blob: 5e259817a7f34cd4a183139fe9c4bf5ee2ab6689 [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 */
Paul Gortmakerbc3b2d72011-07-15 11:47:34 -040081#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <net/sock.h>
83#include <net/sctp/sctp.h>
84#include <net/sctp/sm.h>
85
86/* WARNING: Please do not remove the SCTP_STATIC attribute to
87 * any of the functions below as they are used to export functions
88 * used by a project regression testsuite.
89 */
90
91/* Forward declarations for internal helper functions. */
92static int sctp_writeable(struct sock *sk);
93static void sctp_wfree(struct sk_buff *skb);
94static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
95 size_t msg_len);
96static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
97static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
98static int sctp_wait_for_accept(struct sock *sk, long timeo);
99static void sctp_wait_for_close(struct sock *sk, long timeo);
100static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
101 union sctp_addr *addr, int len);
102static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
103static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
104static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
105static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
106static int sctp_send_asconf(struct sctp_association *asoc,
107 struct sctp_chunk *chunk);
108static int sctp_do_bind(struct sock *, union sctp_addr *, int);
109static int sctp_autobind(struct sock *sk);
110static void sctp_sock_migrate(struct sock *, struct sock *,
111 struct sctp_association *, sctp_socket_type_t);
112static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
113
Neil Horman4d93df02007-08-15 16:07:44 -0700114extern struct kmem_cache *sctp_bucket_cachep;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000115extern long sysctl_sctp_mem[3];
Neil Horman4d93df02007-08-15 16:07:44 -0700116extern int sysctl_sctp_rmem[3];
117extern int sysctl_sctp_wmem[3];
118
Adrian Bunkb6fa1a42007-09-12 15:18:00 +0200119static int sctp_memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000120static atomic_long_t sctp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800121struct percpu_counter sctp_sockets_allocated;
Neil Horman4d93df02007-08-15 16:07:44 -0700122
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700123static void sctp_enter_memory_pressure(struct sock *sk)
Neil Horman4d93df02007-08-15 16:07:44 -0700124{
125 sctp_memory_pressure = 1;
126}
127
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129/* Get the sndbuf space available at the time on the association. */
130static inline int sctp_wspace(struct sctp_association *asoc)
131{
Neil Horman4d93df02007-08-15 16:07:44 -0700132 int amt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
Neil Horman4d93df02007-08-15 16:07:44 -0700134 if (asoc->ep->sndbuf_policy)
135 amt = asoc->sndbuf_used;
136 else
Eric Dumazet31e6d362009-06-17 19:05:41 -0700137 amt = sk_wmem_alloc_get(asoc->base.sk);
Neil Horman4d93df02007-08-15 16:07:44 -0700138
139 if (amt >= asoc->base.sk->sk_sndbuf) {
140 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
141 amt = 0;
142 else {
143 amt = sk_stream_wspace(asoc->base.sk);
144 if (amt < 0)
145 amt = 0;
146 }
Neil Horman4eb701d2005-04-28 12:02:04 -0700147 } else {
Neil Horman4d93df02007-08-15 16:07:44 -0700148 amt = asoc->base.sk->sk_sndbuf - amt;
Neil Horman4eb701d2005-04-28 12:02:04 -0700149 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 return amt;
151}
152
153/* Increment the used sndbuf space count of the corresponding association by
154 * the size of the outgoing data chunk.
155 * Also, set the skb destructor for sndbuf accounting later.
156 *
157 * Since it is always 1-1 between chunk and skb, and also a new skb is always
158 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
159 * destructor in the data chunk skb for the purpose of the sndbuf space
160 * tracking.
161 */
162static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
163{
164 struct sctp_association *asoc = chunk->asoc;
165 struct sock *sk = asoc->base.sk;
166
167 /* The sndbuf space is tracked per association. */
168 sctp_association_hold(asoc);
169
Neil Horman4eb701d2005-04-28 12:02:04 -0700170 skb_set_owner_w(chunk->skb, sk);
171
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 chunk->skb->destructor = sctp_wfree;
173 /* Save the chunk pointer in skb for sctp_wfree to use later. */
174 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
175
Neil Horman4eb701d2005-04-28 12:02:04 -0700176 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
177 sizeof(struct sk_buff) +
178 sizeof(struct sctp_chunk);
179
Neil Horman4eb701d2005-04-28 12:02:04 -0700180 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800181 sk->sk_wmem_queued += chunk->skb->truesize;
182 sk_mem_charge(sk, chunk->skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183}
184
185/* Verify that this is a valid address. */
186static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
187 int len)
188{
189 struct sctp_af *af;
190
191 /* Verify basic sockaddr. */
192 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
193 if (!af)
194 return -EINVAL;
195
196 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700197 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 return -EINVAL;
199
200 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
201 return -EINVAL;
202
203 return 0;
204}
205
206/* Look up the association by its id. If this is not a UDP-style
207 * socket, the ID field is always ignored.
208 */
209struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
210{
211 struct sctp_association *asoc = NULL;
212
213 /* If this is not a UDP-style socket, assoc id should be ignored. */
214 if (!sctp_style(sk, UDP)) {
215 /* Return NULL if the socket state is not ESTABLISHED. It
216 * could be a TCP-style listening socket or a socket which
217 * hasn't yet called connect() to establish an association.
218 */
219 if (!sctp_sstate(sk, ESTABLISHED))
220 return NULL;
221
222 /* Get the first and the only association from the list. */
223 if (!list_empty(&sctp_sk(sk)->ep->asocs))
224 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
225 struct sctp_association, asocs);
226 return asoc;
227 }
228
229 /* Otherwise this is a UDP-style socket. */
230 if (!id || (id == (sctp_assoc_t)-1))
231 return NULL;
232
233 spin_lock_bh(&sctp_assocs_id_lock);
234 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
235 spin_unlock_bh(&sctp_assocs_id_lock);
236
237 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
238 return NULL;
239
240 return asoc;
241}
242
243/* Look up the transport from an address and an assoc id. If both address and
244 * id are specified, the associations matching the address and the id should be
245 * the same.
246 */
247static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
248 struct sockaddr_storage *addr,
249 sctp_assoc_t id)
250{
251 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
252 struct sctp_transport *transport;
253 union sctp_addr *laddr = (union sctp_addr *)addr;
254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
Al Virocd4ff032006-11-20 17:11:33 -0800256 laddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259 if (!addr_asoc)
260 return NULL;
261
262 id_asoc = sctp_id2assoc(sk, id);
263 if (id_asoc && (id_asoc != addr_asoc))
264 return NULL;
265
266 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
267 (union sctp_addr *)addr);
268
269 return transport;
270}
271
272/* API 3.1.2 bind() - UDP Style Syntax
273 * The syntax of bind() is,
274 *
275 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 *
277 * sd - the socket descriptor returned by socket().
278 * addr - the address structure (struct sockaddr_in or struct
279 * sockaddr_in6 [RFC 2553]),
280 * addr_len - the size of the address structure.
281 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700282SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283{
284 int retval = 0;
285
286 sctp_lock_sock(sk);
287
Frank Filz3f7a87d2005-06-20 13:14:57 -0700288 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
289 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290
291 /* Disallow binding twice. */
292 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700293 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 addr_len);
295 else
296 retval = -EINVAL;
297
298 sctp_release_sock(sk);
299
300 return retval;
301}
302
303static long sctp_get_port_local(struct sock *, union sctp_addr *);
304
305/* Verify this is a valid sockaddr. */
306static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
307 union sctp_addr *addr, int len)
308{
309 struct sctp_af *af;
310
311 /* Check minimum size. */
312 if (len < sizeof (struct sockaddr))
313 return NULL;
314
Vlad Yasevich7dab83d2008-07-18 23:05:40 -0700315 /* V4 mapped address are really of AF_INET family */
316 if (addr->sa.sa_family == AF_INET6 &&
317 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
318 if (!opt->pf->af_supported(AF_INET, opt))
319 return NULL;
320 } else {
321 /* Does this PF support this AF? */
322 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
323 return NULL;
324 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
326 /* If we get this far, af is valid. */
327 af = sctp_get_af_specific(addr->sa.sa_family);
328
329 if (len < af->sockaddr_len)
330 return NULL;
331
332 return af;
333}
334
335/* Bind a local address either to an endpoint or to an association. */
336SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
337{
338 struct sctp_sock *sp = sctp_sk(sk);
339 struct sctp_endpoint *ep = sp->ep;
340 struct sctp_bind_addr *bp = &ep->base.bind_addr;
341 struct sctp_af *af;
342 unsigned short snum;
343 int ret = 0;
344
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 /* Common sockaddr verification. */
346 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700347 if (!af) {
348 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
349 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700351 }
352
353 snum = ntohs(addr->v4.sin_port);
354
355 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
356 ", port: %d, new port: %d, len: %d)\n",
357 sk,
358 addr,
359 bp->port, snum,
360 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
362 /* PF specific bind() address verification. */
363 if (!sp->pf->bind_verify(sp, addr))
364 return -EADDRNOTAVAIL;
365
Vlad Yasevich8b358052007-05-15 17:14:58 -0400366 /* We must either be unbound, or bind to the same port.
367 * It's OK to allow 0 ports if we are already bound.
368 * We'll just inhert an already bound port in this case
369 */
370 if (bp->port) {
371 if (!snum)
372 snum = bp->port;
373 else if (snum != bp->port) {
374 SCTP_DEBUG_PRINTK("sctp_do_bind:"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 " New port %d does not match existing port "
376 "%d.\n", snum, bp->port);
Vlad Yasevich8b358052007-05-15 17:14:58 -0400377 return -EINVAL;
378 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 }
380
381 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
382 return -EACCES;
383
Vlad Yasevich4e540642008-07-18 23:06:32 -0700384 /* See if the address matches any of the addresses we may have
385 * already bound before checking against other endpoints.
386 */
387 if (sctp_bind_addr_match(bp, addr, sp))
388 return -EINVAL;
389
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 /* Make sure we are allowed to bind here.
391 * The function sctp_get_port_local() does duplicate address
392 * detection.
393 */
Vlad Yasevich2772b492007-08-21 14:24:30 +0900394 addr->v4.sin_port = htons(snum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 if ((ret = sctp_get_port_local(sk, addr))) {
Vlad Yasevich4e540642008-07-18 23:06:32 -0700396 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 }
398
399 /* Refresh ephemeral port. */
400 if (!bp->port)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000401 bp->port = inet_sk(sk)->inet_num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
Vlad Yasevich559cf712007-09-16 16:03:28 -0700403 /* Add the address to the bind address list.
404 * Use GFP_ATOMIC since BHs will be disabled.
405 */
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800406 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408 /* Copy back into socket for getsockname() use. */
409 if (!ret) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000410 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 af->to_sk_saddr(addr, sk);
412 }
413
414 return ret;
415}
416
417 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900419 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 * at any one time. If a sender, after sending an ASCONF chunk, decides
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900421 * it needs to transfer another ASCONF Chunk, it MUST wait until the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900423 * subsequent ASCONF. Note this restriction binds each side, so at any
424 * time two ASCONF may be in-transit on any given association (one sent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 * from each endpoint).
426 */
427static int sctp_send_asconf(struct sctp_association *asoc,
428 struct sctp_chunk *chunk)
429{
430 int retval = 0;
431
432 /* If there is an outstanding ASCONF chunk, queue it for later
433 * transmission.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900434 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700436 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900437 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 }
439
440 /* Hold the chunk until an ASCONF_ACK is received. */
441 sctp_chunk_hold(chunk);
442 retval = sctp_primitive_ASCONF(asoc, chunk);
443 if (retval)
444 sctp_chunk_free(chunk);
445 else
446 asoc->addip_last_asconf = chunk;
447
448out:
449 return retval;
450}
451
452/* Add a list of addresses as bind addresses to local endpoint or
453 * association.
454 *
455 * Basically run through each address specified in the addrs/addrcnt
456 * array/length pair, determine if it is IPv6 or IPv4 and call
457 * sctp_do_bind() on it.
458 *
459 * If any of them fails, then the operation will be reversed and the
460 * ones that were added will be removed.
461 *
462 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200464static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465{
466 int cnt;
467 int retval = 0;
468 void *addr_buf;
469 struct sockaddr *sa_addr;
470 struct sctp_af *af;
471
472 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
473 sk, addrs, addrcnt);
474
475 addr_buf = addrs;
476 for (cnt = 0; cnt < addrcnt; cnt++) {
477 /* The list may contain either IPv4 or IPv6 address;
478 * determine the address length for walking thru the list.
479 */
Joe Perchesea110732011-06-13 16:21:26 +0000480 sa_addr = addr_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 af = sctp_get_af_specific(sa_addr->sa_family);
482 if (!af) {
483 retval = -EINVAL;
484 goto err_bindx_add;
485 }
486
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900487 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 af->sockaddr_len);
489
490 addr_buf += af->sockaddr_len;
491
492err_bindx_add:
493 if (retval < 0) {
494 /* Failed. Cleanup the ones that have been added */
495 if (cnt > 0)
496 sctp_bindx_rem(sk, addrs, cnt);
497 return retval;
498 }
499 }
500
501 return retval;
502}
503
504/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
505 * associations that are part of the endpoint indicating that a list of local
506 * addresses are added to the endpoint.
507 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900508 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 * association, we do not send the chunk for that association. But it will not
510 * affect other associations.
511 *
512 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 */
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900514static int sctp_send_asconf_add_ip(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 struct sockaddr *addrs,
516 int addrcnt)
517{
518 struct sctp_sock *sp;
519 struct sctp_endpoint *ep;
520 struct sctp_association *asoc;
521 struct sctp_bind_addr *bp;
522 struct sctp_chunk *chunk;
523 struct sctp_sockaddr_entry *laddr;
524 union sctp_addr *addr;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700525 union sctp_addr saveaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 void *addr_buf;
527 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 struct list_head *p;
529 int i;
530 int retval = 0;
531
532 if (!sctp_addip_enable)
533 return retval;
534
535 sp = sctp_sk(sk);
536 ep = sp->ep;
537
538 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800539 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700541 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543 if (!asoc->peer.asconf_capable)
544 continue;
545
546 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
547 continue;
548
549 if (!sctp_state(asoc, ESTABLISHED))
550 continue;
551
552 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900553 * in the bind address list of the association. If so,
554 * do not send the asconf chunk to its peer, but continue with
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 * other associations.
556 */
557 addr_buf = addrs;
558 for (i = 0; i < addrcnt; i++) {
Joe Perchesea110732011-06-13 16:21:26 +0000559 addr = addr_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 af = sctp_get_af_specific(addr->v4.sin_family);
561 if (!af) {
562 retval = -EINVAL;
563 goto out;
564 }
565
566 if (sctp_assoc_lookup_laddr(asoc, addr))
567 break;
568
569 addr_buf += af->sockaddr_len;
570 }
571 if (i < addrcnt)
572 continue;
573
Vlad Yasevich559cf712007-09-16 16:03:28 -0700574 /* Use the first valid address in bind addr list of
575 * association as Address Parameter of ASCONF CHUNK.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 bp = &asoc->base.bind_addr;
578 p = bp->address_list.next;
579 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
Al Viro5ae955c2006-11-20 17:22:08 -0800580 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 addrcnt, SCTP_PARAM_ADD_IP);
582 if (!chunk) {
583 retval = -ENOMEM;
584 goto out;
585 }
586
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700590 addr_buf = addrs;
591 for (i = 0; i < addrcnt; i++) {
Joe Perchesea110732011-06-13 16:21:26 +0000592 addr = addr_buf;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700597 addr_buf += af->sockaddr_len;
598 }
Michio Honda8a07eb02011-04-26 20:19:36 +0900599 if (asoc->src_out_of_asoc_ok) {
600 struct sctp_transport *trans;
601
602 list_for_each_entry(trans,
603 &asoc->peer.transport_addr_list, transports) {
604 /* Clear the source and route cache */
605 dst_release(trans->dst);
606 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
607 2*asoc->pathmtu, 4380));
608 trans->ssthresh = asoc->peer.i.a_rwnd;
609 trans->rto = asoc->rto_initial;
610 trans->rtt = trans->srtt = trans->rttvar = 0;
611 sctp_transport_route(trans, NULL,
612 sctp_sk(asoc->base.sk));
613 }
614 }
615 retval = sctp_send_asconf(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 }
617
618out:
619 return retval;
620}
621
622/* Remove a list of addresses from bind addresses list. Do not remove the
623 * last address.
624 *
625 * Basically run through each address specified in the addrs/addrcnt
626 * array/length pair, determine if it is IPv6 or IPv4 and call
627 * sctp_del_bind() on it.
628 *
629 * If any of them fails, then the operation will be reversed and the
630 * ones that were removed will be added back.
631 *
632 * At least one address has to be left; if only one address is
633 * available, the operation will return -EBUSY.
634 *
635 * Only sctp_setsockopt_bindx() is supposed to call this function.
636 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200637static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
639 struct sctp_sock *sp = sctp_sk(sk);
640 struct sctp_endpoint *ep = sp->ep;
641 int cnt;
642 struct sctp_bind_addr *bp = &ep->base.bind_addr;
643 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 void *addr_buf;
Al Viroc9a08502006-11-20 17:07:48 -0800645 union sctp_addr *sa_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 struct sctp_af *af;
647
648 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
649 sk, addrs, addrcnt);
650
651 addr_buf = addrs;
652 for (cnt = 0; cnt < addrcnt; cnt++) {
653 /* If the bind address list is empty or if there is only one
654 * bind address, there is nothing more to be removed (we need
655 * at least one address here).
656 */
657 if (list_empty(&bp->address_list) ||
658 (sctp_list_single_entry(&bp->address_list))) {
659 retval = -EBUSY;
660 goto err_bindx_rem;
661 }
662
Joe Perchesea110732011-06-13 16:21:26 +0000663 sa_addr = addr_buf;
Al Viroc9a08502006-11-20 17:07:48 -0800664 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 if (!af) {
666 retval = -EINVAL;
667 goto err_bindx_rem;
668 }
Paolo Galtieri0304ff8a2007-04-17 12:52:36 -0700669
670 if (!af->addr_valid(sa_addr, sp, NULL)) {
671 retval = -EADDRNOTAVAIL;
672 goto err_bindx_rem;
673 }
674
Vlad Yasevichee9cbac2011-04-18 19:14:47 +0000675 if (sa_addr->v4.sin_port &&
676 sa_addr->v4.sin_port != htons(bp->port)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 retval = -EINVAL;
678 goto err_bindx_rem;
679 }
680
Vlad Yasevichee9cbac2011-04-18 19:14:47 +0000681 if (!sa_addr->v4.sin_port)
682 sa_addr->v4.sin_port = htons(bp->port);
683
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 /* FIXME - There is probably a need to check if sk->sk_saddr and
685 * sk->sk_rcv_addr are currently set to one of the addresses to
686 * be removed. This is something which needs to be looked into
687 * when we are fixing the outstanding issues with multi-homing
688 * socket routing and failover schemes. Refer to comments in
689 * sctp_do_bind(). -daisy
690 */
Vlad Yasevich0ed90fb2007-10-24 16:10:00 -0400691 retval = sctp_del_bind_addr(bp, sa_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
693 addr_buf += af->sockaddr_len;
694err_bindx_rem:
695 if (retval < 0) {
696 /* Failed. Add the ones that has been removed back */
697 if (cnt > 0)
698 sctp_bindx_add(sk, addrs, cnt);
699 return retval;
700 }
701 }
702
703 return retval;
704}
705
706/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
707 * the associations that are part of the endpoint indicating that a list of
708 * local addresses are removed from the endpoint.
709 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900710 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 * association, we do not send the chunk for that association. But it will not
712 * affect other associations.
713 *
714 * Only sctp_setsockopt_bindx() is supposed to call this function.
715 */
716static int sctp_send_asconf_del_ip(struct sock *sk,
717 struct sockaddr *addrs,
718 int addrcnt)
719{
720 struct sctp_sock *sp;
721 struct sctp_endpoint *ep;
722 struct sctp_association *asoc;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700723 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 struct sctp_bind_addr *bp;
725 struct sctp_chunk *chunk;
726 union sctp_addr *laddr;
727 void *addr_buf;
728 struct sctp_af *af;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700729 struct sctp_sockaddr_entry *saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 int i;
731 int retval = 0;
Michio Honda8a07eb02011-04-26 20:19:36 +0900732 int stored = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733
Michio Honda8a07eb02011-04-26 20:19:36 +0900734 chunk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 if (!sctp_addip_enable)
736 return retval;
737
738 sp = sctp_sk(sk);
739 ep = sp->ep;
740
741 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800742 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700744 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
746 if (!asoc->peer.asconf_capable)
747 continue;
748
749 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
750 continue;
751
752 if (!sctp_state(asoc, ESTABLISHED))
753 continue;
754
755 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900756 * not present in the bind address list of the association.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 * If so, do not send the asconf chunk to its peer, but
758 * continue with other associations.
759 */
760 addr_buf = addrs;
761 for (i = 0; i < addrcnt; i++) {
Joe Perchesea110732011-06-13 16:21:26 +0000762 laddr = addr_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 af = sctp_get_af_specific(laddr->v4.sin_family);
764 if (!af) {
765 retval = -EINVAL;
766 goto out;
767 }
768
769 if (!sctp_assoc_lookup_laddr(asoc, laddr))
770 break;
771
772 addr_buf += af->sockaddr_len;
773 }
774 if (i < addrcnt)
775 continue;
776
777 /* Find one address in the association's bind address list
778 * that is not in the packed array of addresses. This is to
779 * make sure that we do not delete all the addresses in the
780 * association.
781 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 bp = &asoc->base.bind_addr;
783 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
784 addrcnt, sp);
Michio Honda8a07eb02011-04-26 20:19:36 +0900785 if ((laddr == NULL) && (addrcnt == 1)) {
786 if (asoc->asconf_addr_del_pending)
787 continue;
788 asoc->asconf_addr_del_pending =
789 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
Michio Honda6d65e5e2011-06-10 16:42:14 +0900790 if (asoc->asconf_addr_del_pending == NULL) {
791 retval = -ENOMEM;
792 goto out;
793 }
Michio Honda8a07eb02011-04-26 20:19:36 +0900794 asoc->asconf_addr_del_pending->sa.sa_family =
795 addrs->sa_family;
796 asoc->asconf_addr_del_pending->v4.sin_port =
797 htons(bp->port);
798 if (addrs->sa_family == AF_INET) {
799 struct sockaddr_in *sin;
800
801 sin = (struct sockaddr_in *)addrs;
802 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
803 } else if (addrs->sa_family == AF_INET6) {
804 struct sockaddr_in6 *sin6;
805
806 sin6 = (struct sockaddr_in6 *)addrs;
Alexey Dobriyan4e3fd7a2011-11-21 03:39:03 +0000807 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
Michio Honda8a07eb02011-04-26 20:19:36 +0900808 }
809 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
810 " at %p\n", asoc, asoc->asconf_addr_del_pending,
811 asoc->asconf_addr_del_pending);
812 asoc->src_out_of_asoc_ok = 1;
813 stored = 1;
814 goto skip_mkasconf;
815 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816
Vlad Yasevich559cf712007-09-16 16:03:28 -0700817 /* We do not need RCU protection throughout this loop
818 * because this is done under a socket lock from the
819 * setsockopt call.
820 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
822 SCTP_PARAM_DEL_IP);
823 if (!chunk) {
824 retval = -ENOMEM;
825 goto out;
826 }
827
Michio Honda8a07eb02011-04-26 20:19:36 +0900828skip_mkasconf:
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700829 /* Reset use_as_src flag for the addresses in the bind address
830 * list that are to be deleted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700832 addr_buf = addrs;
833 for (i = 0; i < addrcnt; i++) {
Joe Perchesea110732011-06-13 16:21:26 +0000834 laddr = addr_buf;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700835 af = sctp_get_af_specific(laddr->v4.sin_family);
Vlad Yasevich559cf712007-09-16 16:03:28 -0700836 list_for_each_entry(saddr, &bp->address_list, list) {
Al Viro5f242a12006-11-20 17:05:23 -0800837 if (sctp_cmp_addr_exact(&saddr->a, laddr))
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800838 saddr->state = SCTP_ADDR_DEL;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700839 }
840 addr_buf += af->sockaddr_len;
841 }
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700842
843 /* Update the route and saddr entries for all the transports
844 * as some of the addresses in the bind address list are
845 * about to be deleted and cannot be used as source addresses.
846 */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700847 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
848 transports) {
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700849 dst_release(transport->dst);
850 sctp_transport_route(transport, NULL,
851 sctp_sk(asoc->base.sk));
852 }
853
Michio Honda8a07eb02011-04-26 20:19:36 +0900854 if (stored)
855 /* We don't need to transmit ASCONF */
856 continue;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700857 retval = sctp_send_asconf(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 }
859out:
860 return retval;
861}
862
Michio Honda9f7d6532011-04-26 19:32:51 +0900863/* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
864int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
865{
866 struct sock *sk = sctp_opt2sk(sp);
867 union sctp_addr *addr;
868 struct sctp_af *af;
869
870 /* It is safe to write port space in caller. */
871 addr = &addrw->a;
872 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
873 af = sctp_get_af_specific(addr->sa.sa_family);
874 if (!af)
875 return -EINVAL;
876 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
877 return -EINVAL;
878
879 if (addrw->state == SCTP_ADDR_NEW)
880 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
881 else
882 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
883}
884
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
886 *
887 * API 8.1
888 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
889 * int flags);
890 *
891 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
892 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
893 * or IPv6 addresses.
894 *
895 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
896 * Section 3.1.2 for this usage.
897 *
898 * addrs is a pointer to an array of one or more socket addresses. Each
899 * address is contained in its appropriate structure (i.e. struct
900 * sockaddr_in or struct sockaddr_in6) the family of the address type
Ville Nuorvala23c435f2006-10-16 22:08:28 -0700901 * must be used to distinguish the address length (note that this
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 * representation is termed a "packed array" of addresses). The caller
903 * specifies the number of addresses in the array with addrcnt.
904 *
905 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
906 * -1, and sets errno to the appropriate error code.
907 *
908 * For SCTP, the port given in each socket address must be the same, or
909 * sctp_bindx() will fail, setting errno to EINVAL.
910 *
911 * The flags parameter is formed from the bitwise OR of zero or more of
912 * the following currently defined flags:
913 *
914 * SCTP_BINDX_ADD_ADDR
915 *
916 * SCTP_BINDX_REM_ADDR
917 *
918 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
919 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
920 * addresses from the association. The two flags are mutually exclusive;
921 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
922 * not remove all addresses from an association; sctp_bindx() will
923 * reject such an attempt with EINVAL.
924 *
925 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
926 * additional addresses with an endpoint after calling bind(). Or use
927 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
928 * socket is associated with so that no new association accepted will be
929 * associated with those addresses. If the endpoint supports dynamic
930 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
931 * endpoint to send the appropriate message to the peer to change the
932 * peers address lists.
933 *
934 * Adding and removing addresses from a connected association is
935 * optional functionality. Implementations that do not support this
936 * functionality should return EOPNOTSUPP.
937 *
938 * Basically do nothing but copying the addresses from user to kernel
939 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700940 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
941 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 *
943 * We don't use copy_from_user() for optimization: we first do the
944 * sanity checks (buffer size -fast- and access check-healthy
945 * pointer); if all of those succeed, then we can alloc the memory
946 * (expensive operation) needed to copy the data to kernel. Then we do
947 * the copying without checking the user space area
948 * (__copy_from_user()).
949 *
950 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
951 * it.
952 *
953 * sk The sk of the socket
954 * addrs The pointer to the addresses in user land
955 * addrssize Size of the addrs buffer
956 * op Operation to perform (add or remove, see the flags of
957 * sctp_bindx)
958 *
959 * Returns 0 if ok, <0 errno code on error.
