (i = 0; i < addrcnt; i++) {
746 laddr = (union sctp_addr *)addr_buf;
747 af = sctp_get_af_specific(laddr->v4.sin_family);
748 if (!af) {
749 retval = -EINVAL;
750 goto out;
751 }
752
753 if (!sctp_assoc_lookup_laddr(asoc, laddr))
754 break;
755
756 addr_buf += af->sockaddr_len;
757 }
758 if (i < addrcnt)
759 continue;
760
761 /* Find one address in the association's bind address list
762 * that is not in the packed array of addresses. This is to
763 * make sure that we do not delete all the addresses in the
764 * association.
765 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 bp = &asoc->base.bind_addr;
767 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
768 addrcnt, sp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 if (!laddr)
770 continue;
771
Vlad Yasevich559cf712007-09-16 16:03:28 -0700772 /* We do not need RCU protection throughout this loop
773 * because this is done under a socket lock from the
774 * setsockopt call.
775 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
777 SCTP_PARAM_DEL_IP);
778 if (!chunk) {
779 retval = -ENOMEM;
780 goto out;
781 }
782
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700783 /* Reset use_as_src flag for the addresses in the bind address
784 * list that are to be deleted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700786 addr_buf = addrs;
787 for (i = 0; i < addrcnt; i++) {
788 laddr = (union sctp_addr *)addr_buf;
789 af = sctp_get_af_specific(laddr->v4.sin_family);
Vlad Yasevich559cf712007-09-16 16:03:28 -0700790 list_for_each_entry(saddr, &bp->address_list, list) {
Al Viro5f242a12006-11-20 17:05:23 -0800791 if (sctp_cmp_addr_exact(&saddr->a, laddr))
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800792 saddr->state = SCTP_ADDR_DEL;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700793 }
794 addr_buf += af->sockaddr_len;
795 }
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700796
797 /* Update the route and saddr entries for all the transports
798 * as some of the addresses in the bind address list are
799 * about to be deleted and cannot be used as source addresses.
800 */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700801 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
802 transports) {
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700803 dst_release(transport->dst);
804 sctp_transport_route(transport, NULL,
805 sctp_sk(asoc->base.sk));
806 }
807
808 retval = sctp_send_asconf(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 }
810out:
811 return retval;
812}
813
814/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
815 *
816 * API 8.1
817 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
818 * int flags);
819 *
820 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
821 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
822 * or IPv6 addresses.
823 *
824 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
825 * Section 3.1.2 for this usage.
826 *
827 * addrs is a pointer to an array of one or more socket addresses. Each
828 * address is contained in its appropriate structure (i.e. struct
829 * sockaddr_in or struct sockaddr_in6) the family of the address type
Ville Nuorvala23c435f2006-10-16 22:08:28 -0700830 * must be used to distinguish the address length (note that this
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 * representation is termed a "packed array" of addresses). The caller
832 * specifies the number of addresses in the array with addrcnt.
833 *
834 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
835 * -1, and sets errno to the appropriate error code.
836 *
837 * For SCTP, the port given in each socket address must be the same, or
838 * sctp_bindx() will fail, setting errno to EINVAL.
839 *
840 * The flags parameter is formed from the bitwise OR of zero or more of
841 * the following currently defined flags:
842 *
843 * SCTP_BINDX_ADD_ADDR
844 *
845 * SCTP_BINDX_REM_ADDR
846 *
847 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
848 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
849 * addresses from the association. The two flags are mutually exclusive;
850 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
851 * not remove all addresses from an association; sctp_bindx() will
852 * reject such an attempt with EINVAL.
853 *
854 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
855 * additional addresses with an endpoint after calling bind(). Or use
856 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
857 * socket is associated with so that no new association accepted will be
858 * associated with those addresses. If the endpoint supports dynamic
859 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
860 * endpoint to send the appropriate message to the peer to change the
861 * peers address lists.
862 *
863 * Adding and removing addresses from a connected association is
864 * optional functionality. Implementations that do not support this
865 * functionality should return EOPNOTSUPP.
866 *
867 * Basically do nothing but copying the addresses from user to kernel
868 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700869 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
870 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 *
872 * We don't use copy_from_user() for optimization: we first do the
873 * sanity checks (buffer size -fast- and access check-healthy
874 * pointer); if all of those succeed, then we can alloc the memory
875 * (expensive operation) needed to copy the data to kernel. Then we do
876 * the copying without checking the user space area
877 * (__copy_from_user()).
878 *
879 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
880 * it.
881 *
882 * sk The sk of the socket
883 * addrs The pointer to the addresses in user land
884 * addrssize Size of the addrs buffer
885 * op Operation to perform (add or remove, see the flags of
886 * sctp_bindx)
887 *
888 * Returns 0 if ok, <0 errno code on error.
889 */
890SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
891 struct sockaddr __user *addrs,
892 int addrs_size, int op)
893{
894 struct sockaddr *kaddrs;
895 int err;
896 int addrcnt = 0;
897 int walk_size = 0;
898 struct sockaddr *sa_addr;
899 void *addr_buf;
900 struct sctp_af *af;
901
902 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
903 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
904
905 if (unlikely(addrs_size <= 0))
906 return -EINVAL;
907
908 /* Check the user passed a healthy pointer. */
909 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
910 return -EFAULT;
911
912 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800913 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 if (unlikely(!kaddrs))
915 return -ENOMEM;
916
917 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
918 kfree(kaddrs);
919 return -EFAULT;
920 }
921
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900922 /* Walk through the addrs buffer and count the number of addresses. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 addr_buf = kaddrs;
924 while (walk_size < addrs_size) {
Dan Rosenbergd7e0d192010-10-01 11:16:58 +0000925 if (walk_size + sizeof(sa_family_t) > addrs_size) {
926 kfree(kaddrs);
927 return -EINVAL;
928 }
929
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 sa_addr = (struct sockaddr *)addr_buf;
931 af = sctp_get_af_specific(sa_addr->sa_family);
932
933 /* If the address family is not supported or if this address
934 * causes the address buffer to overflow return EINVAL.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900935 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
937 kfree(kaddrs);
938 return -EINVAL;
939 }
940 addrcnt++;
941 addr_buf += af->sockaddr_len;
942 walk_size += af->sockaddr_len;
943 }
944
945 /* Do the work. */
946 switch (op) {
947 case SCTP_BINDX_ADD_ADDR:
948 err = sctp_bindx_add(sk, kaddrs, addrcnt);
949 if (err)
950 goto out;
951 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
952 break;
953
954 case SCTP_BINDX_REM_ADDR:
955 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
956 if (err)
957 goto out;
958 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
959 break;
960
961 default:
962 err = -EINVAL;
963 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700964 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
966out:
967 kfree(kaddrs);
968
969 return err;
970}
971
Frank Filz3f7a87d2005-06-20 13:14:57 -0700972/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
973 *
974 * Common routine for handling connect() and sctp_connectx().
975 * Connect will come in with just a single address.
976 */
977static int __sctp_connect(struct sock* sk,
978 struct sockaddr *kaddrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -0700979 int addrs_size,
980 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700981{
982 struct sctp_sock *sp;
983 struct sctp_endpoint *ep;
984 struct sctp_association *asoc = NULL;
985 struct sctp_association *asoc2;
986 struct sctp_transport *transport;
987 union sctp_addr to;
988 struct sctp_af *af;
989 sctp_scope_t scope;
990 long timeo;
991 int err = 0;
992 int addrcnt = 0;
993 int walk_size = 0;
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -0400994 union sctp_addr *sa_addr = NULL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700995 void *addr_buf;
Vlad Yasevich16d00fb2007-05-04 13:34:09 -0700996 unsigned short port;
Vlad Yasevichf50f95c2007-07-03 12:47:40 -0400997 unsigned int f_flags = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700998
999 sp = sctp_sk(sk);
1000 ep = sp->ep;
1001
1002 /* connect() cannot be done on a socket that is already in ESTABLISHED
1003 * state - UDP-style peeled off socket or a TCP-style socket that
1004 * is already connected.
1005 * It cannot be done even on a TCP-style listening socket.
1006 */
1007 if (sctp_sstate(sk, ESTABLISHED) ||
1008 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1009 err = -EISCONN;
1010 goto out_free;
1011 }
1012
1013 /* Walk through the addrs buffer and count the number of addresses. */
1014 addr_buf = kaddrs;
1015 while (walk_size < addrs_size) {
Dan Rosenbergd7e0d192010-10-01 11:16:58 +00001016 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1017 err = -EINVAL;
1018 goto out_free;
1019 }
1020
Al Viro4bdf4b52006-11-20 17:10:20 -08001021 sa_addr = (union sctp_addr *)addr_buf;
1022 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001023
1024 /* If the address family is not supported or if this address
1025 * causes the address buffer to overflow return EINVAL.
1026 */
1027 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1028 err = -EINVAL;
1029 goto out_free;
1030 }
1031
Dan Rosenbergd7e0d192010-10-01 11:16:58 +00001032 port = ntohs(sa_addr->v4.sin_port);
1033
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001034 /* Save current address so we can work with it */
1035 memcpy(&to, sa_addr, af->sockaddr_len);
1036
1037 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001038 if (err)
1039 goto out_free;
1040
Vlad Yasevich16d00fb2007-05-04 13:34:09 -07001041 /* Make sure the destination port is correctly set
1042 * in all addresses.
1043 */
1044 if (asoc && asoc->peer.port && asoc->peer.port != port)
1045 goto out_free;
1046
Frank Filz3f7a87d2005-06-20 13:14:57 -07001047
1048 /* Check if there already is a matching association on the
1049 * endpoint (other than the one created here).
1050 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001051 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001052 if (asoc2 && asoc2 != asoc) {
1053 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1054 err = -EISCONN;
1055 else
1056 err = -EALREADY;
1057 goto out_free;
1058 }
1059
1060 /* If we could not find a matching association on the endpoint,
1061 * make sure that there is no peeled-off association matching
1062 * the peer address even on another socket.
1063 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001064 if (sctp_endpoint_is_peeled_off(ep, &to)) {
Frank Filz3f7a87d2005-06-20 13:14:57 -07001065 err = -EADDRNOTAVAIL;
1066 goto out_free;
1067 }
1068
1069 if (!asoc) {
1070 /* If a bind() or sctp_bindx() is not called prior to
1071 * an sctp_connectx() call, the system picks an
1072 * ephemeral port and will choose an address set
1073 * equivalent to binding with a wildcard address.
1074 */
1075 if (!ep->base.bind_addr.port) {
1076 if (sctp_autobind(sk)) {
1077 err = -EAGAIN;
1078 goto out_free;
1079 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001080 } else {
1081 /*
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001082 * If an unprivileged user inherits a 1-many
1083 * style socket with open associations on a
1084 * privileged port, it MAY be permitted to
1085 * accept new associations, but it SHOULD NOT
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001086 * be permitted to open new associations.
1087 */
1088 if (ep->base.bind_addr.port < PROT_SOCK &&
1089 !capable(CAP_NET_BIND_SERVICE)) {
1090 err = -EACCES;
1091 goto out_free;
1092 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001093 }
1094
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001095 scope = sctp_scope(&to);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001096 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1097 if (!asoc) {
1098 err = -ENOMEM;
1099 goto out_free;
1100 }
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001101
1102 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1103 GFP_KERNEL);
1104 if (err < 0) {
1105 goto out_free;
1106 }
1107
Frank Filz3f7a87d2005-06-20 13:14:57 -07001108 }
1109
1110 /* Prime the peer's transport structures. */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001111 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001112 SCTP_UNKNOWN);
1113 if (!transport) {
1114 err = -ENOMEM;
1115 goto out_free;
1116 }
1117
1118 addrcnt++;
1119 addr_buf += af->sockaddr_len;
1120 walk_size += af->sockaddr_len;
1121 }
1122
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001123 /* In case the user of sctp_connectx() wants an association
1124 * id back, assign one now.
1125 */
1126 if (assoc_id) {
1127 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1128 if (err < 0)
1129 goto out_free;
1130 }
1131
Frank Filz3f7a87d2005-06-20 13:14:57 -07001132 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1133 if (err < 0) {
1134 goto out_free;
1135 }
1136
1137 /* Initialize sk's dport and daddr for getpeername() */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001138 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001139 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1140 af->to_sk_daddr(sa_addr, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001141 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001142
Vlad Yasevichf50f95c2007-07-03 12:47:40 -04001143 /* in-kernel sockets don't generally have a file allocated to them
1144 * if all they do is call sock_create_kern().
1145 */
1146 if (sk->sk_socket->file)
1147 f_flags = sk->sk_socket->file->f_flags;
1148
1149 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1150
Frank Filz3f7a87d2005-06-20 13:14:57 -07001151 err = sctp_wait_for_connect(asoc, &timeo);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001152 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001153 *assoc_id = asoc->assoc_id;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001154
1155 /* Don't free association on exit. */
1156 asoc = NULL;
1157
1158out_free:
1159
1160 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001161 " kaddrs: %p err: %d\n",
1162 asoc, kaddrs, err);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001163 if (asoc)
1164 sctp_association_free(asoc);
1165 return err;
1166}
1167
1168/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1169 *
1170 * API 8.9
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001171 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1172 * sctp_assoc_t *asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001173 *
1174 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1175 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1176 * or IPv6 addresses.
1177 *
1178 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1179 * Section 3.1.2 for this usage.
1180 *
1181 * addrs is a pointer to an array of one or more socket addresses. Each
1182 * address is contained in its appropriate structure (i.e. struct
1183 * sockaddr_in or struct sockaddr_in6) the family of the address type
1184 * must be used to distengish the address length (note that this
1185 * representation is termed a "packed array" of addresses). The caller
1186 * specifies the number of addresses in the array with addrcnt.
1187 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001188 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1189 * the association id of the new association. On failure, sctp_connectx()
1190 * returns -1, and sets errno to the appropriate error code. The assoc_id
1191 * is not touched by the kernel.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001192 *
1193 * For SCTP, the port given in each socket address must be the same, or
1194 * sctp_connectx() will fail, setting errno to EINVAL.
1195 *
1196 * An application can use sctp_connectx to initiate an association with
1197 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1198 * allows a caller to specify multiple addresses at which a peer can be
1199 * reached. The way the SCTP stack uses the list of addresses to set up
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001200 * the association is implementation dependent. This function only
Frank Filz3f7a87d2005-06-20 13:14:57 -07001201 * specifies that the stack will try to make use of all the addresses in
1202 * the list when needed.
1203 *
1204 * Note that the list of addresses passed in is only used for setting up
1205 * the association. It does not necessarily equal the set of addresses
1206 * the peer uses for the resulting association. If the caller wants to
1207 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1208 * retrieve them after the association has been set up.
1209 *
1210 * Basically do nothing but copying the addresses from user to kernel
1211 * land and invoking either sctp_connectx(). This is used for tunneling
1212 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1213 *
1214 * We don't use copy_from_user() for optimization: we first do the
1215 * sanity checks (buffer size -fast- and access check-healthy
1216 * pointer); if all of those succeed, then we can alloc the memory
1217 * (expensive operation) needed to copy the data to kernel. Then we do
1218 * the copying without checking the user space area
1219 * (__copy_from_user()).
1220 *
1221 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1222 * it.
1223 *
1224 * sk The sk of the socket
1225 * addrs The pointer to the addresses in user land
1226 * addrssize Size of the addrs buffer
1227 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001228 * Returns >=0 if ok, <0 errno code on error.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001229 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001230SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001231 struct sockaddr __user *addrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001232 int addrs_size,
1233 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -07001234{
1235 int err = 0;
1236 struct sockaddr *kaddrs;
1237
1238 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001239 __func__, sk, addrs, addrs_size);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001240
1241 if (unlikely(addrs_size <= 0))
1242 return -EINVAL;
1243
1244 /* Check the user passed a healthy pointer. */
1245 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1246 return -EFAULT;
1247
1248 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001249 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001250 if (unlikely(!kaddrs))
1251 return -ENOMEM;
1252
1253 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1254 err = -EFAULT;
1255 } else {
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001256 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001257 }
1258
1259 kfree(kaddrs);
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001260
Frank Filz3f7a87d2005-06-20 13:14:57 -07001261 return err;
1262}
1263
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001264/*
1265 * This is an older interface. It's kept for backward compatibility
1266 * to the option that doesn't provide association id.
1267 */
1268SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1269 struct sockaddr __user *addrs,
1270 int addrs_size)
1271{
1272 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1273}
1274
1275/*
1276 * New interface for the API. The since the API is done with a socket
1277 * option, to make it simple we feed back the association id is as a return
1278 * indication to the call. Error is always negative and association id is
1279 * always positive.
1280 */
1281SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1282 struct sockaddr __user *addrs,
1283 int addrs_size)
1284{
1285 sctp_assoc_t assoc_id = 0;
1286 int err = 0;
1287
1288 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1289
1290 if (err)
1291 return err;
1292 else
1293 return assoc_id;
1294}
1295
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001296/*
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001297 * New (hopefully final) interface for the API.
1298 * We use the sctp_getaddrs_old structure so that use-space library
1299 * can avoid any unnecessary allocations. The only defferent part
1300 * is that we store the actual length of the address buffer into the
1301 * addrs_num structure member. That way we can re-use the existing
1302 * code.
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001303 */
1304SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1305 char __user *optval,
1306 int __user *optlen)
1307{
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001308 struct sctp_getaddrs_old param;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001309 sctp_assoc_t assoc_id = 0;
1310 int err = 0;
1311
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001312 if (len < sizeof(param))
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001313 return -EINVAL;
1314
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001315 if (copy_from_user(&param, optval, sizeof(param)))
1316 return -EFAULT;
1317
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001318 err = __sctp_setsockopt_connectx(sk,
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001319 (struct sockaddr __user *)param.addrs,
1320 param.addr_num, &assoc_id);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001321
1322 if (err == 0 || err == -EINPROGRESS) {
1323 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1324 return -EFAULT;
1325 if (put_user(sizeof(assoc_id), optlen))
1326 return -EFAULT;
1327 }
1328
1329 return err;
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332/* API 3.1.4 close() - UDP Style Syntax
1333 * Applications use close() to perform graceful shutdown (as described in
1334 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1335 * by a UDP-style socket.
1336 *
1337 * The syntax is
1338 *
1339 * ret = close(int sd);
1340 *
1341 * sd - the socket descriptor of the associations to be closed.
1342 *
1343 * To gracefully shutdown a specific association represented by the
1344 * UDP-style socket, an application should use the sendmsg() call,
1345 * passing no user data, but including the appropriate flag in the
1346 * ancillary data (see Section xxxx).
1347 *
1348 * If sd in the close() call is a branched-off socket representing only
1349 * one association, the shutdown is performed on that association only.
1350 *
1351 * 4.1.6 close() - TCP Style Syntax
1352 *
1353 * Applications use close() to gracefully close down an association.
1354 *
1355 * The syntax is:
1356 *
1357 * int close(int sd);
1358 *
1359 * sd - the socket descriptor of the association to be closed.
1360 *
1361 * After an application calls close() on a socket descriptor, no further
1362 * socket operations will succeed on that descriptor.
1363 *
1364 * API 7.1.4 SO_LINGER
1365 *
1366 * An application using the TCP-style socket can use this option to
1367 * perform the SCTP ABORT primitive. The linger option structure is:
1368 *
1369 * struct linger {
1370 * int l_onoff; // option on/off
1371 * int l_linger; // linger time
1372 * };
1373 *
1374 * To enable the option, set l_onoff to 1. If the l_linger value is set
1375 * to 0, calling close() is the same as the ABORT primitive. If the
1376 * value is set to a negative value, the setsockopt() call will return
1377 * an error. If the value is set to a positive value linger_time, the
1378 * close() can be blocked for at most linger_time ms. If the graceful
1379 * shutdown phase does not finish during this period, close() will
1380 * return but the graceful shutdown phase continues in the system.
1381 */
1382SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1383{
1384 struct sctp_endpoint *ep;
1385 struct sctp_association *asoc;
1386 struct list_head *pos, *temp;
1387
1388 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1389
1390 sctp_lock_sock(sk);
1391 sk->sk_shutdown = SHUTDOWN_MASK;
Vlad Yasevichbec96402009-07-30 18:08:28 -04001392 sk->sk_state = SCTP_SS_CLOSING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
1394 ep = sctp_sk(sk)->ep;
1395
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001396 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 list_for_each_safe(pos, temp, &ep->asocs) {
1398 asoc = list_entry(pos, struct sctp_association, asocs);
1399
1400 if (sctp_style(sk, TCP)) {
1401 /* A closed association can still be in the list if
1402 * it belongs to a TCP-style listening socket that is
1403 * not yet accepted. If so, free it. If not, send an
1404 * ABORT or SHUTDOWN based on the linger options.
1405 */
1406 if (sctp_state(asoc, CLOSED)) {
1407 sctp_unhash_established(asoc);
1408 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001409 continue;
1410 }
1411 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
Sridhar Samudralab9ac8672006-08-28 13:53:01 -07001413 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1414 struct sctp_chunk *chunk;
1415
1416 chunk = sctp_make_abort_user(asoc, NULL, 0);
1417 if (chunk)
1418 sctp_primitive_ABORT(asoc, chunk);
1419 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 sctp_primitive_SHUTDOWN(asoc, NULL);
1421 }
1422
1423 /* Clean up any skbs sitting on the receive queue. */
1424 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1425 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1426
1427 /* On a TCP-style socket, block for at most linger_time if set. */
1428 if (sctp_style(sk, TCP) && timeout)
1429 sctp_wait_for_close(sk, timeout);
1430
1431 /* This will run the backlog queue. */
1432 sctp_release_sock(sk);
1433
1434 /* Supposedly, no process has access to the socket, but
1435 * the net layers still may.
1436 */
1437 sctp_local_bh_disable();
1438 sctp_bh_lock_sock(sk);
1439
1440 /* Hold the sock, since sk_common_release() will put sock_put()
1441 * and we have just a little more cleanup.
1442 */
1443 sock_hold(sk);
1444 sk_common_release(sk);
1445
1446 sctp_bh_unlock_sock(sk);
1447 sctp_local_bh_enable();
1448
1449 sock_put(sk);
1450
1451 SCTP_DBG_OBJCNT_DEC(sock);
1452}
1453
1454/* Handle EPIPE error. */
1455static int sctp_error(struct sock *sk, int flags, int err)
1456{
1457 if (err == -EPIPE)
1458 err = sock_error(sk) ? : -EPIPE;
1459 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1460 send_sig(SIGPIPE, current, 0);
1461 return err;
1462}
1463
1464/* API 3.1.3 sendmsg() - UDP Style Syntax
1465 *
1466 * An application uses sendmsg() and recvmsg() calls to transmit data to
1467 * and receive data from its peer.
1468 *
1469 * ssize_t sendmsg(int socket, const struct msghdr *message,
1470 * int flags);
1471 *
1472 * socket - the socket descriptor of the endpoint.
1473 * message - pointer to the msghdr structure which contains a single
1474 * user message and possibly some ancillary data.
1475 *
1476 * See Section 5 for complete description of the data
1477 * structures.
1478 *
1479 * flags - flags sent or received with the user message, see Section
1480 * 5 for complete description of the flags.
1481 *
1482 * Note: This function could use a rewrite especially when explicit
1483 * connect support comes in.