960 */
961SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
962 struct sockaddr __user *addrs,
963 int addrs_size, int op)
964{
965 struct sockaddr *kaddrs;
966 int err;
967 int addrcnt = 0;
968 int walk_size = 0;
969 struct sockaddr *sa_addr;
970 void *addr_buf;
971 struct sctp_af *af;
972
973 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
974 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
975
976 if (unlikely(addrs_size <= 0))
977 return -EINVAL;
978
979 /* Check the user passed a healthy pointer. */
980 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
981 return -EFAULT;
982
983 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800984 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 if (unlikely(!kaddrs))
986 return -ENOMEM;
987
988 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
989 kfree(kaddrs);
990 return -EFAULT;
991 }
992
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900993 /* Walk through the addrs buffer and count the number of addresses. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 addr_buf = kaddrs;
995 while (walk_size < addrs_size) {
Dan Rosenbergd7e0d192010-10-01 11:16:58 +0000996 if (walk_size + sizeof(sa_family_t) > addrs_size) {
997 kfree(kaddrs);
998 return -EINVAL;
999 }
1000
Joe Perchesea110732011-06-13 16:21:26 +00001001 sa_addr = addr_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 af = sctp_get_af_specific(sa_addr->sa_family);
1003
1004 /* If the address family is not supported or if this address
1005 * causes the address buffer to overflow return EINVAL.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001006 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1008 kfree(kaddrs);
1009 return -EINVAL;
1010 }
1011 addrcnt++;
1012 addr_buf += af->sockaddr_len;
1013 walk_size += af->sockaddr_len;
1014 }
1015
1016 /* Do the work. */
1017 switch (op) {
1018 case SCTP_BINDX_ADD_ADDR:
1019 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1020 if (err)
1021 goto out;
1022 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1023 break;
1024
1025 case SCTP_BINDX_REM_ADDR:
1026 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1027 if (err)
1028 goto out;
1029 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1030 break;
1031
1032 default:
1033 err = -EINVAL;
1034 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07001035 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
1037out:
1038 kfree(kaddrs);
1039
1040 return err;
1041}
1042
Frank Filz3f7a87d2005-06-20 13:14:57 -07001043/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1044 *
1045 * Common routine for handling connect() and sctp_connectx().
1046 * Connect will come in with just a single address.
1047 */
1048static int __sctp_connect(struct sock* sk,
1049 struct sockaddr *kaddrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001050 int addrs_size,
1051 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -07001052{
1053 struct sctp_sock *sp;
1054 struct sctp_endpoint *ep;
1055 struct sctp_association *asoc = NULL;
1056 struct sctp_association *asoc2;
1057 struct sctp_transport *transport;
1058 union sctp_addr to;
1059 struct sctp_af *af;
1060 sctp_scope_t scope;
1061 long timeo;
1062 int err = 0;
1063 int addrcnt = 0;
1064 int walk_size = 0;
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001065 union sctp_addr *sa_addr = NULL;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001066 void *addr_buf;
Vlad Yasevich16d00fb2007-05-04 13:34:09 -07001067 unsigned short port;
Vlad Yasevichf50f95c2007-07-03 12:47:40 -04001068 unsigned int f_flags = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001069
1070 sp = sctp_sk(sk);
1071 ep = sp->ep;
1072
1073 /* connect() cannot be done on a socket that is already in ESTABLISHED
1074 * state - UDP-style peeled off socket or a TCP-style socket that
1075 * is already connected.
1076 * It cannot be done even on a TCP-style listening socket.
1077 */
1078 if (sctp_sstate(sk, ESTABLISHED) ||
1079 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1080 err = -EISCONN;
1081 goto out_free;
1082 }
1083
1084 /* Walk through the addrs buffer and count the number of addresses. */
1085 addr_buf = kaddrs;
1086 while (walk_size < addrs_size) {
Dan Rosenbergd7e0d192010-10-01 11:16:58 +00001087 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1088 err = -EINVAL;
1089 goto out_free;
1090 }
1091
Joe Perchesea110732011-06-13 16:21:26 +00001092 sa_addr = addr_buf;
Al Viro4bdf4b52006-11-20 17:10:20 -08001093 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001094
1095 /* If the address family is not supported or if this address
1096 * causes the address buffer to overflow return EINVAL.
1097 */
1098 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1099 err = -EINVAL;
1100 goto out_free;
1101 }
1102
Dan Rosenbergd7e0d192010-10-01 11:16:58 +00001103 port = ntohs(sa_addr->v4.sin_port);
1104
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001105 /* Save current address so we can work with it */
1106 memcpy(&to, sa_addr, af->sockaddr_len);
1107
1108 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001109 if (err)
1110 goto out_free;
1111
Vlad Yasevich16d00fb2007-05-04 13:34:09 -07001112 /* Make sure the destination port is correctly set
1113 * in all addresses.
1114 */
1115 if (asoc && asoc->peer.port && asoc->peer.port != port)
1116 goto out_free;
1117
Frank Filz3f7a87d2005-06-20 13:14:57 -07001118
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1121 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001122 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001123 if (asoc2 && asoc2 != asoc) {
1124 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1125 err = -EISCONN;
1126 else
1127 err = -EALREADY;
1128 goto out_free;
1129 }
1130
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1134 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001135 if (sctp_endpoint_is_peeled_off(ep, &to)) {
Frank Filz3f7a87d2005-06-20 13:14:57 -07001136 err = -EADDRNOTAVAIL;
1137 goto out_free;
1138 }
1139
1140 if (!asoc) {
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1145 */
1146 if (!ep->base.bind_addr.port) {
1147 if (sctp_autobind(sk)) {
1148 err = -EAGAIN;
1149 goto out_free;
1150 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001151 } else {
1152 /*
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001157 * be permitted to open new associations.
1158 */
1159 if (ep->base.bind_addr.port < PROT_SOCK &&
1160 !capable(CAP_NET_BIND_SERVICE)) {
1161 err = -EACCES;
1162 goto out_free;
1163 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001164 }
1165
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001166 scope = sctp_scope(&to);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001167 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1168 if (!asoc) {
1169 err = -ENOMEM;
1170 goto out_free;
1171 }
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001172
1173 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1174 GFP_KERNEL);
1175 if (err < 0) {
1176 goto out_free;
1177 }
1178
Frank Filz3f7a87d2005-06-20 13:14:57 -07001179 }
1180
1181 /* Prime the peer's transport structures. */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001182 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001183 SCTP_UNKNOWN);
1184 if (!transport) {
1185 err = -ENOMEM;
1186 goto out_free;
1187 }
1188
1189 addrcnt++;
1190 addr_buf += af->sockaddr_len;
1191 walk_size += af->sockaddr_len;
1192 }
1193
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1196 */
1197 if (assoc_id) {
1198 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 if (err < 0)
1200 goto out_free;
1201 }
1202
Frank Filz3f7a87d2005-06-20 13:14:57 -07001203 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1204 if (err < 0) {
1205 goto out_free;
1206 }
1207
1208 /* Initialize sk's dport and daddr for getpeername() */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001209 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001210 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1211 af->to_sk_daddr(sa_addr, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001212 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001213
Vlad Yasevichf50f95c2007-07-03 12:47:40 -04001214 /* in-kernel sockets don't generally have a file allocated to them
1215 * if all they do is call sock_create_kern().
1216 */
1217 if (sk->sk_socket->file)
1218 f_flags = sk->sk_socket->file->f_flags;
1219
1220 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1221
Frank Filz3f7a87d2005-06-20 13:14:57 -07001222 err = sctp_wait_for_connect(asoc, &timeo);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001223 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001224 *assoc_id = asoc->assoc_id;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001225
1226 /* Don't free association on exit. */
1227 asoc = NULL;
1228
1229out_free:
1230
1231 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001232 " kaddrs: %p err: %d\n",
1233 asoc, kaddrs, err);
Neil Horman2eebc1e2012-07-16 09:13:51 +00001234 if (asoc) {
1235 /* sctp_primitive_ASSOCIATE may have added this association
1236 * To the hash table, try to unhash it, just in case, its a noop
1237 * if it wasn't hashed so we're safe
1238 */
1239 sctp_unhash_established(asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001240 sctp_association_free(asoc);
Neil Horman2eebc1e2012-07-16 09:13:51 +00001241 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001242 return err;
1243}
1244
1245/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1246 *
1247 * API 8.9
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001248 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1249 * sctp_assoc_t *asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001250 *
1251 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1252 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1253 * or IPv6 addresses.
1254 *
1255 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1256 * Section 3.1.2 for this usage.
1257 *
1258 * addrs is a pointer to an array of one or more socket addresses. Each
1259 * address is contained in its appropriate structure (i.e. struct
1260 * sockaddr_in or struct sockaddr_in6) the family of the address type
1261 * must be used to distengish the address length (note that this
1262 * representation is termed a "packed array" of addresses). The caller
1263 * specifies the number of addresses in the array with addrcnt.
1264 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001265 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1266 * the association id of the new association. On failure, sctp_connectx()
1267 * returns -1, and sets errno to the appropriate error code. The assoc_id
1268 * is not touched by the kernel.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001269 *
1270 * For SCTP, the port given in each socket address must be the same, or
1271 * sctp_connectx() will fail, setting errno to EINVAL.
1272 *
1273 * An application can use sctp_connectx to initiate an association with
1274 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1275 * allows a caller to specify multiple addresses at which a peer can be
1276 * reached. The way the SCTP stack uses the list of addresses to set up
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001277 * the association is implementation dependent. This function only
Frank Filz3f7a87d2005-06-20 13:14:57 -07001278 * specifies that the stack will try to make use of all the addresses in
1279 * the list when needed.
1280 *
1281 * Note that the list of addresses passed in is only used for setting up
1282 * the association. It does not necessarily equal the set of addresses
1283 * the peer uses for the resulting association. If the caller wants to
1284 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1285 * retrieve them after the association has been set up.
1286 *
1287 * Basically do nothing but copying the addresses from user to kernel
1288 * land and invoking either sctp_connectx(). This is used for tunneling
1289 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1290 *
1291 * We don't use copy_from_user() for optimization: we first do the
1292 * sanity checks (buffer size -fast- and access check-healthy
1293 * pointer); if all of those succeed, then we can alloc the memory
1294 * (expensive operation) needed to copy the data to kernel. Then we do
1295 * the copying without checking the user space area
1296 * (__copy_from_user()).
1297 *
1298 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1299 * it.
1300 *
1301 * sk The sk of the socket
1302 * addrs The pointer to the addresses in user land
1303 * addrssize Size of the addrs buffer
1304 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001305 * Returns >=0 if ok, <0 errno code on error.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001306 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001307SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001308 struct sockaddr __user *addrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001309 int addrs_size,
1310 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -07001311{
1312 int err = 0;
1313 struct sockaddr *kaddrs;
1314
1315 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001316 __func__, sk, addrs, addrs_size);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001317
1318 if (unlikely(addrs_size <= 0))
1319 return -EINVAL;
1320
1321 /* Check the user passed a healthy pointer. */
1322 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1323 return -EFAULT;
1324
1325 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001326 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001327 if (unlikely(!kaddrs))
1328 return -ENOMEM;
1329
1330 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1331 err = -EFAULT;
1332 } else {
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001333 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001334 }
1335
1336 kfree(kaddrs);
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001337
Frank Filz3f7a87d2005-06-20 13:14:57 -07001338 return err;
1339}
1340
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001341/*
1342 * This is an older interface. It's kept for backward compatibility
1343 * to the option that doesn't provide association id.
1344 */
1345SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1346 struct sockaddr __user *addrs,
1347 int addrs_size)
1348{
1349 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1350}
1351
1352/*
1353 * New interface for the API. The since the API is done with a socket
1354 * option, to make it simple we feed back the association id is as a return
1355 * indication to the call. Error is always negative and association id is
1356 * always positive.
1357 */
1358SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1359 struct sockaddr __user *addrs,
1360 int addrs_size)
1361{
1362 sctp_assoc_t assoc_id = 0;
1363 int err = 0;
1364
1365 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1366
1367 if (err)
1368 return err;
1369 else
1370 return assoc_id;
1371}
1372
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001373/*
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001374 * New (hopefully final) interface for the API.
1375 * We use the sctp_getaddrs_old structure so that use-space library
1376 * can avoid any unnecessary allocations. The only defferent part
1377 * is that we store the actual length of the address buffer into the
1378 * addrs_num structure member. That way we can re-use the existing
1379 * code.
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001380 */
1381SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1382 char __user *optval,
1383 int __user *optlen)
1384{
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001385 struct sctp_getaddrs_old param;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001386 sctp_assoc_t assoc_id = 0;
1387 int err = 0;
1388
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001389 if (len < sizeof(param))
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001390 return -EINVAL;
1391
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001392 if (copy_from_user(&param, optval, sizeof(param)))
1393 return -EFAULT;
1394
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001395 err = __sctp_setsockopt_connectx(sk,
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001396 (struct sockaddr __user *)param.addrs,
1397 param.addr_num, &assoc_id);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001398
1399 if (err == 0 || err == -EINPROGRESS) {
1400 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1401 return -EFAULT;
1402 if (put_user(sizeof(assoc_id), optlen))
1403 return -EFAULT;
1404 }
1405
1406 return err;
1407}
1408
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409/* API 3.1.4 close() - UDP Style Syntax
1410 * Applications use close() to perform graceful shutdown (as described in
1411 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1412 * by a UDP-style socket.
1413 *
1414 * The syntax is
1415 *
1416 * ret = close(int sd);
1417 *
1418 * sd - the socket descriptor of the associations to be closed.
1419 *
1420 * To gracefully shutdown a specific association represented by the
1421 * UDP-style socket, an application should use the sendmsg() call,
1422 * passing no user data, but including the appropriate flag in the
1423 * ancillary data (see Section xxxx).
1424 *
1425 * If sd in the close() call is a branched-off socket representing only
1426 * one association, the shutdown is performed on that association only.
1427 *
1428 * 4.1.6 close() - TCP Style Syntax
1429 *
1430 * Applications use close() to gracefully close down an association.
1431 *
1432 * The syntax is:
1433 *
1434 * int close(int sd);
1435 *
1436 * sd - the socket descriptor of the association to be closed.
1437 *
1438 * After an application calls close() on a socket descriptor, no further
1439 * socket operations will succeed on that descriptor.
1440 *
1441 * API 7.1.4 SO_LINGER
1442 *
1443 * An application using the TCP-style socket can use this option to
1444 * perform the SCTP ABORT primitive. The linger option structure is:
1445 *
1446 * struct linger {
1447 * int l_onoff; // option on/off
1448 * int l_linger; // linger time
1449 * };
1450 *
1451 * To enable the option, set l_onoff to 1. If the l_linger value is set
1452 * to 0, calling close() is the same as the ABORT primitive. If the
1453 * value is set to a negative value, the setsockopt() call will return
1454 * an error. If the value is set to a positive value linger_time, the
1455 * close() can be blocked for at most linger_time ms. If the graceful
1456 * shutdown phase does not finish during this period, close() will
1457 * return but the graceful shutdown phase continues in the system.
1458 */
1459SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1460{
1461 struct sctp_endpoint *ep;
1462 struct sctp_association *asoc;
1463 struct list_head *pos, *temp;
Thomas Grafcd4fcc72011-07-08 04:37:46 +00001464 unsigned int data_was_unread;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1467
1468 sctp_lock_sock(sk);
1469 sk->sk_shutdown = SHUTDOWN_MASK;
Vlad Yasevichbec96402009-07-30 18:08:28 -04001470 sk->sk_state = SCTP_SS_CLOSING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
1472 ep = sctp_sk(sk)->ep;
1473
Thomas Grafcd4fcc72011-07-08 04:37:46 +00001474 /* Clean up any skbs sitting on the receive queue. */
1475 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1476 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1477
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001478 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 list_for_each_safe(pos, temp, &ep->asocs) {
1480 asoc = list_entry(pos, struct sctp_association, asocs);
1481
1482 if (sctp_style(sk, TCP)) {
1483 /* A closed association can still be in the list if
1484 * it belongs to a TCP-style listening socket that is
1485 * not yet accepted. If so, free it. If not, send an
1486 * ABORT or SHUTDOWN based on the linger options.
1487 */
1488 if (sctp_state(asoc, CLOSED)) {
1489 sctp_unhash_established(asoc);
1490 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001491 continue;
1492 }
1493 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494
Thomas Grafcd4fcc72011-07-08 04:37:46 +00001495 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1496 !skb_queue_empty(&asoc->ulpq.reasm) ||
1497 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
Sridhar Samudralab9ac8672006-08-28 13:53:01 -07001498 struct sctp_chunk *chunk;
1499
1500 chunk = sctp_make_abort_user(asoc, NULL, 0);
1501 if (chunk)
1502 sctp_primitive_ABORT(asoc, chunk);
1503 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 sctp_primitive_SHUTDOWN(asoc, NULL);
1505 }
1506
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 /* On a TCP-style socket, block for at most linger_time if set. */
1508 if (sctp_style(sk, TCP) && timeout)
1509 sctp_wait_for_close(sk, timeout);
1510
1511 /* This will run the backlog queue. */
1512 sctp_release_sock(sk);
1513
1514 /* Supposedly, no process has access to the socket, but
1515 * the net layers still may.
1516 */
1517 sctp_local_bh_disable();
1518 sctp_bh_lock_sock(sk);
1519
1520 /* Hold the sock, since sk_common_release() will put sock_put()
1521 * and we have just a little more cleanup.
1522 */
1523 sock_hold(sk);
1524 sk_common_release(sk);
1525
1526 sctp_bh_unlock_sock(sk);
1527 sctp_local_bh_enable();
1528
1529 sock_put(sk);
1530
1531 SCTP_DBG_OBJCNT_DEC(sock);
1532}
1533
1534/* Handle EPIPE error. */
1535static int sctp_error(struct sock *sk, int flags, int err)
1536{
1537 if (err == -EPIPE)
1538 err = sock_error(sk) ? : -EPIPE;
1539 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1540 send_sig(SIGPIPE, current, 0);
1541 return err;
1542}
1543
1544/* API 3.1.3 sendmsg() - UDP Style Syntax
1545 *
1546 * An application uses sendmsg() and recvmsg() calls to transmit data to
1547 * and receive data from its peer.
1548 *
1549 * ssize_t sendmsg(int socket, const struct msghdr *message,
1550 * int flags);
1551 *
1552 * socket - the socket descriptor of the endpoint.
1553 * message - pointer to the msghdr structure which contains a single
1554 * user message and possibly some ancillary data.
1555 *
1556 * See Section 5 for complete description of the data
1557 * structures.
1558 *
1559 * flags - flags sent or received with the user message, see Section
1560 * 5 for complete description of the flags.
1561 *
1562 * Note: This function could use a rewrite especially when explicit
1563 * connect support comes in.
1564 */
1565/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1566
1567SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1568
1569SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1570 struct msghdr *msg, size_t msg_len)
1571{
1572 struct sctp_sock *sp;
1573 struct sctp_endpoint *ep;
1574 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1575 struct sctp_transport *transport, *chunk_tp;
1576 struct sctp_chunk *chunk;
Al Virodce116a2006-11-20 17:25:15 -08001577 union sctp_addr to;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 struct sockaddr *msg_name = NULL;
Joe Perches517aa0b2011-05-12 11:27:20 +00001579 struct sctp_sndrcvinfo default_sinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 struct sctp_sndrcvinfo *sinfo;
1581 struct sctp_initmsg *sinit;
1582 sctp_assoc_t associd = 0;
1583 sctp_cmsgs_t cmsgs = { NULL };
1584 int err;
1585 sctp_scope_t scope;
1586 long timeo;
1587 __u16 sinfo_flags = 0;
1588 struct sctp_datamsg *datamsg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 int msg_flags = msg->msg_flags;
1590
1591 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1592 sk, msg, msg_len);
1593
1594 err = 0;
1595 sp = sctp_sk(sk);
1596 ep = sp->ep;
1597
Frank Filz3f7a87d2005-06-20 13:14:57 -07001598 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599
1600 /* We cannot send a message over a TCP-style listening socket. */
1601 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1602 err = -EPIPE;
1603 goto out_nounlock;
1604 }
1605
1606 /* Parse out the SCTP CMSGs. */
1607 err = sctp_msghdr_parse(msg, &cmsgs);
1608
1609 if (err) {
1610 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1611 goto out_nounlock;
1612 }
1613
1614 /* Fetch the destination address for this packet. This
1615 * address only selects the association--it is not necessarily
1616 * the address we will send to.
1617 * For a peeled-off socket, msg_name is ignored.
1618 */
1619 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1620 int msg_namelen = msg->msg_namelen;
1621
1622 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1623 msg_namelen);
1624 if (err)
1625 return err;
1626
1627 if (msg_namelen > sizeof(to))
1628 msg_namelen = sizeof(to);
1629 memcpy(&to, msg->msg_name, msg_namelen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 msg_name = msg->msg_name;
1631 }
1632
1633 sinfo = cmsgs.info;
1634 sinit = cmsgs.init;
1635
1636 /* Did the user specify SNDRCVINFO? */
1637 if (sinfo) {
1638 sinfo_flags = sinfo->sinfo_flags;
1639 associd = sinfo->sinfo_assoc_id;
1640 }
1641
1642 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1643 msg_len, sinfo_flags);
1644
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001645 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1646 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 err = -EINVAL;
1648 goto out_nounlock;
1649 }
1650
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001651 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1652 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1653 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 * the msg_iov set to the user abort reason.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001655 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001656 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1657 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 err = -EINVAL;
1659 goto out_nounlock;
1660 }
1661
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001662 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 * specified in msg_name.
1664 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001665 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 err = -EINVAL;
1667 goto out_nounlock;
1668 }
1669
1670 transport = NULL;
1671
1672 SCTP_DEBUG_PRINTK("About to look up association.\n");
1673
1674 sctp_lock_sock(sk);
1675
1676 /* If a msg_name has been specified, assume this is to be used. */
1677 if (msg_name) {
1678 /* Look for a matching association on the endpoint. */
Al Virodce116a2006-11-20 17:25:15 -08001679 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 if (!asoc) {
1681 /* If we could not find a matching association on the
1682 * endpoint, make sure that it is not a TCP-style
1683 * socket that already has an association or there is
1684 * no peeled-off association on another socket.
1685 */
1686 if ((sctp_style(sk, TCP) &&
1687 sctp_sstate(sk, ESTABLISHED)) ||
Al Virodce116a2006-11-20 17:25:15 -08001688 sctp_endpoint_is_peeled_off(ep, &to)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 err = -EADDRNOTAVAIL;
1690 goto out_unlock;
1691 }
1692 }
1693 } else {
1694 asoc = sctp_id2assoc(sk, associd);
1695 if (!asoc) {
1696 err = -EPIPE;
1697 goto out_unlock;
1698 }
1699 }
1700
1701 if (asoc) {
1702 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1703
1704 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1705 * socket that has an association in CLOSED state. This can
1706 * happen when an accepted socket has an association that is
1707 * already CLOSED.
1708 */
1709 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1710 err = -EPIPE;
1711 goto out_unlock;
1712 }
1713
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001714 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1716 asoc);
1717 sctp_primitive_SHUTDOWN(asoc, NULL);
1718 err = 0;
1719 goto out_unlock;
1720 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001721 if (sinfo_flags & SCTP_ABORT) {
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001722
1723 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1724 if (!chunk) {
1725 err = -ENOMEM;
1726 goto out_unlock;
1727 }
1728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001730 sctp_primitive_ABORT(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 err = 0;
1732 goto out_unlock;
1733 }
1734 }
1735
1736 /* Do we need to create the association? */
1737 if (!asoc) {
1738 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1739
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001740 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 err = -EINVAL;
1742 goto out_unlock;
1743 }
1744
1745 /* Check for invalid stream against the stream counts,
1746 * either the default or the user specified stream counts.
1747 */
1748 if (sinfo) {
1749 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1750 /* Check against the defaults. */
1751 if (sinfo->sinfo_stream >=
1752 sp->initmsg.sinit_num_ostreams) {
1753 err = -EINVAL;
1754 goto out_unlock;
1755 }
1756 } else {
1757 /* Check against the requested. */
1758 if (sinfo->sinfo_stream >=
1759 sinit->sinit_num_ostreams) {
1760 err = -EINVAL;
1761 goto out_unlock;
1762 }
1763 }
1764 }
1765
1766 /*
1767 * API 3.1.2 bind() - UDP Style Syntax
1768 * If a bind() or sctp_bindx() is not called prior to a
1769 * sendmsg() call that initiates a new association, the
1770 * system picks an ephemeral port and will choose an address
1771 * set equivalent to binding with a wildcard address.
1772 */
1773 if (!ep->base.bind_addr.port) {
1774 if (sctp_autobind(sk)) {
1775 err = -EAGAIN;
1776 goto out_unlock;
1777 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001778 } else {
1779 /*
1780 * If an unprivileged user inherits a one-to-many
1781 * style socket with open associations on a privileged
1782 * port, it MAY be permitted to accept new associations,
1783 * but it SHOULD NOT be permitted to open new
1784 * associations.
1785 */
1786 if (ep->base.bind_addr.port < PROT_SOCK &&
1787 !capable(CAP_NET_BIND_SERVICE)) {
1788 err = -EACCES;
1789 goto out_unlock;
1790 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 }
1792
1793 scope = sctp_scope(&to);
1794 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1795 if (!new_asoc) {
1796 err = -ENOMEM;
1797 goto out_unlock;
1798 }
1799 asoc = new_asoc;
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001800 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1801 if (err < 0) {
1802 err = -ENOMEM;
1803 goto out_free;
1804 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805
1806 /* If the SCTP_INIT ancillary data is specified, set all
1807 * the association init values accordingly.
1808 */
1809 if (sinit) {
1810 if (sinit->sinit_num_ostreams) {
1811 asoc->c.sinit_num_ostreams =
1812 sinit->sinit_num_ostreams;
1813 }
1814 if (sinit->sinit_max_instreams) {
1815 asoc->c.sinit_max_instreams =
1816 sinit->sinit_max_instreams;
1817 }
1818 if (sinit->sinit_max_attempts) {
1819 asoc->max_init_attempts
1820 = sinit->sinit_max_attempts;
1821 }
1822 if (sinit->sinit_max_init_timeo) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001823 asoc->max_init_timeo =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1825 }
1826 }
1827
1828 /* Prime the peer's transport structures. */
Al Virodce116a2006-11-20 17:25:15 -08001829 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 if (!transport) {
1831 err = -ENOMEM;
1832 goto out_free;
1833 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 }
1835
1836 /* ASSERT: we have a valid association at this point. */
1837 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1838
1839 if (!sinfo) {
1840 /* If the user didn't specify SNDRCVINFO, make up one with
1841 * some defaults.
1842 */
Joe Perches517aa0b2011-05-12 11:27:20 +00001843 memset(&default_sinfo, 0, sizeof(default_sinfo));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 default_sinfo.sinfo_stream = asoc->default_stream;
1845 default_sinfo.sinfo_flags = asoc->default_flags;
1846 default_sinfo.sinfo_ppid = asoc->default_ppid;
1847 default_sinfo.sinfo_context = asoc->default_context;
1848 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1849 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1850 sinfo = &default_sinfo;
1851 }
1852
1853 /* API 7.1.7, the sndbuf size per association bounds the
1854 * maximum size of data that can be sent in a single send call.
1855 */
1856 if (msg_len > sk->sk_sndbuf) {
1857 err = -EMSGSIZE;
1858 goto out_free;
1859 }
1860
Vlad Yasevich8a479492007-06-07 14:21:05 -04001861 if (asoc->pmtu_pending)
David S. Miller02f3d4c2012-07-16 03:57:14 -07001862 sctp_assoc_pending_pmtu(sk, asoc);
Vlad Yasevich8a479492007-06-07 14:21:05 -04001863
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 /* If fragmentation is disabled and the message length exceeds the
1865 * association fragmentation point, return EMSGSIZE. The I-D
1866 * does not specify what this error is, but this looks like
1867 * a great fit.
1868 */
1869 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1870 err = -EMSGSIZE;
1871 goto out_free;
1872 }
1873
Joe Perchesafd76142011-05-12 09:19:10 +00001874 /* Check for invalid stream. */
1875 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1876 err = -EINVAL;
1877 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 }
1879
1880 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1881 if (!sctp_wspace(asoc)) {
1882 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1883 if (err)
1884 goto out_free;
1885 }
1886
1887 /* If an address is passed with the sendto/sendmsg call, it is used
1888 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001889 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 */
1891 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001892 (sinfo_flags & SCTP_ADDR_OVER)) {
Al Virodce116a2006-11-20 17:25:15 -08001893 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 if (!chunk_tp) {
1895 err = -EINVAL;
1896 goto out_free;
1897 }
1898 } else
1899 chunk_tp = NULL;
1900
1901 /* Auto-connect, if we aren't connected already. */
1902 if (sctp_state(asoc, CLOSED)) {
1903 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1904 if (err < 0)
1905 goto out_free;
1906 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1907 }
1908
1909 /* Break the message into multiple chunks of maximum size. */
1910 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1911 if (!datamsg) {
1912 err = -ENOMEM;
1913 goto out_free;
1914 }
1915
1916 /* Now send the (possibly) fragmented message. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001917 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
Florian Westphal80445cf2008-03-23 22:47:08 -07001918 sctp_chunk_hold(chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
1920 /* Do accounting for the write space. */
1921 sctp_set_owner_w(chunk);
1922
1923 chunk->transport = chunk_tp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 }
1925
Vlad Yasevich9c5c62b2009-08-10 13:51:03 -04001926 /* Send it to the lower layers. Note: all chunks
1927 * must either fail or succeed. The lower layer
1928 * works that way today. Keep it that way or this
1929 * breaks.
1930 */
1931 err = sctp_primitive_SEND(asoc, datamsg);
1932 /* Did the lower layer accept the chunk? */
1933 if (err)
1934 sctp_datamsg_free(datamsg);
1935 else
1936 sctp_datamsg_put(datamsg);
1937
1938 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1939
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 if (err)
1941 goto out_free;
1942 else
1943 err = msg_len;
1944
1945 /* If we are already past ASSOCIATE, the lower
1946 * layers are responsible for association cleanup.
1947 */
1948 goto out_unlock;
1949
1950out_free:
Neil Horman2eebc1e2012-07-16 09:13:51 +00001951 if (new_asoc) {
1952 sctp_unhash_established(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 sctp_association_free(asoc);
Neil Horman2eebc1e2012-07-16 09:13:51 +00001954 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955out_unlock:
1956 sctp_release_sock(sk);
1957
1958out_nounlock:
1959 return sctp_error(sk, msg_flags, err);
1960
1961#if 0
1962do_sock_err:
1963 if (msg_len)
1964 err = msg_len;
1965 else
1966 err = sock_error(sk);
1967 goto out;
1968
1969do_interrupted:
1970 if (msg_len)
1971 err = msg_len;
1972 goto out;
1973#endif /* 0 */
1974}
1975
1976/* This is an extended version of skb_pull() that removes the data from the
1977 * start of a skb even when data is spread across the list of skb's in the
1978 * frag_list. len specifies the total amount of data that needs to be removed.
1979 * when 'len' bytes could be removed from the skb, it returns 0.
1980 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1981 * could not be removed.
1982 */
1983static int sctp_skb_pull(struct sk_buff *skb, int len)
1984{
1985 struct sk_buff *list;
1986 int skb_len = skb_headlen(skb);
1987 int rlen;
1988
1989 if (len <= skb_len) {
1990 __skb_pull(skb, len);
1991 return 0;
1992 }
1993 len -= skb_len;
1994 __skb_pull(skb, skb_len);
1995
David S. Miller1b003be2009-06-09 00:22:35 -07001996 skb_walk_frags(skb, list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 rlen = sctp_skb_pull(list, len);
1998 skb->len -= (len-rlen);
1999 skb->data_len -= (len-rlen);
2000
2001 if (!rlen)
2002 return 0;
2003
2004 len = rlen;
2005 }
2006
2007 return len;
2008}
2009
2010/* API 3.1.3 recvmsg() - UDP Style Syntax
2011 *
2012 * ssize_t recvmsg(int socket, struct msghdr *message,
2013 * int flags);
2014 *
2015 * socket - the socket descriptor of the endpoint.
2016 * message - pointer to the msghdr structure which contains a single
2017 * user message and possibly some ancillary data.
2018 *
2019 * See Section 5 for complete description of the data
2020 * structures.
2021 *
2022 * flags - flags sent or received with the user message, see Section
2023 * 5 for complete description of the flags.
2024 */
2025static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2026
2027SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2028 struct msghdr *msg, size_t len, int noblock,
2029 int flags, int *addr_len)
2030{
2031 struct sctp_ulpevent *event = NULL;
2032 struct sctp_sock *sp = sctp_sk(sk);
2033 struct sk_buff *skb;
2034 int copied;
2035 int err = 0;
2036 int skb_len;
2037
2038 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2039 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2040 "len", len, "knoblauch", noblock,
2041 "flags", flags, "addr_len", addr_len);
2042
2043 sctp_lock_sock(sk);
2044
2045 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2046 err = -ENOTCONN;
2047 goto out;
2048 }
2049
2050 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2051 if (!skb)
2052 goto out;
2053
2054 /* Get the total length of the skb including any skb's in the
2055 * frag_list.
2056 */
2057 skb_len = skb->len;
2058
2059 copied = skb_len;
2060 if (copied > len)
2061 copied = len;
2062
2063 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2064
2065 event = sctp_skb2event(skb);
2066
2067 if (err)
2068 goto out_free;
2069
Neil Horman3b885782009-10-12 13:26:31 -07002070 sock_recv_ts_and_drops(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 if (sctp_ulpevent_is_notification(event)) {
2072 msg->msg_flags |= MSG_NOTIFICATION;
2073 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2074 } else {
2075 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2076 }
2077
2078 /* Check if we allow SCTP_SNDRCVINFO. */
2079 if (sp->subscribe.sctp_data_io_event)
2080 sctp_ulpevent_read_sndrcvinfo(event, msg);
2081#if 0
2082 /* FIXME: we should be calling IP/IPv6 layers. */
2083 if (sk->sk_protinfo.af_inet.cmsg_flags)
2084 ip_cmsg_recv(msg, skb);
2085#endif
2086
2087 err = copied;
2088
2089 /* If skb's length exceeds the user's buffer, update the skb and
2090 * push it back to the receive_queue so that the next call to
2091 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2092 */
2093 if (skb_len > copied) {
2094 msg->msg_flags &= ~MSG_EOR;
2095 if (flags & MSG_PEEK)
2096 goto out_free;
2097 sctp_skb_pull(skb, copied);
2098 skb_queue_head(&sk->sk_receive_queue, skb);
2099
2100 /* When only partial message is copied to the user, increase
2101 * rwnd by that amount. If all the data in the skb is read,
2102 * rwnd is updated when the event is freed.
2103 */
Vlad Yasevich0eca8fe2008-01-11 10:12:56 -05002104 if (!sctp_ulpevent_is_notification(event))
2105 sctp_assoc_rwnd_increase(event->asoc, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 goto out;
2107 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2108 (event->msg_flags & MSG_EOR))
2109 msg->msg_flags |= MSG_EOR;
2110 else
2111 msg->msg_flags &= ~MSG_EOR;
2112
2113out_free:
2114 if (flags & MSG_PEEK) {
2115 /* Release the skb reference acquired after peeking the skb in
2116 * sctp_skb_recv_datagram().
2117 */
2118 kfree_skb(skb);
2119 } else {
2120 /* Free the event which includes releasing the reference to
2121 * the owner of the skb, freeing the skb and updating the
2122 * rwnd.
2123 */
2124 sctp_ulpevent_free(event);
2125 }
2126out:
2127 sctp_release_sock(sk);
2128 return err;
2129}
2130
2131/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2132 *
2133 * This option is a on/off flag. If enabled no SCTP message
2134 * fragmentation will be performed. Instead if a message being sent
2135 * exceeds the current PMTU size, the message will NOT be sent and
2136 * instead a error will be indicated to the user.
2137 */
2138static int sctp_setsockopt_disable_fragments(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002139 char __user *optval,
2140 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141{
2142 int val;
2143
2144 if (optlen < sizeof(int))
2145 return -EINVAL;
2146
2147 if (get_user(val, (int __user *)optval))
2148 return -EFAULT;
2149
2150 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2151
2152 return 0;
2153}
2154
2155static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002156 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157{
Wei Yongjun94912302011-07-02 09:28:04 +00002158 struct sctp_association *asoc;
2159 struct sctp_ulpevent *event;
2160
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05002161 if (optlen > sizeof(struct sctp_event_subscribe))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162 return -EINVAL;
2163 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2164 return -EFAULT;
Wei Yongjun94912302011-07-02 09:28:04 +00002165
2166 /*
2167 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2168 * if there is no data to be sent or retransmit, the stack will
2169 * immediately send up this notification.
2170 */
2171 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2172 &sctp_sk(sk)->subscribe)) {
2173 asoc = sctp_id2assoc(sk, 0);
2174
2175 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2176 event = sctp_ulpevent_make_sender_dry_event(asoc,
2177 GFP_ATOMIC);
2178 if (!event)
2179 return -ENOMEM;
2180
2181 sctp_ulpq_tail_event(&asoc->ulpq, event);
2182 }
2183 }
2184
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185 return 0;
2186}
2187
2188/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2189 *
2190 * This socket option is applicable to the UDP-style socket only. When
2191 * set it will cause associations that are idle for more than the
2192 * specified number of seconds to automatically close. An association
2193 * being idle is defined an association that has NOT sent or received
2194 * user data. The special value of '0' indicates that no automatic
2195 * close of any associations should be performed. The option expects an
2196 * integer defining the number of seconds of idle time before an
2197 * association is closed.
2198 */
2199static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002200 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201{
2202 struct sctp_sock *sp = sctp_sk(sk);
2203
2204 /* Applicable to UDP-style socket only */
2205 if (sctp_style(sk, TCP))
2206 return -EOPNOTSUPP;
2207 if (optlen != sizeof(int))
2208 return -EINVAL;
2209 if (copy_from_user(&sp->autoclose, optval, optlen))
2210 return -EFAULT;
2211
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 return 0;
2213}
2214
2215/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2216 *
2217 * Applications can enable or disable heartbeats for any peer address of
2218 * an association, modify an address's heartbeat interval, force a
2219 * heartbeat to be sent immediately, and adjust the address's maximum
2220 * number of retransmissions sent before an address is considered
2221 * unreachable. The following structure is used to access and modify an
2222 * address's parameters:
2223 *
2224 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002225 * sctp_assoc_t spp_assoc_id;
2226 * struct sockaddr_storage spp_address;
2227 * uint32_t spp_hbinterval;
2228 * uint16_t spp_pathmaxrxt;
2229 * uint32_t spp_pathmtu;
2230 * uint32_t spp_sackdelay;
2231 * uint32_t spp_flags;
2232 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002234 * spp_assoc_id - (one-to-many style socket) This is filled in the
2235 * application, and identifies the association for
2236 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 * spp_address - This specifies which address is of interest.
2238 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08002239 * in milliseconds. If a value of zero
2240 * is present in this field then no changes are to
2241 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 * spp_pathmaxrxt - This contains the maximum number of
2243 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08002244 * considered unreachable. If a value of zero
2245 * is present in this field then no changes are to
2246 * be made to this parameter.
2247 * spp_pathmtu - When Path MTU discovery is disabled the value
2248 * specified here will be the "fixed" path mtu.
2249 * Note that if the spp_address field is empty
2250 * then all associations on this address will
2251 * have this fixed path mtu set upon them.
2252 *
2253 * spp_sackdelay - When delayed sack is enabled, this value specifies
2254 * the number of milliseconds that sacks will be delayed
2255 * for. This value will apply to all addresses of an
2256 * association if the spp_address field is empty. Note
2257 * also, that if delayed sack is enabled and this
2258 * value is set to 0, no change is made to the last
2259 * recorded delayed sack timer value.
2260 *
2261 * spp_flags - These flags are used to control various features
2262 * on an association. The flag field may contain
2263 * zero or more of the following options.
2264 *
2265 * SPP_HB_ENABLE - Enable heartbeats on the
2266 * specified address. Note that if the address
2267 * field is empty all addresses for the association
2268 * have heartbeats enabled upon them.
2269 *
2270 * SPP_HB_DISABLE - Disable heartbeats on the
2271 * speicifed address. Note that if the address
2272 * field is empty all addresses for the association
2273 * will have their heartbeats disabled. Note also
2274 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2275 * mutually exclusive, only one of these two should
2276 * be specified. Enabling both fields will have
2277 * undetermined results.
2278 *
2279 * SPP_HB_DEMAND - Request a user initiated heartbeat
2280 * to be made immediately.
2281 *
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002282 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2283 * heartbeat delayis to be set to the value of 0
2284 * milliseconds.
2285 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002286 * SPP_PMTUD_ENABLE - This field will enable PMTU
2287 * discovery upon the specified address. Note that
2288 * if the address feild is empty then all addresses
2289 * on the association are effected.
2290 *
2291 * SPP_PMTUD_DISABLE - This field will disable PMTU
2292 * discovery upon the specified address. Note that
2293 * if the address feild is empty then all addresses
2294 * on the association are effected. Not also that
2295 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2296 * exclusive. Enabling both will have undetermined
2297 * results.
2298 *
2299 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2300 * on delayed sack. The time specified in spp_sackdelay
2301 * is used to specify the sack delay for this address. Note
2302 * that if spp_address is empty then all addresses will
2303 * enable delayed sack and take on the sack delay
2304 * value specified in spp_sackdelay.
2305 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2306 * off delayed sack. If the spp_address field is blank then
2307 * delayed sack is disabled for the entire association. Note
2308 * also that this field is mutually exclusive to
2309 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2310 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 */
Adrian Bunk16164362006-09-18 00:40:38 -07002312static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2313 struct sctp_transport *trans,
2314 struct sctp_association *asoc,
2315 struct sctp_sock *sp,
2316 int hb_change,
2317 int pmtud_change,
2318 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320 int error;
2321
Frank Filz52ccb8e2005-12-22 11:36:46 -08002322 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2324 if (error)
2325 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 }
2327
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002328 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2329 * this field is ignored. Note also that a value of zero indicates
2330 * the current setting should be left unchanged.
2331 */
2332 if (params->spp_flags & SPP_HB_ENABLE) {
2333
2334 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2335 * set. This lets us use 0 value when this flag
2336 * is set.
2337 */
2338 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2339 params->spp_hbinterval = 0;
2340
2341 if (params->spp_hbinterval ||
2342 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2343 if (trans) {
2344 trans->hbinterval =
2345 msecs_to_jiffies(params->spp_hbinterval);
2346 } else if (asoc) {
2347 asoc->hbinterval =
2348 msecs_to_jiffies(params->spp_hbinterval);
2349 } else {
2350 sp->hbinterval = params->spp_hbinterval;
2351 }
Frank Filz52ccb8e2005-12-22 11:36:46 -08002352 }
2353 }
2354
2355 if (hb_change) {
2356 if (trans) {
2357 trans->param_flags =
2358 (trans->param_flags & ~SPP_HB) | hb_change;
2359 } else if (asoc) {
2360 asoc->param_flags =
2361 (asoc->param_flags & ~SPP_HB) | hb_change;
2362 } else {
2363 sp->param_flags =
2364 (sp->param_flags & ~SPP_HB) | hb_change;
2365 }
2366 }
2367
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002368 /* When Path MTU discovery is disabled the value specified here will
2369 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2370 * include the flag SPP_PMTUD_DISABLE for this field to have any
2371 * effect).
2372 */
2373 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002374 if (trans) {
2375 trans->pathmtu = params->spp_pathmtu;
David S. Miller02f3d4c2012-07-16 03:57:14 -07002376 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002377 } else if (asoc) {
2378 asoc->pathmtu = params->spp_pathmtu;
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04002379 sctp_frag_point(asoc, params->spp_pathmtu);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002380 } else {
2381 sp->pathmtu = params->spp_pathmtu;
2382 }
2383 }
2384
2385 if (pmtud_change) {
2386 if (trans) {
2387 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2388 (params->spp_flags & SPP_PMTUD_ENABLE);
2389 trans->param_flags =
2390 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2391 if (update) {
Vlad Yasevich9914ae32011-04-26 21:51:31 +00002392 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
David S. Miller02f3d4c2012-07-16 03:57:14 -07002393 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002394 }
2395 } else if (asoc) {
2396 asoc->param_flags =
2397 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2398 } else {
2399 sp->param_flags =
2400 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2401 }
2402 }
2403
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002404 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2405 * value of this field is ignored. Note also that a value of zero
2406 * indicates the current setting should be left unchanged.
2407 */
2408 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002409 if (trans) {
2410 trans->sackdelay =
2411 msecs_to_jiffies(params->spp_sackdelay);
2412 } else if (asoc) {
2413 asoc->sackdelay =
2414 msecs_to_jiffies(params->spp_sackdelay);
2415 } else {
2416 sp->sackdelay = params->spp_sackdelay;
2417 }
2418 }
2419
2420 if (sackdelay_change) {
2421 if (trans) {
2422 trans->param_flags =
2423 (trans->param_flags & ~SPP_SACKDELAY) |
2424 sackdelay_change;
2425 } else if (asoc) {
2426 asoc->param_flags =
2427 (asoc->param_flags & ~SPP_SACKDELAY) |
2428 sackdelay_change;
2429 } else {
2430 sp->param_flags =
2431 (sp->param_flags & ~SPP_SACKDELAY) |
2432 sackdelay_change;
2433 }
2434 }
2435
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002436 /* Note that a value of zero indicates the current setting should be
2437 left unchanged.
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002438 */
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002439 if (params->spp_pathmaxrxt) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002440 if (trans) {
2441 trans->pathmaxrxt = params->spp_pathmaxrxt;
2442 } else if (asoc) {
2443 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2444 } else {
2445 sp->pathmaxrxt = params->spp_pathmaxrxt;
2446 }
2447 }
2448
2449 return 0;
2450}
2451
2452static int sctp_setsockopt_peer_addr_params(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002453 char __user *optval,
2454 unsigned int optlen)
Frank Filz52ccb8e2005-12-22 11:36:46 -08002455{
2456 struct sctp_paddrparams params;
2457 struct sctp_transport *trans = NULL;
2458 struct sctp_association *asoc = NULL;
2459 struct sctp_sock *sp = sctp_sk(sk);
2460 int error;
2461 int hb_change, pmtud_change, sackdelay_change;
2462
2463 if (optlen != sizeof(struct sctp_paddrparams))
2464 return - EINVAL;
2465
2466 if (copy_from_user(&params, optval, optlen))
2467 return -EFAULT;
2468
2469 /* Validate flags and value parameters. */
2470 hb_change = params.spp_flags & SPP_HB;
2471 pmtud_change = params.spp_flags & SPP_PMTUD;
2472 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2473
2474 if (hb_change == SPP_HB ||
2475 pmtud_change == SPP_PMTUD ||
2476 sackdelay_change == SPP_SACKDELAY ||
2477 params.spp_sackdelay > 500 ||
Joe Perchesf64f9e72009-11-29 16:55:45 -08002478 (params.spp_pathmtu &&
2479 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
Frank Filz52ccb8e2005-12-22 11:36:46 -08002480 return -EINVAL;
2481
2482 /* If an address other than INADDR_ANY is specified, and
2483 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04002485 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002486 trans = sctp_addr_id2transport(sk, &params.spp_address,
2487 params.spp_assoc_id);
2488 if (!trans)
2489 return -EINVAL;
2490 }
2491
2492 /* Get association, if assoc_id != 0 and the socket is a one
2493 * to many style socket, and an association was not found, then
2494 * the id was invalid.
2495 */
2496 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2497 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2498 return -EINVAL;
2499
2500 /* Heartbeat demand can only be sent on a transport or
2501 * association, but not a socket.
2502 */
2503 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2504 return -EINVAL;
2505
2506 /* Process parameters. */
2507 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2508 hb_change, pmtud_change,
2509 sackdelay_change);
2510
2511 if (error)
2512 return error;
2513
2514 /* If changes are for association, also apply parameters to each
2515 * transport.
2516 */
2517 if (!trans && asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002518 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2519 transports) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002520 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2521 hb_change, pmtud_change,
2522 sackdelay_change);
2523 }
2524 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525
2526 return 0;
2527}
2528
Wei Yongjund364d922008-05-09 15:13:26 -07002529/*
2530 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08002531 *
Wei Yongjund364d922008-05-09 15:13:26 -07002532 * This option will effect the way delayed acks are performed. This
2533 * option allows you to get or set the delayed ack time, in
2534 * milliseconds. It also allows changing the delayed ack frequency.
2535 * Changing the frequency to 1 disables the delayed sack algorithm. If
2536 * the assoc_id is 0, then this sets or gets the endpoints default
2537 * values. If the assoc_id field is non-zero, then the set or get
2538 * effects the specified association for the one to many model (the
2539 * assoc_id field is ignored by the one to one model). Note that if
2540 * sack_delay or sack_freq are 0 when setting this option, then the
2541 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08002542 *
Wei Yongjund364d922008-05-09 15:13:26 -07002543 * struct sctp_sack_info {
2544 * sctp_assoc_t sack_assoc_id;
2545 * uint32_t sack_delay;
2546 * uint32_t sack_freq;
2547 * };
Frank Filz77086102005-12-22 11:37:30 -08002548 *
Wei Yongjund364d922008-05-09 15:13:26 -07002549 * sack_assoc_id - This parameter, indicates which association the user
2550 * is performing an action upon. Note that if this field's value is
2551 * zero then the endpoints default value is changed (effecting future
2552 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08002553 *
Wei Yongjund364d922008-05-09 15:13:26 -07002554 * sack_delay - This parameter contains the number of milliseconds that
2555 * the user is requesting the delayed ACK timer be set to. Note that
2556 * this value is defined in the standard to be between 200 and 500
2557 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08002558 *
Wei Yongjund364d922008-05-09 15:13:26 -07002559 * sack_freq - This parameter contains the number of packets that must
2560 * be received before a sack is sent without waiting for the delay
2561 * timer to expire. The default value for this is 2, setting this
2562 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08002563 */
2564
Wei Yongjund364d922008-05-09 15:13:26 -07002565static int sctp_setsockopt_delayed_ack(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002566 char __user *optval, unsigned int optlen)
Frank Filz77086102005-12-22 11:37:30 -08002567{
Wei Yongjund364d922008-05-09 15:13:26 -07002568 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08002569 struct sctp_transport *trans = NULL;
2570 struct sctp_association *asoc = NULL;
2571 struct sctp_sock *sp = sctp_sk(sk);
2572
Wei Yongjund364d922008-05-09 15:13:26 -07002573 if (optlen == sizeof(struct sctp_sack_info)) {
2574 if (copy_from_user(&params, optval, optlen))
2575 return -EFAULT;
2576
2577 if (params.sack_delay == 0 && params.sack_freq == 0)
2578 return 0;
2579 } else if (optlen == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00002580 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2581 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07002582 if (copy_from_user(&params, optval, optlen))
2583 return -EFAULT;
2584
2585 if (params.sack_delay == 0)
2586 params.sack_freq = 1;
2587 else
2588 params.sack_freq = 0;
2589 } else
Frank Filz77086102005-12-22 11:37:30 -08002590 return - EINVAL;
2591
Frank Filz77086102005-12-22 11:37:30 -08002592 /* Validate value parameter. */
Wei Yongjund364d922008-05-09 15:13:26 -07002593 if (params.sack_delay > 500)
Frank Filz77086102005-12-22 11:37:30 -08002594 return -EINVAL;
2595
Wei Yongjund364d922008-05-09 15:13:26 -07002596 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08002597 * to many style socket, and an association was not found, then
2598 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002599 */
Wei Yongjund364d922008-05-09 15:13:26 -07002600 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2601 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08002602 return -EINVAL;
2603
Wei Yongjund364d922008-05-09 15:13:26 -07002604 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002605 if (asoc) {
2606 asoc->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002607 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002608 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002609 (asoc->param_flags & ~SPP_SACKDELAY) |
2610 SPP_SACKDELAY_ENABLE;
2611 } else {
Wei Yongjund364d922008-05-09 15:13:26 -07002612 sp->sackdelay = params.sack_delay;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002613 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002614 (sp->param_flags & ~SPP_SACKDELAY) |
2615 SPP_SACKDELAY_ENABLE;
2616 }
Wei Yongjund364d922008-05-09 15:13:26 -07002617 }
2618
2619 if (params.sack_freq == 1) {
Frank Filz77086102005-12-22 11:37:30 -08002620 if (asoc) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002621 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002622 (asoc->param_flags & ~SPP_SACKDELAY) |
2623 SPP_SACKDELAY_DISABLE;
2624 } else {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002625 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002626 (sp->param_flags & ~SPP_SACKDELAY) |
2627 SPP_SACKDELAY_DISABLE;
2628 }
Wei Yongjund364d922008-05-09 15:13:26 -07002629 } else if (params.sack_freq > 1) {
2630 if (asoc) {
2631 asoc->sackfreq = params.sack_freq;
2632 asoc->param_flags =
2633 (asoc->param_flags & ~SPP_SACKDELAY) |
2634 SPP_SACKDELAY_ENABLE;
2635 } else {
2636 sp->sackfreq = params.sack_freq;
2637 sp->param_flags =
2638 (sp->param_flags & ~SPP_SACKDELAY) |
2639 SPP_SACKDELAY_ENABLE;
2640 }
Frank Filz77086102005-12-22 11:37:30 -08002641 }
2642
2643 /* If change is for association, also apply to each transport. */
2644 if (asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002645 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2646 transports) {
Wei Yongjund364d922008-05-09 15:13:26 -07002647 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002648 trans->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002649 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002650 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002651 (trans->param_flags & ~SPP_SACKDELAY) |
2652 SPP_SACKDELAY_ENABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002653 }
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07002654 if (params.sack_freq == 1) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002655 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002656 (trans->param_flags & ~SPP_SACKDELAY) |
2657 SPP_SACKDELAY_DISABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002658 } else if (params.sack_freq > 1) {
2659 trans->sackfreq = params.sack_freq;
2660 trans->param_flags =
2661 (trans->param_flags & ~SPP_SACKDELAY) |
2662 SPP_SACKDELAY_ENABLE;
Frank Filz77086102005-12-22 11:37:30 -08002663 }
2664 }
2665 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002666
Frank Filz77086102005-12-22 11:37:30 -08002667 return 0;
2668}
2669
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2671 *
2672 * Applications can specify protocol parameters for the default association
2673 * initialization. The option name argument to setsockopt() and getsockopt()
2674 * is SCTP_INITMSG.
2675 *
2676 * Setting initialization parameters is effective only on an unconnected
2677 * socket (for UDP-style sockets only future associations are effected
2678 * by the change). With TCP-style sockets, this option is inherited by
2679 * sockets derived from a listener socket.
2680 */
David S. Millerb7058842009-09-30 16:12:20 -07002681static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002682{
2683 struct sctp_initmsg sinit;
2684 struct sctp_sock *sp = sctp_sk(sk);
2685
2686 if (optlen != sizeof(struct sctp_initmsg))
2687 return -EINVAL;
2688 if (copy_from_user(&sinit, optval, optlen))
2689 return -EFAULT;
2690
2691 if (sinit.sinit_num_ostreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002692 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002693 if (sinit.sinit_max_instreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002694 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002695 if (sinit.sinit_max_attempts)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002696 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697 if (sinit.sinit_max_init_timeo)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002698 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002699
2700 return 0;
2701}
2702
2703/*
2704 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2705 *
2706 * Applications that wish to use the sendto() system call may wish to
2707 * specify a default set of parameters that would normally be supplied
2708 * through the inclusion of ancillary data. This socket option allows
2709 * such an application to set the default sctp_sndrcvinfo structure.