1484 */
1485/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1486
1487SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1488
1489SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1490 struct msghdr *msg, size_t msg_len)
1491{
1492 struct sctp_sock *sp;
1493 struct sctp_endpoint *ep;
1494 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1495 struct sctp_transport *transport, *chunk_tp;
1496 struct sctp_chunk *chunk;
Al Virodce116a2006-11-20 17:25:15 -08001497 union sctp_addr to;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 struct sockaddr *msg_name = NULL;
1499 struct sctp_sndrcvinfo default_sinfo = { 0 };
1500 struct sctp_sndrcvinfo *sinfo;
1501 struct sctp_initmsg *sinit;
1502 sctp_assoc_t associd = 0;
1503 sctp_cmsgs_t cmsgs = { NULL };
1504 int err;
1505 sctp_scope_t scope;
1506 long timeo;
1507 __u16 sinfo_flags = 0;
1508 struct sctp_datamsg *datamsg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 int msg_flags = msg->msg_flags;
1510
1511 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1512 sk, msg, msg_len);
1513
1514 err = 0;
1515 sp = sctp_sk(sk);
1516 ep = sp->ep;
1517
Frank Filz3f7a87d2005-06-20 13:14:57 -07001518 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
1520 /* We cannot send a message over a TCP-style listening socket. */
1521 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1522 err = -EPIPE;
1523 goto out_nounlock;
1524 }
1525
1526 /* Parse out the SCTP CMSGs. */
1527 err = sctp_msghdr_parse(msg, &cmsgs);
1528
1529 if (err) {
1530 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1531 goto out_nounlock;
1532 }
1533
1534 /* Fetch the destination address for this packet. This
1535 * address only selects the association--it is not necessarily
1536 * the address we will send to.
1537 * For a peeled-off socket, msg_name is ignored.
1538 */
1539 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1540 int msg_namelen = msg->msg_namelen;
1541
1542 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1543 msg_namelen);
1544 if (err)
1545 return err;
1546
1547 if (msg_namelen > sizeof(to))
1548 msg_namelen = sizeof(to);
1549 memcpy(&to, msg->msg_name, msg_namelen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 msg_name = msg->msg_name;
1551 }
1552
1553 sinfo = cmsgs.info;
1554 sinit = cmsgs.init;
1555
1556 /* Did the user specify SNDRCVINFO? */
1557 if (sinfo) {
1558 sinfo_flags = sinfo->sinfo_flags;
1559 associd = sinfo->sinfo_assoc_id;
1560 }
1561
1562 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1563 msg_len, sinfo_flags);
1564
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001565 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1566 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 err = -EINVAL;
1568 goto out_nounlock;
1569 }
1570
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001571 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1572 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1573 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 * the msg_iov set to the user abort reason.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001575 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001576 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1577 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 err = -EINVAL;
1579 goto out_nounlock;
1580 }
1581
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001582 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 * specified in msg_name.
1584 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001585 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 err = -EINVAL;
1587 goto out_nounlock;
1588 }
1589
1590 transport = NULL;
1591
1592 SCTP_DEBUG_PRINTK("About to look up association.\n");
1593
1594 sctp_lock_sock(sk);
1595
1596 /* If a msg_name has been specified, assume this is to be used. */
1597 if (msg_name) {
1598 /* Look for a matching association on the endpoint. */
Al Virodce116a2006-11-20 17:25:15 -08001599 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 if (!asoc) {
1601 /* If we could not find a matching association on the
1602 * endpoint, make sure that it is not a TCP-style
1603 * socket that already has an association or there is
1604 * no peeled-off association on another socket.
1605 */
1606 if ((sctp_style(sk, TCP) &&
1607 sctp_sstate(sk, ESTABLISHED)) ||
Al Virodce116a2006-11-20 17:25:15 -08001608 sctp_endpoint_is_peeled_off(ep, &to)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 err = -EADDRNOTAVAIL;
1610 goto out_unlock;
1611 }
1612 }
1613 } else {
1614 asoc = sctp_id2assoc(sk, associd);
1615 if (!asoc) {
1616 err = -EPIPE;
1617 goto out_unlock;
1618 }
1619 }
1620
1621 if (asoc) {
1622 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1623
1624 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1625 * socket that has an association in CLOSED state. This can
1626 * happen when an accepted socket has an association that is
1627 * already CLOSED.
1628 */
1629 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1630 err = -EPIPE;
1631 goto out_unlock;
1632 }
1633
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001634 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1636 asoc);
1637 sctp_primitive_SHUTDOWN(asoc, NULL);
1638 err = 0;
1639 goto out_unlock;
1640 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001641 if (sinfo_flags & SCTP_ABORT) {
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001642
1643 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1644 if (!chunk) {
1645 err = -ENOMEM;
1646 goto out_unlock;
1647 }
1648
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001650 sctp_primitive_ABORT(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 err = 0;
1652 goto out_unlock;
1653 }
1654 }
1655
1656 /* Do we need to create the association? */
1657 if (!asoc) {
1658 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1659
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001660 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 err = -EINVAL;
1662 goto out_unlock;
1663 }
1664
1665 /* Check for invalid stream against the stream counts,
1666 * either the default or the user specified stream counts.
1667 */
1668 if (sinfo) {
1669 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1670 /* Check against the defaults. */
1671 if (sinfo->sinfo_stream >=
1672 sp->initmsg.sinit_num_ostreams) {
1673 err = -EINVAL;
1674 goto out_unlock;
1675 }
1676 } else {
1677 /* Check against the requested. */
1678 if (sinfo->sinfo_stream >=
1679 sinit->sinit_num_ostreams) {
1680 err = -EINVAL;
1681 goto out_unlock;
1682 }
1683 }
1684 }
1685
1686 /*
1687 * API 3.1.2 bind() - UDP Style Syntax
1688 * If a bind() or sctp_bindx() is not called prior to a
1689 * sendmsg() call that initiates a new association, the
1690 * system picks an ephemeral port and will choose an address
1691 * set equivalent to binding with a wildcard address.
1692 */
1693 if (!ep->base.bind_addr.port) {
1694 if (sctp_autobind(sk)) {
1695 err = -EAGAIN;
1696 goto out_unlock;
1697 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001698 } else {
1699 /*
1700 * If an unprivileged user inherits a one-to-many
1701 * style socket with open associations on a privileged
1702 * port, it MAY be permitted to accept new associations,
1703 * but it SHOULD NOT be permitted to open new
1704 * associations.
1705 */
1706 if (ep->base.bind_addr.port < PROT_SOCK &&
1707 !capable(CAP_NET_BIND_SERVICE)) {
1708 err = -EACCES;
1709 goto out_unlock;
1710 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 }
1712
1713 scope = sctp_scope(&to);
1714 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1715 if (!new_asoc) {
1716 err = -ENOMEM;
1717 goto out_unlock;
1718 }
1719 asoc = new_asoc;
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001720 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1721 if (err < 0) {
1722 err = -ENOMEM;
1723 goto out_free;
1724 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725
1726 /* If the SCTP_INIT ancillary data is specified, set all
1727 * the association init values accordingly.
1728 */
1729 if (sinit) {
1730 if (sinit->sinit_num_ostreams) {
1731 asoc->c.sinit_num_ostreams =
1732 sinit->sinit_num_ostreams;
1733 }
1734 if (sinit->sinit_max_instreams) {
1735 asoc->c.sinit_max_instreams =
1736 sinit->sinit_max_instreams;
1737 }
1738 if (sinit->sinit_max_attempts) {
1739 asoc->max_init_attempts
1740 = sinit->sinit_max_attempts;
1741 }
1742 if (sinit->sinit_max_init_timeo) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001743 asoc->max_init_timeo =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1745 }
1746 }
1747
1748 /* Prime the peer's transport structures. */
Al Virodce116a2006-11-20 17:25:15 -08001749 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (!transport) {
1751 err = -ENOMEM;
1752 goto out_free;
1753 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 }
1755
1756 /* ASSERT: we have a valid association at this point. */
1757 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1758
1759 if (!sinfo) {
1760 /* If the user didn't specify SNDRCVINFO, make up one with
1761 * some defaults.
1762 */
1763 default_sinfo.sinfo_stream = asoc->default_stream;
1764 default_sinfo.sinfo_flags = asoc->default_flags;
1765 default_sinfo.sinfo_ppid = asoc->default_ppid;
1766 default_sinfo.sinfo_context = asoc->default_context;
1767 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1768 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1769 sinfo = &default_sinfo;
1770 }
1771
1772 /* API 7.1.7, the sndbuf size per association bounds the
1773 * maximum size of data that can be sent in a single send call.
1774 */
1775 if (msg_len > sk->sk_sndbuf) {
1776 err = -EMSGSIZE;
1777 goto out_free;
1778 }
1779
Vlad Yasevich8a479492007-06-07 14:21:05 -04001780 if (asoc->pmtu_pending)
1781 sctp_assoc_pending_pmtu(asoc);
1782
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 /* If fragmentation is disabled and the message length exceeds the
1784 * association fragmentation point, return EMSGSIZE. The I-D
1785 * does not specify what this error is, but this looks like
1786 * a great fit.
1787 */
1788 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1789 err = -EMSGSIZE;
1790 goto out_free;
1791 }
1792
1793 if (sinfo) {
1794 /* Check for invalid stream. */
1795 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1796 err = -EINVAL;
1797 goto out_free;
1798 }
1799 }
1800
1801 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1802 if (!sctp_wspace(asoc)) {
1803 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1804 if (err)
1805 goto out_free;
1806 }
1807
1808 /* If an address is passed with the sendto/sendmsg call, it is used
1809 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001810 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 */
1812 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001813 (sinfo_flags & SCTP_ADDR_OVER)) {
Al Virodce116a2006-11-20 17:25:15 -08001814 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 if (!chunk_tp) {
1816 err = -EINVAL;
1817 goto out_free;
1818 }
1819 } else
1820 chunk_tp = NULL;
1821
1822 /* Auto-connect, if we aren't connected already. */
1823 if (sctp_state(asoc, CLOSED)) {
1824 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1825 if (err < 0)
1826 goto out_free;
1827 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1828 }
1829
1830 /* Break the message into multiple chunks of maximum size. */
1831 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1832 if (!datamsg) {
1833 err = -ENOMEM;
1834 goto out_free;
1835 }
1836
1837 /* Now send the (possibly) fragmented message. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001838 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
Florian Westphal80445cf2008-03-23 22:47:08 -07001839 sctp_chunk_hold(chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
1841 /* Do accounting for the write space. */
1842 sctp_set_owner_w(chunk);
1843
1844 chunk->transport = chunk_tp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 }
1846
Vlad Yasevich9c5c62b2009-08-10 13:51:03 -04001847 /* Send it to the lower layers. Note: all chunks
1848 * must either fail or succeed. The lower layer
1849 * works that way today. Keep it that way or this
1850 * breaks.
1851 */
1852 err = sctp_primitive_SEND(asoc, datamsg);
1853 /* Did the lower layer accept the chunk? */
1854 if (err)
1855 sctp_datamsg_free(datamsg);
1856 else
1857 sctp_datamsg_put(datamsg);
1858
1859 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 if (err)
1862 goto out_free;
1863 else
1864 err = msg_len;
1865
1866 /* If we are already past ASSOCIATE, the lower
1867 * layers are responsible for association cleanup.
1868 */
1869 goto out_unlock;
1870
1871out_free:
1872 if (new_asoc)
1873 sctp_association_free(asoc);
1874out_unlock:
1875 sctp_release_sock(sk);
1876
1877out_nounlock:
1878 return sctp_error(sk, msg_flags, err);
1879
1880#if 0
1881do_sock_err:
1882 if (msg_len)
1883 err = msg_len;
1884 else
1885 err = sock_error(sk);
1886 goto out;
1887
1888do_interrupted:
1889 if (msg_len)
1890 err = msg_len;
1891 goto out;
1892#endif /* 0 */
1893}
1894
1895/* This is an extended version of skb_pull() that removes the data from the
1896 * start of a skb even when data is spread across the list of skb's in the
1897 * frag_list. len specifies the total amount of data that needs to be removed.
1898 * when 'len' bytes could be removed from the skb, it returns 0.
1899 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1900 * could not be removed.
1901 */
1902static int sctp_skb_pull(struct sk_buff *skb, int len)
1903{
1904 struct sk_buff *list;
1905 int skb_len = skb_headlen(skb);
1906 int rlen;
1907
1908 if (len <= skb_len) {
1909 __skb_pull(skb, len);
1910 return 0;
1911 }
1912 len -= skb_len;
1913 __skb_pull(skb, skb_len);
1914
David S. Miller1b003be2009-06-09 00:22:35 -07001915 skb_walk_frags(skb, list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 rlen = sctp_skb_pull(list, len);
1917 skb->len -= (len-rlen);
1918 skb->data_len -= (len-rlen);
1919
1920 if (!rlen)
1921 return 0;
1922
1923 len = rlen;
1924 }
1925
1926 return len;
1927}
1928
1929/* API 3.1.3 recvmsg() - UDP Style Syntax
1930 *
1931 * ssize_t recvmsg(int socket, struct msghdr *message,
1932 * int flags);
1933 *
1934 * socket - the socket descriptor of the endpoint.
1935 * message - pointer to the msghdr structure which contains a single
1936 * user message and possibly some ancillary data.
1937 *
1938 * See Section 5 for complete description of the data
1939 * structures.
1940 *
1941 * flags - flags sent or received with the user message, see Section
1942 * 5 for complete description of the flags.
1943 */
1944static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1945
1946SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1947 struct msghdr *msg, size_t len, int noblock,
1948 int flags, int *addr_len)
1949{
1950 struct sctp_ulpevent *event = NULL;
1951 struct sctp_sock *sp = sctp_sk(sk);
1952 struct sk_buff *skb;
1953 int copied;
1954 int err = 0;
1955 int skb_len;
1956
1957 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1958 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1959 "len", len, "knoblauch", noblock,
1960 "flags", flags, "addr_len", addr_len);
1961
1962 sctp_lock_sock(sk);
1963
1964 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1965 err = -ENOTCONN;
1966 goto out;
1967 }
1968
1969 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1970 if (!skb)
1971 goto out;
1972
1973 /* Get the total length of the skb including any skb's in the
1974 * frag_list.
1975 */
1976 skb_len = skb->len;
1977
1978 copied = skb_len;
1979 if (copied > len)
1980 copied = len;
1981
1982 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1983
1984 event = sctp_skb2event(skb);
1985
1986 if (err)
1987 goto out_free;
1988
Neil Horman3b885782009-10-12 13:26:31 -07001989 sock_recv_ts_and_drops(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 if (sctp_ulpevent_is_notification(event)) {
1991 msg->msg_flags |= MSG_NOTIFICATION;
1992 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1993 } else {
1994 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1995 }
1996
1997 /* Check if we allow SCTP_SNDRCVINFO. */
1998 if (sp->subscribe.sctp_data_io_event)
1999 sctp_ulpevent_read_sndrcvinfo(event, msg);
2000#if 0
2001 /* FIXME: we should be calling IP/IPv6 layers. */
2002 if (sk->sk_protinfo.af_inet.cmsg_flags)
2003 ip_cmsg_recv(msg, skb);
2004#endif
2005
2006 err = copied;
2007
2008 /* If skb's length exceeds the user's buffer, update the skb and
2009 * push it back to the receive_queue so that the next call to
2010 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2011 */
2012 if (skb_len > copied) {
2013 msg->msg_flags &= ~MSG_EOR;
2014 if (flags & MSG_PEEK)
2015 goto out_free;
2016 sctp_skb_pull(skb, copied);
2017 skb_queue_head(&sk->sk_receive_queue, skb);
2018
2019 /* When only partial message is copied to the user, increase
2020 * rwnd by that amount. If all the data in the skb is read,
2021 * rwnd is updated when the event is freed.
2022 */
Vlad Yasevich0eca8fe2008-01-11 10:12:56 -05002023 if (!sctp_ulpevent_is_notification(event))
2024 sctp_assoc_rwnd_increase(event->asoc, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 goto out;
2026 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2027 (event->msg_flags & MSG_EOR))
2028 msg->msg_flags |= MSG_EOR;
2029 else
2030 msg->msg_flags &= ~MSG_EOR;
2031
2032out_free:
2033 if (flags & MSG_PEEK) {
2034 /* Release the skb reference acquired after peeking the skb in
2035 * sctp_skb_recv_datagram().
2036 */
2037 kfree_skb(skb);
2038 } else {
2039 /* Free the event which includes releasing the reference to
2040 * the owner of the skb, freeing the skb and updating the
2041 * rwnd.
2042 */
2043 sctp_ulpevent_free(event);
2044 }
2045out:
2046 sctp_release_sock(sk);
2047 return err;
2048}
2049
2050/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2051 *
2052 * This option is a on/off flag. If enabled no SCTP message
2053 * fragmentation will be performed. Instead if a message being sent
2054 * exceeds the current PMTU size, the message will NOT be sent and
2055 * instead a error will be indicated to the user.
2056 */
2057static int sctp_setsockopt_disable_fragments(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002058 char __user *optval,
2059 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060{
2061 int val;
2062
2063 if (optlen < sizeof(int))
2064 return -EINVAL;
2065
2066 if (get_user(val, (int __user *)optval))
2067 return -EFAULT;
2068
2069 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2070
2071 return 0;
2072}
2073
2074static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002075 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076{
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05002077 if (optlen > sizeof(struct sctp_event_subscribe))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 return -EINVAL;
2079 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2080 return -EFAULT;
2081 return 0;
2082}
2083
2084/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2085 *
2086 * This socket option is applicable to the UDP-style socket only. When
2087 * set it will cause associations that are idle for more than the
2088 * specified number of seconds to automatically close. An association
2089 * being idle is defined an association that has NOT sent or received
2090 * user data. The special value of '0' indicates that no automatic
2091 * close of any associations should be performed. The option expects an
2092 * integer defining the number of seconds of idle time before an
2093 * association is closed.
2094 */
2095static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002096 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097{
2098 struct sctp_sock *sp = sctp_sk(sk);
2099
2100 /* Applicable to UDP-style socket only */
2101 if (sctp_style(sk, TCP))
2102 return -EOPNOTSUPP;
2103 if (optlen != sizeof(int))
2104 return -EINVAL;
2105 if (copy_from_user(&sp->autoclose, optval, optlen))
2106 return -EFAULT;
Andrei Pelinescu-Onciulf6778aa2009-11-23 15:54:01 -05002107 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
Andrew Morton8ffd3202009-12-21 14:25:06 +00002108 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 return 0;
2111}
2112
2113/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2114 *
2115 * Applications can enable or disable heartbeats for any peer address of
2116 * an association, modify an address's heartbeat interval, force a
2117 * heartbeat to be sent immediately, and adjust the address's maximum
2118 * number of retransmissions sent before an address is considered
2119 * unreachable. The following structure is used to access and modify an
2120 * address's parameters:
2121 *
2122 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002123 * sctp_assoc_t spp_assoc_id;
2124 * struct sockaddr_storage spp_address;
2125 * uint32_t spp_hbinterval;
2126 * uint16_t spp_pathmaxrxt;
2127 * uint32_t spp_pathmtu;
2128 * uint32_t spp_sackdelay;
2129 * uint32_t spp_flags;
2130 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002132 * spp_assoc_id - (one-to-many style socket) This is filled in the
2133 * application, and identifies the association for
2134 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135 * spp_address - This specifies which address is of interest.
2136 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08002137 * in milliseconds. If a value of zero
2138 * is present in this field then no changes are to
2139 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 * spp_pathmaxrxt - This contains the maximum number of
2141 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08002142 * considered unreachable. If a value of zero
2143 * is present in this field then no changes are to
2144 * be made to this parameter.
2145 * spp_pathmtu - When Path MTU discovery is disabled the value
2146 * specified here will be the "fixed" path mtu.
2147 * Note that if the spp_address field is empty
2148 * then all associations on this address will
2149 * have this fixed path mtu set upon them.
2150 *
2151 * spp_sackdelay - When delayed sack is enabled, this value specifies
2152 * the number of milliseconds that sacks will be delayed
2153 * for. This value will apply to all addresses of an
2154 * association if the spp_address field is empty. Note
2155 * also, that if delayed sack is enabled and this
2156 * value is set to 0, no change is made to the last
2157 * recorded delayed sack timer value.
2158 *
2159 * spp_flags - These flags are used to control various features
2160 * on an association. The flag field may contain
2161 * zero or more of the following options.
2162 *
2163 * SPP_HB_ENABLE - Enable heartbeats on the
2164 * specified address. Note that if the address
2165 * field is empty all addresses for the association
2166 * have heartbeats enabled upon them.
2167 *
2168 * SPP_HB_DISABLE - Disable heartbeats on the
2169 * speicifed address. Note that if the address
2170 * field is empty all addresses for the association
2171 * will have their heartbeats disabled. Note also
2172 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2173 * mutually exclusive, only one of these two should
2174 * be specified. Enabling both fields will have
2175 * undetermined results.
2176 *
2177 * SPP_HB_DEMAND - Request a user initiated heartbeat
2178 * to be made immediately.
2179 *
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002180 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2181 * heartbeat delayis to be set to the value of 0
2182 * milliseconds.
2183 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002184 * SPP_PMTUD_ENABLE - This field will enable PMTU
2185 * discovery upon the specified address. Note that
2186 * if the address feild is empty then all addresses
2187 * on the association are effected.
2188 *
2189 * SPP_PMTUD_DISABLE - This field will disable PMTU
2190 * discovery upon the specified address. Note that
2191 * if the address feild is empty then all addresses
2192 * on the association are effected. Not also that
2193 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2194 * exclusive. Enabling both will have undetermined
2195 * results.
2196 *
2197 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2198 * on delayed sack. The time specified in spp_sackdelay
2199 * is used to specify the sack delay for this address. Note
2200 * that if spp_address is empty then all addresses will
2201 * enable delayed sack and take on the sack delay
2202 * value specified in spp_sackdelay.
2203 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2204 * off delayed sack. If the spp_address field is blank then
2205 * delayed sack is disabled for the entire association. Note
2206 * also that this field is mutually exclusive to
2207 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2208 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 */
Adrian Bunk16164362006-09-18 00:40:38 -07002210static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2211 struct sctp_transport *trans,
2212 struct sctp_association *asoc,
2213 struct sctp_sock *sp,
2214 int hb_change,
2215 int pmtud_change,
2216 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 int error;
2219
Frank Filz52ccb8e2005-12-22 11:36:46 -08002220 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2222 if (error)
2223 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 }
2225
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002226 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2227 * this field is ignored. Note also that a value of zero indicates
2228 * the current setting should be left unchanged.
2229 */
2230 if (params->spp_flags & SPP_HB_ENABLE) {
2231
2232 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2233 * set. This lets us use 0 value when this flag
2234 * is set.
2235 */
2236 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2237 params->spp_hbinterval = 0;
2238
2239 if (params->spp_hbinterval ||
2240 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2241 if (trans) {
2242 trans->hbinterval =
2243 msecs_to_jiffies(params->spp_hbinterval);
2244 } else if (asoc) {
2245 asoc->hbinterval =
2246 msecs_to_jiffies(params->spp_hbinterval);
2247 } else {
2248 sp->hbinterval = params->spp_hbinterval;
2249 }
Frank Filz52ccb8e2005-12-22 11:36:46 -08002250 }
2251 }
2252
2253 if (hb_change) {
2254 if (trans) {
2255 trans->param_flags =
2256 (trans->param_flags & ~SPP_HB) | hb_change;
2257 } else if (asoc) {
2258 asoc->param_flags =
2259 (asoc->param_flags & ~SPP_HB) | hb_change;
2260 } else {
2261 sp->param_flags =
2262 (sp->param_flags & ~SPP_HB) | hb_change;
2263 }
2264 }
2265
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002266 /* When Path MTU discovery is disabled the value specified here will
2267 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2268 * include the flag SPP_PMTUD_DISABLE for this field to have any
2269 * effect).
2270 */
2271 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002272 if (trans) {
2273 trans->pathmtu = params->spp_pathmtu;
2274 sctp_assoc_sync_pmtu(asoc);
2275 } else if (asoc) {
2276 asoc->pathmtu = params->spp_pathmtu;
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04002277 sctp_frag_point(asoc, params->spp_pathmtu);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002278 } else {
2279 sp->pathmtu = params->spp_pathmtu;
2280 }
2281 }
2282
2283 if (pmtud_change) {
2284 if (trans) {
2285 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2286 (params->spp_flags & SPP_PMTUD_ENABLE);
2287 trans->param_flags =
2288 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2289 if (update) {
2290 sctp_transport_pmtu(trans);
2291 sctp_assoc_sync_pmtu(asoc);
2292 }
2293 } else if (asoc) {
2294 asoc->param_flags =
2295 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2296 } else {
2297 sp->param_flags =
2298 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2299 }
2300 }
2301
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002302 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2303 * value of this field is ignored. Note also that a value of zero
2304 * indicates the current setting should be left unchanged.
2305 */
2306 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002307 if (trans) {
2308 trans->sackdelay =
2309 msecs_to_jiffies(params->spp_sackdelay);
2310 } else if (asoc) {
2311 asoc->sackdelay =
2312 msecs_to_jiffies(params->spp_sackdelay);
2313 } else {
2314 sp->sackdelay = params->spp_sackdelay;
2315 }
2316 }
2317
2318 if (sackdelay_change) {
2319 if (trans) {
2320 trans->param_flags =
2321 (trans->param_flags & ~SPP_SACKDELAY) |
2322 sackdelay_change;
2323 } else if (asoc) {
2324 asoc->param_flags =
2325 (asoc->param_flags & ~SPP_SACKDELAY) |
2326 sackdelay_change;
2327 } else {
2328 sp->param_flags =
2329 (sp->param_flags & ~SPP_SACKDELAY) |
2330 sackdelay_change;
2331 }
2332 }
2333
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002334 /* Note that a value of zero indicates the current setting should be
2335 left unchanged.
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002336 */
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002337 if (params->spp_pathmaxrxt) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002338 if (trans) {
2339 trans->pathmaxrxt = params->spp_pathmaxrxt;
2340 } else if (asoc) {
2341 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2342 } else {
2343 sp->pathmaxrxt = params->spp_pathmaxrxt;
2344 }
2345 }
2346
2347 return 0;
2348}
2349
2350static int sctp_setsockopt_peer_addr_params(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002351 char __user *optval,
2352 unsigned int optlen)
Frank Filz52ccb8e2005-12-22 11:36:46 -08002353{
2354 struct sctp_paddrparams params;
2355 struct sctp_transport *trans = NULL;
2356 struct sctp_association *asoc = NULL;
2357 struct sctp_sock *sp = sctp_sk(sk);
2358 int error;
2359 int hb_change, pmtud_change, sackdelay_change;
2360
2361 if (optlen != sizeof(struct sctp_paddrparams))
2362 return - EINVAL;
2363
2364 if (copy_from_user(&params, optval, optlen))
2365 return -EFAULT;
2366
2367 /* Validate flags and value parameters. */
2368 hb_change = params.spp_flags & SPP_HB;
2369 pmtud_change = params.spp_flags & SPP_PMTUD;
2370 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2371
2372 if (hb_change == SPP_HB ||
2373 pmtud_change == SPP_PMTUD ||
2374 sackdelay_change == SPP_SACKDELAY ||
2375 params.spp_sackdelay > 500 ||
Joe Perchesf64f9e72009-11-29 16:55:45 -08002376 (params.spp_pathmtu &&
2377 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
Frank Filz52ccb8e2005-12-22 11:36:46 -08002378 return -EINVAL;
2379
2380 /* If an address other than INADDR_ANY is specified, and
2381 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04002383 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002384 trans = sctp_addr_id2transport(sk, &params.spp_address,
2385 params.spp_assoc_id);
2386 if (!trans)
2387 return -EINVAL;
2388 }
2389
2390 /* Get association, if assoc_id != 0 and the socket is a one
2391 * to many style socket, and an association was not found, then
2392 * the id was invalid.
2393 */
2394 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2395 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2396 return -EINVAL;
2397
2398 /* Heartbeat demand can only be sent on a transport or
2399 * association, but not a socket.
2400 */
2401 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2402 return -EINVAL;
2403
2404 /* Process parameters. */
2405 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2406 hb_change, pmtud_change,
2407 sackdelay_change);
2408
2409 if (error)
2410 return error;
2411
2412 /* If changes are for association, also apply parameters to each
2413 * transport.
2414 */
2415 if (!trans && asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002416 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2417 transports) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002418 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2419 hb_change, pmtud_change,
2420 sackdelay_change);
2421 }
2422 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423
2424 return 0;
2425}
2426
Wei Yongjund364d922008-05-09 15:13:26 -07002427/*
2428 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08002429 *
Wei Yongjund364d922008-05-09 15:13:26 -07002430 * This option will effect the way delayed acks are performed. This
2431 * option allows you to get or set the delayed ack time, in
2432 * milliseconds. It also allows changing the delayed ack frequency.
2433 * Changing the frequency to 1 disables the delayed sack algorithm. If
2434 * the assoc_id is 0, then this sets or gets the endpoints default
2435 * values. If the assoc_id field is non-zero, then the set or get
2436 * effects the specified association for the one to many model (the
2437 * assoc_id field is ignored by the one to one model). Note that if
2438 * sack_delay or sack_freq are 0 when setting this option, then the
2439 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08002440 *
Wei Yongjund364d922008-05-09 15:13:26 -07002441 * struct sctp_sack_info {
2442 * sctp_assoc_t sack_assoc_id;
2443 * uint32_t sack_delay;
2444 * uint32_t sack_freq;
2445 * };
Frank Filz77086102005-12-22 11:37:30 -08002446 *
Wei Yongjund364d922008-05-09 15:13:26 -07002447 * sack_assoc_id - This parameter, indicates which association the user
2448 * is performing an action upon. Note that if this field's value is
2449 * zero then the endpoints default value is changed (effecting future
2450 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08002451 *
Wei Yongjund364d922008-05-09 15:13:26 -07002452 * sack_delay - This parameter contains the number of milliseconds that
2453 * the user is requesting the delayed ACK timer be set to. Note that
2454 * this value is defined in the standard to be between 200 and 500
2455 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08002456 *
Wei Yongjund364d922008-05-09 15:13:26 -07002457 * sack_freq - This parameter contains the number of packets that must
2458 * be received before a sack is sent without waiting for the delay
2459 * timer to expire. The default value for this is 2, setting this
2460 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08002461 */
2462
Wei Yongjund364d922008-05-09 15:13:26 -07002463static int sctp_setsockopt_delayed_ack(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002464 char __user *optval, unsigned int optlen)
Frank Filz77086102005-12-22 11:37:30 -08002465{
Wei Yongjund364d922008-05-09 15:13:26 -07002466 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08002467 struct sctp_transport *trans = NULL;
2468 struct sctp_association *asoc = NULL;
2469 struct sctp_sock *sp = sctp_sk(sk);
2470
Wei Yongjund364d922008-05-09 15:13:26 -07002471 if (optlen == sizeof(struct sctp_sack_info)) {
2472 if (copy_from_user(&params, optval, optlen))
2473 return -EFAULT;
2474
2475 if (params.sack_delay == 0 && params.sack_freq == 0)
2476 return 0;
2477 } else if (optlen == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00002478 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2479 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07002480 if (copy_from_user(&params, optval, optlen))
2481 return -EFAULT;
2482
2483 if (params.sack_delay == 0)
2484 params.sack_freq = 1;
2485 else
2486 params.sack_freq = 0;
2487 } else
Frank Filz77086102005-12-22 11:37:30 -08002488 return - EINVAL;
2489
Frank Filz77086102005-12-22 11:37:30 -08002490 /* Validate value parameter. */
Wei Yongjund364d922008-05-09 15:13:26 -07002491 if (params.sack_delay > 500)
Frank Filz77086102005-12-22 11:37:30 -08002492 return -EINVAL;
2493
Wei Yongjund364d922008-05-09 15:13:26 -07002494 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08002495 * to many style socket, and an association was not found, then
2496 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002497 */
Wei Yongjund364d922008-05-09 15:13:26 -07002498 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2499 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08002500 return -EINVAL;
2501
Wei Yongjund364d922008-05-09 15:13:26 -07002502 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002503 if (asoc) {
2504 asoc->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002505 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002506 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002507 (asoc->param_flags & ~SPP_SACKDELAY) |
2508 SPP_SACKDELAY_ENABLE;
2509 } else {
Wei Yongjund364d922008-05-09 15:13:26 -07002510 sp->sackdelay = params.sack_delay;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002511 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002512 (sp->param_flags & ~SPP_SACKDELAY) |
2513 SPP_SACKDELAY_ENABLE;
2514 }
Wei Yongjund364d922008-05-09 15:13:26 -07002515 }
2516
2517 if (params.sack_freq == 1) {
Frank Filz77086102005-12-22 11:37:30 -08002518 if (asoc) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002519 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002520 (asoc->param_flags & ~SPP_SACKDELAY) |
2521 SPP_SACKDELAY_DISABLE;
2522 } else {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002523 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002524 (sp->param_flags & ~SPP_SACKDELAY) |
2525 SPP_SACKDELAY_DISABLE;
2526 }
Wei Yongjund364d922008-05-09 15:13:26 -07002527 } else if (params.sack_freq > 1) {
2528 if (asoc) {
2529 asoc->sackfreq = params.sack_freq;
2530 asoc->param_flags =
2531 (asoc->param_flags & ~SPP_SACKDELAY) |
2532 SPP_SACKDELAY_ENABLE;
2533 } else {
2534 sp->sackfreq = params.sack_freq;
2535 sp->param_flags =
2536 (sp->param_flags & ~SPP_SACKDELAY) |
2537 SPP_SACKDELAY_ENABLE;
2538 }
Frank Filz77086102005-12-22 11:37:30 -08002539 }
2540
2541 /* If change is for association, also apply to each transport. */
2542 if (asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002543 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2544 transports) {
Wei Yongjund364d922008-05-09 15:13:26 -07002545 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002546 trans->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002547 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002548 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002549 (trans->param_flags & ~SPP_SACKDELAY) |
2550 SPP_SACKDELAY_ENABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002551 }
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07002552 if (params.sack_freq == 1) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002553 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002554 (trans->param_flags & ~SPP_SACKDELAY) |
2555 SPP_SACKDELAY_DISABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002556 } else if (params.sack_freq > 1) {
2557 trans->sackfreq = params.sack_freq;
2558 trans->param_flags =
2559 (trans->param_flags & ~SPP_SACKDELAY) |
2560 SPP_SACKDELAY_ENABLE;
Frank Filz77086102005-12-22 11:37:30 -08002561 }
2562 }
2563 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002564
Frank Filz77086102005-12-22 11:37:30 -08002565 return 0;
2566}
2567
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2569 *
2570 * Applications can specify protocol parameters for the default association
2571 * initialization. The option name argument to setsockopt() and getsockopt()
2572 * is SCTP_INITMSG.
2573 *
2574 * Setting initialization parameters is effective only on an unconnected
2575 * socket (for UDP-style sockets only future associations are effected
2576 * by the change). With TCP-style sockets, this option is inherited by
2577 * sockets derived from a listener socket.
2578 */
David S. Millerb7058842009-09-30 16:12:20 -07002579static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580{
2581 struct sctp_initmsg sinit;
2582 struct sctp_sock *sp = sctp_sk(sk);
2583
2584 if (optlen != sizeof(struct sctp_initmsg))
2585 return -EINVAL;
2586 if (copy_from_user(&sinit, optval, optlen))
2587 return -EFAULT;
2588
2589 if (sinit.sinit_num_ostreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002590 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591 if (sinit.sinit_max_instreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002592 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593 if (sinit.sinit_max_attempts)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002594 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002595 if (sinit.sinit_max_init_timeo)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002596 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597
2598 return 0;
2599}
2600
2601/*
2602 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2603 *
2604 * Applications that wish to use the sendto() system call may wish to
2605 * specify a default set of parameters that would normally be supplied
2606 * through the inclusion of ancillary data. This socket option allows
2607 * such an application to set the default sctp_sndrcvinfo structure.
2608 * The application that wishes to use this socket option simply passes
2609 * in to this call the sctp_sndrcvinfo structure defined in Section
2610 * 5.2.2) The input parameters accepted by this call include
2611 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2612 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2613 * to this call if the caller is using the UDP model.
2614 */
2615static int sctp_setsockopt_default_send_param(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002616 char __user *optval,
2617 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618{
2619 struct sctp_sndrcvinfo info;
2620 struct sctp_association *asoc;
2621 struct sctp_sock *sp = sctp_sk(sk);
2622
2623 if (optlen != sizeof(struct sctp_sndrcvinfo))
2624 return -EINVAL;
2625 if (copy_from_user(&info, optval, optlen))
2626 return -EFAULT;
2627
2628 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2629 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2630 return -EINVAL;
2631
2632 if (asoc) {
2633 asoc->default_stream = info.sinfo_stream;
2634 asoc->default_flags = info.sinfo_flags;
2635 asoc->default_ppid = info.sinfo_ppid;
2636 asoc->default_context = info.sinfo_context;
2637 asoc->default_timetolive = info.sinfo_timetolive;
2638 } else {
2639 sp->default_stream = info.sinfo_stream;
2640 sp->default_flags = info.sinfo_flags;
2641 sp->default_ppid = info.sinfo_ppid;
2642 sp->default_context = info.sinfo_context;
2643 sp->default_timetolive = info.sinfo_timetolive;
2644 }
2645
2646 return 0;
2647}
2648
2649/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2650 *
2651 * Requests that the local SCTP stack use the enclosed peer address as
2652 * the association primary. The enclosed address must be one of the
2653 * association peer's addresses.
2654 */
2655static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002656 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657{
2658 struct sctp_prim prim;
2659 struct sctp_transport *trans;
2660
2661 if (optlen != sizeof(struct sctp_prim))
2662 return -EINVAL;
2663
2664 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2665 return -EFAULT;
2666
2667 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2668 if (!trans)
2669 return -EINVAL;
2670
2671 sctp_assoc_set_primary(trans->asoc, trans);
2672
2673 return 0;
2674}
2675
2676/*
2677 * 7.1.5 SCTP_NODELAY
2678 *
2679 * Turn on/off any Nagle-like algorithm. This means that packets are
2680 * generally sent as soon as possible and no unnecessary delays are
2681 * introduced, at the cost of more packets in the network. Expects an
2682 * integer boolean flag.
2683 */
2684static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002685 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002686{
2687 int val;
2688
2689 if (optlen < sizeof(int))
2690 return -EINVAL;
2691 if (get_user(val, (int __user *)optval))
2692 return -EFAULT;
2693
2694 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2695 return 0;
2696}
2697
2698/*
2699 *
2700 * 7.1.1 SCTP_RTOINFO
2701 *
2702 * The protocol parameters used to initialize and bound retransmission
2703 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2704 * and modify these parameters.
2705 * All parameters are time values, in milliseconds. A value of 0, when
2706 * modifying the parameters, indicates that the current value should not
2707 * be changed.
2708 *
2709 */
David S. Millerb7058842009-09-30 16:12:20 -07002710static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2711{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712 struct sctp_rtoinfo rtoinfo;
2713 struct sctp_association *asoc;
2714
2715 if (optlen != sizeof (struct sctp_rtoinfo))
2716 return -EINVAL;
2717
2718 if (copy_from_user(&rtoinfo, optval, optlen))
2719 return -EFAULT;
2720
2721 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2722
2723 /* Set the values to the specific association */
2724 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2725 return -EINVAL;
2726
2727 if (asoc) {
2728 if (rtoinfo.srto_initial != 0)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002729 asoc->rto_initial =
Linus Torvalds1da177e2005-04-16 15:20:36 -07002730 msecs_to_jiffies(rtoinfo.srto_initial);
2731 if (rtoinfo.srto_max != 0)
2732 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2733 if (rtoinfo.srto_min != 0)
2734 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2735 } else {
2736 /* If there is no association or the association-id = 0
2737 * set the values to the endpoint.
2738 */
2739 struct sctp_sock *sp = sctp_sk(sk);
2740
2741 if (rtoinfo.srto_initial != 0)
2742 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2743 if (rtoinfo.srto_max != 0)
2744 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2745 if (rtoinfo.srto_min != 0)
2746 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2747 }
2748
2749 return 0;
2750}
2751
2752/*
2753 *
2754 * 7.1.2 SCTP_ASSOCINFO
2755 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02002756 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07002757 * of the association.
2758 * Returns an error if the new association retransmission value is
2759 * greater than the sum of the retransmission value of the peer.
2760 * See [SCTP] for more information.
2761 *
2762 */
David S. Millerb7058842009-09-30 16:12:20 -07002763static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764{
2765
2766 struct sctp_assocparams assocparams;
2767 struct sctp_association *asoc;
2768
2769 if (optlen != sizeof(struct sctp_assocparams))
2770 return -EINVAL;
2771 if (copy_from_user(&assocparams, optval, optlen))
2772 return -EFAULT;
2773
2774 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2775
2776 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2777 return -EINVAL;
2778
2779 /* Set the values to the specific association */
2780 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002781 if (assocparams.sasoc_asocmaxrxt != 0) {
2782 __u32 path_sum = 0;
2783 int paths = 0;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002784 struct sctp_transport *peer_addr;
2785
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002786 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2787 transports) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002788 path_sum += peer_addr->pathmaxrxt;
2789 paths++;
2790 }
2791
Frederik Schwarzer025dfda2008-10-16 19:02:37 +02002792 /* Only validate asocmaxrxt if we have more than
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002793 * one path/transport. We do this because path
2794 * retransmissions are only counted when we have more
2795 * then one path.
2796 */
2797 if (paths > 1 &&
2798 assocparams.sasoc_asocmaxrxt > path_sum)
2799 return -EINVAL;
2800
Linus Torvalds1da177e2005-04-16 15:20:36 -07002801 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002802 }
2803
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804 if (assocparams.sasoc_cookie_life != 0) {
2805 asoc->cookie_life.tv_sec =
2806 assocparams.sasoc_cookie_life / 1000;
2807 asoc->cookie_life.tv_usec =
2808 (assocparams.sasoc_cookie_life % 1000)
2809 * 1000;
2810 }
2811 } else {
2812 /* Set the values to the endpoint */
2813 struct sctp_sock *sp = sctp_sk(sk);
2814
2815 if (assocparams.sasoc_asocmaxrxt != 0)
2816 sp->assocparams.sasoc_asocmaxrxt =
2817 assocparams.sasoc_asocmaxrxt;
2818 if (assocparams.sasoc_cookie_life != 0)
2819 sp->assocparams.sasoc_cookie_life =
2820 assocparams.sasoc_cookie_life;
2821 }
2822 return 0;
2823}
2824
2825/*
2826 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2827 *
2828 * This socket option is a boolean flag which turns on or off mapped V4
2829 * addresses. If this option is turned on and the socket is type
2830 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2831 * If this option is turned off, then no mapping will be done of V4
2832 * addresses and a user will receive both PF_INET6 and PF_INET type
2833 * addresses on the socket.
2834 */
David S. Millerb7058842009-09-30 16:12:20 -07002835static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836{
2837 int val;
2838 struct sctp_sock *sp = sctp_sk(sk);
2839
2840 if (optlen < sizeof(int))
2841 return -EINVAL;
2842 if (get_user(val, (int __user *)optval))
2843 return -EFAULT;
2844 if (val)
2845 sp->v4mapped = 1;
2846 else
2847 sp->v4mapped = 0;
2848
2849 return 0;
2850}
2851
2852/*
Wei Yongjune89c2092008-12-25 16:54:58 -08002853 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2854 * This option will get or set the maximum size to put in any outgoing
2855 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856 * fragmented by SCTP into the specified size. Note that the underlying
2857 * SCTP implementation may fragment into smaller sized chunks when the
2858 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08002859 * the user. The default value for this option is '0' which indicates
2860 * the user is NOT limiting fragmentation and only the PMTU will effect
2861 * SCTP's choice of DATA chunk size. Note also that values set larger
2862 * than the maximum size of an IP datagram will effectively let SCTP
2863 * control fragmentation (i.e. the same as setting this option to 0).
2864 *
2865 * The following structure is used to access and modify this parameter:
2866 *
2867 * struct sctp_assoc_value {
2868 * sctp_assoc_t assoc_id;
2869 * uint32_t assoc_value;
2870 * };
2871 *
2872 * assoc_id: This parameter is ignored for one-to-one style sockets.
2873 * For one-to-many style sockets this parameter indicates which
2874 * association the user is performing an action upon. Note that if
2875 * this field's value is zero then the endpoints default value is
2876 * changed (effecting future associations only).
2877 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878 */
David S. Millerb7058842009-09-30 16:12:20 -07002879static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880{
Wei Yongjune89c2092008-12-25 16:54:58 -08002881 struct sctp_assoc_value params;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002883 struct sctp_sock *sp = sctp_sk(sk);
2884 int val;
2885
Wei Yongjune89c2092008-12-25 16:54:58 -08002886 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00002887 pr_warn("Use of int in maxseg socket option deprecated\n");
2888 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08002889 if (copy_from_user(&val, optval, optlen))
2890 return -EFAULT;
2891 params.assoc_id = 0;
2892 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2893 if (copy_from_user(&params, optval, optlen))
2894 return -EFAULT;
2895 val = params.assoc_value;
2896 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002897 return -EINVAL;
Wei Yongjune89c2092008-12-25 16:54:58 -08002898
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002899 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002901
Wei Yongjune89c2092008-12-25 16:54:58 -08002902 asoc = sctp_id2assoc(sk, params.assoc_id);
2903 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2904 return -EINVAL;
2905
2906 if (asoc) {
2907 if (val == 0) {
2908 val = asoc->pathmtu;
2909 val -= sp->pf->af->net_header_len;
2910 val -= sizeof(struct sctphdr) +
2911 sizeof(struct sctp_data_chunk);
2912 }
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04002913 asoc->user_frag = val;
2914 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
Wei Yongjune89c2092008-12-25 16:54:58 -08002915 } else {
2916 sp->user_frag = val;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002917 }
2918
2919 return 0;
2920}
2921
2922
2923/*
2924 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2925 *
2926 * Requests that the peer mark the enclosed address as the association
2927 * primary. The enclosed address must be one of the association's
2928 * locally bound addresses. The following structure is used to make a
2929 * set primary request:
2930 */
2931static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002932 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002933{
2934 struct sctp_sock *sp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002935 struct sctp_association *asoc = NULL;
2936 struct sctp_setpeerprim prim;
2937 struct sctp_chunk *chunk;
Wei Yongjun40a01032010-12-07 17:11:09 +00002938 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002939 int err;
2940
2941 sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002942
2943 if (!sctp_addip_enable)
2944 return -EPERM;
2945
2946 if (optlen != sizeof(struct sctp_setpeerprim))
2947 return -EINVAL;
2948
2949 if (copy_from_user(&prim, optval, optlen))
2950 return -EFAULT;
2951
2952 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002953 if (!asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002954 return -EINVAL;
2955
2956 if (!asoc->peer.asconf_capable)
2957 return -EPERM;
2958
2959 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2960 return -EPERM;
2961
2962 if (!sctp_state(asoc, ESTABLISHED))
2963 return -ENOTCONN;
2964
Wei Yongjun40a01032010-12-07 17:11:09 +00002965 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
2966 if (!af)
2967 return -EINVAL;
2968
2969 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
2970 return -EADDRNOTAVAIL;
2971
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2973 return -EADDRNOTAVAIL;
2974
2975 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2976 chunk = sctp_make_asconf_set_prim(asoc,
2977 (union sctp_addr *)&prim.sspp_addr);
2978 if (!chunk)
2979 return -ENOMEM;
2980
2981 err = sctp_send_asconf(asoc, chunk);
2982
2983 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2984
2985 return err;
2986}
2987
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002988static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002989 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002990{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002991 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002992
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002993 if (optlen != sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002994 return -EINVAL;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002995 if (copy_from_user(&adaptation, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002996 return -EFAULT;
2997
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002998 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999
3000 return 0;
3001}
3002
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003003/*
3004 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3005 *
3006 * The context field in the sctp_sndrcvinfo structure is normally only
3007 * used when a failed message is retrieved holding the value that was
3008 * sent down on the actual send call. This option allows the setting of
3009 * a default context on an association basis that will be received on
3010 * reading messages from the peer. This is especially helpful in the
3011 * one-2-many model for an application to keep some reference to an
3012 * internal state machine that is processing messages on the
3013 * association. Note that the setting of this value only effects
3014 * received messages from the peer and does not effect the value that is
3015 * saved with outbound messages.