2710 * The application that wishes to use this socket option simply passes
2711 * in to this call the sctp_sndrcvinfo structure defined in Section
2712 * 5.2.2) The input parameters accepted by this call include
2713 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2714 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2715 * to this call if the caller is using the UDP model.
2716 */
2717static int sctp_setsockopt_default_send_param(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002718 char __user *optval,
2719 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002720{
2721 struct sctp_sndrcvinfo info;
2722 struct sctp_association *asoc;
2723 struct sctp_sock *sp = sctp_sk(sk);
2724
2725 if (optlen != sizeof(struct sctp_sndrcvinfo))
2726 return -EINVAL;
2727 if (copy_from_user(&info, optval, optlen))
2728 return -EFAULT;
2729
2730 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2731 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2732 return -EINVAL;
2733
2734 if (asoc) {
2735 asoc->default_stream = info.sinfo_stream;
2736 asoc->default_flags = info.sinfo_flags;
2737 asoc->default_ppid = info.sinfo_ppid;
2738 asoc->default_context = info.sinfo_context;
2739 asoc->default_timetolive = info.sinfo_timetolive;
2740 } else {
2741 sp->default_stream = info.sinfo_stream;
2742 sp->default_flags = info.sinfo_flags;
2743 sp->default_ppid = info.sinfo_ppid;
2744 sp->default_context = info.sinfo_context;
2745 sp->default_timetolive = info.sinfo_timetolive;
2746 }
2747
2748 return 0;
2749}
2750
2751/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2752 *
2753 * Requests that the local SCTP stack use the enclosed peer address as
2754 * the association primary. The enclosed address must be one of the
2755 * association peer's addresses.
2756 */
2757static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002758 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002759{
2760 struct sctp_prim prim;
2761 struct sctp_transport *trans;
2762
2763 if (optlen != sizeof(struct sctp_prim))
2764 return -EINVAL;
2765
2766 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2767 return -EFAULT;
2768
2769 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2770 if (!trans)
2771 return -EINVAL;
2772
2773 sctp_assoc_set_primary(trans->asoc, trans);
2774
2775 return 0;
2776}
2777
2778/*
2779 * 7.1.5 SCTP_NODELAY
2780 *
2781 * Turn on/off any Nagle-like algorithm. This means that packets are
2782 * generally sent as soon as possible and no unnecessary delays are
2783 * introduced, at the cost of more packets in the network. Expects an
2784 * integer boolean flag.
2785 */
2786static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002787 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002788{
2789 int val;
2790
2791 if (optlen < sizeof(int))
2792 return -EINVAL;
2793 if (get_user(val, (int __user *)optval))
2794 return -EFAULT;
2795
2796 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2797 return 0;
2798}
2799
2800/*
2801 *
2802 * 7.1.1 SCTP_RTOINFO
2803 *
2804 * The protocol parameters used to initialize and bound retransmission
2805 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2806 * and modify these parameters.
2807 * All parameters are time values, in milliseconds. A value of 0, when
2808 * modifying the parameters, indicates that the current value should not
2809 * be changed.
2810 *
2811 */
David S. Millerb7058842009-09-30 16:12:20 -07002812static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2813{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002814 struct sctp_rtoinfo rtoinfo;
2815 struct sctp_association *asoc;
2816
2817 if (optlen != sizeof (struct sctp_rtoinfo))
2818 return -EINVAL;
2819
2820 if (copy_from_user(&rtoinfo, optval, optlen))
2821 return -EFAULT;
2822
2823 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2824
2825 /* Set the values to the specific association */
2826 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2827 return -EINVAL;
2828
2829 if (asoc) {
2830 if (rtoinfo.srto_initial != 0)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002831 asoc->rto_initial =
Linus Torvalds1da177e2005-04-16 15:20:36 -07002832 msecs_to_jiffies(rtoinfo.srto_initial);
2833 if (rtoinfo.srto_max != 0)
2834 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2835 if (rtoinfo.srto_min != 0)
2836 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2837 } else {
2838 /* If there is no association or the association-id = 0
2839 * set the values to the endpoint.
2840 */
2841 struct sctp_sock *sp = sctp_sk(sk);
2842
2843 if (rtoinfo.srto_initial != 0)
2844 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2845 if (rtoinfo.srto_max != 0)
2846 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2847 if (rtoinfo.srto_min != 0)
2848 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2849 }
2850
2851 return 0;
2852}
2853
2854/*
2855 *
2856 * 7.1.2 SCTP_ASSOCINFO
2857 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02002858 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859 * of the association.
2860 * Returns an error if the new association retransmission value is
2861 * greater than the sum of the retransmission value of the peer.
2862 * See [SCTP] for more information.
2863 *
2864 */
David S. Millerb7058842009-09-30 16:12:20 -07002865static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002866{
2867
2868 struct sctp_assocparams assocparams;
2869 struct sctp_association *asoc;
2870
2871 if (optlen != sizeof(struct sctp_assocparams))
2872 return -EINVAL;
2873 if (copy_from_user(&assocparams, optval, optlen))
2874 return -EFAULT;
2875
2876 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2877
2878 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2879 return -EINVAL;
2880
2881 /* Set the values to the specific association */
2882 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002883 if (assocparams.sasoc_asocmaxrxt != 0) {
2884 __u32 path_sum = 0;
2885 int paths = 0;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002886 struct sctp_transport *peer_addr;
2887
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002888 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2889 transports) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002890 path_sum += peer_addr->pathmaxrxt;
2891 paths++;
2892 }
2893
Frederik Schwarzer025dfda2008-10-16 19:02:37 +02002894 /* Only validate asocmaxrxt if we have more than
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002895 * one path/transport. We do this because path
2896 * retransmissions are only counted when we have more
2897 * then one path.
2898 */
2899 if (paths > 1 &&
2900 assocparams.sasoc_asocmaxrxt > path_sum)
2901 return -EINVAL;
2902
Linus Torvalds1da177e2005-04-16 15:20:36 -07002903 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002904 }
2905
Linus Torvalds1da177e2005-04-16 15:20:36 -07002906 if (assocparams.sasoc_cookie_life != 0) {
2907 asoc->cookie_life.tv_sec =
2908 assocparams.sasoc_cookie_life / 1000;
2909 asoc->cookie_life.tv_usec =
2910 (assocparams.sasoc_cookie_life % 1000)
2911 * 1000;
2912 }
2913 } else {
2914 /* Set the values to the endpoint */
2915 struct sctp_sock *sp = sctp_sk(sk);
2916
2917 if (assocparams.sasoc_asocmaxrxt != 0)
2918 sp->assocparams.sasoc_asocmaxrxt =
2919 assocparams.sasoc_asocmaxrxt;
2920 if (assocparams.sasoc_cookie_life != 0)
2921 sp->assocparams.sasoc_cookie_life =
2922 assocparams.sasoc_cookie_life;
2923 }
2924 return 0;
2925}
2926
2927/*
2928 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2929 *
2930 * This socket option is a boolean flag which turns on or off mapped V4
2931 * addresses. If this option is turned on and the socket is type
2932 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2933 * If this option is turned off, then no mapping will be done of V4
2934 * addresses and a user will receive both PF_INET6 and PF_INET type
2935 * addresses on the socket.
2936 */
David S. Millerb7058842009-09-30 16:12:20 -07002937static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002938{
2939 int val;
2940 struct sctp_sock *sp = sctp_sk(sk);
2941
2942 if (optlen < sizeof(int))
2943 return -EINVAL;
2944 if (get_user(val, (int __user *)optval))
2945 return -EFAULT;
2946 if (val)
2947 sp->v4mapped = 1;
2948 else
2949 sp->v4mapped = 0;
2950
2951 return 0;
2952}
2953
2954/*
Wei Yongjune89c2092008-12-25 16:54:58 -08002955 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2956 * This option will get or set the maximum size to put in any outgoing
2957 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07002958 * fragmented by SCTP into the specified size. Note that the underlying
2959 * SCTP implementation may fragment into smaller sized chunks when the
2960 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08002961 * the user. The default value for this option is '0' which indicates
2962 * the user is NOT limiting fragmentation and only the PMTU will effect
2963 * SCTP's choice of DATA chunk size. Note also that values set larger
2964 * than the maximum size of an IP datagram will effectively let SCTP
2965 * control fragmentation (i.e. the same as setting this option to 0).
2966 *
2967 * The following structure is used to access and modify this parameter:
2968 *
2969 * struct sctp_assoc_value {
2970 * sctp_assoc_t assoc_id;
2971 * uint32_t assoc_value;
2972 * };
2973 *
2974 * assoc_id: This parameter is ignored for one-to-one style sockets.
2975 * For one-to-many style sockets this parameter indicates which
2976 * association the user is performing an action upon. Note that if
2977 * this field's value is zero then the endpoints default value is
2978 * changed (effecting future associations only).
2979 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980 */
David S. Millerb7058842009-09-30 16:12:20 -07002981static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002982{
Wei Yongjune89c2092008-12-25 16:54:58 -08002983 struct sctp_assoc_value params;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002984 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002985 struct sctp_sock *sp = sctp_sk(sk);
2986 int val;
2987
Wei Yongjune89c2092008-12-25 16:54:58 -08002988 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00002989 pr_warn("Use of int in maxseg socket option deprecated\n");
2990 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08002991 if (copy_from_user(&val, optval, optlen))
2992 return -EFAULT;
2993 params.assoc_id = 0;
2994 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2995 if (copy_from_user(&params, optval, optlen))
2996 return -EFAULT;
2997 val = params.assoc_value;
2998 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999 return -EINVAL;
Wei Yongjune89c2092008-12-25 16:54:58 -08003000
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07003001 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003002 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003
Wei Yongjune89c2092008-12-25 16:54:58 -08003004 asoc = sctp_id2assoc(sk, params.assoc_id);
3005 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3006 return -EINVAL;
3007
3008 if (asoc) {
3009 if (val == 0) {
3010 val = asoc->pathmtu;
3011 val -= sp->pf->af->net_header_len;
3012 val -= sizeof(struct sctphdr) +
3013 sizeof(struct sctp_data_chunk);
3014 }
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04003015 asoc->user_frag = val;
3016 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
Wei Yongjune89c2092008-12-25 16:54:58 -08003017 } else {
3018 sp->user_frag = val;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019 }
3020
3021 return 0;
3022}
3023
3024
3025/*
3026 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3027 *
3028 * Requests that the peer mark the enclosed address as the association
3029 * primary. The enclosed address must be one of the association's
3030 * locally bound addresses. The following structure is used to make a
3031 * set primary request:
3032 */
3033static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003034 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003035{
3036 struct sctp_sock *sp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003037 struct sctp_association *asoc = NULL;
3038 struct sctp_setpeerprim prim;
3039 struct sctp_chunk *chunk;
Wei Yongjun40a01032010-12-07 17:11:09 +00003040 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003041 int err;
3042
3043 sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044
3045 if (!sctp_addip_enable)
3046 return -EPERM;
3047
3048 if (optlen != sizeof(struct sctp_setpeerprim))
3049 return -EINVAL;
3050
3051 if (copy_from_user(&prim, optval, optlen))
3052 return -EFAULT;
3053
3054 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003055 if (!asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003056 return -EINVAL;
3057
3058 if (!asoc->peer.asconf_capable)
3059 return -EPERM;
3060
3061 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3062 return -EPERM;
3063
3064 if (!sctp_state(asoc, ESTABLISHED))
3065 return -ENOTCONN;
3066
Wei Yongjun40a01032010-12-07 17:11:09 +00003067 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3068 if (!af)
3069 return -EINVAL;
3070
3071 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3072 return -EADDRNOTAVAIL;
3073
Linus Torvalds1da177e2005-04-16 15:20:36 -07003074 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3075 return -EADDRNOTAVAIL;
3076
3077 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3078 chunk = sctp_make_asconf_set_prim(asoc,
3079 (union sctp_addr *)&prim.sspp_addr);
3080 if (!chunk)
3081 return -ENOMEM;
3082
3083 err = sctp_send_asconf(asoc, chunk);
3084
3085 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3086
3087 return err;
3088}
3089
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003090static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003091 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003092{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003093 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003094
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003095 if (optlen != sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003096 return -EINVAL;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003097 if (copy_from_user(&adaptation, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003098 return -EFAULT;
3099
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003100 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003101
3102 return 0;
3103}
3104
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003105/*
3106 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3107 *
3108 * The context field in the sctp_sndrcvinfo structure is normally only
3109 * used when a failed message is retrieved holding the value that was
3110 * sent down on the actual send call. This option allows the setting of
3111 * a default context on an association basis that will be received on
3112 * reading messages from the peer. This is especially helpful in the
3113 * one-2-many model for an application to keep some reference to an
3114 * internal state machine that is processing messages on the
3115 * association. Note that the setting of this value only effects
3116 * received messages from the peer and does not effect the value that is
3117 * saved with outbound messages.
3118 */
3119static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003120 unsigned int optlen)
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003121{
3122 struct sctp_assoc_value params;
3123 struct sctp_sock *sp;
3124 struct sctp_association *asoc;
3125
3126 if (optlen != sizeof(struct sctp_assoc_value))
3127 return -EINVAL;
3128 if (copy_from_user(&params, optval, optlen))
3129 return -EFAULT;
3130
3131 sp = sctp_sk(sk);
3132
3133 if (params.assoc_id != 0) {
3134 asoc = sctp_id2assoc(sk, params.assoc_id);
3135 if (!asoc)
3136 return -EINVAL;
3137 asoc->default_rcv_context = params.assoc_value;
3138 } else {
3139 sp->default_rcv_context = params.assoc_value;
3140 }
3141
3142 return 0;
3143}
3144
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003145/*
3146 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3147 *
3148 * This options will at a minimum specify if the implementation is doing
3149 * fragmented interleave. Fragmented interleave, for a one to many
3150 * socket, is when subsequent calls to receive a message may return
3151 * parts of messages from different associations. Some implementations
3152 * may allow you to turn this value on or off. If so, when turned off,
3153 * no fragment interleave will occur (which will cause a head of line
3154 * blocking amongst multiple associations sharing the same one to many
3155 * socket). When this option is turned on, then each receive call may
3156 * come from a different association (thus the user must receive data
3157 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3158 * association each receive belongs to.
3159 *
3160 * This option takes a boolean value. A non-zero value indicates that
3161 * fragmented interleave is on. A value of zero indicates that
3162 * fragmented interleave is off.
3163 *
3164 * Note that it is important that an implementation that allows this
3165 * option to be turned on, have it off by default. Otherwise an unaware
3166 * application using the one to many model may become confused and act
3167 * incorrectly.
3168 */
3169static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3170 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003171 unsigned int optlen)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003172{
3173 int val;
3174
3175 if (optlen != sizeof(int))
3176 return -EINVAL;
3177 if (get_user(val, (int __user *)optval))
3178 return -EFAULT;
3179
3180 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3181
3182 return 0;
3183}
3184
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003185/*
Wei Yongjun8510b932008-12-25 16:59:03 -08003186 * 8.1.21. Set or Get the SCTP Partial Delivery Point
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003187 * (SCTP_PARTIAL_DELIVERY_POINT)
Wei Yongjun8510b932008-12-25 16:59:03 -08003188 *
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003189 * This option will set or get the SCTP partial delivery point. This
3190 * point is the size of a message where the partial delivery API will be
3191 * invoked to help free up rwnd space for the peer. Setting this to a
Wei Yongjun8510b932008-12-25 16:59:03 -08003192 * lower value will cause partial deliveries to happen more often. The
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003193 * calls argument is an integer that sets or gets the partial delivery
Wei Yongjun8510b932008-12-25 16:59:03 -08003194 * point. Note also that the call will fail if the user attempts to set
3195 * this value larger than the socket receive buffer size.
3196 *
3197 * Note that any single message having a length smaller than or equal to
3198 * the SCTP partial delivery point will be delivered in one single read
3199 * call as long as the user provided buffer is large enough to hold the
3200 * message.
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003201 */
3202static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3203 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003204 unsigned int optlen)
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003205{
3206 u32 val;
3207
3208 if (optlen != sizeof(u32))
3209 return -EINVAL;
3210 if (get_user(val, (int __user *)optval))
3211 return -EFAULT;
3212
Wei Yongjun8510b932008-12-25 16:59:03 -08003213 /* Note: We double the receive buffer from what the user sets
3214 * it to be, also initial rwnd is based on rcvbuf/2.
3215 */
3216 if (val > (sk->sk_rcvbuf >> 1))
3217 return -EINVAL;
3218
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003219 sctp_sk(sk)->pd_point = val;
3220
3221 return 0; /* is this the right error code? */
3222}
3223
Vlad Yasevich70331572007-03-23 11:34:36 -07003224/*
3225 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3226 *
3227 * This option will allow a user to change the maximum burst of packets
3228 * that can be emitted by this association. Note that the default value
3229 * is 4, and some implementations may restrict this setting so that it
3230 * can only be lowered.
3231 *
3232 * NOTE: This text doesn't seem right. Do this on a socket basis with
3233 * future associations inheriting the socket value.
3234 */
3235static int sctp_setsockopt_maxburst(struct sock *sk,
3236 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003237 unsigned int optlen)
Vlad Yasevich70331572007-03-23 11:34:36 -07003238{
Neil Horman219b99a2008-03-05 13:44:46 -08003239 struct sctp_assoc_value params;
3240 struct sctp_sock *sp;
3241 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07003242 int val;
Neil Horman219b99a2008-03-05 13:44:46 -08003243 int assoc_id = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003244
Neil Horman219b99a2008-03-05 13:44:46 -08003245 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00003246 pr_warn("Use of int in max_burst socket option deprecated\n");
3247 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08003248 if (copy_from_user(&val, optval, optlen))
3249 return -EFAULT;
3250 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3251 if (copy_from_user(&params, optval, optlen))
3252 return -EFAULT;
3253 val = params.assoc_value;
3254 assoc_id = params.assoc_id;
3255 } else
3256 return -EINVAL;
3257
3258 sp = sctp_sk(sk);
3259
3260 if (assoc_id != 0) {
3261 asoc = sctp_id2assoc(sk, assoc_id);
3262 if (!asoc)
3263 return -EINVAL;
3264 asoc->max_burst = val;
3265 } else
3266 sp->max_burst = val;
Vlad Yasevich70331572007-03-23 11:34:36 -07003267
3268 return 0;
3269}
3270
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003271/*
3272 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3273 *
3274 * This set option adds a chunk type that the user is requesting to be
3275 * received only in an authenticated way. Changes to the list of chunks
3276 * will only effect future associations on the socket.
3277 */
3278static int sctp_setsockopt_auth_chunk(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003279 char __user *optval,
3280 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003281{
3282 struct sctp_authchunk val;
3283
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003284 if (!sctp_auth_enable)
3285 return -EACCES;
3286
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003287 if (optlen != sizeof(struct sctp_authchunk))
3288 return -EINVAL;
3289 if (copy_from_user(&val, optval, optlen))
3290 return -EFAULT;
3291
3292 switch (val.sauth_chunk) {
Joe Perches7fd71b12011-07-01 09:43:11 +00003293 case SCTP_CID_INIT:
3294 case SCTP_CID_INIT_ACK:
3295 case SCTP_CID_SHUTDOWN_COMPLETE:
3296 case SCTP_CID_AUTH:
3297 return -EINVAL;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003298 }
3299
3300 /* add this chunk id to the endpoint */
3301 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3302}
3303
3304/*
3305 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3306 *
3307 * This option gets or sets the list of HMAC algorithms that the local
3308 * endpoint requires the peer to use.
3309 */
3310static int sctp_setsockopt_hmac_ident(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003311 char __user *optval,
3312 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003313{
3314 struct sctp_hmacalgo *hmacs;
Vlad Yasevichd9724052008-08-27 16:09:49 -07003315 u32 idents;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003316 int err;
3317
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003318 if (!sctp_auth_enable)
3319 return -EACCES;
3320
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003321 if (optlen < sizeof(struct sctp_hmacalgo))
3322 return -EINVAL;
3323
Shan Wei934253a2011-04-18 19:13:18 +00003324 hmacs= memdup_user(optval, optlen);
3325 if (IS_ERR(hmacs))
3326 return PTR_ERR(hmacs);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003327
Vlad Yasevichd9724052008-08-27 16:09:49 -07003328 idents = hmacs->shmac_num_idents;
3329 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3330 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003331 err = -EINVAL;
3332 goto out;
3333 }
3334
3335 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3336out:
3337 kfree(hmacs);
3338 return err;
3339}
3340
3341/*
3342 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3343 *
3344 * This option will set a shared secret key which is used to build an
3345 * association shared key.
3346 */
3347static int sctp_setsockopt_auth_key(struct sock *sk,
3348 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003349 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003350{
3351 struct sctp_authkey *authkey;
3352 struct sctp_association *asoc;
3353 int ret;
3354
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003355 if (!sctp_auth_enable)
3356 return -EACCES;
3357
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003358 if (optlen <= sizeof(struct sctp_authkey))
3359 return -EINVAL;
3360
Shan Wei934253a2011-04-18 19:13:18 +00003361 authkey= memdup_user(optval, optlen);
3362 if (IS_ERR(authkey))
3363 return PTR_ERR(authkey);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003364
Vlad Yasevich328fc472008-08-27 16:08:54 -07003365 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
Vlad Yasevich30c22352008-08-25 15:16:19 -07003366 ret = -EINVAL;
3367 goto out;
3368 }
3369
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003370 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3371 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3372 ret = -EINVAL;
3373 goto out;
3374 }
3375
3376 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3377out:
3378 kfree(authkey);
3379 return ret;
3380}
3381
3382/*
3383 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3384 *
3385 * This option will get or set the active shared key to be used to build
3386 * the association shared key.
3387 */
3388static int sctp_setsockopt_active_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003389 char __user *optval,
3390 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003391{
3392 struct sctp_authkeyid val;
3393 struct sctp_association *asoc;
3394
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003395 if (!sctp_auth_enable)
3396 return -EACCES;
3397
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003398 if (optlen != sizeof(struct sctp_authkeyid))
3399 return -EINVAL;
3400 if (copy_from_user(&val, optval, optlen))
3401 return -EFAULT;
3402
3403 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3404 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3405 return -EINVAL;
3406
3407 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3408 val.scact_keynumber);
3409}
3410
3411/*
3412 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3413 *
3414 * This set option will delete a shared secret key from use.
3415 */
3416static int sctp_setsockopt_del_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003417 char __user *optval,
3418 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003419{
3420 struct sctp_authkeyid val;
3421 struct sctp_association *asoc;
3422
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003423 if (!sctp_auth_enable)
3424 return -EACCES;
3425
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003426 if (optlen != sizeof(struct sctp_authkeyid))
3427 return -EINVAL;
3428 if (copy_from_user(&val, optval, optlen))
3429 return -EFAULT;
3430
3431 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3432 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3433 return -EINVAL;
3434
3435 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3436 val.scact_keynumber);
3437
3438}
3439
Michio Honda7dc04d72011-04-26 20:16:31 +09003440/*
3441 * 8.1.23 SCTP_AUTO_ASCONF
3442 *
3443 * This option will enable or disable the use of the automatic generation of
3444 * ASCONF chunks to add and delete addresses to an existing association. Note
3445 * that this option has two caveats namely: a) it only affects sockets that
3446 * are bound to all addresses available to the SCTP stack, and b) the system
3447 * administrator may have an overriding control that turns the ASCONF feature
3448 * off no matter what setting the socket option may have.
3449 * This option expects an integer boolean flag, where a non-zero value turns on
3450 * the option, and a zero value turns off the option.
3451 * Note. In this implementation, socket operation overrides default parameter
3452 * being set by sysctl as well as FreeBSD implementation
3453 */
3454static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3455 unsigned int optlen)
3456{
3457 int val;
3458 struct sctp_sock *sp = sctp_sk(sk);
3459
3460 if (optlen < sizeof(int))
3461 return -EINVAL;
3462 if (get_user(val, (int __user *)optval))
3463 return -EFAULT;
3464 if (!sctp_is_ep_boundall(sk) && val)
3465 return -EINVAL;
3466 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3467 return 0;
3468
3469 if (val == 0 && sp->do_auto_asconf) {
3470 list_del(&sp->auto_asconf_list);
3471 sp->do_auto_asconf = 0;
3472 } else if (val && !sp->do_auto_asconf) {
3473 list_add_tail(&sp->auto_asconf_list,
3474 &sctp_auto_asconf_splist);
3475 sp->do_auto_asconf = 1;
3476 }
3477 return 0;
3478}
3479
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003480
Neil Horman5aa93bc2012-07-21 07:56:07 +00003481/*
3482 * SCTP_PEER_ADDR_THLDS
3483 *
3484 * This option allows us to alter the partially failed threshold for one or all
3485 * transports in an association. See Section 6.1 of:
3486 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3487 */
3488static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3489 char __user *optval,
3490 unsigned int optlen)
3491{
3492 struct sctp_paddrthlds val;
3493 struct sctp_transport *trans;
3494 struct sctp_association *asoc;
3495
3496 if (optlen < sizeof(struct sctp_paddrthlds))
3497 return -EINVAL;
3498 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3499 sizeof(struct sctp_paddrthlds)))
3500 return -EFAULT;
3501
3502
3503 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3504 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3505 if (!asoc)
3506 return -ENOENT;
3507 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3508 transports) {
3509 if (val.spt_pathmaxrxt)
3510 trans->pathmaxrxt = val.spt_pathmaxrxt;
3511 trans->pf_retrans = val.spt_pathpfthld;
3512 }
3513
3514 if (val.spt_pathmaxrxt)
3515 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3516 asoc->pf_retrans = val.spt_pathpfthld;
3517 } else {
3518 trans = sctp_addr_id2transport(sk, &val.spt_address,
3519 val.spt_assoc_id);
3520 if (!trans)
3521 return -ENOENT;
3522
3523 if (val.spt_pathmaxrxt)
3524 trans->pathmaxrxt = val.spt_pathmaxrxt;
3525 trans->pf_retrans = val.spt_pathpfthld;
3526 }
3527
3528 return 0;
3529}
3530
Linus Torvalds1da177e2005-04-16 15:20:36 -07003531/* API 6.2 setsockopt(), getsockopt()
3532 *
3533 * Applications use setsockopt() and getsockopt() to set or retrieve
3534 * socket options. Socket options are used to change the default
3535 * behavior of sockets calls. They are described in Section 7.
3536 *
3537 * The syntax is:
3538 *
3539 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3540 * int __user *optlen);
3541 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3542 * int optlen);
3543 *
3544 * sd - the socket descript.
3545 * level - set to IPPROTO_SCTP for all SCTP options.
3546 * optname - the option name.
3547 * optval - the buffer to store the value of the option.
3548 * optlen - the size of the buffer.
3549 */
3550SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003551 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003552{
3553 int retval = 0;
3554
3555 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3556 sk, optname);
3557
3558 /* I can hardly begin to describe how wrong this is. This is
3559 * so broken as to be worse than useless. The API draft
3560 * REALLY is NOT helpful here... I am not convinced that the
3561 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3562 * are at all well-founded.