3016 */
3017static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003018 unsigned int optlen)
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003019{
3020 struct sctp_assoc_value params;
3021 struct sctp_sock *sp;
3022 struct sctp_association *asoc;
3023
3024 if (optlen != sizeof(struct sctp_assoc_value))
3025 return -EINVAL;
3026 if (copy_from_user(&params, optval, optlen))
3027 return -EFAULT;
3028
3029 sp = sctp_sk(sk);
3030
3031 if (params.assoc_id != 0) {
3032 asoc = sctp_id2assoc(sk, params.assoc_id);
3033 if (!asoc)
3034 return -EINVAL;
3035 asoc->default_rcv_context = params.assoc_value;
3036 } else {
3037 sp->default_rcv_context = params.assoc_value;
3038 }
3039
3040 return 0;
3041}
3042
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003043/*
3044 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3045 *
3046 * This options will at a minimum specify if the implementation is doing
3047 * fragmented interleave. Fragmented interleave, for a one to many
3048 * socket, is when subsequent calls to receive a message may return
3049 * parts of messages from different associations. Some implementations
3050 * may allow you to turn this value on or off. If so, when turned off,
3051 * no fragment interleave will occur (which will cause a head of line
3052 * blocking amongst multiple associations sharing the same one to many
3053 * socket). When this option is turned on, then each receive call may
3054 * come from a different association (thus the user must receive data
3055 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3056 * association each receive belongs to.
3057 *
3058 * This option takes a boolean value. A non-zero value indicates that
3059 * fragmented interleave is on. A value of zero indicates that
3060 * fragmented interleave is off.
3061 *
3062 * Note that it is important that an implementation that allows this
3063 * option to be turned on, have it off by default. Otherwise an unaware
3064 * application using the one to many model may become confused and act
3065 * incorrectly.
3066 */
3067static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3068 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003069 unsigned int optlen)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003070{
3071 int val;
3072
3073 if (optlen != sizeof(int))
3074 return -EINVAL;
3075 if (get_user(val, (int __user *)optval))
3076 return -EFAULT;
3077
3078 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3079
3080 return 0;
3081}
3082
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003083/*
Wei Yongjun8510b932008-12-25 16:59:03 -08003084 * 8.1.21. Set or Get the SCTP Partial Delivery Point
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003085 * (SCTP_PARTIAL_DELIVERY_POINT)
Wei Yongjun8510b932008-12-25 16:59:03 -08003086 *
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003087 * This option will set or get the SCTP partial delivery point. This
3088 * point is the size of a message where the partial delivery API will be
3089 * invoked to help free up rwnd space for the peer. Setting this to a
Wei Yongjun8510b932008-12-25 16:59:03 -08003090 * lower value will cause partial deliveries to happen more often. The
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003091 * calls argument is an integer that sets or gets the partial delivery
Wei Yongjun8510b932008-12-25 16:59:03 -08003092 * point. Note also that the call will fail if the user attempts to set
3093 * this value larger than the socket receive buffer size.
3094 *
3095 * Note that any single message having a length smaller than or equal to
3096 * the SCTP partial delivery point will be delivered in one single read
3097 * call as long as the user provided buffer is large enough to hold the
3098 * message.
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003099 */
3100static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3101 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003102 unsigned int optlen)
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003103{
3104 u32 val;
3105
3106 if (optlen != sizeof(u32))
3107 return -EINVAL;
3108 if (get_user(val, (int __user *)optval))
3109 return -EFAULT;
3110
Wei Yongjun8510b932008-12-25 16:59:03 -08003111 /* Note: We double the receive buffer from what the user sets
3112 * it to be, also initial rwnd is based on rcvbuf/2.
3113 */
3114 if (val > (sk->sk_rcvbuf >> 1))
3115 return -EINVAL;
3116
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003117 sctp_sk(sk)->pd_point = val;
3118
3119 return 0; /* is this the right error code? */
3120}
3121
Vlad Yasevich70331572007-03-23 11:34:36 -07003122/*
3123 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3124 *
3125 * This option will allow a user to change the maximum burst of packets
3126 * that can be emitted by this association. Note that the default value
3127 * is 4, and some implementations may restrict this setting so that it
3128 * can only be lowered.
3129 *
3130 * NOTE: This text doesn't seem right. Do this on a socket basis with
3131 * future associations inheriting the socket value.
3132 */
3133static int sctp_setsockopt_maxburst(struct sock *sk,
3134 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003135 unsigned int optlen)
Vlad Yasevich70331572007-03-23 11:34:36 -07003136{
Neil Horman219b99a2008-03-05 13:44:46 -08003137 struct sctp_assoc_value params;
3138 struct sctp_sock *sp;
3139 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07003140 int val;
Neil Horman219b99a2008-03-05 13:44:46 -08003141 int assoc_id = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003142
Neil Horman219b99a2008-03-05 13:44:46 -08003143 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00003144 pr_warn("Use of int in max_burst socket option deprecated\n");
3145 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08003146 if (copy_from_user(&val, optval, optlen))
3147 return -EFAULT;
3148 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3149 if (copy_from_user(&params, optval, optlen))
3150 return -EFAULT;
3151 val = params.assoc_value;
3152 assoc_id = params.assoc_id;
3153 } else
3154 return -EINVAL;
3155
3156 sp = sctp_sk(sk);
3157
3158 if (assoc_id != 0) {
3159 asoc = sctp_id2assoc(sk, assoc_id);
3160 if (!asoc)
3161 return -EINVAL;
3162 asoc->max_burst = val;
3163 } else
3164 sp->max_burst = val;
Vlad Yasevich70331572007-03-23 11:34:36 -07003165
3166 return 0;
3167}
3168
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003169/*
3170 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3171 *
3172 * This set option adds a chunk type that the user is requesting to be
3173 * received only in an authenticated way. Changes to the list of chunks
3174 * will only effect future associations on the socket.
3175 */
3176static int sctp_setsockopt_auth_chunk(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003177 char __user *optval,
3178 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003179{
3180 struct sctp_authchunk val;
3181
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003182 if (!sctp_auth_enable)
3183 return -EACCES;
3184
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003185 if (optlen != sizeof(struct sctp_authchunk))
3186 return -EINVAL;
3187 if (copy_from_user(&val, optval, optlen))
3188 return -EFAULT;
3189
3190 switch (val.sauth_chunk) {
3191 case SCTP_CID_INIT:
3192 case SCTP_CID_INIT_ACK:
3193 case SCTP_CID_SHUTDOWN_COMPLETE:
3194 case SCTP_CID_AUTH:
3195 return -EINVAL;
3196 }
3197
3198 /* add this chunk id to the endpoint */
3199 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3200}
3201
3202/*
3203 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3204 *
3205 * This option gets or sets the list of HMAC algorithms that the local
3206 * endpoint requires the peer to use.
3207 */
3208static int sctp_setsockopt_hmac_ident(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003209 char __user *optval,
3210 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003211{
3212 struct sctp_hmacalgo *hmacs;
Vlad Yasevichd9724052008-08-27 16:09:49 -07003213 u32 idents;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003214 int err;
3215
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003216 if (!sctp_auth_enable)
3217 return -EACCES;
3218
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003219 if (optlen < sizeof(struct sctp_hmacalgo))
3220 return -EINVAL;
3221
Shan Wei934253a2011-04-18 19:13:18 +00003222 hmacs= memdup_user(optval, optlen);
3223 if (IS_ERR(hmacs))
3224 return PTR_ERR(hmacs);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003225
Vlad Yasevichd9724052008-08-27 16:09:49 -07003226 idents = hmacs->shmac_num_idents;
3227 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3228 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003229 err = -EINVAL;
3230 goto out;
3231 }
3232
3233 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3234out:
3235 kfree(hmacs);
3236 return err;
3237}
3238
3239/*
3240 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3241 *
3242 * This option will set a shared secret key which is used to build an
3243 * association shared key.
3244 */
3245static int sctp_setsockopt_auth_key(struct sock *sk,
3246 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003247 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003248{
3249 struct sctp_authkey *authkey;
3250 struct sctp_association *asoc;
3251 int ret;
3252
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003253 if (!sctp_auth_enable)
3254 return -EACCES;
3255
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003256 if (optlen <= sizeof(struct sctp_authkey))
3257 return -EINVAL;
3258
Shan Wei934253a2011-04-18 19:13:18 +00003259 authkey= memdup_user(optval, optlen);
3260 if (IS_ERR(authkey))
3261 return PTR_ERR(authkey);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003262
Vlad Yasevich328fc472008-08-27 16:08:54 -07003263 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
Vlad Yasevich30c22352008-08-25 15:16:19 -07003264 ret = -EINVAL;
3265 goto out;
3266 }
3267
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003268 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3269 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3270 ret = -EINVAL;
3271 goto out;
3272 }
3273
3274 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3275out:
3276 kfree(authkey);
3277 return ret;
3278}
3279
3280/*
3281 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3282 *
3283 * This option will get or set the active shared key to be used to build
3284 * the association shared key.
3285 */
3286static int sctp_setsockopt_active_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003287 char __user *optval,
3288 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003289{
3290 struct sctp_authkeyid val;
3291 struct sctp_association *asoc;
3292
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003293 if (!sctp_auth_enable)
3294 return -EACCES;
3295
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003296 if (optlen != sizeof(struct sctp_authkeyid))
3297 return -EINVAL;
3298 if (copy_from_user(&val, optval, optlen))
3299 return -EFAULT;
3300
3301 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3302 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3303 return -EINVAL;
3304
3305 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3306 val.scact_keynumber);
3307}
3308
3309/*
3310 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3311 *
3312 * This set option will delete a shared secret key from use.
3313 */
3314static int sctp_setsockopt_del_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003315 char __user *optval,
3316 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003317{
3318 struct sctp_authkeyid val;
3319 struct sctp_association *asoc;
3320
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003321 if (!sctp_auth_enable)
3322 return -EACCES;
3323
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003324 if (optlen != sizeof(struct sctp_authkeyid))
3325 return -EINVAL;
3326 if (copy_from_user(&val, optval, optlen))
3327 return -EFAULT;
3328
3329 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3330 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3331 return -EINVAL;
3332
3333 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3334 val.scact_keynumber);
3335
3336}
3337
3338
Linus Torvalds1da177e2005-04-16 15:20:36 -07003339/* API 6.2 setsockopt(), getsockopt()
3340 *
3341 * Applications use setsockopt() and getsockopt() to set or retrieve
3342 * socket options. Socket options are used to change the default
3343 * behavior of sockets calls. They are described in Section 7.
3344 *
3345 * The syntax is:
3346 *
3347 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3348 * int __user *optlen);
3349 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3350 * int optlen);
3351 *
3352 * sd - the socket descript.
3353 * level - set to IPPROTO_SCTP for all SCTP options.
3354 * optname - the option name.
3355 * optval - the buffer to store the value of the option.
3356 * optlen - the size of the buffer.
3357 */
3358SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003359 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360{
3361 int retval = 0;
3362
3363 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3364 sk, optname);
3365
3366 /* I can hardly begin to describe how wrong this is. This is
3367 * so broken as to be worse than useless. The API draft
3368 * REALLY is NOT helpful here... I am not convinced that the
3369 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3370 * are at all well-founded.
3371 */
3372 if (level != SOL_SCTP) {
3373 struct sctp_af *af = sctp_sk(sk)->pf->af;
3374 retval = af->setsockopt(sk, level, optname, optval, optlen);
3375 goto out_nounlock;
3376 }
3377
3378 sctp_lock_sock(sk);
3379
3380 switch (optname) {
3381 case SCTP_SOCKOPT_BINDX_ADD:
3382 /* 'optlen' is the size of the addresses buffer. */
3383 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3384 optlen, SCTP_BINDX_ADD_ADDR);
3385 break;
3386
3387 case SCTP_SOCKOPT_BINDX_REM:
3388 /* 'optlen' is the size of the addresses buffer. */
3389 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3390 optlen, SCTP_BINDX_REM_ADDR);
3391 break;
3392
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003393 case SCTP_SOCKOPT_CONNECTX_OLD:
3394 /* 'optlen' is the size of the addresses buffer. */
3395 retval = sctp_setsockopt_connectx_old(sk,
3396 (struct sockaddr __user *)optval,
3397 optlen);
3398 break;
3399
Frank Filz3f7a87d2005-06-20 13:14:57 -07003400 case SCTP_SOCKOPT_CONNECTX:
3401 /* 'optlen' is the size of the addresses buffer. */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003402 retval = sctp_setsockopt_connectx(sk,
3403 (struct sockaddr __user *)optval,
3404 optlen);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003405 break;
3406
Linus Torvalds1da177e2005-04-16 15:20:36 -07003407 case SCTP_DISABLE_FRAGMENTS:
3408 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3409 break;
3410
3411 case SCTP_EVENTS:
3412 retval = sctp_setsockopt_events(sk, optval, optlen);
3413 break;
3414
3415 case SCTP_AUTOCLOSE:
3416 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3417 break;
3418
3419 case SCTP_PEER_ADDR_PARAMS:
3420 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3421 break;
3422
Shan Wei4580ccc2011-01-18 22:39:00 +00003423 case SCTP_DELAYED_SACK:
Wei Yongjund364d922008-05-09 15:13:26 -07003424 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
Frank Filz77086102005-12-22 11:37:30 -08003425 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003426 case SCTP_PARTIAL_DELIVERY_POINT:
3427 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3428 break;
Frank Filz77086102005-12-22 11:37:30 -08003429
Linus Torvalds1da177e2005-04-16 15:20:36 -07003430 case SCTP_INITMSG:
3431 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3432 break;
3433 case SCTP_DEFAULT_SEND_PARAM:
3434 retval = sctp_setsockopt_default_send_param(sk, optval,
3435 optlen);
3436 break;
3437 case SCTP_PRIMARY_ADDR:
3438 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3439 break;
3440 case SCTP_SET_PEER_PRIMARY_ADDR:
3441 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3442 break;
3443 case SCTP_NODELAY:
3444 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3445 break;
3446 case SCTP_RTOINFO:
3447 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3448 break;
3449 case SCTP_ASSOCINFO:
3450 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3451 break;
3452 case SCTP_I_WANT_MAPPED_V4_ADDR:
3453 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3454 break;
3455 case SCTP_MAXSEG:
3456 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3457 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003458 case SCTP_ADAPTATION_LAYER:
3459 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003460 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003461 case SCTP_CONTEXT:
3462 retval = sctp_setsockopt_context(sk, optval, optlen);
3463 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003464 case SCTP_FRAGMENT_INTERLEAVE:
3465 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3466 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07003467 case SCTP_MAX_BURST:
3468 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3469 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003470 case SCTP_AUTH_CHUNK:
3471 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3472 break;
3473 case SCTP_HMAC_IDENT:
3474 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3475 break;
3476 case SCTP_AUTH_KEY:
3477 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3478 break;
3479 case SCTP_AUTH_ACTIVE_KEY:
3480 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3481 break;
3482 case SCTP_AUTH_DELETE_KEY:
3483 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3484 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003485 default:
3486 retval = -ENOPROTOOPT;
3487 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07003488 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003489
3490 sctp_release_sock(sk);
3491
3492out_nounlock:
3493 return retval;
3494}
3495
3496/* API 3.1.6 connect() - UDP Style Syntax
3497 *
3498 * An application may use the connect() call in the UDP model to initiate an
3499 * association without sending data.
3500 *
3501 * The syntax is:
3502 *
3503 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3504 *
3505 * sd: the socket descriptor to have a new association added to.
3506 *
3507 * nam: the address structure (either struct sockaddr_in or struct
3508 * sockaddr_in6 defined in RFC2553 [7]).
3509 *
3510 * len: the size of the address.
3511 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07003512SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003513 int addr_len)
3514{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003515 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07003516 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003517
3518 sctp_lock_sock(sk);
3519
Frank Filz3f7a87d2005-06-20 13:14:57 -07003520 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08003521 __func__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003522
Frank Filz3f7a87d2005-06-20 13:14:57 -07003523 /* Validate addr_len before calling common connect/connectx routine. */
3524 af = sctp_get_af_specific(addr->sa_family);
3525 if (!af || addr_len < af->sockaddr_len) {
3526 err = -EINVAL;
3527 } else {
3528 /* Pass correct addr len to common routine (so it knows there
3529 * is only one address being passed.
3530 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003531 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003532 }
3533
Linus Torvalds1da177e2005-04-16 15:20:36 -07003534 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003535 return err;
3536}
3537
3538/* FIXME: Write comments. */
3539SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3540{
3541 return -EOPNOTSUPP; /* STUB */
3542}
3543
3544/* 4.1.4 accept() - TCP Style Syntax
3545 *
3546 * Applications use accept() call to remove an established SCTP
3547 * association from the accept queue of the endpoint. A new socket
3548 * descriptor will be returned from accept() to represent the newly
3549 * formed association.
3550 */
3551SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3552{
3553 struct sctp_sock *sp;
3554 struct sctp_endpoint *ep;
3555 struct sock *newsk = NULL;
3556 struct sctp_association *asoc;
3557 long timeo;
3558 int error = 0;
3559
3560 sctp_lock_sock(sk);
3561
3562 sp = sctp_sk(sk);
3563 ep = sp->ep;
3564
3565 if (!sctp_style(sk, TCP)) {
3566 error = -EOPNOTSUPP;
3567 goto out;
3568 }
3569
3570 if (!sctp_sstate(sk, LISTENING)) {
3571 error = -EINVAL;
3572 goto out;
3573 }
3574
Sridhar Samudrala8abfedd2006-08-22 00:24:09 -07003575 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003576
3577 error = sctp_wait_for_accept(sk, timeo);
3578 if (error)
3579 goto out;
3580
3581 /* We treat the list of associations on the endpoint as the accept
3582 * queue and pick the first association on the list.
3583 */
3584 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3585
3586 newsk = sp->pf->create_accept_sk(sk, asoc);
3587 if (!newsk) {
3588 error = -ENOMEM;
3589 goto out;
3590 }
3591
3592 /* Populate the fields of the newsk from the oldsk and migrate the
3593 * asoc to the newsk.
3594 */
3595 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3596
3597out:
3598 sctp_release_sock(sk);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003599 *err = error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003600 return newsk;
3601}
3602
3603/* The SCTP ioctl handler. */
3604SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3605{
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003606 int rc = -ENOTCONN;
3607
3608 sctp_lock_sock(sk);
3609
3610 /*
3611 * SEQPACKET-style sockets in LISTENING state are valid, for
3612 * SCTP, so only discard TCP-style sockets in LISTENING state.
3613 */
3614 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3615 goto out;
3616
3617 switch (cmd) {
3618 case SIOCINQ: {
3619 struct sk_buff *skb;
3620 unsigned int amount = 0;
3621
3622 skb = skb_peek(&sk->sk_receive_queue);
3623 if (skb != NULL) {
3624 /*
3625 * We will only return the amount of this packet since
3626 * that is all that will be read.
3627 */
3628 amount = skb->len;
3629 }
3630 rc = put_user(amount, (int __user *)arg);
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003631 break;
David S. Miller9a7241c2010-10-03 22:14:37 -07003632 }
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003633 default:
3634 rc = -ENOIOCTLCMD;
3635 break;
3636 }
3637out:
3638 sctp_release_sock(sk);
3639 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003640}
3641
3642/* This is the function which gets called during socket creation to
3643 * initialized the SCTP-specific portion of the sock.
3644 * The sock structure should already be zero-filled memory.
3645 */
3646SCTP_STATIC int sctp_init_sock(struct sock *sk)
3647{
3648 struct sctp_endpoint *ep;
3649 struct sctp_sock *sp;
3650
3651 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3652
3653 sp = sctp_sk(sk);
3654
3655 /* Initialize the SCTP per socket area. */
3656 switch (sk->sk_type) {
3657 case SOCK_SEQPACKET:
3658 sp->type = SCTP_SOCKET_UDP;
3659 break;
3660 case SOCK_STREAM:
3661 sp->type = SCTP_SOCKET_TCP;
3662 break;
3663 default:
3664 return -ESOCKTNOSUPPORT;
3665 }
3666
3667 /* Initialize default send parameters. These parameters can be
3668 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3669 */
3670 sp->default_stream = 0;
3671 sp->default_ppid = 0;
3672 sp->default_flags = 0;
3673 sp->default_context = 0;
3674 sp->default_timetolive = 0;
3675
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003676 sp->default_rcv_context = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003677 sp->max_burst = sctp_max_burst;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003678
Linus Torvalds1da177e2005-04-16 15:20:36 -07003679 /* Initialize default setup parameters. These parameters
3680 * can be modified with the SCTP_INITMSG socket option or
3681 * overridden by the SCTP_INIT CMSG.
3682 */
3683 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3684 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3685 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003686 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003687
3688 /* Initialize default RTO related parameters. These parameters can
3689 * be modified for with the SCTP_RTOINFO socket option.
3690 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003691 sp->rtoinfo.srto_initial = sctp_rto_initial;
3692 sp->rtoinfo.srto_max = sctp_rto_max;
3693 sp->rtoinfo.srto_min = sctp_rto_min;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003694
3695 /* Initialize default association related parameters. These parameters
3696 * can be modified with the SCTP_ASSOCINFO socket option.
3697 */
3698 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3699 sp->assocparams.sasoc_number_peer_destinations = 0;
3700 sp->assocparams.sasoc_peer_rwnd = 0;
3701 sp->assocparams.sasoc_local_rwnd = 0;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003702 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003703
3704 /* Initialize default event subscriptions. By default, all the
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003705 * options are off.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003706 */
3707 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3708
3709 /* Default Peer Address Parameters. These defaults can
3710 * be modified via SCTP_PEER_ADDR_PARAMS
3711 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003712 sp->hbinterval = sctp_hb_interval;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003713 sp->pathmaxrxt = sctp_max_retrans_path;
3714 sp->pathmtu = 0; // allow default discovery
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003715 sp->sackdelay = sctp_sack_timeout;
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07003716 sp->sackfreq = 2;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003717 sp->param_flags = SPP_HB_ENABLE |
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003718 SPP_PMTUD_ENABLE |
3719 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003720
3721 /* If enabled no SCTP message fragmentation will be performed.
3722 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3723 */
3724 sp->disable_fragments = 0;
3725
Sridhar Samudrala208edef2006-09-29 17:08:01 -07003726 /* Enable Nagle algorithm by default. */
3727 sp->nodelay = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003728
3729 /* Enable by default. */
3730 sp->v4mapped = 1;
3731
3732 /* Auto-close idle associations after the configured
3733 * number of seconds. A value of 0 disables this
3734 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3735 * for UDP-style sockets only.
3736 */
3737 sp->autoclose = 0;
3738
3739 /* User specified fragmentation limit. */
3740 sp->user_frag = 0;
3741
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003742 sp->adaptation_ind = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003743
3744 sp->pf = sctp_get_pf_specific(sk->sk_family);
3745
3746 /* Control variables for partial data delivery. */
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003747 atomic_set(&sp->pd_mode, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003748 skb_queue_head_init(&sp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003749 sp->frag_interleave = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003750
3751 /* Create a per socket endpoint structure. Even if we
3752 * change the data structure relationships, this may still
3753 * be useful for storing pre-connect address information.