3563 */
3564 if (level != SOL_SCTP) {
3565 struct sctp_af *af = sctp_sk(sk)->pf->af;
3566 retval = af->setsockopt(sk, level, optname, optval, optlen);
3567 goto out_nounlock;
3568 }
3569
3570 sctp_lock_sock(sk);
3571
3572 switch (optname) {
3573 case SCTP_SOCKOPT_BINDX_ADD:
3574 /* 'optlen' is the size of the addresses buffer. */
3575 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3576 optlen, SCTP_BINDX_ADD_ADDR);
3577 break;
3578
3579 case SCTP_SOCKOPT_BINDX_REM:
3580 /* 'optlen' is the size of the addresses buffer. */
3581 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3582 optlen, SCTP_BINDX_REM_ADDR);
3583 break;
3584
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003585 case SCTP_SOCKOPT_CONNECTX_OLD:
3586 /* 'optlen' is the size of the addresses buffer. */
3587 retval = sctp_setsockopt_connectx_old(sk,
3588 (struct sockaddr __user *)optval,
3589 optlen);
3590 break;
3591
Frank Filz3f7a87d2005-06-20 13:14:57 -07003592 case SCTP_SOCKOPT_CONNECTX:
3593 /* 'optlen' is the size of the addresses buffer. */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003594 retval = sctp_setsockopt_connectx(sk,
3595 (struct sockaddr __user *)optval,
3596 optlen);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003597 break;
3598
Linus Torvalds1da177e2005-04-16 15:20:36 -07003599 case SCTP_DISABLE_FRAGMENTS:
3600 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3601 break;
3602
3603 case SCTP_EVENTS:
3604 retval = sctp_setsockopt_events(sk, optval, optlen);
3605 break;
3606
3607 case SCTP_AUTOCLOSE:
3608 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3609 break;
3610
3611 case SCTP_PEER_ADDR_PARAMS:
3612 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3613 break;
3614
Shan Wei4580ccc2011-01-18 22:39:00 +00003615 case SCTP_DELAYED_SACK:
Wei Yongjund364d922008-05-09 15:13:26 -07003616 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
Frank Filz77086102005-12-22 11:37:30 -08003617 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003618 case SCTP_PARTIAL_DELIVERY_POINT:
3619 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3620 break;
Frank Filz77086102005-12-22 11:37:30 -08003621
Linus Torvalds1da177e2005-04-16 15:20:36 -07003622 case SCTP_INITMSG:
3623 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3624 break;
3625 case SCTP_DEFAULT_SEND_PARAM:
3626 retval = sctp_setsockopt_default_send_param(sk, optval,
3627 optlen);
3628 break;
3629 case SCTP_PRIMARY_ADDR:
3630 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3631 break;
3632 case SCTP_SET_PEER_PRIMARY_ADDR:
3633 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3634 break;
3635 case SCTP_NODELAY:
3636 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3637 break;
3638 case SCTP_RTOINFO:
3639 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3640 break;
3641 case SCTP_ASSOCINFO:
3642 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3643 break;
3644 case SCTP_I_WANT_MAPPED_V4_ADDR:
3645 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3646 break;
3647 case SCTP_MAXSEG:
3648 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3649 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003650 case SCTP_ADAPTATION_LAYER:
3651 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003652 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003653 case SCTP_CONTEXT:
3654 retval = sctp_setsockopt_context(sk, optval, optlen);
3655 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003656 case SCTP_FRAGMENT_INTERLEAVE:
3657 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3658 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07003659 case SCTP_MAX_BURST:
3660 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3661 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003662 case SCTP_AUTH_CHUNK:
3663 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3664 break;
3665 case SCTP_HMAC_IDENT:
3666 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3667 break;
3668 case SCTP_AUTH_KEY:
3669 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3670 break;
3671 case SCTP_AUTH_ACTIVE_KEY:
3672 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3673 break;
3674 case SCTP_AUTH_DELETE_KEY:
3675 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3676 break;
Michio Honda7dc04d72011-04-26 20:16:31 +09003677 case SCTP_AUTO_ASCONF:
3678 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3679 break;
Neil Horman5aa93bc2012-07-21 07:56:07 +00003680 case SCTP_PEER_ADDR_THLDS:
3681 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3682 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003683 default:
3684 retval = -ENOPROTOOPT;
3685 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07003686 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003687
3688 sctp_release_sock(sk);
3689
3690out_nounlock:
3691 return retval;
3692}
3693
3694/* API 3.1.6 connect() - UDP Style Syntax
3695 *
3696 * An application may use the connect() call in the UDP model to initiate an
3697 * association without sending data.
3698 *
3699 * The syntax is:
3700 *
3701 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3702 *
3703 * sd: the socket descriptor to have a new association added to.
3704 *
3705 * nam: the address structure (either struct sockaddr_in or struct
3706 * sockaddr_in6 defined in RFC2553 [7]).
3707 *
3708 * len: the size of the address.
3709 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07003710SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003711 int addr_len)
3712{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003713 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07003714 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003715
3716 sctp_lock_sock(sk);
3717
Frank Filz3f7a87d2005-06-20 13:14:57 -07003718 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08003719 __func__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003720
Frank Filz3f7a87d2005-06-20 13:14:57 -07003721 /* Validate addr_len before calling common connect/connectx routine. */
3722 af = sctp_get_af_specific(addr->sa_family);
3723 if (!af || addr_len < af->sockaddr_len) {
3724 err = -EINVAL;
3725 } else {
3726 /* Pass correct addr len to common routine (so it knows there
3727 * is only one address being passed.
3728 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003729 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003730 }
3731
Linus Torvalds1da177e2005-04-16 15:20:36 -07003732 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003733 return err;
3734}
3735
3736/* FIXME: Write comments. */
3737SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3738{
3739 return -EOPNOTSUPP; /* STUB */
3740}
3741
3742/* 4.1.4 accept() - TCP Style Syntax
3743 *
3744 * Applications use accept() call to remove an established SCTP
3745 * association from the accept queue of the endpoint. A new socket
3746 * descriptor will be returned from accept() to represent the newly
3747 * formed association.
3748 */
3749SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3750{
3751 struct sctp_sock *sp;
3752 struct sctp_endpoint *ep;
3753 struct sock *newsk = NULL;
3754 struct sctp_association *asoc;
3755 long timeo;
3756 int error = 0;
3757
3758 sctp_lock_sock(sk);
3759
3760 sp = sctp_sk(sk);
3761 ep = sp->ep;
3762
3763 if (!sctp_style(sk, TCP)) {
3764 error = -EOPNOTSUPP;
3765 goto out;
3766 }
3767
3768 if (!sctp_sstate(sk, LISTENING)) {
3769 error = -EINVAL;
3770 goto out;
3771 }
3772
Sridhar Samudrala8abfedd2006-08-22 00:24:09 -07003773 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003774
3775 error = sctp_wait_for_accept(sk, timeo);
3776 if (error)
3777 goto out;
3778
3779 /* We treat the list of associations on the endpoint as the accept
3780 * queue and pick the first association on the list.
3781 */
3782 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3783
3784 newsk = sp->pf->create_accept_sk(sk, asoc);
3785 if (!newsk) {
3786 error = -ENOMEM;
3787 goto out;
3788 }
3789
3790 /* Populate the fields of the newsk from the oldsk and migrate the
3791 * asoc to the newsk.
3792 */
3793 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3794
3795out:
3796 sctp_release_sock(sk);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003797 *err = error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003798 return newsk;
3799}
3800
3801/* The SCTP ioctl handler. */
3802SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3803{
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003804 int rc = -ENOTCONN;
3805
3806 sctp_lock_sock(sk);
3807
3808 /*
3809 * SEQPACKET-style sockets in LISTENING state are valid, for
3810 * SCTP, so only discard TCP-style sockets in LISTENING state.
3811 */
3812 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3813 goto out;
3814
3815 switch (cmd) {
3816 case SIOCINQ: {
3817 struct sk_buff *skb;
3818 unsigned int amount = 0;
3819
3820 skb = skb_peek(&sk->sk_receive_queue);
3821 if (skb != NULL) {
3822 /*
3823 * We will only return the amount of this packet since
3824 * that is all that will be read.
3825 */
3826 amount = skb->len;
3827 }
3828 rc = put_user(amount, (int __user *)arg);
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003829 break;
David S. Miller9a7241c2010-10-03 22:14:37 -07003830 }
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003831 default:
3832 rc = -ENOIOCTLCMD;
3833 break;
3834 }
3835out:
3836 sctp_release_sock(sk);
3837 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003838}
3839
3840/* This is the function which gets called during socket creation to
3841 * initialized the SCTP-specific portion of the sock.
3842 * The sock structure should already be zero-filled memory.
3843 */
3844SCTP_STATIC int sctp_init_sock(struct sock *sk)
3845{
3846 struct sctp_endpoint *ep;
3847 struct sctp_sock *sp;
3848
3849 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3850
3851 sp = sctp_sk(sk);
3852
3853 /* Initialize the SCTP per socket area. */
3854 switch (sk->sk_type) {
3855 case SOCK_SEQPACKET:
3856 sp->type = SCTP_SOCKET_UDP;
3857 break;
3858 case SOCK_STREAM:
3859 sp->type = SCTP_SOCKET_TCP;
3860 break;
3861 default:
3862 return -ESOCKTNOSUPPORT;
3863 }
3864
3865 /* Initialize default send parameters. These parameters can be
3866 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3867 */
3868 sp->default_stream = 0;
3869 sp->default_ppid = 0;
3870 sp->default_flags = 0;
3871 sp->default_context = 0;
3872 sp->default_timetolive = 0;
3873
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003874 sp->default_rcv_context = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003875 sp->max_burst = sctp_max_burst;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003876
Linus Torvalds1da177e2005-04-16 15:20:36 -07003877 /* Initialize default setup parameters. These parameters
3878 * can be modified with the SCTP_INITMSG socket option or
3879 * overridden by the SCTP_INIT CMSG.
3880 */
3881 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3882 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3883 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003884 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003885
3886 /* Initialize default RTO related parameters. These parameters can
3887 * be modified for with the SCTP_RTOINFO socket option.
3888 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003889 sp->rtoinfo.srto_initial = sctp_rto_initial;
3890 sp->rtoinfo.srto_max = sctp_rto_max;
3891 sp->rtoinfo.srto_min = sctp_rto_min;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003892
3893 /* Initialize default association related parameters. These parameters
3894 * can be modified with the SCTP_ASSOCINFO socket option.
3895 */
3896 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3897 sp->assocparams.sasoc_number_peer_destinations = 0;
3898 sp->assocparams.sasoc_peer_rwnd = 0;
3899 sp->assocparams.sasoc_local_rwnd = 0;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003900 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003901
3902 /* Initialize default event subscriptions. By default, all the
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003903 * options are off.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003904 */
3905 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3906
3907 /* Default Peer Address Parameters. These defaults can
3908 * be modified via SCTP_PEER_ADDR_PARAMS
3909 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003910 sp->hbinterval = sctp_hb_interval;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003911 sp->pathmaxrxt = sctp_max_retrans_path;
3912 sp->pathmtu = 0; // allow default discovery
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003913 sp->sackdelay = sctp_sack_timeout;
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07003914 sp->sackfreq = 2;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003915 sp->param_flags = SPP_HB_ENABLE |
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003916 SPP_PMTUD_ENABLE |
3917 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003918
3919 /* If enabled no SCTP message fragmentation will be performed.
3920 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3921 */
3922 sp->disable_fragments = 0;
3923
Sridhar Samudrala208edef2006-09-29 17:08:01 -07003924 /* Enable Nagle algorithm by default. */
3925 sp->nodelay = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003926
3927 /* Enable by default. */
3928 sp->v4mapped = 1;
3929
3930 /* Auto-close idle associations after the configured
3931 * number of seconds. A value of 0 disables this
3932 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3933 * for UDP-style sockets only.
3934 */
3935 sp->autoclose = 0;
3936
3937 /* User specified fragmentation limit. */
3938 sp->user_frag = 0;
3939
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003940 sp->adaptation_ind = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003941
3942 sp->pf = sctp_get_pf_specific(sk->sk_family);
3943
3944 /* Control variables for partial data delivery. */
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003945 atomic_set(&sp->pd_mode, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003946 skb_queue_head_init(&sp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003947 sp->frag_interleave = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003948
3949 /* Create a per socket endpoint structure. Even if we
3950 * change the data structure relationships, this may still
3951 * be useful for storing pre-connect address information.
3952 */
3953 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3954 if (!ep)
3955 return -ENOMEM;
3956
3957 sp->ep = ep;
3958 sp->hmac = NULL;
3959
3960 SCTP_DBG_OBJCNT_INC(sock);
David S. Miller6f756a82008-11-23 17:34:03 -08003961
3962 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003963 percpu_counter_inc(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003964 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Michio Honda9f7d6532011-04-26 19:32:51 +09003965 if (sctp_default_auto_asconf) {
3966 list_add_tail(&sp->auto_asconf_list,
3967 &sctp_auto_asconf_splist);
3968 sp->do_auto_asconf = 1;
3969 } else
3970 sp->do_auto_asconf = 0;
David S. Miller6f756a82008-11-23 17:34:03 -08003971 local_bh_enable();
3972
Linus Torvalds1da177e2005-04-16 15:20:36 -07003973 return 0;
3974}
3975
3976/* Cleanup any SCTP per socket resources. */
Brian Haley7d06b2e2008-06-14 17:04:49 -07003977SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003978{
Michio Honda9f7d6532011-04-26 19:32:51 +09003979 struct sctp_sock *sp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003980
3981 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3982
3983 /* Release our hold on the endpoint. */
Michio Honda9f7d6532011-04-26 19:32:51 +09003984 sp = sctp_sk(sk);
3985 if (sp->do_auto_asconf) {
3986 sp->do_auto_asconf = 0;
3987 list_del(&sp->auto_asconf_list);
3988 }
3989 sctp_endpoint_free(sp->ep);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003990 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003991 percpu_counter_dec(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003992 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003993 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003994}
3995
3996/* API 4.1.7 shutdown() - TCP Style Syntax
3997 * int shutdown(int socket, int how);
3998 *
3999 * sd - the socket descriptor of the association to be closed.
4000 * how - Specifies the type of shutdown. The values are
4001 * as follows:
4002 * SHUT_RD
4003 * Disables further receive operations. No SCTP
4004 * protocol action is taken.
4005 * SHUT_WR
4006 * Disables further send operations, and initiates
4007 * the SCTP shutdown sequence.
4008 * SHUT_RDWR
4009 * Disables further send and receive operations
4010 * and initiates the SCTP shutdown sequence.
4011 */
4012SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
4013{
4014 struct sctp_endpoint *ep;
4015 struct sctp_association *asoc;
4016
4017 if (!sctp_style(sk, TCP))
4018 return;
4019
4020 if (how & SEND_SHUTDOWN) {
4021 ep = sctp_sk(sk)->ep;
4022 if (!list_empty(&ep->asocs)) {
4023 asoc = list_entry(ep->asocs.next,
4024 struct sctp_association, asocs);
4025 sctp_primitive_SHUTDOWN(asoc, NULL);
4026 }
4027 }
4028}
4029
4030/* 7.2.1 Association Status (SCTP_STATUS)
4031
4032 * Applications can retrieve current status information about an
4033 * association, including association state, peer receiver window size,
4034 * number of unacked data chunks, and number of data chunks pending
4035 * receipt. This information is read-only.
4036 */
4037static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4038 char __user *optval,
4039 int __user *optlen)
4040{
4041 struct sctp_status status;
4042 struct sctp_association *asoc = NULL;
4043 struct sctp_transport *transport;
4044 sctp_assoc_t associd;
4045 int retval = 0;
4046
Neil Horman408f22e2007-06-16 14:03:45 -04004047 if (len < sizeof(status)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004048 retval = -EINVAL;
4049 goto out;
4050 }
4051
Neil Horman408f22e2007-06-16 14:03:45 -04004052 len = sizeof(status);
4053 if (copy_from_user(&status, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004054 retval = -EFAULT;
4055 goto out;
4056 }
4057
4058 associd = status.sstat_assoc_id;
4059 asoc = sctp_id2assoc(sk, associd);
4060 if (!asoc) {
4061 retval = -EINVAL;
4062 goto out;
4063 }
4064
4065 transport = asoc->peer.primary_path;
4066
4067 status.sstat_assoc_id = sctp_assoc2id(asoc);
4068 status.sstat_state = asoc->state;
4069 status.sstat_rwnd = asoc->peer.rwnd;
4070 status.sstat_unackdata = asoc->unack_data;
4071
4072 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4073 status.sstat_instrms = asoc->c.sinit_max_instreams;
4074 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4075 status.sstat_fragmentation_point = asoc->frag_point;
4076 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Al Viro8cec6b82006-11-20 17:23:01 -08004077 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4078 transport->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004079 /* Map ipv4 address into v4-mapped-on-v6 address. */
4080 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4081 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07004082 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004083 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4084 status.sstat_primary.spinfo_srtt = transport->srtt;
4085 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08004086 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004087
Frank Filz3f7a87d2005-06-20 13:14:57 -07004088 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4089 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4090
Linus Torvalds1da177e2005-04-16 15:20:36 -07004091 if (put_user(len, optlen)) {
4092 retval = -EFAULT;
4093 goto out;
4094 }
4095
4096 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4097 len, status.sstat_state, status.sstat_rwnd,
4098 status.sstat_assoc_id);
4099
4100 if (copy_to_user(optval, &status, len)) {
4101 retval = -EFAULT;
4102 goto out;
4103 }
4104
4105out:
Eric Dumazeta02cec22010-09-22 20:43:57 +00004106 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004107}
4108
4109
4110/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4111 *
4112 * Applications can retrieve information about a specific peer address
4113 * of an association, including its reachability state, congestion
4114 * window, and retransmission timer values. This information is
4115 * read-only.
4116 */
4117static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4118 char __user *optval,
4119 int __user *optlen)
4120{
4121 struct sctp_paddrinfo pinfo;
4122 struct sctp_transport *transport;
4123 int retval = 0;
4124
Neil Horman408f22e2007-06-16 14:03:45 -04004125 if (len < sizeof(pinfo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004126 retval = -EINVAL;
4127 goto out;
4128 }
4129
Neil Horman408f22e2007-06-16 14:03:45 -04004130 len = sizeof(pinfo);
4131 if (copy_from_user(&pinfo, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004132 retval = -EFAULT;
4133 goto out;
4134 }
4135
4136 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4137 pinfo.spinfo_assoc_id);
4138 if (!transport)
4139 return -EINVAL;
4140
4141 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07004142 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004143 pinfo.spinfo_cwnd = transport->cwnd;
4144 pinfo.spinfo_srtt = transport->srtt;
4145 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08004146 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004147
Frank Filz3f7a87d2005-06-20 13:14:57 -07004148 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4149 pinfo.spinfo_state = SCTP_ACTIVE;
4150
Linus Torvalds1da177e2005-04-16 15:20:36 -07004151 if (put_user(len, optlen)) {
4152 retval = -EFAULT;
4153 goto out;
4154 }
4155
4156 if (copy_to_user(optval, &pinfo, len)) {
4157 retval = -EFAULT;
4158 goto out;
4159 }
4160
4161out:
Eric Dumazeta02cec22010-09-22 20:43:57 +00004162 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004163}
4164
4165/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4166 *
4167 * This option is a on/off flag. If enabled no SCTP message
4168 * fragmentation will be performed. Instead if a message being sent
4169 * exceeds the current PMTU size, the message will NOT be sent and
4170 * instead a error will be indicated to the user.
4171 */
4172static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4173 char __user *optval, int __user *optlen)
4174{
4175 int val;
4176
4177 if (len < sizeof(int))
4178 return -EINVAL;
4179
4180 len = sizeof(int);
4181 val = (sctp_sk(sk)->disable_fragments == 1);
4182 if (put_user(len, optlen))
4183 return -EFAULT;
4184 if (copy_to_user(optval, &val, len))
4185 return -EFAULT;
4186 return 0;
4187}
4188
4189/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4190 *
4191 * This socket option is used to specify various notifications and
4192 * ancillary data the user wishes to receive.
4193 */
4194static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4195 int __user *optlen)
4196{
Thomas Grafacdd5982012-04-03 22:17:53 +00004197 if (len <= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004198 return -EINVAL;
Thomas Grafacdd5982012-04-03 22:17:53 +00004199 if (len > sizeof(struct sctp_event_subscribe))
4200 len = sizeof(struct sctp_event_subscribe);
Neil Horman408f22e2007-06-16 14:03:45 -04004201 if (put_user(len, optlen))
4202 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004203 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4204 return -EFAULT;
4205 return 0;
4206}
4207
4208/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4209 *
4210 * This socket option is applicable to the UDP-style socket only. When
4211 * set it will cause associations that are idle for more than the
4212 * specified number of seconds to automatically close. An association
4213 * being idle is defined an association that has NOT sent or received
4214 * user data. The special value of '0' indicates that no automatic
4215 * close of any associations should be performed. The option expects an
4216 * integer defining the number of seconds of idle time before an
4217 * association is closed.
4218 */
4219static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4220{
4221 /* Applicable to UDP-style socket only */
4222 if (sctp_style(sk, TCP))
4223 return -EOPNOTSUPP;
Neil Horman408f22e2007-06-16 14:03:45 -04004224 if (len < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004225 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004226 len = sizeof(int);
4227 if (put_user(len, optlen))
4228 return -EFAULT;
4229 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004230 return -EFAULT;
4231 return 0;
4232}
4233
4234/* Helper routine to branch off an association to a new socket. */
Benjamin Poirier0343c552012-03-08 05:55:58 +00004235int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004236{
Benjamin Poirier0343c552012-03-08 05:55:58 +00004237 struct sctp_association *asoc = sctp_id2assoc(sk, id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004238 struct socket *sock;
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004239 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004240 int err = 0;
4241
Benjamin Poirier0343c552012-03-08 05:55:58 +00004242 if (!asoc)
4243 return -EINVAL;
4244
Linus Torvalds1da177e2005-04-16 15:20:36 -07004245 /* An association cannot be branched off from an already peeled-off
4246 * socket, nor is this supported for tcp style sockets.
4247 */
4248 if (!sctp_style(sk, UDP))
4249 return -EINVAL;
4250
4251 /* Create a new socket. */
4252 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4253 if (err < 0)
4254 return err;
4255
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004256 sctp_copy_sock(sock->sk, sk, asoc);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004257
4258 /* Make peeled-off sockets more like 1-1 accepted sockets.
4259 * Set the daddr and initialize id to something more random
4260 */
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004261 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4262 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004263
4264 /* Populate the fields of the newsk from the oldsk and migrate the
4265 * asoc to the newsk.
4266 */
4267 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004268
Linus Torvalds1da177e2005-04-16 15:20:36 -07004269 *sockp = sock;
4270
4271 return err;
4272}
Benjamin Poirier0343c552012-03-08 05:55:58 +00004273EXPORT_SYMBOL(sctp_do_peeloff);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004274
4275static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4276{
4277 sctp_peeloff_arg_t peeloff;
4278 struct socket *newsock;
4279 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004280
Neil Horman408f22e2007-06-16 14:03:45 -04004281 if (len < sizeof(sctp_peeloff_arg_t))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004282 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004283 len = sizeof(sctp_peeloff_arg_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004284 if (copy_from_user(&peeloff, optval, len))
4285 return -EFAULT;
4286
Benjamin Poirier0343c552012-03-08 05:55:58 +00004287 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004288 if (retval < 0)
4289 goto out;
4290
4291 /* Map the socket to an unused fd that can be returned to the user. */
Ulrich Dreppera677a032008-07-23 21:29:17 -07004292 retval = sock_map_fd(newsock, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004293 if (retval < 0) {
4294 sock_release(newsock);
4295 goto out;
4296 }
4297
Benjamin Poirier0343c552012-03-08 05:55:58 +00004298 SCTP_DEBUG_PRINTK("%s: sk: %p newsk: %p sd: %d\n",
4299 __func__, sk, newsock->sk, retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004300
4301 /* Return the fd mapped to the new socket. */
4302 peeloff.sd = retval;
Neil Horman408f22e2007-06-16 14:03:45 -04004303 if (put_user(len, optlen))
4304 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004305 if (copy_to_user(optval, &peeloff, len))
4306 retval = -EFAULT;
4307
4308out:
4309 return retval;
4310}
4311
4312/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4313 *
4314 * Applications can enable or disable heartbeats for any peer address of
4315 * an association, modify an address's heartbeat interval, force a
4316 * heartbeat to be sent immediately, and adjust the address's maximum
4317 * number of retransmissions sent before an address is considered
4318 * unreachable. The following structure is used to access and modify an
4319 * address's parameters:
4320 *
4321 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004322 * sctp_assoc_t spp_assoc_id;
4323 * struct sockaddr_storage spp_address;
4324 * uint32_t spp_hbinterval;
4325 * uint16_t spp_pathmaxrxt;
4326 * uint32_t spp_pathmtu;
4327 * uint32_t spp_sackdelay;
4328 * uint32_t spp_flags;
4329 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07004330 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08004331 * spp_assoc_id - (one-to-many style socket) This is filled in the
4332 * application, and identifies the association for
4333 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004334 * spp_address - This specifies which address is of interest.
4335 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004336 * in milliseconds. If a value of zero
4337 * is present in this field then no changes are to
4338 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004339 * spp_pathmaxrxt - This contains the maximum number of
4340 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08004341 * considered unreachable. If a value of zero
4342 * is present in this field then no changes are to
4343 * be made to this parameter.
4344 * spp_pathmtu - When Path MTU discovery is disabled the value
4345 * specified here will be the "fixed" path mtu.
4346 * Note that if the spp_address field is empty
4347 * then all associations on this address will
4348 * have this fixed path mtu set upon them.
4349 *
4350 * spp_sackdelay - When delayed sack is enabled, this value specifies
4351 * the number of milliseconds that sacks will be delayed
4352 * for. This value will apply to all addresses of an
4353 * association if the spp_address field is empty. Note
4354 * also, that if delayed sack is enabled and this
4355 * value is set to 0, no change is made to the last
4356 * recorded delayed sack timer value.
4357 *
4358 * spp_flags - These flags are used to control various features
4359 * on an association. The flag field may contain
4360 * zero or more of the following options.
4361 *
4362 * SPP_HB_ENABLE - Enable heartbeats on the
4363 * specified address. Note that if the address
4364 * field is empty all addresses for the association
4365 * have heartbeats enabled upon them.
4366 *
4367 * SPP_HB_DISABLE - Disable heartbeats on the
4368 * speicifed address. Note that if the address
4369 * field is empty all addresses for the association
4370 * will have their heartbeats disabled. Note also
4371 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4372 * mutually exclusive, only one of these two should
4373 * be specified. Enabling both fields will have
4374 * undetermined results.
4375 *
4376 * SPP_HB_DEMAND - Request a user initiated heartbeat
4377 * to be made immediately.
4378 *
4379 * SPP_PMTUD_ENABLE - This field will enable PMTU
4380 * discovery upon the specified address. Note that
4381 * if the address feild is empty then all addresses
4382 * on the association are effected.
4383 *
4384 * SPP_PMTUD_DISABLE - This field will disable PMTU
4385 * discovery upon the specified address. Note that
4386 * if the address feild is empty then all addresses
4387 * on the association are effected. Not also that
4388 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4389 * exclusive. Enabling both will have undetermined
4390 * results.
4391 *
4392 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4393 * on delayed sack. The time specified in spp_sackdelay
4394 * is used to specify the sack delay for this address. Note
4395 * that if spp_address is empty then all addresses will
4396 * enable delayed sack and take on the sack delay
4397 * value specified in spp_sackdelay.