3754 */
3755 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3756 if (!ep)
3757 return -ENOMEM;
3758
3759 sp->ep = ep;
3760 sp->hmac = NULL;
3761
3762 SCTP_DBG_OBJCNT_INC(sock);
David S. Miller6f756a82008-11-23 17:34:03 -08003763
3764 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003765 percpu_counter_inc(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003766 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
David S. Miller6f756a82008-11-23 17:34:03 -08003767 local_bh_enable();
3768
Linus Torvalds1da177e2005-04-16 15:20:36 -07003769 return 0;
3770}
3771
3772/* Cleanup any SCTP per socket resources. */
Brian Haley7d06b2e2008-06-14 17:04:49 -07003773SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003774{
3775 struct sctp_endpoint *ep;
3776
3777 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3778
3779 /* Release our hold on the endpoint. */
3780 ep = sctp_sk(sk)->ep;
3781 sctp_endpoint_free(ep);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003782 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003783 percpu_counter_dec(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003784 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003785 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003786}
3787
3788/* API 4.1.7 shutdown() - TCP Style Syntax
3789 * int shutdown(int socket, int how);
3790 *
3791 * sd - the socket descriptor of the association to be closed.
3792 * how - Specifies the type of shutdown. The values are
3793 * as follows:
3794 * SHUT_RD
3795 * Disables further receive operations. No SCTP
3796 * protocol action is taken.
3797 * SHUT_WR
3798 * Disables further send operations, and initiates
3799 * the SCTP shutdown sequence.
3800 * SHUT_RDWR
3801 * Disables further send and receive operations
3802 * and initiates the SCTP shutdown sequence.
3803 */
3804SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3805{
3806 struct sctp_endpoint *ep;
3807 struct sctp_association *asoc;
3808
3809 if (!sctp_style(sk, TCP))
3810 return;
3811
3812 if (how & SEND_SHUTDOWN) {
3813 ep = sctp_sk(sk)->ep;
3814 if (!list_empty(&ep->asocs)) {
3815 asoc = list_entry(ep->asocs.next,
3816 struct sctp_association, asocs);
3817 sctp_primitive_SHUTDOWN(asoc, NULL);
3818 }
3819 }
3820}
3821
3822/* 7.2.1 Association Status (SCTP_STATUS)
3823
3824 * Applications can retrieve current status information about an
3825 * association, including association state, peer receiver window size,
3826 * number of unacked data chunks, and number of data chunks pending
3827 * receipt. This information is read-only.
3828 */
3829static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3830 char __user *optval,
3831 int __user *optlen)
3832{
3833 struct sctp_status status;
3834 struct sctp_association *asoc = NULL;
3835 struct sctp_transport *transport;
3836 sctp_assoc_t associd;
3837 int retval = 0;
3838
Neil Horman408f22e2007-06-16 14:03:45 -04003839 if (len < sizeof(status)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003840 retval = -EINVAL;
3841 goto out;
3842 }
3843
Neil Horman408f22e2007-06-16 14:03:45 -04003844 len = sizeof(status);
3845 if (copy_from_user(&status, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003846 retval = -EFAULT;
3847 goto out;
3848 }
3849
3850 associd = status.sstat_assoc_id;
3851 asoc = sctp_id2assoc(sk, associd);
3852 if (!asoc) {
3853 retval = -EINVAL;
3854 goto out;
3855 }
3856
3857 transport = asoc->peer.primary_path;
3858
3859 status.sstat_assoc_id = sctp_assoc2id(asoc);
3860 status.sstat_state = asoc->state;
3861 status.sstat_rwnd = asoc->peer.rwnd;
3862 status.sstat_unackdata = asoc->unack_data;
3863
3864 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3865 status.sstat_instrms = asoc->c.sinit_max_instreams;
3866 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3867 status.sstat_fragmentation_point = asoc->frag_point;
3868 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Al Viro8cec6b82006-11-20 17:23:01 -08003869 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3870 transport->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003871 /* Map ipv4 address into v4-mapped-on-v6 address. */
3872 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3873 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003874 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003875 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3876 status.sstat_primary.spinfo_srtt = transport->srtt;
3877 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003878 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003879
Frank Filz3f7a87d2005-06-20 13:14:57 -07003880 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3881 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3882
Linus Torvalds1da177e2005-04-16 15:20:36 -07003883 if (put_user(len, optlen)) {
3884 retval = -EFAULT;
3885 goto out;
3886 }
3887
3888 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3889 len, status.sstat_state, status.sstat_rwnd,
3890 status.sstat_assoc_id);
3891
3892 if (copy_to_user(optval, &status, len)) {
3893 retval = -EFAULT;
3894 goto out;
3895 }
3896
3897out:
Eric Dumazeta02cec22010-09-22 20:43:57 +00003898 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003899}
3900
3901
3902/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3903 *
3904 * Applications can retrieve information about a specific peer address
3905 * of an association, including its reachability state, congestion
3906 * window, and retransmission timer values. This information is
3907 * read-only.
3908 */
3909static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3910 char __user *optval,
3911 int __user *optlen)
3912{
3913 struct sctp_paddrinfo pinfo;
3914 struct sctp_transport *transport;
3915 int retval = 0;
3916
Neil Horman408f22e2007-06-16 14:03:45 -04003917 if (len < sizeof(pinfo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003918 retval = -EINVAL;
3919 goto out;
3920 }
3921
Neil Horman408f22e2007-06-16 14:03:45 -04003922 len = sizeof(pinfo);
3923 if (copy_from_user(&pinfo, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003924 retval = -EFAULT;
3925 goto out;
3926 }
3927
3928 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3929 pinfo.spinfo_assoc_id);
3930 if (!transport)
3931 return -EINVAL;
3932
3933 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003934 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003935 pinfo.spinfo_cwnd = transport->cwnd;
3936 pinfo.spinfo_srtt = transport->srtt;
3937 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003938 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003939
Frank Filz3f7a87d2005-06-20 13:14:57 -07003940 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3941 pinfo.spinfo_state = SCTP_ACTIVE;
3942
Linus Torvalds1da177e2005-04-16 15:20:36 -07003943 if (put_user(len, optlen)) {
3944 retval = -EFAULT;
3945 goto out;
3946 }
3947
3948 if (copy_to_user(optval, &pinfo, len)) {
3949 retval = -EFAULT;
3950 goto out;
3951 }
3952
3953out:
Eric Dumazeta02cec22010-09-22 20:43:57 +00003954 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003955}
3956
3957/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3958 *
3959 * This option is a on/off flag. If enabled no SCTP message
3960 * fragmentation will be performed. Instead if a message being sent
3961 * exceeds the current PMTU size, the message will NOT be sent and
3962 * instead a error will be indicated to the user.
3963 */
3964static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3965 char __user *optval, int __user *optlen)
3966{
3967 int val;
3968
3969 if (len < sizeof(int))
3970 return -EINVAL;
3971
3972 len = sizeof(int);
3973 val = (sctp_sk(sk)->disable_fragments == 1);
3974 if (put_user(len, optlen))
3975 return -EFAULT;
3976 if (copy_to_user(optval, &val, len))
3977 return -EFAULT;
3978 return 0;
3979}
3980
3981/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3982 *
3983 * This socket option is used to specify various notifications and
3984 * ancillary data the user wishes to receive.
3985 */
3986static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3987 int __user *optlen)
3988{
Neil Horman408f22e2007-06-16 14:03:45 -04003989 if (len < sizeof(struct sctp_event_subscribe))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003990 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04003991 len = sizeof(struct sctp_event_subscribe);
3992 if (put_user(len, optlen))
3993 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003994 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3995 return -EFAULT;
3996 return 0;
3997}
3998
3999/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4000 *
4001 * This socket option is applicable to the UDP-style socket only. When
4002 * set it will cause associations that are idle for more than the
4003 * specified number of seconds to automatically close. An association
4004 * being idle is defined an association that has NOT sent or received
4005 * user data. The special value of '0' indicates that no automatic
4006 * close of any associations should be performed. The option expects an
4007 * integer defining the number of seconds of idle time before an
4008 * association is closed.
4009 */
4010static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4011{
4012 /* Applicable to UDP-style socket only */
4013 if (sctp_style(sk, TCP))
4014 return -EOPNOTSUPP;
Neil Horman408f22e2007-06-16 14:03:45 -04004015 if (len < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004016 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004017 len = sizeof(int);
4018 if (put_user(len, optlen))
4019 return -EFAULT;
4020 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004021 return -EFAULT;
4022 return 0;
4023}
4024
4025/* Helper routine to branch off an association to a new socket. */
4026SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4027 struct socket **sockp)
4028{
4029 struct sock *sk = asoc->base.sk;
4030 struct socket *sock;
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004031 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004032 int err = 0;
4033
4034 /* An association cannot be branched off from an already peeled-off
4035 * socket, nor is this supported for tcp style sockets.
4036 */
4037 if (!sctp_style(sk, UDP))
4038 return -EINVAL;
4039
4040 /* Create a new socket. */
4041 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4042 if (err < 0)
4043 return err;
4044
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004045 sctp_copy_sock(sock->sk, sk, asoc);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004046
4047 /* Make peeled-off sockets more like 1-1 accepted sockets.
4048 * Set the daddr and initialize id to something more random
4049 */
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004050 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4051 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004052
4053 /* Populate the fields of the newsk from the oldsk and migrate the
4054 * asoc to the newsk.
4055 */
4056 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004057
Linus Torvalds1da177e2005-04-16 15:20:36 -07004058 *sockp = sock;
4059
4060 return err;
4061}
4062
4063static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4064{
4065 sctp_peeloff_arg_t peeloff;
4066 struct socket *newsock;
4067 int retval = 0;
4068 struct sctp_association *asoc;
4069
Neil Horman408f22e2007-06-16 14:03:45 -04004070 if (len < sizeof(sctp_peeloff_arg_t))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004071 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004072 len = sizeof(sctp_peeloff_arg_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004073 if (copy_from_user(&peeloff, optval, len))
4074 return -EFAULT;
4075
4076 asoc = sctp_id2assoc(sk, peeloff.associd);
4077 if (!asoc) {
4078 retval = -EINVAL;
4079 goto out;
4080 }
4081
Harvey Harrison0dc47872008-03-05 20:47:47 -08004082 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004083
4084 retval = sctp_do_peeloff(asoc, &newsock);
4085 if (retval < 0)
4086 goto out;
4087
4088 /* Map the socket to an unused fd that can be returned to the user. */
Ulrich Dreppera677a032008-07-23 21:29:17 -07004089 retval = sock_map_fd(newsock, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004090 if (retval < 0) {
4091 sock_release(newsock);
4092 goto out;
4093 }
4094
4095 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08004096 __func__, sk, asoc, newsock->sk, retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004097
4098 /* Return the fd mapped to the new socket. */
4099 peeloff.sd = retval;
Neil Horman408f22e2007-06-16 14:03:45 -04004100 if (put_user(len, optlen))
4101 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004102 if (copy_to_user(optval, &peeloff, len))
4103 retval = -EFAULT;
4104
4105out:
4106 return retval;
4107}
4108
4109/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4110 *
4111 * Applications can enable or disable heartbeats for any peer address of
4112 * an association, modify an address's heartbeat interval, force a
4113 * heartbeat to be sent immediately, and adjust the address's maximum
4114 * number of retransmissions sent before an address is considered
4115 * unreachable. The following structure is used to access and modify an
4116 * address's parameters:
4117 *
4118 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004119 * sctp_assoc_t spp_assoc_id;
4120 * struct sockaddr_storage spp_address;
4121 * uint32_t spp_hbinterval;
4122 * uint16_t spp_pathmaxrxt;
4123 * uint32_t spp_pathmtu;
4124 * uint32_t spp_sackdelay;
4125 * uint32_t spp_flags;
4126 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07004127 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08004128 * spp_assoc_id - (one-to-many style socket) This is filled in the
4129 * application, and identifies the association for
4130 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004131 * spp_address - This specifies which address is of interest.
4132 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004133 * in milliseconds. If a value of zero
4134 * is present in this field then no changes are to
4135 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004136 * spp_pathmaxrxt - This contains the maximum number of
4137 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08004138 * considered unreachable. If a value of zero
4139 * is present in this field then no changes are to
4140 * be made to this parameter.
4141 * spp_pathmtu - When Path MTU discovery is disabled the value
4142 * specified here will be the "fixed" path mtu.
4143 * Note that if the spp_address field is empty
4144 * then all associations on this address will
4145 * have this fixed path mtu set upon them.
4146 *
4147 * spp_sackdelay - When delayed sack is enabled, this value specifies
4148 * the number of milliseconds that sacks will be delayed
4149 * for. This value will apply to all addresses of an
4150 * association if the spp_address field is empty. Note
4151 * also, that if delayed sack is enabled and this
4152 * value is set to 0, no change is made to the last
4153 * recorded delayed sack timer value.
4154 *
4155 * spp_flags - These flags are used to control various features
4156 * on an association. The flag field may contain
4157 * zero or more of the following options.
4158 *
4159 * SPP_HB_ENABLE - Enable heartbeats on the
4160 * specified address. Note that if the address
4161 * field is empty all addresses for the association
4162 * have heartbeats enabled upon them.
4163 *
4164 * SPP_HB_DISABLE - Disable heartbeats on the
4165 * speicifed address. Note that if the address
4166 * field is empty all addresses for the association
4167 * will have their heartbeats disabled. Note also
4168 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4169 * mutually exclusive, only one of these two should
4170 * be specified. Enabling both fields will have
4171 * undetermined results.
4172 *
4173 * SPP_HB_DEMAND - Request a user initiated heartbeat
4174 * to be made immediately.
4175 *
4176 * SPP_PMTUD_ENABLE - This field will enable PMTU
4177 * discovery upon the specified address. Note that
4178 * if the address feild is empty then all addresses
4179 * on the association are effected.
4180 *
4181 * SPP_PMTUD_DISABLE - This field will disable PMTU
4182 * discovery upon the specified address. Note that
4183 * if the address feild is empty then all addresses
4184 * on the association are effected. Not also that
4185 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4186 * exclusive. Enabling both will have undetermined
4187 * results.
4188 *
4189 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4190 * on delayed sack. The time specified in spp_sackdelay
4191 * is used to specify the sack delay for this address. Note
4192 * that if spp_address is empty then all addresses will
4193 * enable delayed sack and take on the sack delay
4194 * value specified in spp_sackdelay.
4195 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4196 * off delayed sack. If the spp_address field is blank then
4197 * delayed sack is disabled for the entire association. Note
4198 * also that this field is mutually exclusive to
4199 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4200 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004201 */
4202static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004203 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004204{
Frank Filz52ccb8e2005-12-22 11:36:46 -08004205 struct sctp_paddrparams params;
4206 struct sctp_transport *trans = NULL;
4207 struct sctp_association *asoc = NULL;
4208 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004209
Neil Horman408f22e2007-06-16 14:03:45 -04004210 if (len < sizeof(struct sctp_paddrparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004211 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004212 len = sizeof(struct sctp_paddrparams);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004213 if (copy_from_user(&params, optval, len))
4214 return -EFAULT;
4215
Frank Filz52ccb8e2005-12-22 11:36:46 -08004216 /* If an address other than INADDR_ANY is specified, and
4217 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004218 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004219 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004220 trans = sctp_addr_id2transport(sk, &params.spp_address,
4221 params.spp_assoc_id);
4222 if (!trans) {
4223 SCTP_DEBUG_PRINTK("Failed no transport\n");
4224 return -EINVAL;
4225 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004226 }
4227
Frank Filz52ccb8e2005-12-22 11:36:46 -08004228 /* Get association, if assoc_id != 0 and the socket is a one
4229 * to many style socket, and an association was not found, then
4230 * the id was invalid.
4231 */
4232 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4233 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4234 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07004235 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08004236 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004237
Frank Filz52ccb8e2005-12-22 11:36:46 -08004238 if (trans) {
4239 /* Fetch transport values. */
4240 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4241 params.spp_pathmtu = trans->pathmtu;
4242 params.spp_pathmaxrxt = trans->pathmaxrxt;
4243 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004244
Frank Filz52ccb8e2005-12-22 11:36:46 -08004245 /*draft-11 doesn't say what to return in spp_flags*/
4246 params.spp_flags = trans->param_flags;
4247 } else if (asoc) {
4248 /* Fetch association values. */
4249 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4250 params.spp_pathmtu = asoc->pathmtu;
4251 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4252 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004253
Frank Filz52ccb8e2005-12-22 11:36:46 -08004254 /*draft-11 doesn't say what to return in spp_flags*/
4255 params.spp_flags = asoc->param_flags;
4256 } else {
4257 /* Fetch socket values. */
4258 params.spp_hbinterval = sp->hbinterval;
4259 params.spp_pathmtu = sp->pathmtu;
4260 params.spp_sackdelay = sp->sackdelay;
4261 params.spp_pathmaxrxt = sp->pathmaxrxt;
4262
4263 /*draft-11 doesn't say what to return in spp_flags*/
4264 params.spp_flags = sp->param_flags;
4265 }
4266
Linus Torvalds1da177e2005-04-16 15:20:36 -07004267 if (copy_to_user(optval, &params, len))
4268 return -EFAULT;
4269
4270 if (put_user(len, optlen))
4271 return -EFAULT;
4272
4273 return 0;
4274}
4275
Wei Yongjund364d922008-05-09 15:13:26 -07004276/*
4277 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08004278 *
Wei Yongjund364d922008-05-09 15:13:26 -07004279 * This option will effect the way delayed acks are performed. This
4280 * option allows you to get or set the delayed ack time, in
4281 * milliseconds. It also allows changing the delayed ack frequency.
4282 * Changing the frequency to 1 disables the delayed sack algorithm. If
4283 * the assoc_id is 0, then this sets or gets the endpoints default
4284 * values. If the assoc_id field is non-zero, then the set or get
4285 * effects the specified association for the one to many model (the
4286 * assoc_id field is ignored by the one to one model). Note that if
4287 * sack_delay or sack_freq are 0 when setting this option, then the
4288 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08004289 *
Wei Yongjund364d922008-05-09 15:13:26 -07004290 * struct sctp_sack_info {
4291 * sctp_assoc_t sack_assoc_id;
4292 * uint32_t sack_delay;
4293 * uint32_t sack_freq;
4294 * };
Frank Filz77086102005-12-22 11:37:30 -08004295 *
Wei Yongjund364d922008-05-09 15:13:26 -07004296 * sack_assoc_id - This parameter, indicates which association the user
4297 * is performing an action upon. Note that if this field's value is
4298 * zero then the endpoints default value is changed (effecting future
4299 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08004300 *
Wei Yongjund364d922008-05-09 15:13:26 -07004301 * sack_delay - This parameter contains the number of milliseconds that
4302 * the user is requesting the delayed ACK timer be set to. Note that
4303 * this value is defined in the standard to be between 200 and 500
4304 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08004305 *
Wei Yongjund364d922008-05-09 15:13:26 -07004306 * sack_freq - This parameter contains the number of packets that must
4307 * be received before a sack is sent without waiting for the delay
4308 * timer to expire. The default value for this is 2, setting this
4309 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08004310 */
Wei Yongjund364d922008-05-09 15:13:26 -07004311static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
Frank Filz77086102005-12-22 11:37:30 -08004312 char __user *optval,
4313 int __user *optlen)
4314{
Wei Yongjund364d922008-05-09 15:13:26 -07004315 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08004316 struct sctp_association *asoc = NULL;
4317 struct sctp_sock *sp = sctp_sk(sk);
4318
Wei Yongjund364d922008-05-09 15:13:26 -07004319 if (len >= sizeof(struct sctp_sack_info)) {
4320 len = sizeof(struct sctp_sack_info);
4321
4322 if (copy_from_user(&params, optval, len))
4323 return -EFAULT;
4324 } else if (len == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00004325 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4326 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07004327 if (copy_from_user(&params, optval, len))
4328 return -EFAULT;
4329 } else
Frank Filz77086102005-12-22 11:37:30 -08004330 return - EINVAL;
4331
Wei Yongjund364d922008-05-09 15:13:26 -07004332 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08004333 * to many style socket, and an association was not found, then
4334 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004335 */
Wei Yongjund364d922008-05-09 15:13:26 -07004336 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4337 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08004338 return -EINVAL;
4339
4340 if (asoc) {
4341 /* Fetch association values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004342 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4343 params.sack_delay = jiffies_to_msecs(
Frank Filz77086102005-12-22 11:37:30 -08004344 asoc->sackdelay);
Wei Yongjund364d922008-05-09 15:13:26 -07004345 params.sack_freq = asoc->sackfreq;
4346
4347 } else {
4348 params.sack_delay = 0;
4349 params.sack_freq = 1;
4350 }
Frank Filz77086102005-12-22 11:37:30 -08004351 } else {
4352 /* Fetch socket values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004353 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4354 params.sack_delay = sp->sackdelay;
4355 params.sack_freq = sp->sackfreq;
4356 } else {
4357 params.sack_delay = 0;
4358 params.sack_freq = 1;
4359 }
Frank Filz77086102005-12-22 11:37:30 -08004360 }
4361
4362 if (copy_to_user(optval, &params, len))
4363 return -EFAULT;
4364
4365 if (put_user(len, optlen))
4366 return -EFAULT;
4367
4368 return 0;
4369}
4370
Linus Torvalds1da177e2005-04-16 15:20:36 -07004371/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4372 *
4373 * Applications can specify protocol parameters for the default association
4374 * initialization. The option name argument to setsockopt() and getsockopt()
4375 * is SCTP_INITMSG.
4376 *
4377 * Setting initialization parameters is effective only on an unconnected
4378 * socket (for UDP-style sockets only future associations are effected
4379 * by the change). With TCP-style sockets, this option is inherited by
4380 * sockets derived from a listener socket.
4381 */
4382static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4383{
Neil Horman408f22e2007-06-16 14:03:45 -04004384 if (len < sizeof(struct sctp_initmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004385 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004386 len = sizeof(struct sctp_initmsg);
4387 if (put_user(len, optlen))
4388 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004389 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4390 return -EFAULT;
4391 return 0;
4392}
4393
Linus Torvalds1da177e2005-04-16 15:20:36 -07004394
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004395static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4396 char __user *optval, int __user *optlen)
4397{
4398 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004399 int cnt = 0;
4400 struct sctp_getaddrs getaddrs;
4401 struct sctp_transport *from;
4402 void __user *to;
4403 union sctp_addr temp;
4404 struct sctp_sock *sp = sctp_sk(sk);
4405 int addrlen;
4406 size_t space_left;
4407 int bytes_copied;
4408
4409 if (len < sizeof(struct sctp_getaddrs))
4410 return -EINVAL;
4411
4412 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4413 return -EFAULT;
4414
4415 /* For UDP-style sockets, id specifies the association to query. */
4416 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4417 if (!asoc)
4418 return -EINVAL;
4419
4420 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004421 space_left = len - offsetof(struct sctp_getaddrs,addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004422
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07004423 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4424 transports) {
Al Virob3f5b3b2006-11-20 17:22:43 -08004425 memcpy(&temp, &from->ipaddr, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004426 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
Vlad Yasevicha5f4cea2010-04-30 21:42:42 -04004427 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004428 if (space_left < addrlen)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004429 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004430 if (copy_to_user(to, &temp, addrlen))
4431 return -EFAULT;
4432 to += addrlen;
4433 cnt++;
4434 space_left -= addrlen;
4435 }
4436
4437 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4438 return -EFAULT;
4439 bytes_copied = ((char __user *)to) - optval;
4440 if (put_user(bytes_copied, optlen))
4441 return -EFAULT;
4442
4443 return 0;
4444}
4445
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004446static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4447 size_t space_left, int *bytes_copied)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004448{
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004449 struct sctp_sockaddr_entry *addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004450 union sctp_addr temp;
4451 int cnt = 0;
4452 int addrlen;
4453
Vlad Yasevich29303542007-09-16 16:02:12 -07004454 rcu_read_lock();
4455 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4456 if (!addr->valid)
4457 continue;
4458
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004459 if ((PF_INET == sk->sk_family) &&
Al Viro6244be42006-11-20 17:21:44 -08004460 (AF_INET6 == addr->a.sa.sa_family))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004461 continue;
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07004462 if ((PF_INET6 == sk->sk_family) &&
4463 inet_v6_ipv6only(sk) &&
4464 (AF_INET == addr->a.sa.sa_family))
4465 continue;
Al Viro6244be42006-11-20 17:21:44 -08004466 memcpy(&temp, &addr->a, sizeof(temp));
Vlad Yasevichb46ae362008-01-28 14:25:36 -05004467 if (!temp.v4.sin_port)
4468 temp.v4.sin_port = htons(port);
4469
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004470 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4471 &temp);
4472 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevich29303542007-09-16 16:02:12 -07004473 if (space_left < addrlen) {
4474 cnt = -ENOMEM;
4475 break;
4476 }
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004477 memcpy(to, &temp, addrlen);
Sridhar Samudrala29c7cf92006-12-13 16:26:26 -08004478
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004479 to += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004480 cnt ++;
4481 space_left -= addrlen;
Vlad Yasevich3663c302007-07-03 12:43:12 -04004482 *bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004483 }
Vlad Yasevich29303542007-09-16 16:02:12 -07004484 rcu_read_unlock();
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004485
4486 return cnt;
4487}
4488
Linus Torvalds1da177e2005-04-16 15:20:36 -07004489
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004490static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4491 char __user *optval, int __user *optlen)
4492{
4493 struct sctp_bind_addr *bp;
4494 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004495 int cnt = 0;
4496 struct sctp_getaddrs getaddrs;
4497 struct sctp_sockaddr_entry *addr;
4498 void __user *to;
4499 union sctp_addr temp;
4500 struct sctp_sock *sp = sctp_sk(sk);
4501 int addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004502 int err = 0;
4503 size_t space_left;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004504 int bytes_copied = 0;
4505 void *addrs;
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004506 void *buf;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004507
Neil Horman408f22e2007-06-16 14:03:45 -04004508 if (len < sizeof(struct sctp_getaddrs))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004509 return -EINVAL;
4510
4511 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4512 return -EFAULT;
4513
4514 /*
4515 * For UDP-style sockets, id specifies the association to query.