4398 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4399 * off delayed sack. If the spp_address field is blank then
4400 * delayed sack is disabled for the entire association. Note
4401 * also that this field is mutually exclusive to
4402 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4403 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004404 */
4405static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004406 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004407{
Frank Filz52ccb8e2005-12-22 11:36:46 -08004408 struct sctp_paddrparams params;
4409 struct sctp_transport *trans = NULL;
4410 struct sctp_association *asoc = NULL;
4411 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004412
Neil Horman408f22e2007-06-16 14:03:45 -04004413 if (len < sizeof(struct sctp_paddrparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004414 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004415 len = sizeof(struct sctp_paddrparams);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004416 if (copy_from_user(&params, optval, len))
4417 return -EFAULT;
4418
Frank Filz52ccb8e2005-12-22 11:36:46 -08004419 /* If an address other than INADDR_ANY is specified, and
4420 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004421 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004422 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004423 trans = sctp_addr_id2transport(sk, &params.spp_address,
4424 params.spp_assoc_id);
4425 if (!trans) {
4426 SCTP_DEBUG_PRINTK("Failed no transport\n");
4427 return -EINVAL;
4428 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004429 }
4430
Frank Filz52ccb8e2005-12-22 11:36:46 -08004431 /* Get association, if assoc_id != 0 and the socket is a one
4432 * to many style socket, and an association was not found, then
4433 * the id was invalid.
4434 */
4435 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4436 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4437 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07004438 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08004439 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004440
Frank Filz52ccb8e2005-12-22 11:36:46 -08004441 if (trans) {
4442 /* Fetch transport values. */
4443 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4444 params.spp_pathmtu = trans->pathmtu;
4445 params.spp_pathmaxrxt = trans->pathmaxrxt;
4446 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004447
Frank Filz52ccb8e2005-12-22 11:36:46 -08004448 /*draft-11 doesn't say what to return in spp_flags*/
4449 params.spp_flags = trans->param_flags;
4450 } else if (asoc) {
4451 /* Fetch association values. */
4452 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4453 params.spp_pathmtu = asoc->pathmtu;
4454 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4455 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004456
Frank Filz52ccb8e2005-12-22 11:36:46 -08004457 /*draft-11 doesn't say what to return in spp_flags*/
4458 params.spp_flags = asoc->param_flags;
4459 } else {
4460 /* Fetch socket values. */
4461 params.spp_hbinterval = sp->hbinterval;
4462 params.spp_pathmtu = sp->pathmtu;
4463 params.spp_sackdelay = sp->sackdelay;
4464 params.spp_pathmaxrxt = sp->pathmaxrxt;
4465
4466 /*draft-11 doesn't say what to return in spp_flags*/
4467 params.spp_flags = sp->param_flags;
4468 }
4469
Linus Torvalds1da177e2005-04-16 15:20:36 -07004470 if (copy_to_user(optval, &params, len))
4471 return -EFAULT;
4472
4473 if (put_user(len, optlen))
4474 return -EFAULT;
4475
4476 return 0;
4477}
4478
Wei Yongjund364d922008-05-09 15:13:26 -07004479/*
4480 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08004481 *
Wei Yongjund364d922008-05-09 15:13:26 -07004482 * This option will effect the way delayed acks are performed. This
4483 * option allows you to get or set the delayed ack time, in
4484 * milliseconds. It also allows changing the delayed ack frequency.
4485 * Changing the frequency to 1 disables the delayed sack algorithm. If
4486 * the assoc_id is 0, then this sets or gets the endpoints default
4487 * values. If the assoc_id field is non-zero, then the set or get
4488 * effects the specified association for the one to many model (the
4489 * assoc_id field is ignored by the one to one model). Note that if
4490 * sack_delay or sack_freq are 0 when setting this option, then the
4491 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08004492 *
Wei Yongjund364d922008-05-09 15:13:26 -07004493 * struct sctp_sack_info {
4494 * sctp_assoc_t sack_assoc_id;
4495 * uint32_t sack_delay;
4496 * uint32_t sack_freq;
4497 * };
Frank Filz77086102005-12-22 11:37:30 -08004498 *
Wei Yongjund364d922008-05-09 15:13:26 -07004499 * sack_assoc_id - This parameter, indicates which association the user
4500 * is performing an action upon. Note that if this field's value is
4501 * zero then the endpoints default value is changed (effecting future
4502 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08004503 *
Wei Yongjund364d922008-05-09 15:13:26 -07004504 * sack_delay - This parameter contains the number of milliseconds that
4505 * the user is requesting the delayed ACK timer be set to. Note that
4506 * this value is defined in the standard to be between 200 and 500
4507 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08004508 *
Wei Yongjund364d922008-05-09 15:13:26 -07004509 * sack_freq - This parameter contains the number of packets that must
4510 * be received before a sack is sent without waiting for the delay
4511 * timer to expire. The default value for this is 2, setting this
4512 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08004513 */
Wei Yongjund364d922008-05-09 15:13:26 -07004514static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
Frank Filz77086102005-12-22 11:37:30 -08004515 char __user *optval,
4516 int __user *optlen)
4517{
Wei Yongjund364d922008-05-09 15:13:26 -07004518 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08004519 struct sctp_association *asoc = NULL;
4520 struct sctp_sock *sp = sctp_sk(sk);
4521
Wei Yongjund364d922008-05-09 15:13:26 -07004522 if (len >= sizeof(struct sctp_sack_info)) {
4523 len = sizeof(struct sctp_sack_info);
4524
4525 if (copy_from_user(&params, optval, len))
4526 return -EFAULT;
4527 } else if (len == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00004528 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4529 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07004530 if (copy_from_user(&params, optval, len))
4531 return -EFAULT;
4532 } else
Frank Filz77086102005-12-22 11:37:30 -08004533 return - EINVAL;
4534
Wei Yongjund364d922008-05-09 15:13:26 -07004535 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08004536 * to many style socket, and an association was not found, then
4537 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004538 */
Wei Yongjund364d922008-05-09 15:13:26 -07004539 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4540 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08004541 return -EINVAL;
4542
4543 if (asoc) {
4544 /* Fetch association values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004545 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4546 params.sack_delay = jiffies_to_msecs(
Frank Filz77086102005-12-22 11:37:30 -08004547 asoc->sackdelay);
Wei Yongjund364d922008-05-09 15:13:26 -07004548 params.sack_freq = asoc->sackfreq;
4549
4550 } else {
4551 params.sack_delay = 0;
4552 params.sack_freq = 1;
4553 }
Frank Filz77086102005-12-22 11:37:30 -08004554 } else {
4555 /* Fetch socket values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004556 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4557 params.sack_delay = sp->sackdelay;
4558 params.sack_freq = sp->sackfreq;
4559 } else {
4560 params.sack_delay = 0;
4561 params.sack_freq = 1;
4562 }
Frank Filz77086102005-12-22 11:37:30 -08004563 }
4564
4565 if (copy_to_user(optval, &params, len))
4566 return -EFAULT;
4567
4568 if (put_user(len, optlen))
4569 return -EFAULT;
4570
4571 return 0;
4572}
4573
Linus Torvalds1da177e2005-04-16 15:20:36 -07004574/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4575 *
4576 * Applications can specify protocol parameters for the default association
4577 * initialization. The option name argument to setsockopt() and getsockopt()
4578 * is SCTP_INITMSG.
4579 *
4580 * Setting initialization parameters is effective only on an unconnected
4581 * socket (for UDP-style sockets only future associations are effected
4582 * by the change). With TCP-style sockets, this option is inherited by
4583 * sockets derived from a listener socket.
4584 */
4585static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4586{
Neil Horman408f22e2007-06-16 14:03:45 -04004587 if (len < sizeof(struct sctp_initmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004588 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004589 len = sizeof(struct sctp_initmsg);
4590 if (put_user(len, optlen))
4591 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004592 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4593 return -EFAULT;
4594 return 0;
4595}
4596
Linus Torvalds1da177e2005-04-16 15:20:36 -07004597
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004598static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4599 char __user *optval, int __user *optlen)
4600{
4601 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004602 int cnt = 0;
4603 struct sctp_getaddrs getaddrs;
4604 struct sctp_transport *from;
4605 void __user *to;
4606 union sctp_addr temp;
4607 struct sctp_sock *sp = sctp_sk(sk);
4608 int addrlen;
4609 size_t space_left;
4610 int bytes_copied;
4611
4612 if (len < sizeof(struct sctp_getaddrs))
4613 return -EINVAL;
4614
4615 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4616 return -EFAULT;
4617
4618 /* For UDP-style sockets, id specifies the association to query. */
4619 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4620 if (!asoc)
4621 return -EINVAL;
4622
4623 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004624 space_left = len - offsetof(struct sctp_getaddrs,addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004625
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07004626 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4627 transports) {
Al Virob3f5b3b2006-11-20 17:22:43 -08004628 memcpy(&temp, &from->ipaddr, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004629 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
Vlad Yasevicha5f4cea2010-04-30 21:42:42 -04004630 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004631 if (space_left < addrlen)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004632 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004633 if (copy_to_user(to, &temp, addrlen))
4634 return -EFAULT;
4635 to += addrlen;
4636 cnt++;
4637 space_left -= addrlen;
4638 }
4639
4640 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4641 return -EFAULT;
4642 bytes_copied = ((char __user *)to) - optval;
4643 if (put_user(bytes_copied, optlen))
4644 return -EFAULT;
4645
4646 return 0;
4647}
4648
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004649static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4650 size_t space_left, int *bytes_copied)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004651{
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004652 struct sctp_sockaddr_entry *addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004653 union sctp_addr temp;
4654 int cnt = 0;
4655 int addrlen;
4656
Vlad Yasevich29303542007-09-16 16:02:12 -07004657 rcu_read_lock();
4658 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4659 if (!addr->valid)
4660 continue;
4661
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004662 if ((PF_INET == sk->sk_family) &&
Al Viro6244be42006-11-20 17:21:44 -08004663 (AF_INET6 == addr->a.sa.sa_family))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004664 continue;
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07004665 if ((PF_INET6 == sk->sk_family) &&
4666 inet_v6_ipv6only(sk) &&
4667 (AF_INET == addr->a.sa.sa_family))
4668 continue;
Al Viro6244be42006-11-20 17:21:44 -08004669 memcpy(&temp, &addr->a, sizeof(temp));
Vlad Yasevichb46ae362008-01-28 14:25:36 -05004670 if (!temp.v4.sin_port)
4671 temp.v4.sin_port = htons(port);
4672
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004673 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4674 &temp);
4675 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevich29303542007-09-16 16:02:12 -07004676 if (space_left < addrlen) {
4677 cnt = -ENOMEM;
4678 break;
4679 }
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004680 memcpy(to, &temp, addrlen);
Sridhar Samudrala29c7cf92006-12-13 16:26:26 -08004681
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004682 to += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004683 cnt ++;
4684 space_left -= addrlen;
Vlad Yasevich3663c302007-07-03 12:43:12 -04004685 *bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004686 }
Vlad Yasevich29303542007-09-16 16:02:12 -07004687 rcu_read_unlock();
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004688
4689 return cnt;
4690}
4691
Linus Torvalds1da177e2005-04-16 15:20:36 -07004692
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004693static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4694 char __user *optval, int __user *optlen)
4695{
4696 struct sctp_bind_addr *bp;
4697 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004698 int cnt = 0;
4699 struct sctp_getaddrs getaddrs;
4700 struct sctp_sockaddr_entry *addr;
4701 void __user *to;
4702 union sctp_addr temp;
4703 struct sctp_sock *sp = sctp_sk(sk);
4704 int addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004705 int err = 0;
4706 size_t space_left;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004707 int bytes_copied = 0;
4708 void *addrs;
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004709 void *buf;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004710
Neil Horman408f22e2007-06-16 14:03:45 -04004711 if (len < sizeof(struct sctp_getaddrs))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004712 return -EINVAL;
4713
4714 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4715 return -EFAULT;
4716
4717 /*
4718 * For UDP-style sockets, id specifies the association to query.
4719 * If the id field is set to the value '0' then the locally bound
4720 * addresses are returned without regard to any particular
4721 * association.
4722 */
4723 if (0 == getaddrs.assoc_id) {
4724 bp = &sctp_sk(sk)->ep->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004725 } else {
4726 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4727 if (!asoc)
4728 return -EINVAL;
4729 bp = &asoc->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004730 }
4731
4732 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004733 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4734
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004735 addrs = kmalloc(space_left, GFP_KERNEL);
4736 if (!addrs)
4737 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004738
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004739 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4740 * addresses from the global local address list.
4741 */
4742 if (sctp_list_single_entry(&bp->address_list)) {
4743 addr = list_entry(bp->address_list.next,
4744 struct sctp_sockaddr_entry, list);
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004745 if (sctp_is_any(sk, &addr->a)) {
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004746 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4747 space_left, &bytes_copied);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004748 if (cnt < 0) {
4749 err = cnt;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004750 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004751 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004752 goto copy_getaddrs;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004753 }
4754 }
4755
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004756 buf = addrs;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004757 /* Protection on the bound address list is not needed since
4758 * in the socket option context we hold a socket lock and
4759 * thus the bound address list can't change.
4760 */
4761 list_for_each_entry(addr, &bp->address_list, list) {
Al Viro6244be42006-11-20 17:21:44 -08004762 memcpy(&temp, &addr->a, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004763 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4764 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004765 if (space_left < addrlen) {
4766 err = -ENOMEM; /*fixme: right error?*/
Vlad Yasevich559cf712007-09-16 16:03:28 -07004767 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004768 }
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004769 memcpy(buf, &temp, addrlen);
4770 buf += addrlen;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004771 bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004772 cnt ++;
4773 space_left -= addrlen;
4774 }
4775
4776copy_getaddrs:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004777 if (copy_to_user(to, addrs, bytes_copied)) {
4778 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004779 goto out;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004780 }
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004781 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4782 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004783 goto out;
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004784 }
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004785 if (put_user(bytes_copied, optlen))
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004786 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004787out:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004788 kfree(addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004789 return err;
4790}
4791
Linus Torvalds1da177e2005-04-16 15:20:36 -07004792/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4793 *
4794 * Requests that the local SCTP stack use the enclosed peer address as
4795 * the association primary. The enclosed address must be one of the
4796 * association peer's addresses.
4797 */
4798static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4799 char __user *optval, int __user *optlen)
4800{
4801 struct sctp_prim prim;
4802 struct sctp_association *asoc;
4803 struct sctp_sock *sp = sctp_sk(sk);
4804
Neil Horman408f22e2007-06-16 14:03:45 -04004805 if (len < sizeof(struct sctp_prim))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004806 return -EINVAL;
4807
Neil Horman408f22e2007-06-16 14:03:45 -04004808 len = sizeof(struct sctp_prim);
4809
4810 if (copy_from_user(&prim, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004811 return -EFAULT;
4812
4813 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4814 if (!asoc)
4815 return -EINVAL;
4816
4817 if (!asoc->peer.primary_path)
4818 return -ENOTCONN;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004819
Al Viro8cec6b82006-11-20 17:23:01 -08004820 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4821 asoc->peer.primary_path->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004822
4823 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4824 (union sctp_addr *)&prim.ssp_addr);
4825
Neil Horman408f22e2007-06-16 14:03:45 -04004826 if (put_user(len, optlen))
4827 return -EFAULT;
4828 if (copy_to_user(optval, &prim, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004829 return -EFAULT;
4830
4831 return 0;
4832}
4833
4834/*
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004835 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004836 *
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004837 * Requests that the local endpoint set the specified Adaptation Layer
Linus Torvalds1da177e2005-04-16 15:20:36 -07004838 * Indication parameter for all future INIT and INIT-ACK exchanges.
4839 */
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004840static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004841 char __user *optval, int __user *optlen)
4842{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004843 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004844
Neil Horman408f22e2007-06-16 14:03:45 -04004845 if (len < sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004846 return -EINVAL;
4847
Neil Horman408f22e2007-06-16 14:03:45 -04004848 len = sizeof(struct sctp_setadaptation);
4849
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004850 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
Neil Horman408f22e2007-06-16 14:03:45 -04004851
4852 if (put_user(len, optlen))
4853 return -EFAULT;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004854 if (copy_to_user(optval, &adaptation, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004855 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004856
Linus Torvalds1da177e2005-04-16 15:20:36 -07004857 return 0;
4858}
4859
4860/*
4861 *
4862 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4863 *
4864 * Applications that wish to use the sendto() system call may wish to
4865 * specify a default set of parameters that would normally be supplied
4866 * through the inclusion of ancillary data. This socket option allows
4867 * such an application to set the default sctp_sndrcvinfo structure.
4868
4869
4870 * The application that wishes to use this socket option simply passes
4871 * in to this call the sctp_sndrcvinfo structure defined in Section
4872 * 5.2.2) The input parameters accepted by this call include
4873 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4874 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4875 * to this call if the caller is using the UDP model.
4876 *
4877 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4878 */
4879static int sctp_getsockopt_default_send_param(struct sock *sk,
4880 int len, char __user *optval,
4881 int __user *optlen)
4882{
4883 struct sctp_sndrcvinfo info;
4884 struct sctp_association *asoc;
4885 struct sctp_sock *sp = sctp_sk(sk);
4886
Neil Horman408f22e2007-06-16 14:03:45 -04004887 if (len < sizeof(struct sctp_sndrcvinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004888 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004889
4890 len = sizeof(struct sctp_sndrcvinfo);
4891
4892 if (copy_from_user(&info, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004893 return -EFAULT;
4894
4895 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4896 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4897 return -EINVAL;
4898
4899 if (asoc) {
4900 info.sinfo_stream = asoc->default_stream;
4901 info.sinfo_flags = asoc->default_flags;
4902 info.sinfo_ppid = asoc->default_ppid;
4903 info.sinfo_context = asoc->default_context;
4904 info.sinfo_timetolive = asoc->default_timetolive;
4905 } else {
4906 info.sinfo_stream = sp->default_stream;
4907 info.sinfo_flags = sp->default_flags;
4908 info.sinfo_ppid = sp->default_ppid;
4909 info.sinfo_context = sp->default_context;
4910 info.sinfo_timetolive = sp->default_timetolive;
4911 }
4912
Neil Horman408f22e2007-06-16 14:03:45 -04004913 if (put_user(len, optlen))
4914 return -EFAULT;
4915 if (copy_to_user(optval, &info, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004916 return -EFAULT;
4917
4918 return 0;
4919}
4920
4921/*
4922 *
4923 * 7.1.5 SCTP_NODELAY
4924 *
4925 * Turn on/off any Nagle-like algorithm. This means that packets are
4926 * generally sent as soon as possible and no unnecessary delays are
4927 * introduced, at the cost of more packets in the network. Expects an
4928 * integer boolean flag.
4929 */
4930
4931static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4932 char __user *optval, int __user *optlen)
4933{
4934 int val;
4935
4936 if (len < sizeof(int))
4937 return -EINVAL;
4938
4939 len = sizeof(int);
4940 val = (sctp_sk(sk)->nodelay == 1);
4941 if (put_user(len, optlen))
4942 return -EFAULT;
4943 if (copy_to_user(optval, &val, len))
4944 return -EFAULT;
4945 return 0;
4946}
4947
4948/*
4949 *
4950 * 7.1.1 SCTP_RTOINFO
4951 *
4952 * The protocol parameters used to initialize and bound retransmission
4953 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4954 * and modify these parameters.
4955 * All parameters are time values, in milliseconds. A value of 0, when
4956 * modifying the parameters, indicates that the current value should not
4957 * be changed.
4958 *
4959 */
4960static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4961 char __user *optval,
4962 int __user *optlen) {
4963 struct sctp_rtoinfo rtoinfo;
4964 struct sctp_association *asoc;
4965
Neil Horman408f22e2007-06-16 14:03:45 -04004966 if (len < sizeof (struct sctp_rtoinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004967 return -EINVAL;
4968
Neil Horman408f22e2007-06-16 14:03:45 -04004969 len = sizeof(struct sctp_rtoinfo);
4970
4971 if (copy_from_user(&rtoinfo, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004972 return -EFAULT;
4973
4974 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4975
4976 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4977 return -EINVAL;
4978
4979 /* Values corresponding to the specific association. */
4980 if (asoc) {
4981 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4982 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4983 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4984 } else {
4985 /* Values corresponding to the endpoint. */
4986 struct sctp_sock *sp = sctp_sk(sk);
4987
4988 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4989 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4990 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4991 }
4992
4993 if (put_user(len, optlen))
4994 return -EFAULT;
4995
4996 if (copy_to_user(optval, &rtoinfo, len))
4997 return -EFAULT;
4998
4999 return 0;
5000}
5001
5002/*
5003 *
5004 * 7.1.2 SCTP_ASSOCINFO
5005 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02005006 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07005007 * of the association.
5008 * Returns an error if the new association retransmission value is
5009 * greater than the sum of the retransmission value of the peer.
5010 * See [SCTP] for more information.
5011 *
5012 */
5013static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5014 char __user *optval,
5015 int __user *optlen)
5016{
5017
5018 struct sctp_assocparams assocparams;
5019 struct sctp_association *asoc;
5020 struct list_head *pos;
5021 int cnt = 0;
5022
Neil Horman408f22e2007-06-16 14:03:45 -04005023 if (len < sizeof (struct sctp_assocparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005024 return -EINVAL;
5025
Neil Horman408f22e2007-06-16 14:03:45 -04005026 len = sizeof(struct sctp_assocparams);
5027
5028 if (copy_from_user(&assocparams, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005029 return -EFAULT;
5030
5031 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5032
5033 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5034 return -EINVAL;
5035
5036 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07005037 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005038 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5039 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5040 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5041 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5042 * 1000) +
5043 (asoc->cookie_life.tv_usec
5044 / 1000);
5045
5046 list_for_each(pos, &asoc->peer.transport_addr_list) {
5047 cnt ++;
5048 }
5049
5050 assocparams.sasoc_number_peer_destinations = cnt;
5051 } else {
5052 /* Values corresponding to the endpoint */
5053 struct sctp_sock *sp = sctp_sk(sk);
5054
5055 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5056 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5057 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5058 assocparams.sasoc_cookie_life =
5059 sp->assocparams.sasoc_cookie_life;
5060 assocparams.sasoc_number_peer_destinations =
5061 sp->assocparams.
5062 sasoc_number_peer_destinations;
5063 }
5064
5065 if (put_user(len, optlen))
5066 return -EFAULT;
5067
5068 if (copy_to_user(optval, &assocparams, len))
5069 return -EFAULT;
5070
5071 return 0;
5072}
5073
5074/*
5075 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5076 *
5077 * This socket option is a boolean flag which turns on or off mapped V4
5078 * addresses. If this option is turned on and the socket is type
5079 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5080 * If this option is turned off, then no mapping will be done of V4
5081 * addresses and a user will receive both PF_INET6 and PF_INET type
5082 * addresses on the socket.
5083 */
5084static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5085 char __user *optval, int __user *optlen)
5086{
5087 int val;
5088 struct sctp_sock *sp = sctp_sk(sk);
5089
5090 if (len < sizeof(int))
5091 return -EINVAL;
5092
5093 len = sizeof(int);
5094 val = sp->v4mapped;
5095 if (put_user(len, optlen))
5096 return -EFAULT;
5097 if (copy_to_user(optval, &val, len))
5098 return -EFAULT;
5099
5100 return 0;
5101}
5102
5103/*
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005104 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5105 * (chapter and verse is quoted at sctp_setsockopt_context())
5106 */
5107static int sctp_getsockopt_context(struct sock *sk, int len,
5108 char __user *optval, int __user *optlen)
5109{
5110 struct sctp_assoc_value params;
5111 struct sctp_sock *sp;
5112 struct sctp_association *asoc;
5113
Neil Horman408f22e2007-06-16 14:03:45 -04005114 if (len < sizeof(struct sctp_assoc_value))
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005115 return -EINVAL;
5116
Neil Horman408f22e2007-06-16 14:03:45 -04005117 len = sizeof(struct sctp_assoc_value);
5118
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005119 if (copy_from_user(&params, optval, len))
5120 return -EFAULT;
5121
5122 sp = sctp_sk(sk);
5123
5124 if (params.assoc_id != 0) {
5125 asoc = sctp_id2assoc(sk, params.assoc_id);
5126 if (!asoc)
5127 return -EINVAL;
5128 params.assoc_value = asoc->default_rcv_context;
5129 } else {
5130 params.assoc_value = sp->default_rcv_context;
5131 }
5132
5133 if (put_user(len, optlen))
5134 return -EFAULT;
5135 if (copy_to_user(optval, &params, len))
5136 return -EFAULT;
5137
5138 return 0;
5139}
5140
5141/*
Wei Yongjune89c2092008-12-25 16:54:58 -08005142 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5143 * This option will get or set the maximum size to put in any outgoing
5144 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07005145 * fragmented by SCTP into the specified size. Note that the underlying
5146 * SCTP implementation may fragment into smaller sized chunks when the
5147 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08005148 * the user. The default value for this option is '0' which indicates
5149 * the user is NOT limiting fragmentation and only the PMTU will effect
5150 * SCTP's choice of DATA chunk size. Note also that values set larger
5151 * than the maximum size of an IP datagram will effectively let SCTP
5152 * control fragmentation (i.e. the same as setting this option to 0).
5153 *
5154 * The following structure is used to access and modify this parameter:
5155 *
5156 * struct sctp_assoc_value {
5157 * sctp_assoc_t assoc_id;
5158 * uint32_t assoc_value;
5159 * };
5160 *
5161 * assoc_id: This parameter is ignored for one-to-one style sockets.
5162 * For one-to-many style sockets this parameter indicates which
5163 * association the user is performing an action upon. Note that if
5164 * this field's value is zero then the endpoints default value is
5165 * changed (effecting future associations only).