4516 * If the id field is set to the value '0' then the locally bound
4517 * addresses are returned without regard to any particular
4518 * association.
4519 */
4520 if (0 == getaddrs.assoc_id) {
4521 bp = &sctp_sk(sk)->ep->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004522 } else {
4523 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4524 if (!asoc)
4525 return -EINVAL;
4526 bp = &asoc->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004527 }
4528
4529 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004530 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4531
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004532 addrs = kmalloc(space_left, GFP_KERNEL);
4533 if (!addrs)
4534 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004535
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004536 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4537 * addresses from the global local address list.
4538 */
4539 if (sctp_list_single_entry(&bp->address_list)) {
4540 addr = list_entry(bp->address_list.next,
4541 struct sctp_sockaddr_entry, list);
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004542 if (sctp_is_any(sk, &addr->a)) {
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004543 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4544 space_left, &bytes_copied);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004545 if (cnt < 0) {
4546 err = cnt;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004547 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004548 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004549 goto copy_getaddrs;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004550 }
4551 }
4552
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004553 buf = addrs;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004554 /* Protection on the bound address list is not needed since
4555 * in the socket option context we hold a socket lock and
4556 * thus the bound address list can't change.
4557 */
4558 list_for_each_entry(addr, &bp->address_list, list) {
Al Viro6244be42006-11-20 17:21:44 -08004559 memcpy(&temp, &addr->a, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004560 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4561 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004562 if (space_left < addrlen) {
4563 err = -ENOMEM; /*fixme: right error?*/
Vlad Yasevich559cf712007-09-16 16:03:28 -07004564 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004565 }
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004566 memcpy(buf, &temp, addrlen);
4567 buf += addrlen;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004568 bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004569 cnt ++;
4570 space_left -= addrlen;
4571 }
4572
4573copy_getaddrs:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004574 if (copy_to_user(to, addrs, bytes_copied)) {
4575 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004576 goto out;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004577 }
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004578 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4579 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004580 goto out;
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004581 }
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004582 if (put_user(bytes_copied, optlen))
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004583 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004584out:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004585 kfree(addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004586 return err;
4587}
4588
Linus Torvalds1da177e2005-04-16 15:20:36 -07004589/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4590 *
4591 * Requests that the local SCTP stack use the enclosed peer address as
4592 * the association primary. The enclosed address must be one of the
4593 * association peer's addresses.
4594 */
4595static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4596 char __user *optval, int __user *optlen)
4597{
4598 struct sctp_prim prim;
4599 struct sctp_association *asoc;
4600 struct sctp_sock *sp = sctp_sk(sk);
4601
Neil Horman408f22e2007-06-16 14:03:45 -04004602 if (len < sizeof(struct sctp_prim))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004603 return -EINVAL;
4604
Neil Horman408f22e2007-06-16 14:03:45 -04004605 len = sizeof(struct sctp_prim);
4606
4607 if (copy_from_user(&prim, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004608 return -EFAULT;
4609
4610 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4611 if (!asoc)
4612 return -EINVAL;
4613
4614 if (!asoc->peer.primary_path)
4615 return -ENOTCONN;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004616
Al Viro8cec6b82006-11-20 17:23:01 -08004617 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4618 asoc->peer.primary_path->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004619
4620 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4621 (union sctp_addr *)&prim.ssp_addr);
4622
Neil Horman408f22e2007-06-16 14:03:45 -04004623 if (put_user(len, optlen))
4624 return -EFAULT;
4625 if (copy_to_user(optval, &prim, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004626 return -EFAULT;
4627
4628 return 0;
4629}
4630
4631/*
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004632 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004633 *
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004634 * Requests that the local endpoint set the specified Adaptation Layer
Linus Torvalds1da177e2005-04-16 15:20:36 -07004635 * Indication parameter for all future INIT and INIT-ACK exchanges.
4636 */
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004637static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004638 char __user *optval, int __user *optlen)
4639{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004640 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004641
Neil Horman408f22e2007-06-16 14:03:45 -04004642 if (len < sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004643 return -EINVAL;
4644
Neil Horman408f22e2007-06-16 14:03:45 -04004645 len = sizeof(struct sctp_setadaptation);
4646
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004647 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
Neil Horman408f22e2007-06-16 14:03:45 -04004648
4649 if (put_user(len, optlen))
4650 return -EFAULT;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004651 if (copy_to_user(optval, &adaptation, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004652 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004653
Linus Torvalds1da177e2005-04-16 15:20:36 -07004654 return 0;
4655}
4656
4657/*
4658 *
4659 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4660 *
4661 * Applications that wish to use the sendto() system call may wish to
4662 * specify a default set of parameters that would normally be supplied
4663 * through the inclusion of ancillary data. This socket option allows
4664 * such an application to set the default sctp_sndrcvinfo structure.
4665
4666
4667 * The application that wishes to use this socket option simply passes
4668 * in to this call the sctp_sndrcvinfo structure defined in Section
4669 * 5.2.2) The input parameters accepted by this call include
4670 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4671 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4672 * to this call if the caller is using the UDP model.
4673 *
4674 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4675 */
4676static int sctp_getsockopt_default_send_param(struct sock *sk,
4677 int len, char __user *optval,
4678 int __user *optlen)
4679{
4680 struct sctp_sndrcvinfo info;
4681 struct sctp_association *asoc;
4682 struct sctp_sock *sp = sctp_sk(sk);
4683
Neil Horman408f22e2007-06-16 14:03:45 -04004684 if (len < sizeof(struct sctp_sndrcvinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004685 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004686
4687 len = sizeof(struct sctp_sndrcvinfo);
4688
4689 if (copy_from_user(&info, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004690 return -EFAULT;
4691
4692 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4693 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4694 return -EINVAL;
4695
4696 if (asoc) {
4697 info.sinfo_stream = asoc->default_stream;
4698 info.sinfo_flags = asoc->default_flags;
4699 info.sinfo_ppid = asoc->default_ppid;
4700 info.sinfo_context = asoc->default_context;
4701 info.sinfo_timetolive = asoc->default_timetolive;
4702 } else {
4703 info.sinfo_stream = sp->default_stream;
4704 info.sinfo_flags = sp->default_flags;
4705 info.sinfo_ppid = sp->default_ppid;
4706 info.sinfo_context = sp->default_context;
4707 info.sinfo_timetolive = sp->default_timetolive;
4708 }
4709
Neil Horman408f22e2007-06-16 14:03:45 -04004710 if (put_user(len, optlen))
4711 return -EFAULT;
4712 if (copy_to_user(optval, &info, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004713 return -EFAULT;
4714
4715 return 0;
4716}
4717
4718/*
4719 *
4720 * 7.1.5 SCTP_NODELAY
4721 *
4722 * Turn on/off any Nagle-like algorithm. This means that packets are
4723 * generally sent as soon as possible and no unnecessary delays are
4724 * introduced, at the cost of more packets in the network. Expects an
4725 * integer boolean flag.
4726 */
4727
4728static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4729 char __user *optval, int __user *optlen)
4730{
4731 int val;
4732
4733 if (len < sizeof(int))
4734 return -EINVAL;
4735
4736 len = sizeof(int);
4737 val = (sctp_sk(sk)->nodelay == 1);
4738 if (put_user(len, optlen))
4739 return -EFAULT;
4740 if (copy_to_user(optval, &val, len))
4741 return -EFAULT;
4742 return 0;
4743}
4744
4745/*
4746 *
4747 * 7.1.1 SCTP_RTOINFO
4748 *
4749 * The protocol parameters used to initialize and bound retransmission
4750 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4751 * and modify these parameters.
4752 * All parameters are time values, in milliseconds. A value of 0, when
4753 * modifying the parameters, indicates that the current value should not
4754 * be changed.
4755 *
4756 */
4757static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4758 char __user *optval,
4759 int __user *optlen) {
4760 struct sctp_rtoinfo rtoinfo;
4761 struct sctp_association *asoc;
4762
Neil Horman408f22e2007-06-16 14:03:45 -04004763 if (len < sizeof (struct sctp_rtoinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004764 return -EINVAL;
4765
Neil Horman408f22e2007-06-16 14:03:45 -04004766 len = sizeof(struct sctp_rtoinfo);
4767
4768 if (copy_from_user(&rtoinfo, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004769 return -EFAULT;
4770
4771 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4772
4773 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4774 return -EINVAL;
4775
4776 /* Values corresponding to the specific association. */
4777 if (asoc) {
4778 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4779 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4780 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4781 } else {
4782 /* Values corresponding to the endpoint. */
4783 struct sctp_sock *sp = sctp_sk(sk);
4784
4785 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4786 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4787 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4788 }
4789
4790 if (put_user(len, optlen))
4791 return -EFAULT;
4792
4793 if (copy_to_user(optval, &rtoinfo, len))
4794 return -EFAULT;
4795
4796 return 0;
4797}
4798
4799/*
4800 *
4801 * 7.1.2 SCTP_ASSOCINFO
4802 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02004803 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07004804 * of the association.
4805 * Returns an error if the new association retransmission value is
4806 * greater than the sum of the retransmission value of the peer.
4807 * See [SCTP] for more information.
4808 *
4809 */
4810static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4811 char __user *optval,
4812 int __user *optlen)
4813{
4814
4815 struct sctp_assocparams assocparams;
4816 struct sctp_association *asoc;
4817 struct list_head *pos;
4818 int cnt = 0;
4819
Neil Horman408f22e2007-06-16 14:03:45 -04004820 if (len < sizeof (struct sctp_assocparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004821 return -EINVAL;
4822
Neil Horman408f22e2007-06-16 14:03:45 -04004823 len = sizeof(struct sctp_assocparams);
4824
4825 if (copy_from_user(&assocparams, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004826 return -EFAULT;
4827
4828 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4829
4830 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4831 return -EINVAL;
4832
4833 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07004834 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004835 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4836 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4837 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4838 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4839 * 1000) +
4840 (asoc->cookie_life.tv_usec
4841 / 1000);
4842
4843 list_for_each(pos, &asoc->peer.transport_addr_list) {
4844 cnt ++;
4845 }
4846
4847 assocparams.sasoc_number_peer_destinations = cnt;
4848 } else {
4849 /* Values corresponding to the endpoint */
4850 struct sctp_sock *sp = sctp_sk(sk);
4851
4852 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4853 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4854 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4855 assocparams.sasoc_cookie_life =
4856 sp->assocparams.sasoc_cookie_life;
4857 assocparams.sasoc_number_peer_destinations =
4858 sp->assocparams.
4859 sasoc_number_peer_destinations;
4860 }
4861
4862 if (put_user(len, optlen))
4863 return -EFAULT;
4864
4865 if (copy_to_user(optval, &assocparams, len))
4866 return -EFAULT;
4867
4868 return 0;
4869}
4870
4871/*
4872 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4873 *
4874 * This socket option is a boolean flag which turns on or off mapped V4
4875 * addresses. If this option is turned on and the socket is type
4876 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4877 * If this option is turned off, then no mapping will be done of V4
4878 * addresses and a user will receive both PF_INET6 and PF_INET type
4879 * addresses on the socket.
4880 */
4881static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4882 char __user *optval, int __user *optlen)
4883{
4884 int val;
4885 struct sctp_sock *sp = sctp_sk(sk);
4886
4887 if (len < sizeof(int))
4888 return -EINVAL;
4889
4890 len = sizeof(int);
4891 val = sp->v4mapped;
4892 if (put_user(len, optlen))
4893 return -EFAULT;
4894 if (copy_to_user(optval, &val, len))
4895 return -EFAULT;
4896
4897 return 0;
4898}
4899
4900/*
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004901 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4902 * (chapter and verse is quoted at sctp_setsockopt_context())
4903 */
4904static int sctp_getsockopt_context(struct sock *sk, int len,
4905 char __user *optval, int __user *optlen)
4906{
4907 struct sctp_assoc_value params;
4908 struct sctp_sock *sp;
4909 struct sctp_association *asoc;
4910
Neil Horman408f22e2007-06-16 14:03:45 -04004911 if (len < sizeof(struct sctp_assoc_value))
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004912 return -EINVAL;
4913
Neil Horman408f22e2007-06-16 14:03:45 -04004914 len = sizeof(struct sctp_assoc_value);
4915
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004916 if (copy_from_user(&params, optval, len))
4917 return -EFAULT;
4918
4919 sp = sctp_sk(sk);
4920
4921 if (params.assoc_id != 0) {
4922 asoc = sctp_id2assoc(sk, params.assoc_id);
4923 if (!asoc)
4924 return -EINVAL;
4925 params.assoc_value = asoc->default_rcv_context;
4926 } else {
4927 params.assoc_value = sp->default_rcv_context;
4928 }
4929
4930 if (put_user(len, optlen))
4931 return -EFAULT;
4932 if (copy_to_user(optval, &params, len))
4933 return -EFAULT;
4934
4935 return 0;
4936}
4937
4938/*
Wei Yongjune89c2092008-12-25 16:54:58 -08004939 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4940 * This option will get or set the maximum size to put in any outgoing
4941 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07004942 * fragmented by SCTP into the specified size. Note that the underlying
4943 * SCTP implementation may fragment into smaller sized chunks when the
4944 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08004945 * the user. The default value for this option is '0' which indicates
4946 * the user is NOT limiting fragmentation and only the PMTU will effect
4947 * SCTP's choice of DATA chunk size. Note also that values set larger
4948 * than the maximum size of an IP datagram will effectively let SCTP
4949 * control fragmentation (i.e. the same as setting this option to 0).
4950 *
4951 * The following structure is used to access and modify this parameter:
4952 *
4953 * struct sctp_assoc_value {
4954 * sctp_assoc_t assoc_id;
4955 * uint32_t assoc_value;
4956 * };
4957 *
4958 * assoc_id: This parameter is ignored for one-to-one style sockets.
4959 * For one-to-many style sockets this parameter indicates which
4960 * association the user is performing an action upon. Note that if
4961 * this field's value is zero then the endpoints default value is
4962 * changed (effecting future associations only).
4963 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004964 */
4965static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4966 char __user *optval, int __user *optlen)
4967{
Wei Yongjune89c2092008-12-25 16:54:58 -08004968 struct sctp_assoc_value params;
4969 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004970
Wei Yongjune89c2092008-12-25 16:54:58 -08004971 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00004972 pr_warn("Use of int in maxseg socket option deprecated\n");
4973 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08004974 params.assoc_id = 0;
4975 } else if (len >= sizeof(struct sctp_assoc_value)) {
4976 len = sizeof(struct sctp_assoc_value);
4977 if (copy_from_user(&params, optval, sizeof(params)))
4978 return -EFAULT;
4979 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07004980 return -EINVAL;
4981
Wei Yongjune89c2092008-12-25 16:54:58 -08004982 asoc = sctp_id2assoc(sk, params.assoc_id);
4983 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4984 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004985
Wei Yongjune89c2092008-12-25 16:54:58 -08004986 if (asoc)
4987 params.assoc_value = asoc->frag_point;
4988 else
4989 params.assoc_value = sctp_sk(sk)->user_frag;
4990
Linus Torvalds1da177e2005-04-16 15:20:36 -07004991 if (put_user(len, optlen))
4992 return -EFAULT;
Wei Yongjune89c2092008-12-25 16:54:58 -08004993 if (len == sizeof(int)) {
4994 if (copy_to_user(optval, &params.assoc_value, len))
4995 return -EFAULT;
4996 } else {
4997 if (copy_to_user(optval, &params, len))
4998 return -EFAULT;
4999 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005000
5001 return 0;
5002}
5003
Vlad Yasevichb6e13312007-04-20 12:23:15 -07005004/*
5005 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5006 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5007 */
5008static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5009 char __user *optval, int __user *optlen)
5010{
5011 int val;
5012
5013 if (len < sizeof(int))
5014 return -EINVAL;
5015
5016 len = sizeof(int);
5017
5018 val = sctp_sk(sk)->frag_interleave;
5019 if (put_user(len, optlen))
5020 return -EFAULT;
5021 if (copy_to_user(optval, &val, len))
5022 return -EFAULT;
5023
5024 return 0;
5025}
5026
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005027/*
5028 * 7.1.25. Set or Get the sctp partial delivery point
5029 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5030 */
5031static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5032 char __user *optval,
5033 int __user *optlen)
5034{
YOSHIFUJI Hideaki9cbcbf42007-07-19 10:44:50 +09005035 u32 val;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005036
5037 if (len < sizeof(u32))
5038 return -EINVAL;
5039
5040 len = sizeof(u32);
5041
5042 val = sctp_sk(sk)->pd_point;
5043 if (put_user(len, optlen))
5044 return -EFAULT;
5045 if (copy_to_user(optval, &val, len))
5046 return -EFAULT;
5047
Wei Yongjun7d743b72010-12-14 16:10:41 +00005048 return 0;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005049}
5050
Vlad Yasevich70331572007-03-23 11:34:36 -07005051/*
5052 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5053 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5054 */
5055static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5056 char __user *optval,
5057 int __user *optlen)
5058{
Neil Horman219b99a2008-03-05 13:44:46 -08005059 struct sctp_assoc_value params;
5060 struct sctp_sock *sp;
5061 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07005062
Neil Horman219b99a2008-03-05 13:44:46 -08005063 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00005064 pr_warn("Use of int in max_burst socket option deprecated\n");
5065 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08005066 params.assoc_id = 0;
Wei Yongjunc6db93a2009-03-02 09:46:12 +00005067 } else if (len >= sizeof(struct sctp_assoc_value)) {
5068 len = sizeof(struct sctp_assoc_value);
Neil Horman219b99a2008-03-05 13:44:46 -08005069 if (copy_from_user(&params, optval, len))
5070 return -EFAULT;
5071 } else
5072 return -EINVAL;
Vlad Yasevich70331572007-03-23 11:34:36 -07005073
Neil Horman219b99a2008-03-05 13:44:46 -08005074 sp = sctp_sk(sk);
Vlad Yasevich70331572007-03-23 11:34:36 -07005075
Neil Horman219b99a2008-03-05 13:44:46 -08005076 if (params.assoc_id != 0) {
5077 asoc = sctp_id2assoc(sk, params.assoc_id);
5078 if (!asoc)
5079 return -EINVAL;
5080 params.assoc_value = asoc->max_burst;
5081 } else
5082 params.assoc_value = sp->max_burst;
5083
5084 if (len == sizeof(int)) {
5085 if (copy_to_user(optval, &params.assoc_value, len))
5086 return -EFAULT;
5087 } else {
5088 if (copy_to_user(optval, &params, len))
5089 return -EFAULT;
5090 }
5091
5092 return 0;
5093
Vlad Yasevich70331572007-03-23 11:34:36 -07005094}
5095
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005096static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5097 char __user *optval, int __user *optlen)
5098{
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005099 struct sctp_hmacalgo __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005100 struct sctp_hmac_algo_param *hmacs;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005101 __u16 data_len = 0;
5102 u32 num_idents;
5103
5104 if (!sctp_auth_enable)
5105 return -EACCES;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005106
5107 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005108 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005109
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005110 if (len < sizeof(struct sctp_hmacalgo) + data_len)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005111 return -EINVAL;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005112
5113 len = sizeof(struct sctp_hmacalgo) + data_len;
5114 num_idents = data_len / sizeof(u16);
5115
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005116 if (put_user(len, optlen))
5117 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005118 if (put_user(num_idents, &p->shmac_num_idents))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005119 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005120 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5121 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005122 return 0;
5123}
5124
5125static int sctp_getsockopt_active_key(struct sock *sk, int len,
5126 char __user *optval, int __user *optlen)
5127{
5128 struct sctp_authkeyid val;
5129 struct sctp_association *asoc;
5130
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005131 if (!sctp_auth_enable)
5132 return -EACCES;
5133
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005134 if (len < sizeof(struct sctp_authkeyid))
5135 return -EINVAL;
5136 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5137 return -EFAULT;
5138
5139 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5140 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5141 return -EINVAL;
5142
5143 if (asoc)
5144 val.scact_keynumber = asoc->active_key_id;
5145 else
5146 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5147
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005148 len = sizeof(struct sctp_authkeyid);
5149 if (put_user(len, optlen))
5150 return -EFAULT;
5151 if (copy_to_user(optval, &val, len))
5152 return -EFAULT;
5153
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005154 return 0;
5155}
5156
5157static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5158 char __user *optval, int __user *optlen)
5159{
Al Viro411223c2007-10-14 19:21:20 +01005160 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005161 struct sctp_authchunks val;
5162 struct sctp_association *asoc;
5163 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005164 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005165 char __user *to;
5166
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005167 if (!sctp_auth_enable)
5168 return -EACCES;
5169
5170 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005171 return -EINVAL;
5172
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005173 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005174 return -EFAULT;
5175
Al Viro411223c2007-10-14 19:21:20 +01005176 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005177 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5178 if (!asoc)
5179 return -EINVAL;
5180
5181 ch = asoc->peer.peer_chunks;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005182 if (!ch)
5183 goto num;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005184
5185 /* See if the user provided enough room for all the data */
Vlad Yasevichb40db682008-02-27 14:40:37 -05005186 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5187 if (len < num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005188 return -EINVAL;
5189
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005190 if (copy_to_user(to, ch->chunks, num_chunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005191 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005192num:
5193 len = sizeof(struct sctp_authchunks) + num_chunks;
5194 if (put_user(len, optlen)) return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005195 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5196 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005197 return 0;
5198}
5199
5200static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5201 char __user *optval, int __user *optlen)
5202{
Al Viro411223c2007-10-14 19:21:20 +01005203 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005204 struct sctp_authchunks val;
5205 struct sctp_association *asoc;
5206 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005207 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005208 char __user *to;
5209
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005210 if (!sctp_auth_enable)
5211 return -EACCES;
5212
5213 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005214 return -EINVAL;
5215
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005216 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005217 return -EFAULT;
5218
Al Viro411223c2007-10-14 19:21:20 +01005219 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005220 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5221 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5222 return -EINVAL;
5223
5224 if (asoc)
5225 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5226 else
5227 ch = sctp_sk(sk)->ep->auth_chunk_list;
5228
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005229 if (!ch)
5230 goto num;
5231
Vlad Yasevichb40db682008-02-27 14:40:37 -05005232 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005233 if (len < sizeof(struct sctp_authchunks) + num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005234 return -EINVAL;
5235
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005236 if (copy_to_user(to, ch->chunks, num_chunks))
5237 return -EFAULT;
5238num:
5239 len = sizeof(struct sctp_authchunks) + num_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005240 if (put_user(len, optlen))
5241 return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005242 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5243 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005244
5245 return 0;
5246}
5247
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005248/*
5249 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5250 * This option gets the current number of associations that are attached
5251 * to a one-to-many style socket. The option value is an uint32_t.
5252 */
5253static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5254 char __user *optval, int __user *optlen)
5255{
5256 struct sctp_sock *sp = sctp_sk(sk);
5257 struct sctp_association *asoc;
5258 u32 val = 0;
5259
5260 if (sctp_style(sk, TCP))
5261 return -EOPNOTSUPP;
5262
5263 if (len < sizeof(u32))
5264 return -EINVAL;
5265
5266 len = sizeof(u32);
5267
5268 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5269 val++;
5270 }
5271
5272 if (put_user(len, optlen))
5273 return -EFAULT;
5274 if (copy_to_user(optval, &val, len))
5275 return -EFAULT;
5276
5277 return 0;
5278}
5279
Wei Yongjun209ba422011-04-17 17:27:08 +00005280/*
5281 * 8.2.6. Get the Current Identifiers of Associations
5282 * (SCTP_GET_ASSOC_ID_LIST)
5283 *
5284 * This option gets the current list of SCTP association identifiers of
5285 * the SCTP associations handled by a one-to-many style socket.
5286 */
5287static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5288 char __user *optval, int __user *optlen)
5289{
5290 struct sctp_sock *sp = sctp_sk(sk);
5291 struct sctp_association *asoc;
5292 struct sctp_assoc_ids *ids;
5293 u32 num = 0;
5294
5295 if (sctp_style(sk, TCP))
5296 return -EOPNOTSUPP;
5297
5298 if (len < sizeof(struct sctp_assoc_ids))
5299 return -EINVAL;
5300
5301 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5302 num++;
5303 }
5304
5305 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5306 return -EINVAL;
5307
5308 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5309
5310 ids = kmalloc(len, GFP_KERNEL);
5311 if (unlikely(!ids))
5312 return -ENOMEM;
5313
5314 ids->gaids_number_of_ids = num;
5315 num = 0;
5316 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5317 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5318 }
5319
5320 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5321 kfree(ids);
5322 return -EFAULT;
5323 }
5324
5325 kfree(ids);
5326 return 0;
5327}
5328
Linus Torvalds1da177e2005-04-16 15:20:36 -07005329SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5330 char __user *optval, int __user *optlen)
5331{
5332 int retval = 0;
5333 int len;
5334
Frank Filz3f7a87d2005-06-20 13:14:57 -07005335 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5336 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005337
5338 /* I can hardly begin to describe how wrong this is. This is
5339 * so broken as to be worse than useless. The API draft
5340 * REALLY is NOT helpful here... I am not convinced that the
5341 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5342 * are at all well-founded.
5343 */
5344 if (level != SOL_SCTP) {
5345 struct sctp_af *af = sctp_sk(sk)->pf->af;
5346
5347 retval = af->getsockopt(sk, level, optname, optval, optlen);
5348 return retval;
5349 }
5350
5351 if (get_user(len, optlen))
5352 return -EFAULT;
5353
5354 sctp_lock_sock(sk);
5355
5356 switch (optname) {
5357 case SCTP_STATUS:
5358 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5359 break;
5360 case SCTP_DISABLE_FRAGMENTS:
5361 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5362 optlen);
5363 break;
5364 case SCTP_EVENTS:
5365 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5366 break;
5367 case SCTP_AUTOCLOSE:
5368 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5369 break;
5370 case SCTP_SOCKOPT_PEELOFF:
5371 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5372 break;
5373 case SCTP_PEER_ADDR_PARAMS:
5374 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5375 optlen);
5376 break;
Shan Wei4580ccc2011-01-18 22:39:00 +00005377 case SCTP_DELAYED_SACK:
Wei Yongjund364d922008-05-09 15:13:26 -07005378 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
Frank Filz77086102005-12-22 11:37:30 -08005379 optlen);
5380 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005381 case SCTP_INITMSG:
5382 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5383 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005384 case SCTP_GET_PEER_ADDRS:
5385 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5386 optlen);
5387 break;
5388 case SCTP_GET_LOCAL_ADDRS:
5389 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5390 optlen);
5391 break;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04005392 case SCTP_SOCKOPT_CONNECTX3:
5393 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5394 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005395 case SCTP_DEFAULT_SEND_PARAM:
5396 retval = sctp_getsockopt_default_send_param(sk, len,
5397 optval, optlen);
5398 break;
5399 case SCTP_PRIMARY_ADDR:
5400 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5401 break;
5402 case SCTP_NODELAY:
5403 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5404 break;
5405 case SCTP_RTOINFO:
5406 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5407 break;
5408 case SCTP_ASSOCINFO:
5409 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5410 break;
5411 case SCTP_I_WANT_MAPPED_V4_ADDR:
5412 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5413 break;
5414 case SCTP_MAXSEG:
5415 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5416 break;
5417 case SCTP_GET_PEER_ADDR_INFO:
5418 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5419 optlen);
5420 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08005421 case SCTP_ADAPTATION_LAYER:
5422 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07005423 optlen);
5424 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005425 case SCTP_CONTEXT:
5426 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5427 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07005428 case SCTP_FRAGMENT_INTERLEAVE:
5429 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5430 optlen);
5431 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005432 case SCTP_PARTIAL_DELIVERY_POINT:
5433 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5434 optlen);
5435 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07005436 case SCTP_MAX_BURST:
5437 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5438 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005439 case SCTP_AUTH_KEY:
5440 case SCTP_AUTH_CHUNK:
5441 case SCTP_AUTH_DELETE_KEY:
5442 retval = -EOPNOTSUPP;
5443 break;
5444 case SCTP_HMAC_IDENT:
5445 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5446 break;
5447 case SCTP_AUTH_ACTIVE_KEY:
5448 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5449 break;
5450 case SCTP_PEER_AUTH_CHUNKS:
5451 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5452 optlen);
5453 break;
5454 case SCTP_LOCAL_AUTH_CHUNKS:
5455 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5456 optlen);
5457 break;
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005458 case SCTP_GET_ASSOC_NUMBER:
5459 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5460 break;
Wei Yongjun209ba422011-04-17 17:27:08 +00005461 case SCTP_GET_ASSOC_ID_LIST:
5462 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5463 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005464 default:
5465 retval = -ENOPROTOOPT;
5466 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005467 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005468
5469 sctp_release_sock(sk);
5470 return retval;
5471}
5472
5473static void sctp_hash(struct sock *sk)
5474{
5475 /* STUB */
5476}
5477
5478static void sctp_unhash(struct sock *sk)
5479{
5480 /* STUB */
5481}
5482
5483/* Check if port is acceptable. Possibly find first available port.
5484 *
5485 * The port hash table (contained in the 'global' SCTP protocol storage
5486 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5487 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5488 * list (the list number is the port number hashed out, so as you
5489 * would expect from a hash function, all the ports in a given list have
5490 * such a number that hashes out to the same list number; you were
5491 * expecting that, right?); so each list has a set of ports, with a
5492 * link to the socket (struct sock) that uses it, the port number and
5493 * a fastreuse flag (FIXME: NPI ipg).
5494 */
5495static struct sctp_bind_bucket *sctp_bucket_create(
5496 struct sctp_bind_hashbucket *head, unsigned short snum);
5497
5498static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5499{
5500 struct sctp_bind_hashbucket *head; /* hash list */
5501 struct sctp_bind_bucket *pp; /* hash list port iterator */
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005502 struct hlist_node *node;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005503 unsigned short snum;
5504 int ret;
5505
Al Viro04afd8b2006-11-20 17:02:01 -08005506 snum = ntohs(addr->v4.sin_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005507
5508 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5509 sctp_local_bh_disable();
5510
5511 if (snum == 0) {
Stephen Hemminger06393002007-10-10 17:30:18 -07005512 /* Search for an available port. */
Stephen Hemminger227b60f2007-10-10 17:30:46 -07005513 int low, high, remaining, index;
5514 unsigned int rover;
5515
5516 inet_get_local_port_range(&low, &high);
5517 remaining = (high - low) + 1;
5518 rover = net_random() % remaining + low;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005519
Linus Torvalds1da177e2005-04-16 15:20:36 -07005520 do {
5521 rover++;
5522 if ((rover < low) || (rover > high))
5523 rover = low;
Amerigo Wange3826f12010-05-05 00:27:06 +00005524 if (inet_is_reserved_local_port(rover))
5525 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005526 index = sctp_phashfn(rover);
5527 head = &sctp_port_hashtable[index];
5528 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005529 sctp_for_each_hentry(pp, node, &head->chain)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005530 if (pp->port == rover)
5531 goto next;
5532 break;
5533 next:
5534 sctp_spin_unlock(&head->lock);
5535 } while (--remaining > 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005536
5537 /* Exhausted local port range during search? */
5538 ret = 1;
5539 if (remaining <= 0)
5540 goto fail;
5541
5542 /* OK, here is the one we will use. HEAD (the port
5543 * hash table list entry) is non-NULL and we hold it's
5544 * mutex.
5545 */
5546 snum = rover;
5547 } else {
5548 /* We are given an specific port number; we verify
5549 * that it is not being used. If it is used, we will
5550 * exahust the search in the hash list corresponding
5551 * to the port number (snum) - we detect that with the
5552 * port iterator, pp being NULL.
5553 */
5554 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5555 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005556 sctp_for_each_hentry(pp, node, &head->chain) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005557 if (pp->port == snum)
5558 goto pp_found;
5559 }
5560 }
5561 pp = NULL;
5562 goto pp_not_found;
5563pp_found:
5564 if (!hlist_empty(&pp->owner)) {
5565 /* We had a port hash table hit - there is an
5566 * available port (pp != NULL) and it is being
5567 * used by other socket (pp->owner not empty); that other
5568 * socket is going to be sk2.
5569 */
5570 int reuse = sk->sk_reuse;
5571 struct sock *sk2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005572
5573 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005574 if (pp->fastreuse && sk->sk_reuse &&
5575 sk->sk_state != SCTP_SS_LISTENING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005576 goto success;
5577
5578 /* Run through the list of sockets bound to the port
5579 * (pp->port) [via the pointers bind_next and
5580 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5581 * we get the endpoint they describe and run through
5582 * the endpoint's list of IP (v4 or v6) addresses,
5583 * comparing each of the addresses with the address of
5584 * the socket sk. If we find a match, then that means
5585 * that this port/socket (sk) combination are already
5586 * in an endpoint.
5587 */
5588 sk_for_each_bound(sk2, node, &pp->owner) {
5589 struct sctp_endpoint *ep2;
5590 ep2 = sctp_sk(sk2)->ep;
5591
Vlad Yasevich4e540642008-07-18 23:06:32 -07005592 if (sk == sk2 ||
5593 (reuse && sk2->sk_reuse &&
5594 sk2->sk_state != SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005595 continue;
5596
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07005597 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5598 sctp_sk(sk2), sctp_sk(sk))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005599 ret = (long)sk2;
5600 goto fail_unlock;
5601 }
5602 }
5603 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5604 }
5605pp_not_found:
5606 /* If there was a hash table miss, create a new port. */
5607 ret = 1;
5608 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5609 goto fail_unlock;
5610
5611 /* In either case (hit or miss), make sure fastreuse is 1 only
5612 * if sk->sk_reuse is too (that is, if the caller requested
5613 * SO_REUSEADDR on this socket -sk-).
5614 */
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005615 if (hlist_empty(&pp->owner)) {
5616 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5617 pp->fastreuse = 1;
5618 else
5619 pp->fastreuse = 0;
5620 } else if (pp->fastreuse &&
5621 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005622 pp->fastreuse = 0;
5623
5624 /* We are set, so fill up all the data in the hash table
5625 * entry, tie the socket list information with the rest of the
5626 * sockets FIXME: Blurry, NPI (ipg).
5627 */
5628success:
Linus Torvalds1da177e2005-04-16 15:20:36 -07005629 if (!sctp_sk(sk)->bind_hash) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005630 inet_sk(sk)->inet_num = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005631 sk_add_bind_node(sk, &pp->owner);
5632 sctp_sk(sk)->bind_hash = pp;
5633 }
5634 ret = 0;
5635
5636fail_unlock:
5637 sctp_spin_unlock(&head->lock);
5638
5639fail:
5640 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07005641 return ret;
5642}
5643
5644/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5645 * port is requested.
5646 */
5647static int sctp_get_port(struct sock *sk, unsigned short snum)
5648{
5649 long ret;
5650 union sctp_addr addr;
5651 struct sctp_af *af = sctp_sk(sk)->pf->af;
5652
5653 /* Set up a dummy address struct from the sk. */
5654 af->from_sk(&addr, sk);
5655 addr.v4.sin_port = htons(snum);
5656
5657 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5658 ret = sctp_get_port_local(sk, &addr);
5659
Eric Dumazeta02cec22010-09-22 20:43:57 +00005660 return ret ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005661}
5662
5663/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07005664 * Move a socket to LISTENING state.
5665 */
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005666SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005667{
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005668 struct sctp_sock *sp = sctp_sk(sk);
5669 struct sctp_endpoint *ep = sp->ep;
Herbert Xu1b489e12006-08-20 15:07:14 +10005670 struct crypto_hash *tfm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005671
5672 /* Allocate HMAC for generating cookie. */
Vlad Yasevich23b29ed2008-07-18 23:06:07 -07005673 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
Herbert Xu1b489e12006-08-20 15:07:14 +10005674 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
Vlad Yasevich8dc49842007-05-09 13:50:20 -07005675 if (IS_ERR(tfm)) {
5676 if (net_ratelimit()) {
Joe Perches145ce502010-08-24 13:21:08 +00005677 pr_info("failed to load transform for %s: %ld\n",
Vlad Yasevich8dc49842007-05-09 13:50:20 -07005678 sctp_hmac_alg, PTR_ERR(tfm));
5679 }
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005680 return -ENOSYS;
5681 }
5682 sctp_sk(sk)->hmac = tfm;
5683 }
5684
5685 /*
5686 * If a bind() or sctp_bindx() is not called prior to a listen()
5687 * call that allows new associations to be accepted, the system
5688 * picks an ephemeral port and will choose an address set equivalent
5689 * to binding with a wildcard address.
5690 *
5691 * This is not currently spelled out in the SCTP sockets
5692 * extensions draft, but follows the practice as seen in TCP
5693 * sockets.
5694 *
5695 */
5696 sk->sk_state = SCTP_SS_LISTENING;
5697 if (!ep->base.bind_addr.port) {
5698 if (sctp_autobind(sk))
5699 return -EAGAIN;
5700 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005701 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005702 sk->sk_state = SCTP_SS_CLOSED;
5703 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005704 }
5705 }
5706
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005707 sk->sk_max_ack_backlog = backlog;
5708 sctp_hash_endpoint(ep);
5709 return 0;
5710}
5711
5712/*
5713 * 4.1.3 / 5.1.3 listen()
5714 *
5715 * By default, new associations are not accepted for UDP style sockets.
5716 * An application uses listen() to mark a socket as being able to
5717 * accept new associations.
5718 *
5719 * On TCP style sockets, applications use listen() to ready the SCTP
5720 * endpoint for accepting inbound associations.
5721 *
5722 * On both types of endpoints a backlog of '0' disables listening.
5723 *
5724 * Move a socket to LISTENING state.
5725 */
5726int sctp_inet_listen(struct socket *sock, int backlog)
5727{
5728 struct sock *sk = sock->sk;
5729 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5730 int err = -EINVAL;
5731
5732 if (unlikely(backlog < 0))
5733 return err;
5734
5735 sctp_lock_sock(sk);
5736
5737 /* Peeled-off sockets are not allowed to listen(). */
5738 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5739 goto out;
5740
5741 if (sock->state != SS_UNCONNECTED)
5742 goto out;
5743
5744 /* If backlog is zero, disable listening. */
5745 if (!backlog) {
5746 if (sctp_sstate(sk, CLOSED))
5747 goto out;
5748
5749 err = 0;
5750 sctp_unhash_endpoint(ep);
5751 sk->sk_state = SCTP_SS_CLOSED;
5752 if (sk->sk_reuse)
5753 sctp_sk(sk)->bind_hash->fastreuse = 1;
5754 goto out;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005755 }
5756
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005757 /* If we are already listening, just update the backlog */
5758 if (sctp_sstate(sk, LISTENING))
5759 sk->sk_max_ack_backlog = backlog;
5760 else {
5761 err = sctp_listen_start(sk, backlog);
5762 if (err)
5763 goto out;
5764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005765
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005766 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005767out:
5768 sctp_release_sock(sk);
5769 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005770}
5771
5772/*
5773 * This function is done by modeling the current datagram_poll() and the
5774 * tcp_poll(). Note that, based on these implementations, we don't
5775 * lock the socket in this function, even though it seems that,
5776 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08005777 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07005778 * in this place. We assume that we don't need locks either until proven
5779 * otherwise.
5780 *
5781 * Another thing to note is that we include the Async I/O support
5782 * here, again, by modeling the current TCP/UDP code. We don't have
5783 * a good way to test with it yet.
5784 */
5785unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5786{
5787 struct sock *sk = sock->sk;
5788 struct sctp_sock *sp = sctp_sk(sk);
5789 unsigned int mask;
5790
Eric Dumazetaa395142010-04-20 13:03:51 +00005791 poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005792
5793 /* A TCP-style listening socket becomes readable when the accept queue
5794 * is not empty.
5795 */
5796 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5797 return (!list_empty(&sp->ep->asocs)) ?
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09005798 (POLLIN | POLLRDNORM) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005799
5800 mask = 0;
5801
5802 /* Is there any exceptional events? */
5803 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5804 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08005805 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00005806 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005807 if (sk->sk_shutdown == SHUTDOWN_MASK)
5808 mask |= POLLHUP;
5809
5810 /* Is it readable? Reconsider this code with TCP-style support. */
Eric Dumazetdb409802010-09-06 11:13:50 +00005811 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005812 mask |= POLLIN | POLLRDNORM;
5813
5814 /* The association is either gone or not ready. */
5815 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5816 return mask;
5817
5818 /* Is it writable? */
5819 if (sctp_writeable(sk)) {
5820 mask |= POLLOUT | POLLWRNORM;
5821 } else {
5822 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5823 /*
5824 * Since the socket is not locked, the buffer
5825 * might be made available after the writeable check and
5826 * before the bit is set. This could cause a lost I/O
5827 * signal. tcp_poll() has a race breaker for this race
5828 * condition. Based on their implementation, we put
5829 * in the following code to cover it as well.
5830 */
5831 if (sctp_writeable(sk))
5832 mask |= POLLOUT | POLLWRNORM;
5833 }
5834 return mask;
5835}
5836
5837/********************************************************************
5838 * 2nd Level Abstractions
5839 ********************************************************************/
5840
5841static struct sctp_bind_bucket *sctp_bucket_create(
5842 struct sctp_bind_hashbucket *head, unsigned short snum)
5843{
5844 struct sctp_bind_bucket *pp;
5845
Christoph Lameter54e6ecb2006-12-06 20:33:16 -08005846 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005847 if (pp) {
Li Zefan935a7f62008-04-10 01:58:06 -07005848 SCTP_DBG_OBJCNT_INC(bind_bucket);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005849 pp->port = snum;
5850 pp->fastreuse = 0;
5851 INIT_HLIST_HEAD(&pp->owner);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005852 hlist_add_head(&pp->node, &head->chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005853 }
5854 return pp;
5855}
5856
5857/* Caller must hold hashbucket lock for this tb with local BH disabled */
5858static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5859{
Sridhar Samudrala37fa6872006-07-21 14:45:47 -07005860 if (pp && hlist_empty(&pp->owner)) {
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005861 __hlist_del(&pp->node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005862 kmem_cache_free(sctp_bucket_cachep, pp);
5863 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5864 }
5865}
5866
5867/* Release this socket's reference to a local port. */
5868static inline void __sctp_put_port(struct sock *sk)
5869{
5870 struct sctp_bind_hashbucket *head =
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005871 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
Linus Torvalds1da177e2005-04-16 15:20:36 -07005872 struct sctp_bind_bucket *pp;
5873
5874 sctp_spin_lock(&head->lock);
5875 pp = sctp_sk(sk)->bind_hash;
5876 __sk_del_bind_node(sk);
5877 sctp_sk(sk)->bind_hash = NULL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005878 inet_sk(sk)->inet_num = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005879 sctp_bucket_destroy(pp);
5880 sctp_spin_unlock(&head->lock);
5881}
5882
5883void sctp_put_port(struct sock *sk)
5884{
5885 sctp_local_bh_disable();
5886 __sctp_put_port(sk);
5887 sctp_local_bh_enable();
5888}
5889
5890/*
5891 * The system picks an ephemeral port and choose an address set equivalent
5892 * to binding with a wildcard address.
5893 * One of those addresses will be the primary address for the association.
5894 * This automatically enables the multihoming capability of SCTP.
5895 */
5896static int sctp_autobind(struct sock *sk)
5897{
5898 union sctp_addr autoaddr;
5899 struct sctp_af *af;
Al Viro6fbfa9f2006-11-20 17:24:53 -08005900 __be16 port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005901
5902 /* Initialize a local sockaddr structure to INADDR_ANY. */
5903 af = sctp_sk(sk)->pf->af;
5904
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005905 port = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005906 af->inaddr_any(&autoaddr, port);
5907
5908 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5909}
5910
5911/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5912 *
5913 * From RFC 2292
5914 * 4.2 The cmsghdr Structure *
5915 *
5916 * When ancillary data is sent or received, any number of ancillary data
5917 * objects can be specified by the msg_control and msg_controllen members of
5918 * the msghdr structure, because each object is preceded by
5919 * a cmsghdr structure defining the object's length (the cmsg_len member).
5920 * Historically Berkeley-derived implementations have passed only one object
5921 * at a time, but this API allows multiple objects to be
5922 * passed in a single call to sendmsg() or recvmsg(). The following example
5923 * shows two ancillary data objects in a control buffer.
5924 *
5925 * |<--------------------------- msg_controllen -------------------------->|
5926 * | |
5927 *
5928 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5929 *
5930 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5931 * | | |
5932 *
5933 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5934 *
5935 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5936 * | | | | |
5937 *
5938 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5939 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5940 *
5941 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5942 *
5943 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5944 * ^
5945 * |
5946 *
5947 * msg_control
5948 * points here
5949 */
5950SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5951 sctp_cmsgs_t *cmsgs)
5952{
5953 struct cmsghdr *cmsg;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07005954 struct msghdr *my_msg = (struct msghdr *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005955
5956 for (cmsg = CMSG_FIRSTHDR(msg);
5957 cmsg != NULL;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07005958 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5959 if (!CMSG_OK(my_msg, cmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005960 return -EINVAL;
5961
5962 /* Should we parse this header or ignore? */
5963 if (cmsg->cmsg_level != IPPROTO_SCTP)
5964 continue;
5965
5966 /* Strictly check lengths following example in SCM code. */
5967 switch (cmsg->cmsg_type) {
5968 case SCTP_INIT:
5969 /* SCTP Socket API Extension
5970 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5971 *
5972 * This cmsghdr structure provides information for
5973 * initializing new SCTP associations with sendmsg().