5166 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07005167 */
5168static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5169 char __user *optval, int __user *optlen)
5170{
Wei Yongjune89c2092008-12-25 16:54:58 -08005171 struct sctp_assoc_value params;
5172 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005173
Wei Yongjune89c2092008-12-25 16:54:58 -08005174 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00005175 pr_warn("Use of int in maxseg socket option deprecated\n");
5176 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08005177 params.assoc_id = 0;
5178 } else if (len >= sizeof(struct sctp_assoc_value)) {
5179 len = sizeof(struct sctp_assoc_value);
5180 if (copy_from_user(&params, optval, sizeof(params)))
5181 return -EFAULT;
5182 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07005183 return -EINVAL;
5184
Wei Yongjune89c2092008-12-25 16:54:58 -08005185 asoc = sctp_id2assoc(sk, params.assoc_id);
5186 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5187 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005188
Wei Yongjune89c2092008-12-25 16:54:58 -08005189 if (asoc)
5190 params.assoc_value = asoc->frag_point;
5191 else
5192 params.assoc_value = sctp_sk(sk)->user_frag;
5193
Linus Torvalds1da177e2005-04-16 15:20:36 -07005194 if (put_user(len, optlen))
5195 return -EFAULT;
Wei Yongjune89c2092008-12-25 16:54:58 -08005196 if (len == sizeof(int)) {
5197 if (copy_to_user(optval, &params.assoc_value, len))
5198 return -EFAULT;
5199 } else {
5200 if (copy_to_user(optval, &params, len))
5201 return -EFAULT;
5202 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005203
5204 return 0;
5205}
5206
Vlad Yasevichb6e13312007-04-20 12:23:15 -07005207/*
5208 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5209 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5210 */
5211static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5212 char __user *optval, int __user *optlen)
5213{
5214 int val;
5215
5216 if (len < sizeof(int))
5217 return -EINVAL;
5218
5219 len = sizeof(int);
5220
5221 val = sctp_sk(sk)->frag_interleave;
5222 if (put_user(len, optlen))
5223 return -EFAULT;
5224 if (copy_to_user(optval, &val, len))
5225 return -EFAULT;
5226
5227 return 0;
5228}
5229
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005230/*
5231 * 7.1.25. Set or Get the sctp partial delivery point
5232 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5233 */
5234static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5235 char __user *optval,
5236 int __user *optlen)
5237{
YOSHIFUJI Hideaki9cbcbf42007-07-19 10:44:50 +09005238 u32 val;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005239
5240 if (len < sizeof(u32))
5241 return -EINVAL;
5242
5243 len = sizeof(u32);
5244
5245 val = sctp_sk(sk)->pd_point;
5246 if (put_user(len, optlen))
5247 return -EFAULT;
5248 if (copy_to_user(optval, &val, len))
5249 return -EFAULT;
5250
Wei Yongjun7d743b72010-12-14 16:10:41 +00005251 return 0;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005252}
5253
Vlad Yasevich70331572007-03-23 11:34:36 -07005254/*
5255 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5256 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5257 */
5258static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5259 char __user *optval,
5260 int __user *optlen)
5261{
Neil Horman219b99a2008-03-05 13:44:46 -08005262 struct sctp_assoc_value params;
5263 struct sctp_sock *sp;
5264 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07005265
Neil Horman219b99a2008-03-05 13:44:46 -08005266 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00005267 pr_warn("Use of int in max_burst socket option deprecated\n");
5268 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08005269 params.assoc_id = 0;
Wei Yongjunc6db93a2009-03-02 09:46:12 +00005270 } else if (len >= sizeof(struct sctp_assoc_value)) {
5271 len = sizeof(struct sctp_assoc_value);
Neil Horman219b99a2008-03-05 13:44:46 -08005272 if (copy_from_user(&params, optval, len))
5273 return -EFAULT;
5274 } else
5275 return -EINVAL;
Vlad Yasevich70331572007-03-23 11:34:36 -07005276
Neil Horman219b99a2008-03-05 13:44:46 -08005277 sp = sctp_sk(sk);
Vlad Yasevich70331572007-03-23 11:34:36 -07005278
Neil Horman219b99a2008-03-05 13:44:46 -08005279 if (params.assoc_id != 0) {
5280 asoc = sctp_id2assoc(sk, params.assoc_id);
5281 if (!asoc)
5282 return -EINVAL;
5283 params.assoc_value = asoc->max_burst;
5284 } else
5285 params.assoc_value = sp->max_burst;
5286
5287 if (len == sizeof(int)) {
5288 if (copy_to_user(optval, &params.assoc_value, len))
5289 return -EFAULT;
5290 } else {
5291 if (copy_to_user(optval, &params, len))
5292 return -EFAULT;
5293 }
5294
5295 return 0;
5296
Vlad Yasevich70331572007-03-23 11:34:36 -07005297}
5298
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005299static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5300 char __user *optval, int __user *optlen)
5301{
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005302 struct sctp_hmacalgo __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005303 struct sctp_hmac_algo_param *hmacs;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005304 __u16 data_len = 0;
5305 u32 num_idents;
5306
5307 if (!sctp_auth_enable)
5308 return -EACCES;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005309
5310 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005311 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005312
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005313 if (len < sizeof(struct sctp_hmacalgo) + data_len)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005314 return -EINVAL;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005315
5316 len = sizeof(struct sctp_hmacalgo) + data_len;
5317 num_idents = data_len / sizeof(u16);
5318
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005319 if (put_user(len, optlen))
5320 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005321 if (put_user(num_idents, &p->shmac_num_idents))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005322 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005323 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5324 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005325 return 0;
5326}
5327
5328static int sctp_getsockopt_active_key(struct sock *sk, int len,
5329 char __user *optval, int __user *optlen)
5330{
5331 struct sctp_authkeyid val;
5332 struct sctp_association *asoc;
5333
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005334 if (!sctp_auth_enable)
5335 return -EACCES;
5336
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005337 if (len < sizeof(struct sctp_authkeyid))
5338 return -EINVAL;
5339 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5340 return -EFAULT;
5341
5342 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5343 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5344 return -EINVAL;
5345
5346 if (asoc)
5347 val.scact_keynumber = asoc->active_key_id;
5348 else
5349 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5350
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005351 len = sizeof(struct sctp_authkeyid);
5352 if (put_user(len, optlen))
5353 return -EFAULT;
5354 if (copy_to_user(optval, &val, len))
5355 return -EFAULT;
5356
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005357 return 0;
5358}
5359
5360static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5361 char __user *optval, int __user *optlen)
5362{
Al Viro411223c2007-10-14 19:21:20 +01005363 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005364 struct sctp_authchunks val;
5365 struct sctp_association *asoc;
5366 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005367 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005368 char __user *to;
5369
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005370 if (!sctp_auth_enable)
5371 return -EACCES;
5372
5373 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005374 return -EINVAL;
5375
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005376 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005377 return -EFAULT;
5378
Al Viro411223c2007-10-14 19:21:20 +01005379 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005380 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5381 if (!asoc)
5382 return -EINVAL;
5383
5384 ch = asoc->peer.peer_chunks;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005385 if (!ch)
5386 goto num;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005387
5388 /* See if the user provided enough room for all the data */
Vlad Yasevichb40db682008-02-27 14:40:37 -05005389 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5390 if (len < num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005391 return -EINVAL;
5392
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005393 if (copy_to_user(to, ch->chunks, num_chunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005394 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005395num:
5396 len = sizeof(struct sctp_authchunks) + num_chunks;
5397 if (put_user(len, optlen)) return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005398 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5399 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005400 return 0;
5401}
5402
5403static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5404 char __user *optval, int __user *optlen)
5405{
Al Viro411223c2007-10-14 19:21:20 +01005406 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005407 struct sctp_authchunks val;
5408 struct sctp_association *asoc;
5409 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005410 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005411 char __user *to;
5412
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005413 if (!sctp_auth_enable)
5414 return -EACCES;
5415
5416 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005417 return -EINVAL;
5418
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005419 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005420 return -EFAULT;
5421
Al Viro411223c2007-10-14 19:21:20 +01005422 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005423 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5424 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5425 return -EINVAL;
5426
5427 if (asoc)
5428 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5429 else
5430 ch = sctp_sk(sk)->ep->auth_chunk_list;
5431
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005432 if (!ch)
5433 goto num;
5434
Vlad Yasevichb40db682008-02-27 14:40:37 -05005435 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005436 if (len < sizeof(struct sctp_authchunks) + num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005437 return -EINVAL;
5438
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005439 if (copy_to_user(to, ch->chunks, num_chunks))
5440 return -EFAULT;
5441num:
5442 len = sizeof(struct sctp_authchunks) + num_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005443 if (put_user(len, optlen))
5444 return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005445 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5446 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005447
5448 return 0;
5449}
5450
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005451/*
5452 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5453 * This option gets the current number of associations that are attached
5454 * to a one-to-many style socket. The option value is an uint32_t.
5455 */
5456static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5457 char __user *optval, int __user *optlen)
5458{
5459 struct sctp_sock *sp = sctp_sk(sk);
5460 struct sctp_association *asoc;
5461 u32 val = 0;
5462
5463 if (sctp_style(sk, TCP))
5464 return -EOPNOTSUPP;
5465
5466 if (len < sizeof(u32))
5467 return -EINVAL;
5468
5469 len = sizeof(u32);
5470
5471 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5472 val++;
5473 }
5474
5475 if (put_user(len, optlen))
5476 return -EFAULT;
5477 if (copy_to_user(optval, &val, len))
5478 return -EFAULT;
5479
5480 return 0;
5481}
5482
Wei Yongjun209ba422011-04-17 17:27:08 +00005483/*
Michio Honda7dc04d72011-04-26 20:16:31 +09005484 * 8.1.23 SCTP_AUTO_ASCONF
5485 * See the corresponding setsockopt entry as description
5486 */
5487static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5488 char __user *optval, int __user *optlen)
5489{
5490 int val = 0;
5491
5492 if (len < sizeof(int))
5493 return -EINVAL;
5494
5495 len = sizeof(int);
5496 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5497 val = 1;
5498 if (put_user(len, optlen))
5499 return -EFAULT;
5500 if (copy_to_user(optval, &val, len))
5501 return -EFAULT;
5502 return 0;
5503}
5504
5505/*
Wei Yongjun209ba422011-04-17 17:27:08 +00005506 * 8.2.6. Get the Current Identifiers of Associations
5507 * (SCTP_GET_ASSOC_ID_LIST)
5508 *
5509 * This option gets the current list of SCTP association identifiers of
5510 * the SCTP associations handled by a one-to-many style socket.
5511 */
5512static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5513 char __user *optval, int __user *optlen)
5514{
5515 struct sctp_sock *sp = sctp_sk(sk);
5516 struct sctp_association *asoc;
5517 struct sctp_assoc_ids *ids;
5518 u32 num = 0;
5519
5520 if (sctp_style(sk, TCP))
5521 return -EOPNOTSUPP;
5522
5523 if (len < sizeof(struct sctp_assoc_ids))
5524 return -EINVAL;
5525
5526 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5527 num++;
5528 }
5529
5530 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5531 return -EINVAL;
5532
5533 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5534
5535 ids = kmalloc(len, GFP_KERNEL);
5536 if (unlikely(!ids))
5537 return -ENOMEM;
5538
5539 ids->gaids_number_of_ids = num;
5540 num = 0;
5541 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5542 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5543 }
5544
5545 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5546 kfree(ids);
5547 return -EFAULT;
5548 }
5549
5550 kfree(ids);
5551 return 0;
5552}
5553
Neil Horman5aa93bc2012-07-21 07:56:07 +00005554/*
5555 * SCTP_PEER_ADDR_THLDS
5556 *
5557 * This option allows us to fetch the partially failed threshold for one or all
5558 * transports in an association. See Section 6.1 of:
5559 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5560 */
5561static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5562 char __user *optval,
5563 int len,
5564 int __user *optlen)
5565{
5566 struct sctp_paddrthlds val;
5567 struct sctp_transport *trans;
5568 struct sctp_association *asoc;
5569
5570 if (len < sizeof(struct sctp_paddrthlds))
5571 return -EINVAL;
5572 len = sizeof(struct sctp_paddrthlds);
5573 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5574 return -EFAULT;
5575
5576 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5577 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5578 if (!asoc)
5579 return -ENOENT;
5580
5581 val.spt_pathpfthld = asoc->pf_retrans;
5582 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5583 } else {
5584 trans = sctp_addr_id2transport(sk, &val.spt_address,
5585 val.spt_assoc_id);
5586 if (!trans)
5587 return -ENOENT;
5588
5589 val.spt_pathmaxrxt = trans->pathmaxrxt;
5590 val.spt_pathpfthld = trans->pf_retrans;
5591 }
5592
5593 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5594 return -EFAULT;
5595
5596 return 0;
5597}
5598
Linus Torvalds1da177e2005-04-16 15:20:36 -07005599SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5600 char __user *optval, int __user *optlen)
5601{
5602 int retval = 0;
5603 int len;
5604
Frank Filz3f7a87d2005-06-20 13:14:57 -07005605 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5606 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005607
5608 /* I can hardly begin to describe how wrong this is. This is
5609 * so broken as to be worse than useless. The API draft
5610 * REALLY is NOT helpful here... I am not convinced that the
5611 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5612 * are at all well-founded.
5613 */
5614 if (level != SOL_SCTP) {
5615 struct sctp_af *af = sctp_sk(sk)->pf->af;
5616
5617 retval = af->getsockopt(sk, level, optname, optval, optlen);
5618 return retval;
5619 }
5620
5621 if (get_user(len, optlen))
5622 return -EFAULT;
5623
5624 sctp_lock_sock(sk);
5625
5626 switch (optname) {
5627 case SCTP_STATUS:
5628 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5629 break;
5630 case SCTP_DISABLE_FRAGMENTS:
5631 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5632 optlen);
5633 break;
5634 case SCTP_EVENTS:
5635 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5636 break;
5637 case SCTP_AUTOCLOSE:
5638 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5639 break;
5640 case SCTP_SOCKOPT_PEELOFF:
5641 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5642 break;
5643 case SCTP_PEER_ADDR_PARAMS:
5644 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5645 optlen);
5646 break;
Shan Wei4580ccc2011-01-18 22:39:00 +00005647 case SCTP_DELAYED_SACK:
Wei Yongjund364d922008-05-09 15:13:26 -07005648 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
Frank Filz77086102005-12-22 11:37:30 -08005649 optlen);
5650 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005651 case SCTP_INITMSG:
5652 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5653 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005654 case SCTP_GET_PEER_ADDRS:
5655 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5656 optlen);
5657 break;
5658 case SCTP_GET_LOCAL_ADDRS:
5659 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5660 optlen);
5661 break;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04005662 case SCTP_SOCKOPT_CONNECTX3:
5663 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5664 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005665 case SCTP_DEFAULT_SEND_PARAM:
5666 retval = sctp_getsockopt_default_send_param(sk, len,
5667 optval, optlen);
5668 break;
5669 case SCTP_PRIMARY_ADDR:
5670 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5671 break;
5672 case SCTP_NODELAY:
5673 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5674 break;
5675 case SCTP_RTOINFO:
5676 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5677 break;
5678 case SCTP_ASSOCINFO:
5679 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5680 break;
5681 case SCTP_I_WANT_MAPPED_V4_ADDR:
5682 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5683 break;
5684 case SCTP_MAXSEG:
5685 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5686 break;
5687 case SCTP_GET_PEER_ADDR_INFO:
5688 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5689 optlen);
5690 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08005691 case SCTP_ADAPTATION_LAYER:
5692 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07005693 optlen);
5694 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005695 case SCTP_CONTEXT:
5696 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5697 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07005698 case SCTP_FRAGMENT_INTERLEAVE:
5699 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5700 optlen);
5701 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005702 case SCTP_PARTIAL_DELIVERY_POINT:
5703 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5704 optlen);
5705 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07005706 case SCTP_MAX_BURST:
5707 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5708 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005709 case SCTP_AUTH_KEY:
5710 case SCTP_AUTH_CHUNK:
5711 case SCTP_AUTH_DELETE_KEY:
5712 retval = -EOPNOTSUPP;
5713 break;
5714 case SCTP_HMAC_IDENT:
5715 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5716 break;
5717 case SCTP_AUTH_ACTIVE_KEY:
5718 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5719 break;
5720 case SCTP_PEER_AUTH_CHUNKS:
5721 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5722 optlen);
5723 break;
5724 case SCTP_LOCAL_AUTH_CHUNKS:
5725 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5726 optlen);
5727 break;
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005728 case SCTP_GET_ASSOC_NUMBER:
5729 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5730 break;
Wei Yongjun209ba422011-04-17 17:27:08 +00005731 case SCTP_GET_ASSOC_ID_LIST:
5732 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5733 break;
Michio Honda7dc04d72011-04-26 20:16:31 +09005734 case SCTP_AUTO_ASCONF:
5735 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5736 break;
Neil Horman5aa93bc2012-07-21 07:56:07 +00005737 case SCTP_PEER_ADDR_THLDS:
5738 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5739 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005740 default:
5741 retval = -ENOPROTOOPT;
5742 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005743 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005744
5745 sctp_release_sock(sk);
5746 return retval;
5747}
5748
5749static void sctp_hash(struct sock *sk)
5750{
5751 /* STUB */
5752}
5753
5754static void sctp_unhash(struct sock *sk)
5755{
5756 /* STUB */
5757}
5758
5759/* Check if port is acceptable. Possibly find first available port.
5760 *
5761 * The port hash table (contained in the 'global' SCTP protocol storage
5762 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5763 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5764 * list (the list number is the port number hashed out, so as you
5765 * would expect from a hash function, all the ports in a given list have
5766 * such a number that hashes out to the same list number; you were
5767 * expecting that, right?); so each list has a set of ports, with a
5768 * link to the socket (struct sock) that uses it, the port number and
5769 * a fastreuse flag (FIXME: NPI ipg).
5770 */
5771static struct sctp_bind_bucket *sctp_bucket_create(
5772 struct sctp_bind_hashbucket *head, unsigned short snum);
5773
5774static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5775{
5776 struct sctp_bind_hashbucket *head; /* hash list */
5777 struct sctp_bind_bucket *pp; /* hash list port iterator */
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005778 struct hlist_node *node;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005779 unsigned short snum;
5780 int ret;
5781
Al Viro04afd8b2006-11-20 17:02:01 -08005782 snum = ntohs(addr->v4.sin_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005783
5784 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5785 sctp_local_bh_disable();
5786
5787 if (snum == 0) {
Stephen Hemminger06393002007-10-10 17:30:18 -07005788 /* Search for an available port. */
Stephen Hemminger227b60f2007-10-10 17:30:46 -07005789 int low, high, remaining, index;
5790 unsigned int rover;
5791
5792 inet_get_local_port_range(&low, &high);
5793 remaining = (high - low) + 1;
5794 rover = net_random() % remaining + low;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005795
Linus Torvalds1da177e2005-04-16 15:20:36 -07005796 do {
5797 rover++;
5798 if ((rover < low) || (rover > high))
5799 rover = low;
Amerigo Wange3826f12010-05-05 00:27:06 +00005800 if (inet_is_reserved_local_port(rover))
5801 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005802 index = sctp_phashfn(rover);
5803 head = &sctp_port_hashtable[index];
5804 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005805 sctp_for_each_hentry(pp, node, &head->chain)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005806 if (pp->port == rover)
5807 goto next;
5808 break;
5809 next:
5810 sctp_spin_unlock(&head->lock);
5811 } while (--remaining > 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005812
5813 /* Exhausted local port range during search? */
5814 ret = 1;
5815 if (remaining <= 0)
5816 goto fail;
5817
5818 /* OK, here is the one we will use. HEAD (the port
5819 * hash table list entry) is non-NULL and we hold it's
5820 * mutex.
5821 */
5822 snum = rover;
5823 } else {
5824 /* We are given an specific port number; we verify
5825 * that it is not being used. If it is used, we will
5826 * exahust the search in the hash list corresponding
5827 * to the port number (snum) - we detect that with the
5828 * port iterator, pp being NULL.
5829 */
5830 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5831 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005832 sctp_for_each_hentry(pp, node, &head->chain) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005833 if (pp->port == snum)
5834 goto pp_found;
5835 }
5836 }
5837 pp = NULL;
5838 goto pp_not_found;
5839pp_found:
5840 if (!hlist_empty(&pp->owner)) {
5841 /* We had a port hash table hit - there is an
5842 * available port (pp != NULL) and it is being
5843 * used by other socket (pp->owner not empty); that other
5844 * socket is going to be sk2.
5845 */
5846 int reuse = sk->sk_reuse;
5847 struct sock *sk2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005848
5849 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005850 if (pp->fastreuse && sk->sk_reuse &&
5851 sk->sk_state != SCTP_SS_LISTENING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005852 goto success;
5853
5854 /* Run through the list of sockets bound to the port
5855 * (pp->port) [via the pointers bind_next and
5856 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5857 * we get the endpoint they describe and run through
5858 * the endpoint's list of IP (v4 or v6) addresses,
5859 * comparing each of the addresses with the address of
5860 * the socket sk. If we find a match, then that means
5861 * that this port/socket (sk) combination are already
5862 * in an endpoint.
5863 */
5864 sk_for_each_bound(sk2, node, &pp->owner) {
5865 struct sctp_endpoint *ep2;
5866 ep2 = sctp_sk(sk2)->ep;
5867
Vlad Yasevich4e540642008-07-18 23:06:32 -07005868 if (sk == sk2 ||
5869 (reuse && sk2->sk_reuse &&
5870 sk2->sk_state != SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005871 continue;
5872
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07005873 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5874 sctp_sk(sk2), sctp_sk(sk))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005875 ret = (long)sk2;
5876 goto fail_unlock;
5877 }
5878 }
5879 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5880 }
5881pp_not_found:
5882 /* If there was a hash table miss, create a new port. */
5883 ret = 1;
5884 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5885 goto fail_unlock;
5886
5887 /* In either case (hit or miss), make sure fastreuse is 1 only
5888 * if sk->sk_reuse is too (that is, if the caller requested
5889 * SO_REUSEADDR on this socket -sk-).
5890 */
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005891 if (hlist_empty(&pp->owner)) {
5892 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5893 pp->fastreuse = 1;
5894 else
5895 pp->fastreuse = 0;
5896 } else if (pp->fastreuse &&
5897 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005898 pp->fastreuse = 0;
5899
5900 /* We are set, so fill up all the data in the hash table
5901 * entry, tie the socket list information with the rest of the
5902 * sockets FIXME: Blurry, NPI (ipg).
5903 */
5904success:
Linus Torvalds1da177e2005-04-16 15:20:36 -07005905 if (!sctp_sk(sk)->bind_hash) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005906 inet_sk(sk)->inet_num = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005907 sk_add_bind_node(sk, &pp->owner);
5908 sctp_sk(sk)->bind_hash = pp;
5909 }
5910 ret = 0;
5911
5912fail_unlock:
5913 sctp_spin_unlock(&head->lock);
5914
5915fail:
5916 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07005917 return ret;
5918}
5919
5920/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5921 * port is requested.
5922 */
5923static int sctp_get_port(struct sock *sk, unsigned short snum)
5924{
5925 long ret;
5926 union sctp_addr addr;
5927 struct sctp_af *af = sctp_sk(sk)->pf->af;
5928
5929 /* Set up a dummy address struct from the sk. */
5930 af->from_sk(&addr, sk);
5931 addr.v4.sin_port = htons(snum);
5932
5933 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5934 ret = sctp_get_port_local(sk, &addr);
5935
Eric Dumazeta02cec22010-09-22 20:43:57 +00005936 return ret ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005937}
5938
5939/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07005940 * Move a socket to LISTENING state.
5941 */
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005942SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005943{
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005944 struct sctp_sock *sp = sctp_sk(sk);
5945 struct sctp_endpoint *ep = sp->ep;
Herbert Xu1b489e12006-08-20 15:07:14 +10005946 struct crypto_hash *tfm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005947
5948 /* Allocate HMAC for generating cookie. */
Vlad Yasevich23b29ed2008-07-18 23:06:07 -07005949 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
Herbert Xu1b489e12006-08-20 15:07:14 +10005950 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
Vlad Yasevich8dc49842007-05-09 13:50:20 -07005951 if (IS_ERR(tfm)) {
Joe Perchese87cc472012-05-13 21:56:26 +00005952 net_info_ratelimited("failed to load transform for %s: %ld\n",
5953 sctp_hmac_alg, PTR_ERR(tfm));
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005954 return -ENOSYS;
5955 }
5956 sctp_sk(sk)->hmac = tfm;
5957 }
5958
5959 /*
5960 * If a bind() or sctp_bindx() is not called prior to a listen()
5961 * call that allows new associations to be accepted, the system
5962 * picks an ephemeral port and will choose an address set equivalent
5963 * to binding with a wildcard address.
5964 *
5965 * This is not currently spelled out in the SCTP sockets
5966 * extensions draft, but follows the practice as seen in TCP
5967 * sockets.
5968 *
5969 */
5970 sk->sk_state = SCTP_SS_LISTENING;
5971 if (!ep->base.bind_addr.port) {
5972 if (sctp_autobind(sk))
5973 return -EAGAIN;
5974 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005975 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005976 sk->sk_state = SCTP_SS_CLOSED;
5977 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005978 }
5979 }
5980
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005981 sk->sk_max_ack_backlog = backlog;
5982 sctp_hash_endpoint(ep);
5983 return 0;
5984}
5985
5986/*
5987 * 4.1.3 / 5.1.3 listen()
5988 *
5989 * By default, new associations are not accepted for UDP style sockets.
5990 * An application uses listen() to mark a socket as being able to
5991 * accept new associations.
5992 *
5993 * On TCP style sockets, applications use listen() to ready the SCTP
5994 * endpoint for accepting inbound associations.
5995 *
5996 * On both types of endpoints a backlog of '0' disables listening.
5997 *
5998 * Move a socket to LISTENING state.
5999 */
6000int sctp_inet_listen(struct socket *sock, int backlog)
6001{
6002 struct sock *sk = sock->sk;
6003 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6004 int err = -EINVAL;
6005
6006 if (unlikely(backlog < 0))
6007 return err;
6008
6009 sctp_lock_sock(sk);
6010
6011 /* Peeled-off sockets are not allowed to listen(). */
6012 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6013 goto out;
6014
6015 if (sock->state != SS_UNCONNECTED)
6016 goto out;
6017
6018 /* If backlog is zero, disable listening. */
6019 if (!backlog) {
6020 if (sctp_sstate(sk, CLOSED))
6021 goto out;
6022
6023 err = 0;
6024 sctp_unhash_endpoint(ep);
6025 sk->sk_state = SCTP_SS_CLOSED;
6026 if (sk->sk_reuse)
6027 sctp_sk(sk)->bind_hash->fastreuse = 1;
6028 goto out;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07006029 }
6030
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00006031 /* If we are already listening, just update the backlog */
6032 if (sctp_sstate(sk, LISTENING))
6033 sk->sk_max_ack_backlog = backlog;
6034 else {
6035 err = sctp_listen_start(sk, backlog);
6036 if (err)
6037 goto out;
6038 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07006039
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00006040 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006041out:
6042 sctp_release_sock(sk);
6043 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006044}
6045
6046/*
6047 * This function is done by modeling the current datagram_poll() and the
6048 * tcp_poll(). Note that, based on these implementations, we don't
6049 * lock the socket in this function, even though it seems that,
6050 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08006051 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07006052 * in this place. We assume that we don't need locks either until proven
6053 * otherwise.
6054 *
6055 * Another thing to note is that we include the Async I/O support
6056 * here, again, by modeling the current TCP/UDP code. We don't have
6057 * a good way to test with it yet.
6058 */
6059unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6060{
6061 struct sock *sk = sock->sk;
6062 struct sctp_sock *sp = sctp_sk(sk);
6063 unsigned int mask;
6064
Eric Dumazetaa395142010-04-20 13:03:51 +00006065 poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006066
6067 /* A TCP-style listening socket becomes readable when the accept queue
6068 * is not empty.
6069 */
6070 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6071 return (!list_empty(&sp->ep->asocs)) ?
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09006072 (POLLIN | POLLRDNORM) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006073
6074 mask = 0;
6075
6076 /* Is there any exceptional events? */
6077 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6078 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08006079 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00006080 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006081 if (sk->sk_shutdown == SHUTDOWN_MASK)
6082 mask |= POLLHUP;
6083
6084 /* Is it readable? Reconsider this code with TCP-style support. */
Eric Dumazetdb409802010-09-06 11:13:50 +00006085 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07006086 mask |= POLLIN | POLLRDNORM;
6087
6088 /* The association is either gone or not ready. */
6089 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6090 return mask;
6091
6092 /* Is it writable? */
6093 if (sctp_writeable(sk)) {
6094 mask |= POLLOUT | POLLWRNORM;
6095 } else {
6096 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6097 /*
6098 * Since the socket is not locked, the buffer
6099 * might be made available after the writeable check and
6100 * before the bit is set. This could cause a lost I/O
6101 * signal. tcp_poll() has a race breaker for this race
6102 * condition. Based on their implementation, we put
6103 * in the following code to cover it as well.
6104 */
6105 if (sctp_writeable(sk))
6106 mask |= POLLOUT | POLLWRNORM;
6107 }
6108 return mask;
6109}
6110
6111/********************************************************************
6112 * 2nd Level Abstractions
6113 ********************************************************************/
6114
6115static struct sctp_bind_bucket *sctp_bucket_create(
6116 struct sctp_bind_hashbucket *head, unsigned short snum)
6117{
6118 struct sctp_bind_bucket *pp;
6119
Christoph Lameter54e6ecb2006-12-06 20:33:16 -08006120 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006121 if (pp) {
Li Zefan935a7f62008-04-10 01:58:06 -07006122 SCTP_DBG_OBJCNT_INC(bind_bucket);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006123 pp->port = snum;
6124 pp->fastreuse = 0;
6125 INIT_HLIST_HEAD(&pp->owner);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05006126 hlist_add_head(&pp->node, &head->chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006127 }
6128 return pp;
6129}
6130
6131/* Caller must hold hashbucket lock for this tb with local BH disabled */
6132static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6133{
Sridhar Samudrala37fa6872006-07-21 14:45:47 -07006134 if (pp && hlist_empty(&pp->owner)) {
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05006135 __hlist_del(&pp->node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006136 kmem_cache_free(sctp_bucket_cachep, pp);
6137 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6138 }
6139}
6140
6141/* Release this socket's reference to a local port. */
6142static inline void __sctp_put_port(struct sock *sk)
6143{
6144 struct sctp_bind_hashbucket *head =
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006145 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
Linus Torvalds1da177e2005-04-16 15:20:36 -07006146 struct sctp_bind_bucket *pp;
6147
6148 sctp_spin_lock(&head->lock);
6149 pp = sctp_sk(sk)->bind_hash;
6150 __sk_del_bind_node(sk);
6151 sctp_sk(sk)->bind_hash = NULL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006152 inet_sk(sk)->inet_num = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006153 sctp_bucket_destroy(pp);
6154 sctp_spin_unlock(&head->lock);
6155}
6156
6157void sctp_put_port(struct sock *sk)
6158{
6159 sctp_local_bh_disable();
6160 __sctp_put_port(sk);
6161 sctp_local_bh_enable();
6162}
6163
6164/*
6165 * The system picks an ephemeral port and choose an address set equivalent
6166 * to binding with a wildcard address.
6167 * One of those addresses will be the primary address for the association.
6168 * This automatically enables the multihoming capability of SCTP.
6169 */
6170static int sctp_autobind(struct sock *sk)
6171{
6172 union sctp_addr autoaddr;
6173 struct sctp_af *af;
Al Viro6fbfa9f2006-11-20 17:24:53 -08006174 __be16 port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006175
6176 /* Initialize a local sockaddr structure to INADDR_ANY. */
6177 af = sctp_sk(sk)->pf->af;
6178
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006179 port = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006180 af->inaddr_any(&autoaddr, port);
6181
6182 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6183}
6184
6185/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6186 *
6187 * From RFC 2292
6188 * 4.2 The cmsghdr Structure *
6189 *
6190 * When ancillary data is sent or received, any number of ancillary data
6191 * objects can be specified by the msg_control and msg_controllen members of
6192 * the msghdr structure, because each object is preceded by
6193 * a cmsghdr structure defining the object's length (the cmsg_len member).
6194 * Historically Berkeley-derived implementations have passed only one object
6195 * at a time, but this API allows multiple objects to be
6196 * passed in a single call to sendmsg() or recvmsg(). The following example
6197 * shows two ancillary data objects in a control buffer.
6198 *
6199 * |<--------------------------- msg_controllen -------------------------->|
6200 * | |
6201 *
6202 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6203 *
6204 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6205 * | | |
6206 *
6207 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6208 *
6209 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6210 * | | | | |
6211 *
6212 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6213 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6214 *
6215 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6216 *
6217 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6218 * ^
6219 * |
6220 *
6221 * msg_control
6222 * points here
6223 */
6224SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6225 sctp_cmsgs_t *cmsgs)
6226{
6227 struct cmsghdr *cmsg;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07006228 struct msghdr *my_msg = (struct msghdr *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006229
6230 for (cmsg = CMSG_FIRSTHDR(msg);
6231 cmsg != NULL;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07006232 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6233 if (!CMSG_OK(my_msg, cmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07006234 return -EINVAL;
6235
6236 /* Should we parse this header or ignore? */
6237 if (cmsg->cmsg_level != IPPROTO_SCTP)
6238 continue;
6239
6240 /* Strictly check lengths following example in SCM code. */
6241 switch (cmsg->cmsg_type) {
6242 case SCTP_INIT:
6243 /* SCTP Socket API Extension
6244 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6245 *
6246 * This cmsghdr structure provides information for
6247 * initializing new SCTP associations with sendmsg().
6248 * The SCTP_INITMSG socket option uses this same data
6249 * structure. This structure is not used for
6250 * recvmsg().
6251 *
6252 * cmsg_level cmsg_type cmsg_data[]
6253 * ------------ ------------ ----------------------
6254 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6255 */
6256 if (cmsg->cmsg_len !=
6257 CMSG_LEN(sizeof(struct sctp_initmsg)))
6258 return -EINVAL;
6259 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6260 break;
6261
6262 case SCTP_SNDRCV:
6263 /* SCTP Socket API Extension
6264 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6265 *
6266 * This cmsghdr structure specifies SCTP options for
6267 * sendmsg() and describes SCTP header information
6268 * about a received message through recvmsg().
6269 *
6270 * cmsg_level cmsg_type cmsg_data[]
6271 * ------------ ------------ ----------------------
6272 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6273 */
6274 if (cmsg->cmsg_len !=
6275 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6276 return -EINVAL;
6277
6278 cmsgs->info =
6279 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6280
6281 /* Minimally, validate the sinfo_flags. */
6282 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07006283 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6284 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07006285 return -EINVAL;
6286 break;
6287
6288 default:
6289 return -EINVAL;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07006290 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07006291 }
6292 return 0;
6293}
6294
6295/*
6296 * Wait for a packet..
6297 * Note: This function is the same function as in core/datagram.c
6298 * with a few modifications to make lksctp work.
6299 */
6300static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6301{
6302 int error;
6303 DEFINE_WAIT(wait);
6304
Eric Dumazetaa395142010-04-20 13:03:51 +00006305 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006306
6307 /* Socket errors? */
6308 error = sock_error(sk);
6309 if (error)
6310 goto out;
6311
6312 if (!skb_queue_empty(&sk->sk_receive_queue))
6313 goto ready;
6314
6315 /* Socket shut down? */
6316 if (sk->sk_shutdown & RCV_SHUTDOWN)
6317 goto out;
6318
6319 /* Sequenced packets can come disconnected. If so we report the
6320 * problem.
6321 */
6322 error = -ENOTCONN;
6323
6324 /* Is there a good reason to think that we may receive some data? */
6325 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6326 goto out;
6327
6328 /* Handle signals. */
6329 if (signal_pending(current))
6330 goto interrupted;
6331
6332 /* Let another process have a go. Since we are going to sleep
6333 * anyway. Note: This may cause odd behaviors if the message
6334 * does not fit in the user's buffer, but this seems to be the
6335 * only way to honor MSG_DONTWAIT realistically.
6336 */
6337 sctp_release_sock(sk);
6338 *timeo_p = schedule_timeout(*timeo_p);
6339 sctp_lock_sock(sk);
6340
6341ready:
Eric Dumazetaa395142010-04-20 13:03:51 +00006342 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006343 return 0;
6344
6345interrupted:
6346 error = sock_intr_errno(*timeo_p);
6347
6348out:
Eric Dumazetaa395142010-04-20 13:03:51 +00006349 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006350 *err = error;
6351 return error;
6352}
6353
6354/* Receive a datagram.
6355 * Note: This is pretty much the same routine as in core/datagram.c
6356 * with a few changes to make lksctp work.
6357 */
6358static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6359 int noblock, int *err)
6360{
6361 int error;
6362 struct sk_buff *skb;
6363 long timeo;
6364
Linus Torvalds1da177e2005-04-16 15:20:36 -07006365 timeo = sock_rcvtimeo(sk, noblock);
6366
6367 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6368 timeo, MAX_SCHEDULE_TIMEOUT);
6369
6370 do {
6371 /* Again only user level code calls this function,
6372 * so nothing interrupt level
6373 * will suddenly eat the receive_queue.
6374 *
6375 * Look at current nfs client by the way...
David Shwatrz8917a3c2010-12-02 09:01:55 +00006376 * However, this function was correct in any case. 8)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006377 */
6378 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07006379 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006380 skb = skb_peek(&sk->sk_receive_queue);
6381 if (skb)
6382 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07006383 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006384 } else {
6385 skb = skb_dequeue(&sk->sk_receive_queue);
6386 }
6387
6388 if (skb)
6389 return skb;
6390
Neil Horman6736dc32005-12-02 20:30:06 -08006391 /* Caller is allowed not to check sk->sk_err before calling. */
6392 error = sock_error(sk);
6393 if (error)
6394 goto no_packet;
6395
Linus Torvalds1da177e2005-04-16 15:20:36 -07006396 if (sk->sk_shutdown & RCV_SHUTDOWN)
6397 break;
6398
6399 /* User doesn't want to wait. */
6400 error = -EAGAIN;
6401 if (!timeo)
6402 goto no_packet;
6403 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6404
6405 return NULL;
6406
6407no_packet:
6408 *err = error;
6409 return NULL;
6410}
6411
6412/* If sndbuf has changed, wake up per association sndbuf waiters. */
6413static void __sctp_write_space(struct sctp_association *asoc)
6414{
6415 struct sock *sk = asoc->base.sk;
6416 struct socket *sock = sk->sk_socket;
6417
6418 if ((sctp_wspace(asoc) > 0) && sock) {
6419 if (waitqueue_active(&asoc->wait))
6420 wake_up_interruptible(&asoc->wait);
6421
6422 if (sctp_writeable(sk)) {
Eric Dumazeteaefd112011-02-18 03:26:36 +00006423 wait_queue_head_t *wq = sk_sleep(sk);
6424
6425 if (wq && waitqueue_active(wq))
6426 wake_up_interruptible(wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006427
6428 /* Note that we try to include the Async I/O support
6429 * here by modeling from the current TCP/UDP code.
6430 * We have not tested with it yet.
6431 */
Eric Dumazeteaefd112011-02-18 03:26:36 +00006432 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08006433 sock_wake_async(sock,
6434 SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006435 }
6436 }
6437}
6438
6439/* Do accounting for the sndbuf space.
6440 * Decrement the used sndbuf space of the corresponding association by the
6441 * data size which was just transmitted(freed).
6442 */
6443static void sctp_wfree(struct sk_buff *skb)
6444{
6445 struct sctp_association *asoc;
6446 struct sctp_chunk *chunk;
6447 struct sock *sk;
6448
6449 /* Get the saved chunk pointer. */
6450 chunk = *((struct sctp_chunk **)(skb->cb));
6451 asoc = chunk->asoc;
6452 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07006453 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6454 sizeof(struct sk_buff) +
6455 sizeof(struct sctp_chunk);
6456
Neil Horman4eb701d2005-04-28 12:02:04 -07006457 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6458
Neil Horman4d93df02007-08-15 16:07:44 -07006459 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006460 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
Neil Horman4d93df02007-08-15 16:07:44 -07006461 */
6462 sk->sk_wmem_queued -= skb->truesize;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006463 sk_mem_uncharge(sk, skb->truesize);
Neil Horman4d93df02007-08-15 16:07:44 -07006464
Neil Horman4eb701d2005-04-28 12:02:04 -07006465 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006466 __sctp_write_space(asoc);
6467
6468 sctp_association_put(asoc);
6469}
6470
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006471/* Do accounting for the receive space on the socket.
6472 * Accounting for the association is done in ulpevent.c
6473 * We set this as a destructor for the cloned data skbs so that
6474 * accounting is done at the correct time.
6475 */
6476void sctp_sock_rfree(struct sk_buff *skb)
6477{
6478 struct sock *sk = skb->sk;
6479 struct sctp_ulpevent *event = sctp_skb2event(skb);
6480
6481 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
Neil Horman4d93df02007-08-15 16:07:44 -07006482
6483 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006484 * Mimic the behavior of sock_rfree
Neil Horman4d93df02007-08-15 16:07:44 -07006485 */
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006486 sk_mem_uncharge(sk, event->rmem_len);
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006487}
6488
6489
Linus Torvalds1da177e2005-04-16 15:20:36 -07006490/* Helper function to wait for space in the sndbuf. */
6491static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6492 size_t msg_len)
6493{
6494 struct sock *sk = asoc->base.sk;
6495 int err = 0;
6496 long current_timeo = *timeo_p;
6497 DEFINE_WAIT(wait);
6498
6499 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09006500 asoc, (long)(*timeo_p), msg_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006501
6502 /* Increment the association's refcnt. */
6503 sctp_association_hold(asoc);
6504
6505 /* Wait on the association specific sndbuf space. */
6506 for (;;) {
6507 prepare_to_wait_exclusive(&asoc->wait, &wait,
6508 TASK_INTERRUPTIBLE);
6509 if (!*timeo_p)
6510 goto do_nonblock;
6511 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6512 asoc->base.dead)
6513 goto do_error;
6514 if (signal_pending(current))
6515 goto do_interrupted;
6516 if (msg_len <= sctp_wspace(asoc))
6517 break;
6518
6519 /* Let another process have a go. Since we are going
6520 * to sleep anyway.
6521 */
6522 sctp_release_sock(sk);
6523 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006524 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006525 sctp_lock_sock(sk);
6526
6527 *timeo_p = current_timeo;
6528 }
6529
6530out:
6531 finish_wait(&asoc->wait, &wait);
6532
6533 /* Release the association's refcnt. */
6534 sctp_association_put(asoc);
6535
6536 return err;
6537
6538do_error:
6539 err = -EPIPE;
6540 goto out;
6541
6542do_interrupted:
6543 err = sock_intr_errno(*timeo_p);
6544 goto out;
6545
6546do_nonblock:
6547 err = -EAGAIN;
6548 goto out;
6549}
6550
Wei Yongjun561b1732010-04-28 08:47:18 +00006551void sctp_data_ready(struct sock *sk, int len)
6552{
David S. Miller7ef52732010-05-02 21:43:40 -07006553 struct socket_wq *wq;
6554
6555 rcu_read_lock();
6556 wq = rcu_dereference(sk->sk_wq);
6557 if (wq_has_sleeper(wq))
6558 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
Wei Yongjun561b1732010-04-28 08:47:18 +00006559 POLLRDNORM | POLLRDBAND);
6560 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
David S. Miller7ef52732010-05-02 21:43:40 -07006561 rcu_read_unlock();
Wei Yongjun561b1732010-04-28 08:47:18 +00006562}
6563
Linus Torvalds1da177e2005-04-16 15:20:36 -07006564/* If socket sndbuf has changed, wake up all per association waiters. */
6565void sctp_write_space(struct sock *sk)
6566{
6567 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006568
6569 /* Wake up the tasks in each wait queue. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07006570 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006571 __sctp_write_space(asoc);
6572 }
6573}
6574
6575/* Is there any sndbuf space available on the socket?
6576 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08006577 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07006578 * associations on the same socket. For a UDP-style socket with
6579 * multiple associations, it is possible for it to be "unwriteable"
6580 * prematurely. I assume that this is acceptable because
6581 * a premature "unwriteable" is better than an accidental "writeable" which
6582 * would cause an unwanted block under certain circumstances. For the 1-1
6583 * UDP-style sockets or TCP-style sockets, this code should work.
6584 * - Daisy
6585 */
6586static int sctp_writeable(struct sock *sk)
6587{
6588 int amt = 0;
6589
Eric Dumazet31e6d362009-06-17 19:05:41 -07006590 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006591 if (amt < 0)
6592 amt = 0;
6593 return amt;
6594}
6595
6596/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6597 * returns immediately with EINPROGRESS.
6598 */
6599static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6600{
6601 struct sock *sk = asoc->base.sk;
6602 int err = 0;
6603 long current_timeo = *timeo_p;
6604 DEFINE_WAIT(wait);
6605
Harvey Harrison0dc47872008-03-05 20:47:47 -08006606 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006607 (long)(*timeo_p));
6608
6609 /* Increment the association's refcnt. */
6610 sctp_association_hold(asoc);
6611
6612 for (;;) {
6613 prepare_to_wait_exclusive(&asoc->wait, &wait,
6614 TASK_INTERRUPTIBLE);
6615 if (!*timeo_p)
6616 goto do_nonblock;
6617 if (sk->sk_shutdown & RCV_SHUTDOWN)
6618 break;
6619 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6620 asoc->base.dead)
6621 goto do_error;
6622 if (signal_pending(current))
6623 goto do_interrupted;
6624
6625 if (sctp_state(asoc, ESTABLISHED))
6626 break;
6627
6628 /* Let another process have a go. Since we are going
6629 * to sleep anyway.
6630 */
6631 sctp_release_sock(sk);
6632 current_timeo = schedule_timeout(current_timeo);
6633 sctp_lock_sock(sk);
6634
6635 *timeo_p = current_timeo;
6636 }
6637
6638out:
6639 finish_wait(&asoc->wait, &wait);
6640
6641 /* Release the association's refcnt. */
6642 sctp_association_put(asoc);
6643
6644 return err;
6645
6646do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08006647 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006648 err = -ETIMEDOUT;
6649 else
6650 err = -ECONNREFUSED;
6651 goto out;
6652
6653do_interrupted:
6654 err = sock_intr_errno(*timeo_p);
6655 goto out;
6656
6657do_nonblock:
6658 err = -EINPROGRESS;
6659 goto out;
6660}
6661
6662static int sctp_wait_for_accept(struct sock *sk, long timeo)
6663{
6664 struct sctp_endpoint *ep;
6665 int err = 0;
6666 DEFINE_WAIT(wait);
6667
6668 ep = sctp_sk(sk)->ep;
6669
6670
6671 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00006672 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006673 TASK_INTERRUPTIBLE);
6674
6675 if (list_empty(&ep->asocs)) {
6676 sctp_release_sock(sk);
6677 timeo = schedule_timeout(timeo);
6678 sctp_lock_sock(sk);
6679 }
6680
6681 err = -EINVAL;
6682 if (!sctp_sstate(sk, LISTENING))
6683 break;
6684
6685 err = 0;
6686 if (!list_empty(&ep->asocs))
6687 break;
6688
6689 err = sock_intr_errno(timeo);
6690 if (signal_pending(current))
6691 break;
6692
6693 err = -EAGAIN;
6694 if (!timeo)
6695 break;
6696 }
6697
Eric Dumazetaa395142010-04-20 13:03:51 +00006698 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006699
6700 return err;
6701}
6702
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +02006703static void sctp_wait_for_close(struct sock *sk, long timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006704{
6705 DEFINE_WAIT(wait);
6706
6707 do {
Eric Dumazetaa395142010-04-20 13:03:51 +00006708 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006709 if (list_empty(&sctp_sk(sk)->ep->asocs))
6710 break;
6711 sctp_release_sock(sk);
6712 timeout = schedule_timeout(timeout);
6713 sctp_lock_sock(sk);
6714 } while (!signal_pending(current) && timeout);
6715
Eric Dumazetaa395142010-04-20 13:03:51 +00006716 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006717}
6718
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006719static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6720{
6721 struct sk_buff *frag;
6722
6723 if (!skb->data_len)
6724 goto done;
6725
6726 /* Don't forget the fragments. */
David S. Miller1b003be2009-06-09 00:22:35 -07006727 skb_walk_frags(skb, frag)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006728 sctp_skb_set_owner_r_frag(frag, sk);
6729
6730done:
6731 sctp_skb_set_owner_r(skb, sk);
6732}
6733
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006734void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6735 struct sctp_association *asoc)
6736{
6737 struct inet_sock *inet = inet_sk(sk);
Julia Lawall09cb47a2010-01-21 02:43:20 -08006738 struct inet_sock *newinet;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006739
6740 newsk->sk_type = sk->sk_type;
6741 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6742 newsk->sk_flags = sk->sk_flags;
6743 newsk->sk_no_check = sk->sk_no_check;
6744 newsk->sk_reuse = sk->sk_reuse;
6745
6746 newsk->sk_shutdown = sk->sk_shutdown;
6747 newsk->sk_destruct = inet_sock_destruct;
6748 newsk->sk_family = sk->sk_family;
6749 newsk->sk_protocol = IPPROTO_SCTP;
6750 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6751 newsk->sk_sndbuf = sk->sk_sndbuf;
6752 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6753 newsk->sk_lingertime = sk->sk_lingertime;
6754 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6755 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6756
6757 newinet = inet_sk(newsk);
6758
6759 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6760 * getsockname() and getpeername()
6761 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006762 newinet->inet_sport = inet->inet_sport;
6763 newinet->inet_saddr = inet->inet_saddr;
6764 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6765 newinet->inet_dport = htons(asoc->peer.port);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006766 newinet->pmtudisc = inet->pmtudisc;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006767 newinet->inet_id = asoc->next_tsn ^ jiffies;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006768
6769 newinet->uc_ttl = inet->uc_ttl;
6770 newinet->mc_loop = 1;
6771 newinet->mc_ttl = 1;
6772 newinet->mc_index = 0;
6773 newinet->mc_list = NULL;
6774}
6775
Linus Torvalds1da177e2005-04-16 15:20:36 -07006776/* Populate the fields of the newsk from the oldsk and migrate the assoc
6777 * and its messages to the newsk.
6778 */
6779static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6780 struct sctp_association *assoc,
6781 sctp_socket_type_t type)
6782{
6783 struct sctp_sock *oldsp = sctp_sk(oldsk);
6784 struct sctp_sock *newsp = sctp_sk(newsk);
6785 struct sctp_bind_bucket *pp; /* hash list port iterator */
6786 struct sctp_endpoint *newep = newsp->ep;
6787 struct sk_buff *skb, *tmp;
6788 struct sctp_ulpevent *event;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006789 struct sctp_bind_hashbucket *head;
Michio Honda9f7d6532011-04-26 19:32:51 +09006790 struct list_head tmplist;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006791
6792 /* Migrate socket buffer sizes and all the socket level options to the
6793 * new socket.
6794 */
6795 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6796 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6797 /* Brute force copy old sctp opt. */
Michio Honda9f7d6532011-04-26 19:32:51 +09006798 if (oldsp->do_auto_asconf) {
6799 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6800 inet_sk_copy_descendant(newsk, oldsk);
6801 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6802 } else
6803 inet_sk_copy_descendant(newsk, oldsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006804
6805 /* Restore the ep value that was overwritten with the above structure
6806 * copy.
6807 */
6808 newsp->ep = newep;
6809 newsp->hmac = NULL;
6810
6811 /* Hook this new socket in to the bind_hash list. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006812 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006813 sctp_local_bh_disable();
6814 sctp_spin_lock(&head->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006815 pp = sctp_sk(oldsk)->bind_hash;
6816 sk_add_bind_node(newsk, &pp->owner);
6817 sctp_sk(newsk)->bind_hash = pp;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006818 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006819 sctp_spin_unlock(&head->lock);
6820 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07006821
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006822 /* Copy the bind_addr list from the original endpoint to the new
6823 * endpoint so that we can handle restarts properly
6824 */
Vlad Yasevich8e71a112007-12-06 22:50:54 -08006825 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6826 &oldsp->ep->base.bind_addr, GFP_KERNEL);
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006827
Linus Torvalds1da177e2005-04-16 15:20:36 -07006828 /* Move any messages in the old socket's receive queue that are for the
6829 * peeled off association to the new socket's receive queue.
6830 */
6831 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6832 event = sctp_skb2event(skb);
6833 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006834 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006835 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006836 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006837 }
6838 }
6839
6840 /* Clean up any messages pending delivery due to partial
6841 * delivery. Three cases:
6842 * 1) No partial deliver; no work.
6843 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6844 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6845 */
6846 skb_queue_head_init(&newsp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006847 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006848
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006849 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006850 struct sk_buff_head *queue;
6851
6852 /* Decide which queue to move pd_lobby skbs to. */
6853 if (assoc->ulpq.pd_mode) {
6854 queue = &newsp->pd_lobby;
6855 } else
6856 queue = &newsk->sk_receive_queue;
6857
6858 /* Walk through the pd_lobby, looking for skbs that
6859 * need moved to the new socket.
6860 */
6861 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6862 event = sctp_skb2event(skb);
6863 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006864 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006865 __skb_queue_tail(queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006866 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006867 }
6868 }
6869
6870 /* Clear up any skbs waiting for the partial
6871 * delivery to finish.
6872 */
6873 if (assoc->ulpq.pd_mode)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006874 sctp_clear_pd(oldsk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006875
6876 }
6877
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006878 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006879 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006880
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006881 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006882 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006883
Linus Torvalds1da177e2005-04-16 15:20:36 -07006884 /* Set the type of socket to indicate that it is peeled off from the
6885 * original UDP-style socket or created with the accept() call on a
6886 * TCP-style socket..
6887 */
6888 newsp->type = type;
6889
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006890 /* Mark the new socket "in-use" by the user so that any packets
6891 * that may arrive on the association after we've moved it are
6892 * queued to the backlog. This prevents a potential race between
6893 * backlog processing on the old socket and new-packet processing
6894 * on the new socket.
Zach Brown5131a182007-06-22 15:14:46 -07006895 *
6896 * The caller has just allocated newsk so we can guarantee that other
6897 * paths won't try to lock it and then oldsk.
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006898 */
Zach Brown5131a182007-06-22 15:14:46 -07006899 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006900 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006901
6902 /* If the association on the newsk is already closed before accept()
6903 * is called, set RCV_SHUTDOWN flag.
6904 */
6905 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6906 newsk->sk_shutdown |= RCV_SHUTDOWN;
6907
6908 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006909 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006910}
6911
Neil Horman4d93df02007-08-15 16:07:44 -07006912
Linus Torvalds1da177e2005-04-16 15:20:36 -07006913/* This proto struct describes the ULP interface for SCTP. */
6914struct proto sctp_prot = {
6915 .name = "SCTP",
6916 .owner = THIS_MODULE,
6917 .close = sctp_close,
6918 .connect = sctp_connect,
6919 .disconnect = sctp_disconnect,
6920 .accept = sctp_accept,
6921 .ioctl = sctp_ioctl,
6922 .init = sctp_init_sock,
6923 .destroy = sctp_destroy_sock,
6924 .shutdown = sctp_shutdown,
6925 .setsockopt = sctp_setsockopt,
6926 .getsockopt = sctp_getsockopt,
6927 .sendmsg = sctp_sendmsg,
6928 .recvmsg = sctp_recvmsg,
6929 .bind = sctp_bind,
6930 .backlog_rcv = sctp_backlog_rcv,
6931 .hash = sctp_hash,
6932 .unhash = sctp_unhash,
6933 .get_port = sctp_get_port,
6934 .obj_size = sizeof(struct sctp_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006935 .sysctl_mem = sysctl_sctp_mem,
6936 .sysctl_rmem = sysctl_sctp_rmem,
6937 .sysctl_wmem = sysctl_sctp_wmem,
6938 .memory_pressure = &sctp_memory_pressure,
6939 .enter_memory_pressure = sctp_enter_memory_pressure,
6940 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006941 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006942};
6943
Eric Dumazetdfd56b82011-12-10 09:48:31 +00006944#if IS_ENABLED(CONFIG_IPV6)
Eric Dumazet8295b6d2007-11-05 23:40:28 -08006945
Linus Torvalds1da177e2005-04-16 15:20:36 -07006946struct proto sctpv6_prot = {
6947 .name = "SCTPv6",
6948 .owner = THIS_MODULE,
6949 .close = sctp_close,
6950 .connect = sctp_connect,
6951 .disconnect = sctp_disconnect,
6952 .accept = sctp_accept,
6953 .ioctl = sctp_ioctl,
6954 .init = sctp_init_sock,
6955 .destroy = sctp_destroy_sock,
6956 .shutdown = sctp_shutdown,
6957 .setsockopt = sctp_setsockopt,
6958 .getsockopt = sctp_getsockopt,
6959 .sendmsg = sctp_sendmsg,
6960 .recvmsg = sctp_recvmsg,
6961 .bind = sctp_bind,
6962 .backlog_rcv = sctp_backlog_rcv,
6963 .hash = sctp_hash,
6964 .unhash = sctp_unhash,
6965 .get_port = sctp_get_port,
6966 .obj_size = sizeof(struct sctp6_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006967 .sysctl_mem = sysctl_sctp_mem,
6968 .sysctl_rmem = sysctl_sctp_rmem,
6969 .sysctl_wmem = sysctl_sctp_wmem,
6970 .memory_pressure = &sctp_memory_pressure,
6971 .enter_memory_pressure = sctp_enter_memory_pressure,
6972 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006973 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006974};
Eric Dumazetdfd56b82011-12-10 09:48:31 +00006975#endif /* IS_ENABLED(CONFIG_IPV6) */