5974 * The SCTP_INITMSG socket option uses this same data
5975 * structure. This structure is not used for
5976 * recvmsg().
5977 *
5978 * cmsg_level cmsg_type cmsg_data[]
5979 * ------------ ------------ ----------------------
5980 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5981 */
5982 if (cmsg->cmsg_len !=
5983 CMSG_LEN(sizeof(struct sctp_initmsg)))
5984 return -EINVAL;
5985 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5986 break;
5987
5988 case SCTP_SNDRCV:
5989 /* SCTP Socket API Extension
5990 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5991 *
5992 * This cmsghdr structure specifies SCTP options for
5993 * sendmsg() and describes SCTP header information
5994 * about a received message through recvmsg().
5995 *
5996 * cmsg_level cmsg_type cmsg_data[]
5997 * ------------ ------------ ----------------------
5998 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5999 */
6000 if (cmsg->cmsg_len !=
6001 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6002 return -EINVAL;
6003
6004 cmsgs->info =
6005 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6006
6007 /* Minimally, validate the sinfo_flags. */
6008 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07006009 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6010 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07006011 return -EINVAL;
6012 break;
6013
6014 default:
6015 return -EINVAL;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07006016 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07006017 }
6018 return 0;
6019}
6020
6021/*
6022 * Wait for a packet..
6023 * Note: This function is the same function as in core/datagram.c
6024 * with a few modifications to make lksctp work.
6025 */
6026static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6027{
6028 int error;
6029 DEFINE_WAIT(wait);
6030
Eric Dumazetaa395142010-04-20 13:03:51 +00006031 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006032
6033 /* Socket errors? */
6034 error = sock_error(sk);
6035 if (error)
6036 goto out;
6037
6038 if (!skb_queue_empty(&sk->sk_receive_queue))
6039 goto ready;
6040
6041 /* Socket shut down? */
6042 if (sk->sk_shutdown & RCV_SHUTDOWN)
6043 goto out;
6044
6045 /* Sequenced packets can come disconnected. If so we report the
6046 * problem.
6047 */
6048 error = -ENOTCONN;
6049
6050 /* Is there a good reason to think that we may receive some data? */
6051 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6052 goto out;
6053
6054 /* Handle signals. */
6055 if (signal_pending(current))
6056 goto interrupted;
6057
6058 /* Let another process have a go. Since we are going to sleep
6059 * anyway. Note: This may cause odd behaviors if the message
6060 * does not fit in the user's buffer, but this seems to be the
6061 * only way to honor MSG_DONTWAIT realistically.
6062 */
6063 sctp_release_sock(sk);
6064 *timeo_p = schedule_timeout(*timeo_p);
6065 sctp_lock_sock(sk);
6066
6067ready:
Eric Dumazetaa395142010-04-20 13:03:51 +00006068 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006069 return 0;
6070
6071interrupted:
6072 error = sock_intr_errno(*timeo_p);
6073
6074out:
Eric Dumazetaa395142010-04-20 13:03:51 +00006075 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006076 *err = error;
6077 return error;
6078}
6079
6080/* Receive a datagram.
6081 * Note: This is pretty much the same routine as in core/datagram.c
6082 * with a few changes to make lksctp work.
6083 */
6084static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6085 int noblock, int *err)
6086{
6087 int error;
6088 struct sk_buff *skb;
6089 long timeo;
6090
Linus Torvalds1da177e2005-04-16 15:20:36 -07006091 timeo = sock_rcvtimeo(sk, noblock);
6092
6093 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6094 timeo, MAX_SCHEDULE_TIMEOUT);
6095
6096 do {
6097 /* Again only user level code calls this function,
6098 * so nothing interrupt level
6099 * will suddenly eat the receive_queue.
6100 *
6101 * Look at current nfs client by the way...
David Shwatrz8917a3c2010-12-02 09:01:55 +00006102 * However, this function was correct in any case. 8)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006103 */
6104 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07006105 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006106 skb = skb_peek(&sk->sk_receive_queue);
6107 if (skb)
6108 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07006109 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006110 } else {
6111 skb = skb_dequeue(&sk->sk_receive_queue);
6112 }
6113
6114 if (skb)
6115 return skb;
6116
Neil Horman6736dc32005-12-02 20:30:06 -08006117 /* Caller is allowed not to check sk->sk_err before calling. */
6118 error = sock_error(sk);
6119 if (error)
6120 goto no_packet;
6121
Linus Torvalds1da177e2005-04-16 15:20:36 -07006122 if (sk->sk_shutdown & RCV_SHUTDOWN)
6123 break;
6124
6125 /* User doesn't want to wait. */
6126 error = -EAGAIN;
6127 if (!timeo)
6128 goto no_packet;
6129 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6130
6131 return NULL;
6132
6133no_packet:
6134 *err = error;
6135 return NULL;
6136}
6137
6138/* If sndbuf has changed, wake up per association sndbuf waiters. */
6139static void __sctp_write_space(struct sctp_association *asoc)
6140{
6141 struct sock *sk = asoc->base.sk;
6142 struct socket *sock = sk->sk_socket;
6143
6144 if ((sctp_wspace(asoc) > 0) && sock) {
6145 if (waitqueue_active(&asoc->wait))
6146 wake_up_interruptible(&asoc->wait);
6147
6148 if (sctp_writeable(sk)) {
Eric Dumazeteaefd112011-02-18 03:26:36 +00006149 wait_queue_head_t *wq = sk_sleep(sk);
6150
6151 if (wq && waitqueue_active(wq))
6152 wake_up_interruptible(wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006153
6154 /* Note that we try to include the Async I/O support
6155 * here by modeling from the current TCP/UDP code.
6156 * We have not tested with it yet.
6157 */
Eric Dumazeteaefd112011-02-18 03:26:36 +00006158 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08006159 sock_wake_async(sock,
6160 SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006161 }
6162 }
6163}
6164
6165/* Do accounting for the sndbuf space.
6166 * Decrement the used sndbuf space of the corresponding association by the
6167 * data size which was just transmitted(freed).
6168 */
6169static void sctp_wfree(struct sk_buff *skb)
6170{
6171 struct sctp_association *asoc;
6172 struct sctp_chunk *chunk;
6173 struct sock *sk;
6174
6175 /* Get the saved chunk pointer. */
6176 chunk = *((struct sctp_chunk **)(skb->cb));
6177 asoc = chunk->asoc;
6178 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07006179 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6180 sizeof(struct sk_buff) +
6181 sizeof(struct sctp_chunk);
6182
Neil Horman4eb701d2005-04-28 12:02:04 -07006183 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6184
Neil Horman4d93df02007-08-15 16:07:44 -07006185 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006186 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
Neil Horman4d93df02007-08-15 16:07:44 -07006187 */
6188 sk->sk_wmem_queued -= skb->truesize;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006189 sk_mem_uncharge(sk, skb->truesize);
Neil Horman4d93df02007-08-15 16:07:44 -07006190
Neil Horman4eb701d2005-04-28 12:02:04 -07006191 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006192 __sctp_write_space(asoc);
6193
6194 sctp_association_put(asoc);
6195}
6196
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006197/* Do accounting for the receive space on the socket.
6198 * Accounting for the association is done in ulpevent.c
6199 * We set this as a destructor for the cloned data skbs so that
6200 * accounting is done at the correct time.
6201 */
6202void sctp_sock_rfree(struct sk_buff *skb)
6203{
6204 struct sock *sk = skb->sk;
6205 struct sctp_ulpevent *event = sctp_skb2event(skb);
6206
6207 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
Neil Horman4d93df02007-08-15 16:07:44 -07006208
6209 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006210 * Mimic the behavior of sock_rfree
Neil Horman4d93df02007-08-15 16:07:44 -07006211 */
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006212 sk_mem_uncharge(sk, event->rmem_len);
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006213}
6214
6215
Linus Torvalds1da177e2005-04-16 15:20:36 -07006216/* Helper function to wait for space in the sndbuf. */
6217static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6218 size_t msg_len)
6219{
6220 struct sock *sk = asoc->base.sk;
6221 int err = 0;
6222 long current_timeo = *timeo_p;
6223 DEFINE_WAIT(wait);
6224
6225 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09006226 asoc, (long)(*timeo_p), msg_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006227
6228 /* Increment the association's refcnt. */
6229 sctp_association_hold(asoc);
6230
6231 /* Wait on the association specific sndbuf space. */
6232 for (;;) {
6233 prepare_to_wait_exclusive(&asoc->wait, &wait,
6234 TASK_INTERRUPTIBLE);
6235 if (!*timeo_p)
6236 goto do_nonblock;
6237 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6238 asoc->base.dead)
6239 goto do_error;
6240 if (signal_pending(current))
6241 goto do_interrupted;
6242 if (msg_len <= sctp_wspace(asoc))
6243 break;
6244
6245 /* Let another process have a go. Since we are going
6246 * to sleep anyway.
6247 */
6248 sctp_release_sock(sk);
6249 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006250 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006251 sctp_lock_sock(sk);
6252
6253 *timeo_p = current_timeo;
6254 }
6255
6256out:
6257 finish_wait(&asoc->wait, &wait);
6258
6259 /* Release the association's refcnt. */
6260 sctp_association_put(asoc);
6261
6262 return err;
6263
6264do_error:
6265 err = -EPIPE;
6266 goto out;
6267
6268do_interrupted:
6269 err = sock_intr_errno(*timeo_p);
6270 goto out;
6271
6272do_nonblock:
6273 err = -EAGAIN;
6274 goto out;
6275}
6276
Wei Yongjun561b1732010-04-28 08:47:18 +00006277void sctp_data_ready(struct sock *sk, int len)
6278{
David S. Miller7ef52732010-05-02 21:43:40 -07006279 struct socket_wq *wq;
6280
6281 rcu_read_lock();
6282 wq = rcu_dereference(sk->sk_wq);
6283 if (wq_has_sleeper(wq))
6284 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
Wei Yongjun561b1732010-04-28 08:47:18 +00006285 POLLRDNORM | POLLRDBAND);
6286 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
David S. Miller7ef52732010-05-02 21:43:40 -07006287 rcu_read_unlock();
Wei Yongjun561b1732010-04-28 08:47:18 +00006288}
6289
Linus Torvalds1da177e2005-04-16 15:20:36 -07006290/* If socket sndbuf has changed, wake up all per association waiters. */
6291void sctp_write_space(struct sock *sk)
6292{
6293 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006294
6295 /* Wake up the tasks in each wait queue. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07006296 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006297 __sctp_write_space(asoc);
6298 }
6299}
6300
6301/* Is there any sndbuf space available on the socket?
6302 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08006303 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07006304 * associations on the same socket. For a UDP-style socket with
6305 * multiple associations, it is possible for it to be "unwriteable"
6306 * prematurely. I assume that this is acceptable because
6307 * a premature "unwriteable" is better than an accidental "writeable" which
6308 * would cause an unwanted block under certain circumstances. For the 1-1
6309 * UDP-style sockets or TCP-style sockets, this code should work.
6310 * - Daisy
6311 */
6312static int sctp_writeable(struct sock *sk)
6313{
6314 int amt = 0;
6315
Eric Dumazet31e6d362009-06-17 19:05:41 -07006316 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006317 if (amt < 0)
6318 amt = 0;
6319 return amt;
6320}
6321
6322/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6323 * returns immediately with EINPROGRESS.
6324 */
6325static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6326{
6327 struct sock *sk = asoc->base.sk;
6328 int err = 0;
6329 long current_timeo = *timeo_p;
6330 DEFINE_WAIT(wait);
6331
Harvey Harrison0dc47872008-03-05 20:47:47 -08006332 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006333 (long)(*timeo_p));
6334
6335 /* Increment the association's refcnt. */
6336 sctp_association_hold(asoc);
6337
6338 for (;;) {
6339 prepare_to_wait_exclusive(&asoc->wait, &wait,
6340 TASK_INTERRUPTIBLE);
6341 if (!*timeo_p)
6342 goto do_nonblock;
6343 if (sk->sk_shutdown & RCV_SHUTDOWN)
6344 break;
6345 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6346 asoc->base.dead)
6347 goto do_error;
6348 if (signal_pending(current))
6349 goto do_interrupted;
6350
6351 if (sctp_state(asoc, ESTABLISHED))
6352 break;
6353
6354 /* Let another process have a go. Since we are going
6355 * to sleep anyway.
6356 */
6357 sctp_release_sock(sk);
6358 current_timeo = schedule_timeout(current_timeo);
6359 sctp_lock_sock(sk);
6360
6361 *timeo_p = current_timeo;
6362 }
6363
6364out:
6365 finish_wait(&asoc->wait, &wait);
6366
6367 /* Release the association's refcnt. */
6368 sctp_association_put(asoc);
6369
6370 return err;
6371
6372do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08006373 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006374 err = -ETIMEDOUT;
6375 else
6376 err = -ECONNREFUSED;
6377 goto out;
6378
6379do_interrupted:
6380 err = sock_intr_errno(*timeo_p);
6381 goto out;
6382
6383do_nonblock:
6384 err = -EINPROGRESS;
6385 goto out;
6386}
6387
6388static int sctp_wait_for_accept(struct sock *sk, long timeo)
6389{
6390 struct sctp_endpoint *ep;
6391 int err = 0;
6392 DEFINE_WAIT(wait);
6393
6394 ep = sctp_sk(sk)->ep;
6395
6396
6397 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00006398 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006399 TASK_INTERRUPTIBLE);
6400
6401 if (list_empty(&ep->asocs)) {
6402 sctp_release_sock(sk);
6403 timeo = schedule_timeout(timeo);
6404 sctp_lock_sock(sk);
6405 }
6406
6407 err = -EINVAL;
6408 if (!sctp_sstate(sk, LISTENING))
6409 break;
6410
6411 err = 0;
6412 if (!list_empty(&ep->asocs))
6413 break;
6414
6415 err = sock_intr_errno(timeo);
6416 if (signal_pending(current))
6417 break;
6418
6419 err = -EAGAIN;
6420 if (!timeo)
6421 break;
6422 }
6423
Eric Dumazetaa395142010-04-20 13:03:51 +00006424 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006425
6426 return err;
6427}
6428
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +02006429static void sctp_wait_for_close(struct sock *sk, long timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006430{
6431 DEFINE_WAIT(wait);
6432
6433 do {
Eric Dumazetaa395142010-04-20 13:03:51 +00006434 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006435 if (list_empty(&sctp_sk(sk)->ep->asocs))
6436 break;
6437 sctp_release_sock(sk);
6438 timeout = schedule_timeout(timeout);
6439 sctp_lock_sock(sk);
6440 } while (!signal_pending(current) && timeout);
6441
Eric Dumazetaa395142010-04-20 13:03:51 +00006442 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006443}
6444
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006445static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6446{
6447 struct sk_buff *frag;
6448
6449 if (!skb->data_len)
6450 goto done;
6451
6452 /* Don't forget the fragments. */
David S. Miller1b003be2009-06-09 00:22:35 -07006453 skb_walk_frags(skb, frag)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006454 sctp_skb_set_owner_r_frag(frag, sk);
6455
6456done:
6457 sctp_skb_set_owner_r(skb, sk);
6458}
6459
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006460void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6461 struct sctp_association *asoc)
6462{
6463 struct inet_sock *inet = inet_sk(sk);
Julia Lawall09cb47a2010-01-21 02:43:20 -08006464 struct inet_sock *newinet;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006465
6466 newsk->sk_type = sk->sk_type;
6467 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6468 newsk->sk_flags = sk->sk_flags;
6469 newsk->sk_no_check = sk->sk_no_check;
6470 newsk->sk_reuse = sk->sk_reuse;
6471
6472 newsk->sk_shutdown = sk->sk_shutdown;
6473 newsk->sk_destruct = inet_sock_destruct;
6474 newsk->sk_family = sk->sk_family;
6475 newsk->sk_protocol = IPPROTO_SCTP;
6476 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6477 newsk->sk_sndbuf = sk->sk_sndbuf;
6478 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6479 newsk->sk_lingertime = sk->sk_lingertime;
6480 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6481 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6482
6483 newinet = inet_sk(newsk);
6484
6485 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6486 * getsockname() and getpeername()
6487 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006488 newinet->inet_sport = inet->inet_sport;
6489 newinet->inet_saddr = inet->inet_saddr;
6490 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6491 newinet->inet_dport = htons(asoc->peer.port);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006492 newinet->pmtudisc = inet->pmtudisc;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006493 newinet->inet_id = asoc->next_tsn ^ jiffies;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006494
6495 newinet->uc_ttl = inet->uc_ttl;
6496 newinet->mc_loop = 1;
6497 newinet->mc_ttl = 1;
6498 newinet->mc_index = 0;
6499 newinet->mc_list = NULL;
6500}
6501
Linus Torvalds1da177e2005-04-16 15:20:36 -07006502/* Populate the fields of the newsk from the oldsk and migrate the assoc
6503 * and its messages to the newsk.
6504 */
6505static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6506 struct sctp_association *assoc,
6507 sctp_socket_type_t type)
6508{
6509 struct sctp_sock *oldsp = sctp_sk(oldsk);
6510 struct sctp_sock *newsp = sctp_sk(newsk);
6511 struct sctp_bind_bucket *pp; /* hash list port iterator */
6512 struct sctp_endpoint *newep = newsp->ep;
6513 struct sk_buff *skb, *tmp;
6514 struct sctp_ulpevent *event;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006515 struct sctp_bind_hashbucket *head;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006516
6517 /* Migrate socket buffer sizes and all the socket level options to the
6518 * new socket.
6519 */
6520 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6521 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6522 /* Brute force copy old sctp opt. */
6523 inet_sk_copy_descendant(newsk, oldsk);
6524
6525 /* Restore the ep value that was overwritten with the above structure
6526 * copy.
6527 */
6528 newsp->ep = newep;
6529 newsp->hmac = NULL;
6530
6531 /* Hook this new socket in to the bind_hash list. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006532 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006533 sctp_local_bh_disable();
6534 sctp_spin_lock(&head->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006535 pp = sctp_sk(oldsk)->bind_hash;
6536 sk_add_bind_node(newsk, &pp->owner);
6537 sctp_sk(newsk)->bind_hash = pp;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006538 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006539 sctp_spin_unlock(&head->lock);
6540 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07006541
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006542 /* Copy the bind_addr list from the original endpoint to the new
6543 * endpoint so that we can handle restarts properly
6544 */
Vlad Yasevich8e71a112007-12-06 22:50:54 -08006545 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6546 &oldsp->ep->base.bind_addr, GFP_KERNEL);
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006547
Linus Torvalds1da177e2005-04-16 15:20:36 -07006548 /* Move any messages in the old socket's receive queue that are for the
6549 * peeled off association to the new socket's receive queue.
6550 */
6551 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6552 event = sctp_skb2event(skb);
6553 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006554 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006555 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006556 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006557 }
6558 }
6559
6560 /* Clean up any messages pending delivery due to partial
6561 * delivery. Three cases:
6562 * 1) No partial deliver; no work.
6563 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6564 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6565 */
6566 skb_queue_head_init(&newsp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006567 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006568
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006569 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006570 struct sk_buff_head *queue;
6571
6572 /* Decide which queue to move pd_lobby skbs to. */
6573 if (assoc->ulpq.pd_mode) {
6574 queue = &newsp->pd_lobby;
6575 } else
6576 queue = &newsk->sk_receive_queue;
6577
6578 /* Walk through the pd_lobby, looking for skbs that
6579 * need moved to the new socket.
6580 */
6581 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6582 event = sctp_skb2event(skb);
6583 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006584 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006585 __skb_queue_tail(queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006586 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006587 }
6588 }
6589
6590 /* Clear up any skbs waiting for the partial
6591 * delivery to finish.
6592 */
6593 if (assoc->ulpq.pd_mode)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006594 sctp_clear_pd(oldsk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006595
6596 }
6597
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006598 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006599 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006600
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006601 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006602 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006603
Linus Torvalds1da177e2005-04-16 15:20:36 -07006604 /* Set the type of socket to indicate that it is peeled off from the
6605 * original UDP-style socket or created with the accept() call on a
6606 * TCP-style socket..
6607 */
6608 newsp->type = type;
6609
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006610 /* Mark the new socket "in-use" by the user so that any packets
6611 * that may arrive on the association after we've moved it are
6612 * queued to the backlog. This prevents a potential race between
6613 * backlog processing on the old socket and new-packet processing
6614 * on the new socket.
Zach Brown5131a182007-06-22 15:14:46 -07006615 *
6616 * The caller has just allocated newsk so we can guarantee that other
6617 * paths won't try to lock it and then oldsk.
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006618 */
Zach Brown5131a182007-06-22 15:14:46 -07006619 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006620 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006621
6622 /* If the association on the newsk is already closed before accept()
6623 * is called, set RCV_SHUTDOWN flag.
6624 */
6625 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6626 newsk->sk_shutdown |= RCV_SHUTDOWN;
6627
6628 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006629 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006630}
6631
Neil Horman4d93df02007-08-15 16:07:44 -07006632
Linus Torvalds1da177e2005-04-16 15:20:36 -07006633/* This proto struct describes the ULP interface for SCTP. */
6634struct proto sctp_prot = {
6635 .name = "SCTP",
6636 .owner = THIS_MODULE,
6637 .close = sctp_close,
6638 .connect = sctp_connect,
6639 .disconnect = sctp_disconnect,
6640 .accept = sctp_accept,
6641 .ioctl = sctp_ioctl,
6642 .init = sctp_init_sock,
6643 .destroy = sctp_destroy_sock,
6644 .shutdown = sctp_shutdown,
6645 .setsockopt = sctp_setsockopt,
6646 .getsockopt = sctp_getsockopt,
6647 .sendmsg = sctp_sendmsg,
6648 .recvmsg = sctp_recvmsg,
6649 .bind = sctp_bind,
6650 .backlog_rcv = sctp_backlog_rcv,
6651 .hash = sctp_hash,
6652 .unhash = sctp_unhash,
6653 .get_port = sctp_get_port,
6654 .obj_size = sizeof(struct sctp_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006655 .sysctl_mem = sysctl_sctp_mem,
6656 .sysctl_rmem = sysctl_sctp_rmem,
6657 .sysctl_wmem = sysctl_sctp_wmem,
6658 .memory_pressure = &sctp_memory_pressure,
6659 .enter_memory_pressure = sctp_enter_memory_pressure,
6660 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006661 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006662};
6663
6664#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
Eric Dumazet8295b6d2007-11-05 23:40:28 -08006665
Linus Torvalds1da177e2005-04-16 15:20:36 -07006666struct proto sctpv6_prot = {
6667 .name = "SCTPv6",
6668 .owner = THIS_MODULE,
6669 .close = sctp_close,
6670 .connect = sctp_connect,
6671 .disconnect = sctp_disconnect,
6672 .accept = sctp_accept,
6673 .ioctl = sctp_ioctl,
6674 .init = sctp_init_sock,
6675 .destroy = sctp_destroy_sock,
6676 .shutdown = sctp_shutdown,
6677 .setsockopt = sctp_setsockopt,
6678 .getsockopt = sctp_getsockopt,
6679 .sendmsg = sctp_sendmsg,
6680 .recvmsg = sctp_recvmsg,
6681 .bind = sctp_bind,
6682 .backlog_rcv = sctp_backlog_rcv,
6683 .hash = sctp_hash,
6684 .unhash = sctp_unhash,
6685 .get_port = sctp_get_port,
6686 .obj_size = sizeof(struct sctp6_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006687 .sysctl_mem = sysctl_sctp_mem,
6688 .sysctl_rmem = sysctl_sctp_rmem,
6689 .sysctl_wmem = sysctl_sctp_wmem,
6690 .memory_pressure = &sctp_memory_pressure,
6691 .enter_memory_pressure = sctp_enter_memory_pressure,
6692 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006693 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006694};
6695#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */