blob: 6a691d84aef4223ddea29759f7bf8f4b8c25aa0e [file] [log] [blame]
Vlad Yasevich60c778b2008-01-11 09:57:09 -05001/* SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -05009 * This file is part of the SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050018 * This SCTP implementation is free software;
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050024 * This SCTP implementation is distributed in the hope that it
Linus Torvalds1da177e2005-04-16 15:20:36 -070025 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
Joe Perches145ce502010-08-24 13:21:08 +000060#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/types.h>
63#include <linux/kernel.h>
64#include <linux/wait.h>
65#include <linux/time.h>
66#include <linux/ip.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080067#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/fcntl.h>
69#include <linux/poll.h>
70#include <linux/init.h>
71#include <linux/crypto.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090072#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74#include <net/ip.h>
75#include <net/icmp.h>
76#include <net/route.h>
77#include <net/ipv6.h>
78#include <net/inet_common.h>
79
80#include <linux/socket.h> /* for sa_family_t */
81#include <net/sock.h>
82#include <net/sctp/sctp.h>
83#include <net/sctp/sm.h>
84
85/* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
88 */
89
90/* Forward declarations for internal helper functions. */
91static int sctp_writeable(struct sock *sk);
92static void sctp_wfree(struct sk_buff *skb);
93static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97static int sctp_wait_for_accept(struct sock *sk, long timeo);
98static void sctp_wait_for_close(struct sock *sk, long timeo);
99static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108static int sctp_autobind(struct sock *sk);
109static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
112
Neil Horman4d93df02007-08-15 16:07:44 -0700113extern struct kmem_cache *sctp_bucket_cachep;
114extern int sysctl_sctp_mem[3];
115extern int sysctl_sctp_rmem[3];
116extern int sysctl_sctp_wmem[3];
117
Adrian Bunkb6fa1a42007-09-12 15:18:00 +0200118static int sctp_memory_pressure;
119static atomic_t sctp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800120struct percpu_counter sctp_sockets_allocated;
Neil Horman4d93df02007-08-15 16:07:44 -0700121
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700122static void sctp_enter_memory_pressure(struct sock *sk)
Neil Horman4d93df02007-08-15 16:07:44 -0700123{
124 sctp_memory_pressure = 1;
125}
126
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128/* Get the sndbuf space available at the time on the association. */
129static inline int sctp_wspace(struct sctp_association *asoc)
130{
Neil Horman4d93df02007-08-15 16:07:44 -0700131 int amt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Neil Horman4d93df02007-08-15 16:07:44 -0700133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
Eric Dumazet31e6d362009-06-17 19:05:41 -0700136 amt = sk_wmem_alloc_get(asoc->base.sk);
Neil Horman4d93df02007-08-15 16:07:44 -0700137
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
145 }
Neil Horman4eb701d2005-04-28 12:02:04 -0700146 } else {
Neil Horman4d93df02007-08-15 16:07:44 -0700147 amt = asoc->base.sk->sk_sndbuf - amt;
Neil Horman4eb701d2005-04-28 12:02:04 -0700148 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 return amt;
150}
151
152/* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
155 *
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
160 */
161static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
162{
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
165
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
168
Neil Horman4eb701d2005-04-28 12:02:04 -0700169 skb_set_owner_w(chunk->skb, sk);
170
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
174
Neil Horman4eb701d2005-04-28 12:02:04 -0700175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
178
Neil Horman4eb701d2005-04-28 12:02:04 -0700179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182}
183
184/* Verify that this is a valid address. */
185static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
186 int len)
187{
188 struct sctp_af *af;
189
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 if (!af)
193 return -EINVAL;
194
195 /* Is this a valid SCTP address? */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return -EINVAL;
198
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
201
202 return 0;
203}
204
205/* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
207 */
208struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
209{
210 struct sctp_association *asoc = NULL;
211
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
217 */
218 if (!sctp_sstate(sk, ESTABLISHED))
219 return NULL;
220
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
225 return asoc;
226 }
227
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
230 return NULL;
231
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
235
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
237 return NULL;
238
239 return asoc;
240}
241
242/* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
244 * the same.
245 */
246static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
248 sctp_assoc_t id)
249{
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
253
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
Al Virocd4ff032006-11-20 17:11:33 -0800255 laddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 if (!addr_asoc)
259 return NULL;
260
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
264
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
267
268 return transport;
269}
270
271/* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
273 *
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
275 *
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
280 */
Frank Filz3f7a87d2005-06-20 13:14:57 -0700281SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282{
283 int retval = 0;
284
285 sctp_lock_sock(sk);
286
Frank Filz3f7a87d2005-06-20 13:14:57 -0700287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 addr_len);
294 else
295 retval = -EINVAL;
296
297 sctp_release_sock(sk);
298
299 return retval;
300}
301
302static long sctp_get_port_local(struct sock *, union sctp_addr *);
303
304/* Verify this is a valid sockaddr. */
305static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
307{
308 struct sctp_af *af;
309
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
312 return NULL;
313
Vlad Yasevich7dab83d2008-07-18 23:05:40 -0700314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
323 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
327
328 if (len < af->sockaddr_len)
329 return NULL;
330
331 return af;
332}
333
334/* Bind a local address either to an endpoint or to an association. */
335SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
336{
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
340 struct sctp_af *af;
341 unsigned short snum;
342 int ret = 0;
343
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 return -EINVAL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700350 }
351
352 snum = ntohs(addr->v4.sin_port);
353
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
356 sk,
357 addr,
358 bp->port, snum,
359 len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
364
Vlad Yasevich8b358052007-05-15 17:14:58 -0400365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
368 */
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
Vlad Yasevich8b358052007-05-15 17:14:58 -0400376 return -EINVAL;
377 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 }
379
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
382
Vlad Yasevich4e540642008-07-18 23:06:32 -0700383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
385 */
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
388
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
392 */
Vlad Yasevich2772b492007-08-21 14:24:30 +0900393 addr->v4.sin_port = htons(snum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 if ((ret = sctp_get_port_local(sk, addr))) {
Vlad Yasevich4e540642008-07-18 23:06:32 -0700395 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 }
397
398 /* Refresh ephemeral port. */
399 if (!bp->port)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000400 bp->port = inet_sk(sk)->inet_num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
Vlad Yasevich559cf712007-09-16 16:03:28 -0700402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
404 */
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 af->to_sk_saddr(addr, sk);
411 }
412
413 return ret;
414}
415
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * at any one time. If a sender, after sending an ASCONF chunk, decides
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * from each endpoint).
425 */
426static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
428{
429 int retval = 0;
430
431 /* If there is an outstanding ASCONF chunk, queue it for later
432 * transmission.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900433 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (asoc->addip_last_asconf) {
David S. Miller79af02c2005-07-08 21:47:49 -0700435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900436 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 }
438
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
446
447out:
448 return retval;
449}
450
451/* Add a list of addresses as bind addresses to local endpoint or
452 * association.
453 *
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
457 *
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
460 *
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
462 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200463static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464{
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
470
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
473
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
478 */
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
484 }
485
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 af->sockaddr_len);
488
489 addr_buf += af->sockaddr_len;
490
491err_bindx_add:
492 if (retval < 0) {
493 /* Failed. Cleanup the ones that have been added */
494 if (cnt > 0)
495 sctp_bindx_rem(sk, addrs, cnt);
496 return retval;
497 }
498 }
499
500 return retval;
501}
502
503/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
506 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900507 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
510 *
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 */
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900513static int sctp_send_asconf_add_ip(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 struct sockaddr *addrs,
515 int addrcnt)
516{
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700524 union sctp_addr saveaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 void *addr_buf;
526 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!sctp_addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800538 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700540 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
Vlad Yasevich559cf712007-09-16 16:03:28 -0700573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
Al Viro5ae955c2006-11-20 17:22:08 -0800579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
586 retval = sctp_send_asconf(asoc, chunk);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700587 if (retval)
588 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700593 addr_buf = addrs;
594 for (i = 0; i < addrcnt; i++) {
595 addr = (union sctp_addr *)addr_buf;
596 af = sctp_get_af_specific(addr->v4.sin_family);
597 memcpy(&saveaddr, addr, af->sockaddr_len);
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800598 retval = sctp_add_bind_addr(bp, &saveaddr,
599 SCTP_ADDR_NEW, GFP_ATOMIC);
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700600 addr_buf += af->sockaddr_len;
601 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 }
603
604out:
605 return retval;
606}
607
608/* Remove a list of addresses from bind addresses list. Do not remove the
609 * last address.
610 *
611 * Basically run through each address specified in the addrs/addrcnt
612 * array/length pair, determine if it is IPv6 or IPv4 and call
613 * sctp_del_bind() on it.
614 *
615 * If any of them fails, then the operation will be reversed and the
616 * ones that were removed will be added back.
617 *
618 * At least one address has to be left; if only one address is
619 * available, the operation will return -EBUSY.
620 *
621 * Only sctp_setsockopt_bindx() is supposed to call this function.
622 */
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +0200623static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 struct sctp_sock *sp = sctp_sk(sk);
626 struct sctp_endpoint *ep = sp->ep;
627 int cnt;
628 struct sctp_bind_addr *bp = &ep->base.bind_addr;
629 int retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 void *addr_buf;
Al Viroc9a08502006-11-20 17:07:48 -0800631 union sctp_addr *sa_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 struct sctp_af *af;
633
634 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
635 sk, addrs, addrcnt);
636
637 addr_buf = addrs;
638 for (cnt = 0; cnt < addrcnt; cnt++) {
639 /* If the bind address list is empty or if there is only one
640 * bind address, there is nothing more to be removed (we need
641 * at least one address here).
642 */
643 if (list_empty(&bp->address_list) ||
644 (sctp_list_single_entry(&bp->address_list))) {
645 retval = -EBUSY;
646 goto err_bindx_rem;
647 }
648
Al Viroc9a08502006-11-20 17:07:48 -0800649 sa_addr = (union sctp_addr *)addr_buf;
650 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 if (!af) {
652 retval = -EINVAL;
653 goto err_bindx_rem;
654 }
Paolo Galtieri0304ff8a2007-04-17 12:52:36 -0700655
656 if (!af->addr_valid(sa_addr, sp, NULL)) {
657 retval = -EADDRNOTAVAIL;
658 goto err_bindx_rem;
659 }
660
Al Viroc9a08502006-11-20 17:07:48 -0800661 if (sa_addr->v4.sin_port != htons(bp->port)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 retval = -EINVAL;
663 goto err_bindx_rem;
664 }
665
666 /* FIXME - There is probably a need to check if sk->sk_saddr and
667 * sk->sk_rcv_addr are currently set to one of the addresses to
668 * be removed. This is something which needs to be looked into
669 * when we are fixing the outstanding issues with multi-homing
670 * socket routing and failover schemes. Refer to comments in
671 * sctp_do_bind(). -daisy
672 */
Vlad Yasevich0ed90fb2007-10-24 16:10:00 -0400673 retval = sctp_del_bind_addr(bp, sa_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
675 addr_buf += af->sockaddr_len;
676err_bindx_rem:
677 if (retval < 0) {
678 /* Failed. Add the ones that has been removed back */
679 if (cnt > 0)
680 sctp_bindx_add(sk, addrs, cnt);
681 return retval;
682 }
683 }
684
685 return retval;
686}
687
688/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
689 * the associations that are part of the endpoint indicating that a list of
690 * local addresses are removed from the endpoint.
691 *
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900692 * If any of the addresses is already in the bind address list of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 * association, we do not send the chunk for that association. But it will not
694 * affect other associations.
695 *
696 * Only sctp_setsockopt_bindx() is supposed to call this function.
697 */
698static int sctp_send_asconf_del_ip(struct sock *sk,
699 struct sockaddr *addrs,
700 int addrcnt)
701{
702 struct sctp_sock *sp;
703 struct sctp_endpoint *ep;
704 struct sctp_association *asoc;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700705 struct sctp_transport *transport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 struct sctp_bind_addr *bp;
707 struct sctp_chunk *chunk;
708 union sctp_addr *laddr;
709 void *addr_buf;
710 struct sctp_af *af;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700711 struct sctp_sockaddr_entry *saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 int i;
713 int retval = 0;
714
715 if (!sctp_addip_enable)
716 return retval;
717
718 sp = sctp_sk(sk);
719 ep = sp->ep;
720
721 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -0800722 __func__, sk, addrs, addrcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700724 list_for_each_entry(asoc, &ep->asocs, asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
726 if (!asoc->peer.asconf_capable)
727 continue;
728
729 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
730 continue;
731
732 if (!sctp_state(asoc, ESTABLISHED))
733 continue;
734
735 /* Check if any address in the packed array of addresses is
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900736 * not present in the bind address list of the association.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 * If so, do not send the asconf chunk to its peer, but
738 * continue with other associations.
739 */
740 addr_buf = addrs;
741 for (i = 0; i < addrcnt; i++) {
742 laddr = (union sctp_addr *)addr_buf;
743 af = sctp_get_af_specific(laddr->v4.sin_family);
744 if (!af) {
745 retval = -EINVAL;
746 goto out;
747 }
748
749 if (!sctp_assoc_lookup_laddr(asoc, laddr))
750 break;
751
752 addr_buf += af->sockaddr_len;
753 }
754 if (i < addrcnt)
755 continue;
756
757 /* Find one address in the association's bind address list
758 * that is not in the packed array of addresses. This is to
759 * make sure that we do not delete all the addresses in the
760 * association.
761 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 bp = &asoc->base.bind_addr;
763 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
764 addrcnt, sp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 if (!laddr)
766 continue;
767
Vlad Yasevich559cf712007-09-16 16:03:28 -0700768 /* We do not need RCU protection throughout this loop
769 * because this is done under a socket lock from the
770 * setsockopt call.
771 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
773 SCTP_PARAM_DEL_IP);
774 if (!chunk) {
775 retval = -ENOMEM;
776 goto out;
777 }
778
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700779 /* Reset use_as_src flag for the addresses in the bind address
780 * list that are to be deleted.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 */
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700782 addr_buf = addrs;
783 for (i = 0; i < addrcnt; i++) {
784 laddr = (union sctp_addr *)addr_buf;
785 af = sctp_get_af_specific(laddr->v4.sin_family);
Vlad Yasevich559cf712007-09-16 16:03:28 -0700786 list_for_each_entry(saddr, &bp->address_list, list) {
Al Viro5f242a12006-11-20 17:05:23 -0800787 if (sctp_cmp_addr_exact(&saddr->a, laddr))
Vlad Yasevichf57d96b2007-12-20 14:12:24 -0800788 saddr->state = SCTP_ADDR_DEL;
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700789 }
790 addr_buf += af->sockaddr_len;
791 }
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700792
793 /* Update the route and saddr entries for all the transports
794 * as some of the addresses in the bind address list are
795 * about to be deleted and cannot be used as source addresses.
796 */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700797 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
798 transports) {
Sridhar Samudraladc022a92006-07-21 14:49:25 -0700799 dst_release(transport->dst);
800 sctp_transport_route(transport, NULL,
801 sctp_sk(asoc->base.sk));
802 }
803
804 retval = sctp_send_asconf(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 }
806out:
807 return retval;
808}
809
810/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
811 *
812 * API 8.1
813 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
814 * int flags);
815 *
816 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
817 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
818 * or IPv6 addresses.
819 *
820 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
821 * Section 3.1.2 for this usage.
822 *
823 * addrs is a pointer to an array of one or more socket addresses. Each
824 * address is contained in its appropriate structure (i.e. struct
825 * sockaddr_in or struct sockaddr_in6) the family of the address type
Ville Nuorvala23c435f2006-10-16 22:08:28 -0700826 * must be used to distinguish the address length (note that this
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 * representation is termed a "packed array" of addresses). The caller
828 * specifies the number of addresses in the array with addrcnt.
829 *
830 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
831 * -1, and sets errno to the appropriate error code.
832 *
833 * For SCTP, the port given in each socket address must be the same, or
834 * sctp_bindx() will fail, setting errno to EINVAL.
835 *
836 * The flags parameter is formed from the bitwise OR of zero or more of
837 * the following currently defined flags:
838 *
839 * SCTP_BINDX_ADD_ADDR
840 *
841 * SCTP_BINDX_REM_ADDR
842 *
843 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
844 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
845 * addresses from the association. The two flags are mutually exclusive;
846 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
847 * not remove all addresses from an association; sctp_bindx() will
848 * reject such an attempt with EINVAL.
849 *
850 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
851 * additional addresses with an endpoint after calling bind(). Or use
852 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
853 * socket is associated with so that no new association accepted will be
854 * associated with those addresses. If the endpoint supports dynamic
855 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
856 * endpoint to send the appropriate message to the peer to change the
857 * peers address lists.
858 *
859 * Adding and removing addresses from a connected association is
860 * optional functionality. Implementations that do not support this
861 * functionality should return EOPNOTSUPP.
862 *
863 * Basically do nothing but copying the addresses from user to kernel
864 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
Frank Filz3f7a87d2005-06-20 13:14:57 -0700865 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
866 * from userspace.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 *
868 * We don't use copy_from_user() for optimization: we first do the
869 * sanity checks (buffer size -fast- and access check-healthy
870 * pointer); if all of those succeed, then we can alloc the memory
871 * (expensive operation) needed to copy the data to kernel. Then we do
872 * the copying without checking the user space area
873 * (__copy_from_user()).
874 *
875 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
876 * it.
877 *
878 * sk The sk of the socket
879 * addrs The pointer to the addresses in user land
880 * addrssize Size of the addrs buffer
881 * op Operation to perform (add or remove, see the flags of
882 * sctp_bindx)
883 *
884 * Returns 0 if ok, <0 errno code on error.
885 */
886SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
887 struct sockaddr __user *addrs,
888 int addrs_size, int op)
889{
890 struct sockaddr *kaddrs;
891 int err;
892 int addrcnt = 0;
893 int walk_size = 0;
894 struct sockaddr *sa_addr;
895 void *addr_buf;
896 struct sctp_af *af;
897
898 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
899 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
900
901 if (unlikely(addrs_size <= 0))
902 return -EINVAL;
903
904 /* Check the user passed a healthy pointer. */
905 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
906 return -EFAULT;
907
908 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800909 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 if (unlikely(!kaddrs))
911 return -ENOMEM;
912
913 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
914 kfree(kaddrs);
915 return -EFAULT;
916 }
917
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900918 /* Walk through the addrs buffer and count the number of addresses. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 addr_buf = kaddrs;
920 while (walk_size < addrs_size) {
921 sa_addr = (struct sockaddr *)addr_buf;
922 af = sctp_get_af_specific(sa_addr->sa_family);
923
924 /* If the address family is not supported or if this address
925 * causes the address buffer to overflow return EINVAL.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900926 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
928 kfree(kaddrs);
929 return -EINVAL;
930 }
931 addrcnt++;
932 addr_buf += af->sockaddr_len;
933 walk_size += af->sockaddr_len;
934 }
935
936 /* Do the work. */
937 switch (op) {
938 case SCTP_BINDX_ADD_ADDR:
939 err = sctp_bindx_add(sk, kaddrs, addrcnt);
940 if (err)
941 goto out;
942 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
943 break;
944
945 case SCTP_BINDX_REM_ADDR:
946 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
947 if (err)
948 goto out;
949 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
950 break;
951
952 default:
953 err = -EINVAL;
954 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700955 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956
957out:
958 kfree(kaddrs);
959
960 return err;
961}
962
Frank Filz3f7a87d2005-06-20 13:14:57 -0700963/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
964 *
965 * Common routine for handling connect() and sctp_connectx().
966 * Connect will come in with just a single address.
967 */
968static int __sctp_connect(struct sock* sk,
969 struct sockaddr *kaddrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -0700970 int addrs_size,
971 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -0700972{
973 struct sctp_sock *sp;
974 struct sctp_endpoint *ep;
975 struct sctp_association *asoc = NULL;
976 struct sctp_association *asoc2;
977 struct sctp_transport *transport;
978 union sctp_addr to;
979 struct sctp_af *af;
980 sctp_scope_t scope;
981 long timeo;
982 int err = 0;
983 int addrcnt = 0;
984 int walk_size = 0;
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -0400985 union sctp_addr *sa_addr = NULL;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700986 void *addr_buf;
Vlad Yasevich16d00fb2007-05-04 13:34:09 -0700987 unsigned short port;
Vlad Yasevichf50f95c2007-07-03 12:47:40 -0400988 unsigned int f_flags = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700989
990 sp = sctp_sk(sk);
991 ep = sp->ep;
992
993 /* connect() cannot be done on a socket that is already in ESTABLISHED
994 * state - UDP-style peeled off socket or a TCP-style socket that
995 * is already connected.
996 * It cannot be done even on a TCP-style listening socket.
997 */
998 if (sctp_sstate(sk, ESTABLISHED) ||
999 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1000 err = -EISCONN;
1001 goto out_free;
1002 }
1003
1004 /* Walk through the addrs buffer and count the number of addresses. */
1005 addr_buf = kaddrs;
1006 while (walk_size < addrs_size) {
Al Viro4bdf4b52006-11-20 17:10:20 -08001007 sa_addr = (union sctp_addr *)addr_buf;
1008 af = sctp_get_af_specific(sa_addr->sa.sa_family);
Vlad Yasevich16d00fb2007-05-04 13:34:09 -07001009 port = ntohs(sa_addr->v4.sin_port);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001010
1011 /* If the address family is not supported or if this address
1012 * causes the address buffer to overflow return EINVAL.
1013 */
1014 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1015 err = -EINVAL;
1016 goto out_free;
1017 }
1018
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001019 /* Save current address so we can work with it */
1020 memcpy(&to, sa_addr, af->sockaddr_len);
1021
1022 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001023 if (err)
1024 goto out_free;
1025
Vlad Yasevich16d00fb2007-05-04 13:34:09 -07001026 /* Make sure the destination port is correctly set
1027 * in all addresses.
1028 */
1029 if (asoc && asoc->peer.port && asoc->peer.port != port)
1030 goto out_free;
1031
Frank Filz3f7a87d2005-06-20 13:14:57 -07001032
1033 /* Check if there already is a matching association on the
1034 * endpoint (other than the one created here).
1035 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001036 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001037 if (asoc2 && asoc2 != asoc) {
1038 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1039 err = -EISCONN;
1040 else
1041 err = -EALREADY;
1042 goto out_free;
1043 }
1044
1045 /* If we could not find a matching association on the endpoint,
1046 * make sure that there is no peeled-off association matching
1047 * the peer address even on another socket.
1048 */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001049 if (sctp_endpoint_is_peeled_off(ep, &to)) {
Frank Filz3f7a87d2005-06-20 13:14:57 -07001050 err = -EADDRNOTAVAIL;
1051 goto out_free;
1052 }
1053
1054 if (!asoc) {
1055 /* If a bind() or sctp_bindx() is not called prior to
1056 * an sctp_connectx() call, the system picks an
1057 * ephemeral port and will choose an address set
1058 * equivalent to binding with a wildcard address.
1059 */
1060 if (!ep->base.bind_addr.port) {
1061 if (sctp_autobind(sk)) {
1062 err = -EAGAIN;
1063 goto out_free;
1064 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001065 } else {
1066 /*
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001067 * If an unprivileged user inherits a 1-many
1068 * style socket with open associations on a
1069 * privileged port, it MAY be permitted to
1070 * accept new associations, but it SHOULD NOT
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001071 * be permitted to open new associations.
1072 */
1073 if (ep->base.bind_addr.port < PROT_SOCK &&
1074 !capable(CAP_NET_BIND_SERVICE)) {
1075 err = -EACCES;
1076 goto out_free;
1077 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001078 }
1079
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001080 scope = sctp_scope(&to);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001081 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1082 if (!asoc) {
1083 err = -ENOMEM;
1084 goto out_free;
1085 }
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001086
1087 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1088 GFP_KERNEL);
1089 if (err < 0) {
1090 goto out_free;
1091 }
1092
Frank Filz3f7a87d2005-06-20 13:14:57 -07001093 }
1094
1095 /* Prime the peer's transport structures. */
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001096 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001097 SCTP_UNKNOWN);
1098 if (!transport) {
1099 err = -ENOMEM;
1100 goto out_free;
1101 }
1102
1103 addrcnt++;
1104 addr_buf += af->sockaddr_len;
1105 walk_size += af->sockaddr_len;
1106 }
1107
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001108 /* In case the user of sctp_connectx() wants an association
1109 * id back, assign one now.
1110 */
1111 if (assoc_id) {
1112 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1113 if (err < 0)
1114 goto out_free;
1115 }
1116
Frank Filz3f7a87d2005-06-20 13:14:57 -07001117 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1118 if (err < 0) {
1119 goto out_free;
1120 }
1121
1122 /* Initialize sk's dport and daddr for getpeername() */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001123 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
Vlad Yaseviche4d1fea2007-08-01 10:56:43 -04001124 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1125 af->to_sk_daddr(sa_addr, sk);
Sridhar Samudrala8de8c872006-05-19 10:58:12 -07001126 sk->sk_err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001127
Vlad Yasevichf50f95c2007-07-03 12:47:40 -04001128 /* in-kernel sockets don't generally have a file allocated to them
1129 * if all they do is call sock_create_kern().
1130 */
1131 if (sk->sk_socket->file)
1132 f_flags = sk->sk_socket->file->f_flags;
1133
1134 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1135
Frank Filz3f7a87d2005-06-20 13:14:57 -07001136 err = sctp_wait_for_connect(asoc, &timeo);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001137 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001138 *assoc_id = asoc->assoc_id;
Frank Filz3f7a87d2005-06-20 13:14:57 -07001139
1140 /* Don't free association on exit. */
1141 asoc = NULL;
1142
1143out_free:
1144
1145 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001146 " kaddrs: %p err: %d\n",
1147 asoc, kaddrs, err);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001148 if (asoc)
1149 sctp_association_free(asoc);
1150 return err;
1151}
1152
1153/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1154 *
1155 * API 8.9
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001156 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1157 * sctp_assoc_t *asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001158 *
1159 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1160 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1161 * or IPv6 addresses.
1162 *
1163 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1164 * Section 3.1.2 for this usage.
1165 *
1166 * addrs is a pointer to an array of one or more socket addresses. Each
1167 * address is contained in its appropriate structure (i.e. struct
1168 * sockaddr_in or struct sockaddr_in6) the family of the address type
1169 * must be used to distengish the address length (note that this
1170 * representation is termed a "packed array" of addresses). The caller
1171 * specifies the number of addresses in the array with addrcnt.
1172 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001173 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1174 * the association id of the new association. On failure, sctp_connectx()
1175 * returns -1, and sets errno to the appropriate error code. The assoc_id
1176 * is not touched by the kernel.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001177 *
1178 * For SCTP, the port given in each socket address must be the same, or
1179 * sctp_connectx() will fail, setting errno to EINVAL.
1180 *
1181 * An application can use sctp_connectx to initiate an association with
1182 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1183 * allows a caller to specify multiple addresses at which a peer can be
1184 * reached. The way the SCTP stack uses the list of addresses to set up
1185 * the association is implementation dependant. This function only
1186 * specifies that the stack will try to make use of all the addresses in
1187 * the list when needed.
1188 *
1189 * Note that the list of addresses passed in is only used for setting up
1190 * the association. It does not necessarily equal the set of addresses
1191 * the peer uses for the resulting association. If the caller wants to
1192 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1193 * retrieve them after the association has been set up.
1194 *
1195 * Basically do nothing but copying the addresses from user to kernel
1196 * land and invoking either sctp_connectx(). This is used for tunneling
1197 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1198 *
1199 * We don't use copy_from_user() for optimization: we first do the
1200 * sanity checks (buffer size -fast- and access check-healthy
1201 * pointer); if all of those succeed, then we can alloc the memory
1202 * (expensive operation) needed to copy the data to kernel. Then we do
1203 * the copying without checking the user space area
1204 * (__copy_from_user()).
1205 *
1206 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1207 * it.
1208 *
1209 * sk The sk of the socket
1210 * addrs The pointer to the addresses in user land
1211 * addrssize Size of the addrs buffer
1212 *
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001213 * Returns >=0 if ok, <0 errno code on error.
Frank Filz3f7a87d2005-06-20 13:14:57 -07001214 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001215SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
Frank Filz3f7a87d2005-06-20 13:14:57 -07001216 struct sockaddr __user *addrs,
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001217 int addrs_size,
1218 sctp_assoc_t *assoc_id)
Frank Filz3f7a87d2005-06-20 13:14:57 -07001219{
1220 int err = 0;
1221 struct sockaddr *kaddrs;
1222
1223 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001224 __func__, sk, addrs, addrs_size);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001225
1226 if (unlikely(addrs_size <= 0))
1227 return -EINVAL;
1228
1229 /* Check the user passed a healthy pointer. */
1230 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1231 return -EFAULT;
1232
1233 /* Alloc space for the address array in kernel memory. */
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001234 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001235 if (unlikely(!kaddrs))
1236 return -ENOMEM;
1237
1238 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1239 err = -EFAULT;
1240 } else {
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001241 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
Frank Filz3f7a87d2005-06-20 13:14:57 -07001242 }
1243
1244 kfree(kaddrs);
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001245
Frank Filz3f7a87d2005-06-20 13:14:57 -07001246 return err;
1247}
1248
Vlad Yasevich88a0a942008-05-09 15:14:11 -07001249/*
1250 * This is an older interface. It's kept for backward compatibility
1251 * to the option that doesn't provide association id.
1252 */
1253SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1254 struct sockaddr __user *addrs,
1255 int addrs_size)
1256{
1257 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1258}
1259
1260/*
1261 * New interface for the API. The since the API is done with a socket
1262 * option, to make it simple we feed back the association id is as a return
1263 * indication to the call. Error is always negative and association id is
1264 * always positive.
1265 */
1266SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1267 struct sockaddr __user *addrs,
1268 int addrs_size)
1269{
1270 sctp_assoc_t assoc_id = 0;
1271 int err = 0;
1272
1273 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1274
1275 if (err)
1276 return err;
1277 else
1278 return assoc_id;
1279}
1280
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001281/*
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001282 * New (hopefully final) interface for the API.
1283 * We use the sctp_getaddrs_old structure so that use-space library
1284 * can avoid any unnecessary allocations. The only defferent part
1285 * is that we store the actual length of the address buffer into the
1286 * addrs_num structure member. That way we can re-use the existing
1287 * code.
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001288 */
1289SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1290 char __user *optval,
1291 int __user *optlen)
1292{
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001293 struct sctp_getaddrs_old param;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001294 sctp_assoc_t assoc_id = 0;
1295 int err = 0;
1296
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001297 if (len < sizeof(param))
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001298 return -EINVAL;
1299
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001300 if (copy_from_user(&param, optval, sizeof(param)))
1301 return -EFAULT;
1302
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001303 err = __sctp_setsockopt_connectx(sk,
Vlad Yasevichf9c67812009-11-11 08:19:24 +00001304 (struct sockaddr __user *)param.addrs,
1305 param.addr_num, &assoc_id);
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001306
1307 if (err == 0 || err == -EINPROGRESS) {
1308 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1309 return -EFAULT;
1310 if (put_user(sizeof(assoc_id), optlen))
1311 return -EFAULT;
1312 }
1313
1314 return err;
1315}
1316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317/* API 3.1.4 close() - UDP Style Syntax
1318 * Applications use close() to perform graceful shutdown (as described in
1319 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1320 * by a UDP-style socket.
1321 *
1322 * The syntax is
1323 *
1324 * ret = close(int sd);
1325 *
1326 * sd - the socket descriptor of the associations to be closed.
1327 *
1328 * To gracefully shutdown a specific association represented by the
1329 * UDP-style socket, an application should use the sendmsg() call,
1330 * passing no user data, but including the appropriate flag in the
1331 * ancillary data (see Section xxxx).
1332 *
1333 * If sd in the close() call is a branched-off socket representing only
1334 * one association, the shutdown is performed on that association only.
1335 *
1336 * 4.1.6 close() - TCP Style Syntax
1337 *
1338 * Applications use close() to gracefully close down an association.
1339 *
1340 * The syntax is:
1341 *
1342 * int close(int sd);
1343 *
1344 * sd - the socket descriptor of the association to be closed.
1345 *
1346 * After an application calls close() on a socket descriptor, no further
1347 * socket operations will succeed on that descriptor.
1348 *
1349 * API 7.1.4 SO_LINGER
1350 *
1351 * An application using the TCP-style socket can use this option to
1352 * perform the SCTP ABORT primitive. The linger option structure is:
1353 *
1354 * struct linger {
1355 * int l_onoff; // option on/off
1356 * int l_linger; // linger time
1357 * };
1358 *
1359 * To enable the option, set l_onoff to 1. If the l_linger value is set
1360 * to 0, calling close() is the same as the ABORT primitive. If the
1361 * value is set to a negative value, the setsockopt() call will return
1362 * an error. If the value is set to a positive value linger_time, the
1363 * close() can be blocked for at most linger_time ms. If the graceful
1364 * shutdown phase does not finish during this period, close() will
1365 * return but the graceful shutdown phase continues in the system.
1366 */
1367SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1368{
1369 struct sctp_endpoint *ep;
1370 struct sctp_association *asoc;
1371 struct list_head *pos, *temp;
1372
1373 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1374
1375 sctp_lock_sock(sk);
1376 sk->sk_shutdown = SHUTDOWN_MASK;
Vlad Yasevichbec96402009-07-30 18:08:28 -04001377 sk->sk_state = SCTP_SS_CLOSING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378
1379 ep = sctp_sk(sk)->ep;
1380
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07001381 /* Walk all associations on an endpoint. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 list_for_each_safe(pos, temp, &ep->asocs) {
1383 asoc = list_entry(pos, struct sctp_association, asocs);
1384
1385 if (sctp_style(sk, TCP)) {
1386 /* A closed association can still be in the list if
1387 * it belongs to a TCP-style listening socket that is
1388 * not yet accepted. If so, free it. If not, send an
1389 * ABORT or SHUTDOWN based on the linger options.
1390 */
1391 if (sctp_state(asoc, CLOSED)) {
1392 sctp_unhash_established(asoc);
1393 sctp_association_free(asoc);
Vladislav Yasevichb89498a2006-05-19 14:32:06 -07001394 continue;
1395 }
1396 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397
Sridhar Samudralab9ac8672006-08-28 13:53:01 -07001398 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1399 struct sctp_chunk *chunk;
1400
1401 chunk = sctp_make_abort_user(asoc, NULL, 0);
1402 if (chunk)
1403 sctp_primitive_ABORT(asoc, chunk);
1404 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 sctp_primitive_SHUTDOWN(asoc, NULL);
1406 }
1407
1408 /* Clean up any skbs sitting on the receive queue. */
1409 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1410 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1411
1412 /* On a TCP-style socket, block for at most linger_time if set. */
1413 if (sctp_style(sk, TCP) && timeout)
1414 sctp_wait_for_close(sk, timeout);
1415
1416 /* This will run the backlog queue. */
1417 sctp_release_sock(sk);
1418
1419 /* Supposedly, no process has access to the socket, but
1420 * the net layers still may.
1421 */
1422 sctp_local_bh_disable();
1423 sctp_bh_lock_sock(sk);
1424
1425 /* Hold the sock, since sk_common_release() will put sock_put()
1426 * and we have just a little more cleanup.
1427 */
1428 sock_hold(sk);
1429 sk_common_release(sk);
1430
1431 sctp_bh_unlock_sock(sk);
1432 sctp_local_bh_enable();
1433
1434 sock_put(sk);
1435
1436 SCTP_DBG_OBJCNT_DEC(sock);
1437}
1438
1439/* Handle EPIPE error. */
1440static int sctp_error(struct sock *sk, int flags, int err)
1441{
1442 if (err == -EPIPE)
1443 err = sock_error(sk) ? : -EPIPE;
1444 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1445 send_sig(SIGPIPE, current, 0);
1446 return err;
1447}
1448
1449/* API 3.1.3 sendmsg() - UDP Style Syntax
1450 *
1451 * An application uses sendmsg() and recvmsg() calls to transmit data to
1452 * and receive data from its peer.
1453 *
1454 * ssize_t sendmsg(int socket, const struct msghdr *message,
1455 * int flags);
1456 *
1457 * socket - the socket descriptor of the endpoint.
1458 * message - pointer to the msghdr structure which contains a single
1459 * user message and possibly some ancillary data.
1460 *
1461 * See Section 5 for complete description of the data
1462 * structures.
1463 *
1464 * flags - flags sent or received with the user message, see Section
1465 * 5 for complete description of the flags.
1466 *
1467 * Note: This function could use a rewrite especially when explicit
1468 * connect support comes in.
1469 */
1470/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1471
1472SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1473
1474SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1475 struct msghdr *msg, size_t msg_len)
1476{
1477 struct sctp_sock *sp;
1478 struct sctp_endpoint *ep;
1479 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1480 struct sctp_transport *transport, *chunk_tp;
1481 struct sctp_chunk *chunk;
Al Virodce116a2006-11-20 17:25:15 -08001482 union sctp_addr to;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 struct sockaddr *msg_name = NULL;
1484 struct sctp_sndrcvinfo default_sinfo = { 0 };
1485 struct sctp_sndrcvinfo *sinfo;
1486 struct sctp_initmsg *sinit;
1487 sctp_assoc_t associd = 0;
1488 sctp_cmsgs_t cmsgs = { NULL };
1489 int err;
1490 sctp_scope_t scope;
1491 long timeo;
1492 __u16 sinfo_flags = 0;
1493 struct sctp_datamsg *datamsg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 int msg_flags = msg->msg_flags;
1495
1496 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1497 sk, msg, msg_len);
1498
1499 err = 0;
1500 sp = sctp_sk(sk);
1501 ep = sp->ep;
1502
Frank Filz3f7a87d2005-06-20 13:14:57 -07001503 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
1505 /* We cannot send a message over a TCP-style listening socket. */
1506 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1507 err = -EPIPE;
1508 goto out_nounlock;
1509 }
1510
1511 /* Parse out the SCTP CMSGs. */
1512 err = sctp_msghdr_parse(msg, &cmsgs);
1513
1514 if (err) {
1515 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1516 goto out_nounlock;
1517 }
1518
1519 /* Fetch the destination address for this packet. This
1520 * address only selects the association--it is not necessarily
1521 * the address we will send to.
1522 * For a peeled-off socket, msg_name is ignored.
1523 */
1524 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1525 int msg_namelen = msg->msg_namelen;
1526
1527 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1528 msg_namelen);
1529 if (err)
1530 return err;
1531
1532 if (msg_namelen > sizeof(to))
1533 msg_namelen = sizeof(to);
1534 memcpy(&to, msg->msg_name, msg_namelen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 msg_name = msg->msg_name;
1536 }
1537
1538 sinfo = cmsgs.info;
1539 sinit = cmsgs.init;
1540
1541 /* Did the user specify SNDRCVINFO? */
1542 if (sinfo) {
1543 sinfo_flags = sinfo->sinfo_flags;
1544 associd = sinfo->sinfo_assoc_id;
1545 }
1546
1547 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1548 msg_len, sinfo_flags);
1549
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001550 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1551 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 err = -EINVAL;
1553 goto out_nounlock;
1554 }
1555
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001556 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1557 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1558 * If SCTP_ABORT is set, the message length could be non zero with
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 * the msg_iov set to the user abort reason.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001560 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001561 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1562 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 err = -EINVAL;
1564 goto out_nounlock;
1565 }
1566
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001567 /* If SCTP_ADDR_OVER is set, there must be an address
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 * specified in msg_name.
1569 */
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001570 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 err = -EINVAL;
1572 goto out_nounlock;
1573 }
1574
1575 transport = NULL;
1576
1577 SCTP_DEBUG_PRINTK("About to look up association.\n");
1578
1579 sctp_lock_sock(sk);
1580
1581 /* If a msg_name has been specified, assume this is to be used. */
1582 if (msg_name) {
1583 /* Look for a matching association on the endpoint. */
Al Virodce116a2006-11-20 17:25:15 -08001584 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 if (!asoc) {
1586 /* If we could not find a matching association on the
1587 * endpoint, make sure that it is not a TCP-style
1588 * socket that already has an association or there is
1589 * no peeled-off association on another socket.
1590 */
1591 if ((sctp_style(sk, TCP) &&
1592 sctp_sstate(sk, ESTABLISHED)) ||
Al Virodce116a2006-11-20 17:25:15 -08001593 sctp_endpoint_is_peeled_off(ep, &to)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 err = -EADDRNOTAVAIL;
1595 goto out_unlock;
1596 }
1597 }
1598 } else {
1599 asoc = sctp_id2assoc(sk, associd);
1600 if (!asoc) {
1601 err = -EPIPE;
1602 goto out_unlock;
1603 }
1604 }
1605
1606 if (asoc) {
1607 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1608
1609 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1610 * socket that has an association in CLOSED state. This can
1611 * happen when an accepted socket has an association that is
1612 * already CLOSED.
1613 */
1614 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1615 err = -EPIPE;
1616 goto out_unlock;
1617 }
1618
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001619 if (sinfo_flags & SCTP_EOF) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1621 asoc);
1622 sctp_primitive_SHUTDOWN(asoc, NULL);
1623 err = 0;
1624 goto out_unlock;
1625 }
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001626 if (sinfo_flags & SCTP_ABORT) {
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001627
1628 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1629 if (!chunk) {
1630 err = -ENOMEM;
1631 goto out_unlock;
1632 }
1633
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
Sridhar Samudralac164a9b2006-08-22 11:50:39 -07001635 sctp_primitive_ABORT(asoc, chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 err = 0;
1637 goto out_unlock;
1638 }
1639 }
1640
1641 /* Do we need to create the association? */
1642 if (!asoc) {
1643 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1644
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001645 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 err = -EINVAL;
1647 goto out_unlock;
1648 }
1649
1650 /* Check for invalid stream against the stream counts,
1651 * either the default or the user specified stream counts.
1652 */
1653 if (sinfo) {
1654 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1655 /* Check against the defaults. */
1656 if (sinfo->sinfo_stream >=
1657 sp->initmsg.sinit_num_ostreams) {
1658 err = -EINVAL;
1659 goto out_unlock;
1660 }
1661 } else {
1662 /* Check against the requested. */
1663 if (sinfo->sinfo_stream >=
1664 sinit->sinit_num_ostreams) {
1665 err = -EINVAL;
1666 goto out_unlock;
1667 }
1668 }
1669 }
1670
1671 /*
1672 * API 3.1.2 bind() - UDP Style Syntax
1673 * If a bind() or sctp_bindx() is not called prior to a
1674 * sendmsg() call that initiates a new association, the
1675 * system picks an ephemeral port and will choose an address
1676 * set equivalent to binding with a wildcard address.
1677 */
1678 if (!ep->base.bind_addr.port) {
1679 if (sctp_autobind(sk)) {
1680 err = -EAGAIN;
1681 goto out_unlock;
1682 }
Ivan Skytte Jorgensen64a0c1c2005-10-28 15:39:02 -07001683 } else {
1684 /*
1685 * If an unprivileged user inherits a one-to-many
1686 * style socket with open associations on a privileged
1687 * port, it MAY be permitted to accept new associations,
1688 * but it SHOULD NOT be permitted to open new
1689 * associations.
1690 */
1691 if (ep->base.bind_addr.port < PROT_SOCK &&
1692 !capable(CAP_NET_BIND_SERVICE)) {
1693 err = -EACCES;
1694 goto out_unlock;
1695 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 }
1697
1698 scope = sctp_scope(&to);
1699 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1700 if (!new_asoc) {
1701 err = -ENOMEM;
1702 goto out_unlock;
1703 }
1704 asoc = new_asoc;
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001705 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1706 if (err < 0) {
1707 err = -ENOMEM;
1708 goto out_free;
1709 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710
1711 /* If the SCTP_INIT ancillary data is specified, set all
1712 * the association init values accordingly.
1713 */
1714 if (sinit) {
1715 if (sinit->sinit_num_ostreams) {
1716 asoc->c.sinit_num_ostreams =
1717 sinit->sinit_num_ostreams;
1718 }
1719 if (sinit->sinit_max_instreams) {
1720 asoc->c.sinit_max_instreams =
1721 sinit->sinit_max_instreams;
1722 }
1723 if (sinit->sinit_max_attempts) {
1724 asoc->max_init_attempts
1725 = sinit->sinit_max_attempts;
1726 }
1727 if (sinit->sinit_max_init_timeo) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001728 asoc->max_init_timeo =
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1730 }
1731 }
1732
1733 /* Prime the peer's transport structures. */
Al Virodce116a2006-11-20 17:25:15 -08001734 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 if (!transport) {
1736 err = -ENOMEM;
1737 goto out_free;
1738 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 }
1740
1741 /* ASSERT: we have a valid association at this point. */
1742 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1743
1744 if (!sinfo) {
1745 /* If the user didn't specify SNDRCVINFO, make up one with
1746 * some defaults.
1747 */
1748 default_sinfo.sinfo_stream = asoc->default_stream;
1749 default_sinfo.sinfo_flags = asoc->default_flags;
1750 default_sinfo.sinfo_ppid = asoc->default_ppid;
1751 default_sinfo.sinfo_context = asoc->default_context;
1752 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1753 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1754 sinfo = &default_sinfo;
1755 }
1756
1757 /* API 7.1.7, the sndbuf size per association bounds the
1758 * maximum size of data that can be sent in a single send call.
1759 */
1760 if (msg_len > sk->sk_sndbuf) {
1761 err = -EMSGSIZE;
1762 goto out_free;
1763 }
1764
Vlad Yasevich8a479492007-06-07 14:21:05 -04001765 if (asoc->pmtu_pending)
1766 sctp_assoc_pending_pmtu(asoc);
1767
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 /* If fragmentation is disabled and the message length exceeds the
1769 * association fragmentation point, return EMSGSIZE. The I-D
1770 * does not specify what this error is, but this looks like
1771 * a great fit.
1772 */
1773 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1774 err = -EMSGSIZE;
1775 goto out_free;
1776 }
1777
1778 if (sinfo) {
1779 /* Check for invalid stream. */
1780 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1781 err = -EINVAL;
1782 goto out_free;
1783 }
1784 }
1785
1786 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1787 if (!sctp_wspace(asoc)) {
1788 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1789 if (err)
1790 goto out_free;
1791 }
1792
1793 /* If an address is passed with the sendto/sendmsg call, it is used
1794 * to override the primary destination address in the TCP model, or
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001795 * when SCTP_ADDR_OVER flag is set in the UDP model.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 */
1797 if ((sctp_style(sk, TCP) && msg_name) ||
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07001798 (sinfo_flags & SCTP_ADDR_OVER)) {
Al Virodce116a2006-11-20 17:25:15 -08001799 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 if (!chunk_tp) {
1801 err = -EINVAL;
1802 goto out_free;
1803 }
1804 } else
1805 chunk_tp = NULL;
1806
1807 /* Auto-connect, if we aren't connected already. */
1808 if (sctp_state(asoc, CLOSED)) {
1809 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1810 if (err < 0)
1811 goto out_free;
1812 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1813 }
1814
1815 /* Break the message into multiple chunks of maximum size. */
1816 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1817 if (!datamsg) {
1818 err = -ENOMEM;
1819 goto out_free;
1820 }
1821
1822 /* Now send the (possibly) fragmented message. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001823 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
Florian Westphal80445cf2008-03-23 22:47:08 -07001824 sctp_chunk_hold(chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825
1826 /* Do accounting for the write space. */
1827 sctp_set_owner_w(chunk);
1828
1829 chunk->transport = chunk_tp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 }
1831
Vlad Yasevich9c5c62b2009-08-10 13:51:03 -04001832 /* Send it to the lower layers. Note: all chunks
1833 * must either fail or succeed. The lower layer
1834 * works that way today. Keep it that way or this
1835 * breaks.
1836 */
1837 err = sctp_primitive_SEND(asoc, datamsg);
1838 /* Did the lower layer accept the chunk? */
1839 if (err)
1840 sctp_datamsg_free(datamsg);
1841 else
1842 sctp_datamsg_put(datamsg);
1843
1844 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1845
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 if (err)
1847 goto out_free;
1848 else
1849 err = msg_len;
1850
1851 /* If we are already past ASSOCIATE, the lower
1852 * layers are responsible for association cleanup.
1853 */
1854 goto out_unlock;
1855
1856out_free:
1857 if (new_asoc)
1858 sctp_association_free(asoc);
1859out_unlock:
1860 sctp_release_sock(sk);
1861
1862out_nounlock:
1863 return sctp_error(sk, msg_flags, err);
1864
1865#if 0
1866do_sock_err:
1867 if (msg_len)
1868 err = msg_len;
1869 else
1870 err = sock_error(sk);
1871 goto out;
1872
1873do_interrupted:
1874 if (msg_len)
1875 err = msg_len;
1876 goto out;
1877#endif /* 0 */
1878}
1879
1880/* This is an extended version of skb_pull() that removes the data from the
1881 * start of a skb even when data is spread across the list of skb's in the
1882 * frag_list. len specifies the total amount of data that needs to be removed.
1883 * when 'len' bytes could be removed from the skb, it returns 0.
1884 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1885 * could not be removed.
1886 */
1887static int sctp_skb_pull(struct sk_buff *skb, int len)
1888{
1889 struct sk_buff *list;
1890 int skb_len = skb_headlen(skb);
1891 int rlen;
1892
1893 if (len <= skb_len) {
1894 __skb_pull(skb, len);
1895 return 0;
1896 }
1897 len -= skb_len;
1898 __skb_pull(skb, skb_len);
1899
David S. Miller1b003be2009-06-09 00:22:35 -07001900 skb_walk_frags(skb, list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 rlen = sctp_skb_pull(list, len);
1902 skb->len -= (len-rlen);
1903 skb->data_len -= (len-rlen);
1904
1905 if (!rlen)
1906 return 0;
1907
1908 len = rlen;
1909 }
1910
1911 return len;
1912}
1913
1914/* API 3.1.3 recvmsg() - UDP Style Syntax
1915 *
1916 * ssize_t recvmsg(int socket, struct msghdr *message,
1917 * int flags);
1918 *
1919 * socket - the socket descriptor of the endpoint.
1920 * message - pointer to the msghdr structure which contains a single
1921 * user message and possibly some ancillary data.
1922 *
1923 * See Section 5 for complete description of the data
1924 * structures.
1925 *
1926 * flags - flags sent or received with the user message, see Section
1927 * 5 for complete description of the flags.
1928 */
1929static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1930
1931SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1932 struct msghdr *msg, size_t len, int noblock,
1933 int flags, int *addr_len)
1934{
1935 struct sctp_ulpevent *event = NULL;
1936 struct sctp_sock *sp = sctp_sk(sk);
1937 struct sk_buff *skb;
1938 int copied;
1939 int err = 0;
1940 int skb_len;
1941
1942 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1943 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1944 "len", len, "knoblauch", noblock,
1945 "flags", flags, "addr_len", addr_len);
1946
1947 sctp_lock_sock(sk);
1948
1949 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1950 err = -ENOTCONN;
1951 goto out;
1952 }
1953
1954 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1955 if (!skb)
1956 goto out;
1957
1958 /* Get the total length of the skb including any skb's in the
1959 * frag_list.
1960 */
1961 skb_len = skb->len;
1962
1963 copied = skb_len;
1964 if (copied > len)
1965 copied = len;
1966
1967 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1968
1969 event = sctp_skb2event(skb);
1970
1971 if (err)
1972 goto out_free;
1973
Neil Horman3b885782009-10-12 13:26:31 -07001974 sock_recv_ts_and_drops(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975 if (sctp_ulpevent_is_notification(event)) {
1976 msg->msg_flags |= MSG_NOTIFICATION;
1977 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1978 } else {
1979 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1980 }
1981
1982 /* Check if we allow SCTP_SNDRCVINFO. */
1983 if (sp->subscribe.sctp_data_io_event)
1984 sctp_ulpevent_read_sndrcvinfo(event, msg);
1985#if 0
1986 /* FIXME: we should be calling IP/IPv6 layers. */
1987 if (sk->sk_protinfo.af_inet.cmsg_flags)
1988 ip_cmsg_recv(msg, skb);
1989#endif
1990
1991 err = copied;
1992
1993 /* If skb's length exceeds the user's buffer, update the skb and
1994 * push it back to the receive_queue so that the next call to
1995 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1996 */
1997 if (skb_len > copied) {
1998 msg->msg_flags &= ~MSG_EOR;
1999 if (flags & MSG_PEEK)
2000 goto out_free;
2001 sctp_skb_pull(skb, copied);
2002 skb_queue_head(&sk->sk_receive_queue, skb);
2003
2004 /* When only partial message is copied to the user, increase
2005 * rwnd by that amount. If all the data in the skb is read,
2006 * rwnd is updated when the event is freed.
2007 */
Vlad Yasevich0eca8fe2008-01-11 10:12:56 -05002008 if (!sctp_ulpevent_is_notification(event))
2009 sctp_assoc_rwnd_increase(event->asoc, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010 goto out;
2011 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2012 (event->msg_flags & MSG_EOR))
2013 msg->msg_flags |= MSG_EOR;
2014 else
2015 msg->msg_flags &= ~MSG_EOR;
2016
2017out_free:
2018 if (flags & MSG_PEEK) {
2019 /* Release the skb reference acquired after peeking the skb in
2020 * sctp_skb_recv_datagram().
2021 */
2022 kfree_skb(skb);
2023 } else {
2024 /* Free the event which includes releasing the reference to
2025 * the owner of the skb, freeing the skb and updating the
2026 * rwnd.
2027 */
2028 sctp_ulpevent_free(event);
2029 }
2030out:
2031 sctp_release_sock(sk);
2032 return err;
2033}
2034
2035/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2036 *
2037 * This option is a on/off flag. If enabled no SCTP message
2038 * fragmentation will be performed. Instead if a message being sent
2039 * exceeds the current PMTU size, the message will NOT be sent and
2040 * instead a error will be indicated to the user.
2041 */
2042static int sctp_setsockopt_disable_fragments(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002043 char __user *optval,
2044 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045{
2046 int val;
2047
2048 if (optlen < sizeof(int))
2049 return -EINVAL;
2050
2051 if (get_user(val, (int __user *)optval))
2052 return -EFAULT;
2053
2054 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2055
2056 return 0;
2057}
2058
2059static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002060 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061{
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05002062 if (optlen > sizeof(struct sctp_event_subscribe))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 return -EINVAL;
2064 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2065 return -EFAULT;
2066 return 0;
2067}
2068
2069/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2070 *
2071 * This socket option is applicable to the UDP-style socket only. When
2072 * set it will cause associations that are idle for more than the
2073 * specified number of seconds to automatically close. An association
2074 * being idle is defined an association that has NOT sent or received
2075 * user data. The special value of '0' indicates that no automatic
2076 * close of any associations should be performed. The option expects an
2077 * integer defining the number of seconds of idle time before an
2078 * association is closed.
2079 */
2080static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002081 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082{
2083 struct sctp_sock *sp = sctp_sk(sk);
2084
2085 /* Applicable to UDP-style socket only */
2086 if (sctp_style(sk, TCP))
2087 return -EOPNOTSUPP;
2088 if (optlen != sizeof(int))
2089 return -EINVAL;
2090 if (copy_from_user(&sp->autoclose, optval, optlen))
2091 return -EFAULT;
Andrei Pelinescu-Onciulf6778aa2009-11-23 15:54:01 -05002092 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
Andrew Morton8ffd3202009-12-21 14:25:06 +00002093 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 return 0;
2096}
2097
2098/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2099 *
2100 * Applications can enable or disable heartbeats for any peer address of
2101 * an association, modify an address's heartbeat interval, force a
2102 * heartbeat to be sent immediately, and adjust the address's maximum
2103 * number of retransmissions sent before an address is considered
2104 * unreachable. The following structure is used to access and modify an
2105 * address's parameters:
2106 *
2107 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002108 * sctp_assoc_t spp_assoc_id;
2109 * struct sockaddr_storage spp_address;
2110 * uint32_t spp_hbinterval;
2111 * uint16_t spp_pathmaxrxt;
2112 * uint32_t spp_pathmtu;
2113 * uint32_t spp_sackdelay;
2114 * uint32_t spp_flags;
2115 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002117 * spp_assoc_id - (one-to-many style socket) This is filled in the
2118 * application, and identifies the association for
2119 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 * spp_address - This specifies which address is of interest.
2121 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08002122 * in milliseconds. If a value of zero
2123 * is present in this field then no changes are to
2124 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 * spp_pathmaxrxt - This contains the maximum number of
2126 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08002127 * considered unreachable. If a value of zero
2128 * is present in this field then no changes are to
2129 * be made to this parameter.
2130 * spp_pathmtu - When Path MTU discovery is disabled the value
2131 * specified here will be the "fixed" path mtu.
2132 * Note that if the spp_address field is empty
2133 * then all associations on this address will
2134 * have this fixed path mtu set upon them.
2135 *
2136 * spp_sackdelay - When delayed sack is enabled, this value specifies
2137 * the number of milliseconds that sacks will be delayed
2138 * for. This value will apply to all addresses of an
2139 * association if the spp_address field is empty. Note
2140 * also, that if delayed sack is enabled and this
2141 * value is set to 0, no change is made to the last
2142 * recorded delayed sack timer value.
2143 *
2144 * spp_flags - These flags are used to control various features
2145 * on an association. The flag field may contain
2146 * zero or more of the following options.
2147 *
2148 * SPP_HB_ENABLE - Enable heartbeats on the
2149 * specified address. Note that if the address
2150 * field is empty all addresses for the association
2151 * have heartbeats enabled upon them.
2152 *
2153 * SPP_HB_DISABLE - Disable heartbeats on the
2154 * speicifed address. Note that if the address
2155 * field is empty all addresses for the association
2156 * will have their heartbeats disabled. Note also
2157 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2158 * mutually exclusive, only one of these two should
2159 * be specified. Enabling both fields will have
2160 * undetermined results.
2161 *
2162 * SPP_HB_DEMAND - Request a user initiated heartbeat
2163 * to be made immediately.
2164 *
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002165 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2166 * heartbeat delayis to be set to the value of 0
2167 * milliseconds.
2168 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08002169 * SPP_PMTUD_ENABLE - This field will enable PMTU
2170 * discovery upon the specified address. Note that
2171 * if the address feild is empty then all addresses
2172 * on the association are effected.
2173 *
2174 * SPP_PMTUD_DISABLE - This field will disable PMTU
2175 * discovery upon the specified address. Note that
2176 * if the address feild is empty then all addresses
2177 * on the association are effected. Not also that
2178 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2179 * exclusive. Enabling both will have undetermined
2180 * results.
2181 *
2182 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2183 * on delayed sack. The time specified in spp_sackdelay
2184 * is used to specify the sack delay for this address. Note
2185 * that if spp_address is empty then all addresses will
2186 * enable delayed sack and take on the sack delay
2187 * value specified in spp_sackdelay.
2188 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2189 * off delayed sack. If the spp_address field is blank then
2190 * delayed sack is disabled for the entire association. Note
2191 * also that this field is mutually exclusive to
2192 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2193 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 */
Adrian Bunk16164362006-09-18 00:40:38 -07002195static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2196 struct sctp_transport *trans,
2197 struct sctp_association *asoc,
2198 struct sctp_sock *sp,
2199 int hb_change,
2200 int pmtud_change,
2201 int sackdelay_change)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 int error;
2204
Frank Filz52ccb8e2005-12-22 11:36:46 -08002205 if (params->spp_flags & SPP_HB_DEMAND && trans) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2207 if (error)
2208 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 }
2210
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002211 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2212 * this field is ignored. Note also that a value of zero indicates
2213 * the current setting should be left unchanged.
2214 */
2215 if (params->spp_flags & SPP_HB_ENABLE) {
2216
2217 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2218 * set. This lets us use 0 value when this flag
2219 * is set.
2220 */
2221 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2222 params->spp_hbinterval = 0;
2223
2224 if (params->spp_hbinterval ||
2225 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2226 if (trans) {
2227 trans->hbinterval =
2228 msecs_to_jiffies(params->spp_hbinterval);
2229 } else if (asoc) {
2230 asoc->hbinterval =
2231 msecs_to_jiffies(params->spp_hbinterval);
2232 } else {
2233 sp->hbinterval = params->spp_hbinterval;
2234 }
Frank Filz52ccb8e2005-12-22 11:36:46 -08002235 }
2236 }
2237
2238 if (hb_change) {
2239 if (trans) {
2240 trans->param_flags =
2241 (trans->param_flags & ~SPP_HB) | hb_change;
2242 } else if (asoc) {
2243 asoc->param_flags =
2244 (asoc->param_flags & ~SPP_HB) | hb_change;
2245 } else {
2246 sp->param_flags =
2247 (sp->param_flags & ~SPP_HB) | hb_change;
2248 }
2249 }
2250
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002251 /* When Path MTU discovery is disabled the value specified here will
2252 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2253 * include the flag SPP_PMTUD_DISABLE for this field to have any
2254 * effect).
2255 */
2256 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002257 if (trans) {
2258 trans->pathmtu = params->spp_pathmtu;
2259 sctp_assoc_sync_pmtu(asoc);
2260 } else if (asoc) {
2261 asoc->pathmtu = params->spp_pathmtu;
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04002262 sctp_frag_point(asoc, params->spp_pathmtu);
Frank Filz52ccb8e2005-12-22 11:36:46 -08002263 } else {
2264 sp->pathmtu = params->spp_pathmtu;
2265 }
2266 }
2267
2268 if (pmtud_change) {
2269 if (trans) {
2270 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2271 (params->spp_flags & SPP_PMTUD_ENABLE);
2272 trans->param_flags =
2273 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2274 if (update) {
2275 sctp_transport_pmtu(trans);
2276 sctp_assoc_sync_pmtu(asoc);
2277 }
2278 } else if (asoc) {
2279 asoc->param_flags =
2280 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2281 } else {
2282 sp->param_flags =
2283 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2284 }
2285 }
2286
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002287 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2288 * value of this field is ignored. Note also that a value of zero
2289 * indicates the current setting should be left unchanged.
2290 */
2291 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002292 if (trans) {
2293 trans->sackdelay =
2294 msecs_to_jiffies(params->spp_sackdelay);
2295 } else if (asoc) {
2296 asoc->sackdelay =
2297 msecs_to_jiffies(params->spp_sackdelay);
2298 } else {
2299 sp->sackdelay = params->spp_sackdelay;
2300 }
2301 }
2302
2303 if (sackdelay_change) {
2304 if (trans) {
2305 trans->param_flags =
2306 (trans->param_flags & ~SPP_SACKDELAY) |
2307 sackdelay_change;
2308 } else if (asoc) {
2309 asoc->param_flags =
2310 (asoc->param_flags & ~SPP_SACKDELAY) |
2311 sackdelay_change;
2312 } else {
2313 sp->param_flags =
2314 (sp->param_flags & ~SPP_SACKDELAY) |
2315 sackdelay_change;
2316 }
2317 }
2318
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002319 /* Note that a value of zero indicates the current setting should be
2320 left unchanged.
Vlad Yasevichbdf30922007-03-23 11:33:12 -07002321 */
Andrei Pelinescu-Onciul37051f72009-11-23 15:53:57 -05002322 if (params->spp_pathmaxrxt) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002323 if (trans) {
2324 trans->pathmaxrxt = params->spp_pathmaxrxt;
2325 } else if (asoc) {
2326 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2327 } else {
2328 sp->pathmaxrxt = params->spp_pathmaxrxt;
2329 }
2330 }
2331
2332 return 0;
2333}
2334
2335static int sctp_setsockopt_peer_addr_params(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002336 char __user *optval,
2337 unsigned int optlen)
Frank Filz52ccb8e2005-12-22 11:36:46 -08002338{
2339 struct sctp_paddrparams params;
2340 struct sctp_transport *trans = NULL;
2341 struct sctp_association *asoc = NULL;
2342 struct sctp_sock *sp = sctp_sk(sk);
2343 int error;
2344 int hb_change, pmtud_change, sackdelay_change;
2345
2346 if (optlen != sizeof(struct sctp_paddrparams))
2347 return - EINVAL;
2348
2349 if (copy_from_user(&params, optval, optlen))
2350 return -EFAULT;
2351
2352 /* Validate flags and value parameters. */
2353 hb_change = params.spp_flags & SPP_HB;
2354 pmtud_change = params.spp_flags & SPP_PMTUD;
2355 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2356
2357 if (hb_change == SPP_HB ||
2358 pmtud_change == SPP_PMTUD ||
2359 sackdelay_change == SPP_SACKDELAY ||
2360 params.spp_sackdelay > 500 ||
Joe Perchesf64f9e72009-11-29 16:55:45 -08002361 (params.spp_pathmtu &&
2362 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
Frank Filz52ccb8e2005-12-22 11:36:46 -08002363 return -EINVAL;
2364
2365 /* If an address other than INADDR_ANY is specified, and
2366 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04002368 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002369 trans = sctp_addr_id2transport(sk, &params.spp_address,
2370 params.spp_assoc_id);
2371 if (!trans)
2372 return -EINVAL;
2373 }
2374
2375 /* Get association, if assoc_id != 0 and the socket is a one
2376 * to many style socket, and an association was not found, then
2377 * the id was invalid.
2378 */
2379 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2380 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2381 return -EINVAL;
2382
2383 /* Heartbeat demand can only be sent on a transport or
2384 * association, but not a socket.
2385 */
2386 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2387 return -EINVAL;
2388
2389 /* Process parameters. */
2390 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2391 hb_change, pmtud_change,
2392 sackdelay_change);
2393
2394 if (error)
2395 return error;
2396
2397 /* If changes are for association, also apply parameters to each
2398 * transport.
2399 */
2400 if (!trans && asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002401 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2402 transports) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08002403 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2404 hb_change, pmtud_change,
2405 sackdelay_change);
2406 }
2407 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408
2409 return 0;
2410}
2411
Wei Yongjund364d922008-05-09 15:13:26 -07002412/*
2413 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08002414 *
Wei Yongjund364d922008-05-09 15:13:26 -07002415 * This option will effect the way delayed acks are performed. This
2416 * option allows you to get or set the delayed ack time, in
2417 * milliseconds. It also allows changing the delayed ack frequency.
2418 * Changing the frequency to 1 disables the delayed sack algorithm. If
2419 * the assoc_id is 0, then this sets or gets the endpoints default
2420 * values. If the assoc_id field is non-zero, then the set or get
2421 * effects the specified association for the one to many model (the
2422 * assoc_id field is ignored by the one to one model). Note that if
2423 * sack_delay or sack_freq are 0 when setting this option, then the
2424 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08002425 *
Wei Yongjund364d922008-05-09 15:13:26 -07002426 * struct sctp_sack_info {
2427 * sctp_assoc_t sack_assoc_id;
2428 * uint32_t sack_delay;
2429 * uint32_t sack_freq;
2430 * };
Frank Filz77086102005-12-22 11:37:30 -08002431 *
Wei Yongjund364d922008-05-09 15:13:26 -07002432 * sack_assoc_id - This parameter, indicates which association the user
2433 * is performing an action upon. Note that if this field's value is
2434 * zero then the endpoints default value is changed (effecting future
2435 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08002436 *
Wei Yongjund364d922008-05-09 15:13:26 -07002437 * sack_delay - This parameter contains the number of milliseconds that
2438 * the user is requesting the delayed ACK timer be set to. Note that
2439 * this value is defined in the standard to be between 200 and 500
2440 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08002441 *
Wei Yongjund364d922008-05-09 15:13:26 -07002442 * sack_freq - This parameter contains the number of packets that must
2443 * be received before a sack is sent without waiting for the delay
2444 * timer to expire. The default value for this is 2, setting this
2445 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08002446 */
2447
Wei Yongjund364d922008-05-09 15:13:26 -07002448static int sctp_setsockopt_delayed_ack(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002449 char __user *optval, unsigned int optlen)
Frank Filz77086102005-12-22 11:37:30 -08002450{
Wei Yongjund364d922008-05-09 15:13:26 -07002451 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08002452 struct sctp_transport *trans = NULL;
2453 struct sctp_association *asoc = NULL;
2454 struct sctp_sock *sp = sctp_sk(sk);
2455
Wei Yongjund364d922008-05-09 15:13:26 -07002456 if (optlen == sizeof(struct sctp_sack_info)) {
2457 if (copy_from_user(&params, optval, optlen))
2458 return -EFAULT;
2459
2460 if (params.sack_delay == 0 && params.sack_freq == 0)
2461 return 0;
2462 } else if (optlen == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00002463 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2464 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07002465 if (copy_from_user(&params, optval, optlen))
2466 return -EFAULT;
2467
2468 if (params.sack_delay == 0)
2469 params.sack_freq = 1;
2470 else
2471 params.sack_freq = 0;
2472 } else
Frank Filz77086102005-12-22 11:37:30 -08002473 return - EINVAL;
2474
Frank Filz77086102005-12-22 11:37:30 -08002475 /* Validate value parameter. */
Wei Yongjund364d922008-05-09 15:13:26 -07002476 if (params.sack_delay > 500)
Frank Filz77086102005-12-22 11:37:30 -08002477 return -EINVAL;
2478
Wei Yongjund364d922008-05-09 15:13:26 -07002479 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08002480 * to many style socket, and an association was not found, then
2481 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002482 */
Wei Yongjund364d922008-05-09 15:13:26 -07002483 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2484 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08002485 return -EINVAL;
2486
Wei Yongjund364d922008-05-09 15:13:26 -07002487 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002488 if (asoc) {
2489 asoc->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002490 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002491 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002492 (asoc->param_flags & ~SPP_SACKDELAY) |
2493 SPP_SACKDELAY_ENABLE;
2494 } else {
Wei Yongjund364d922008-05-09 15:13:26 -07002495 sp->sackdelay = params.sack_delay;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002496 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002497 (sp->param_flags & ~SPP_SACKDELAY) |
2498 SPP_SACKDELAY_ENABLE;
2499 }
Wei Yongjund364d922008-05-09 15:13:26 -07002500 }
2501
2502 if (params.sack_freq == 1) {
Frank Filz77086102005-12-22 11:37:30 -08002503 if (asoc) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002504 asoc->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002505 (asoc->param_flags & ~SPP_SACKDELAY) |
2506 SPP_SACKDELAY_DISABLE;
2507 } else {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002508 sp->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002509 (sp->param_flags & ~SPP_SACKDELAY) |
2510 SPP_SACKDELAY_DISABLE;
2511 }
Wei Yongjund364d922008-05-09 15:13:26 -07002512 } else if (params.sack_freq > 1) {
2513 if (asoc) {
2514 asoc->sackfreq = params.sack_freq;
2515 asoc->param_flags =
2516 (asoc->param_flags & ~SPP_SACKDELAY) |
2517 SPP_SACKDELAY_ENABLE;
2518 } else {
2519 sp->sackfreq = params.sack_freq;
2520 sp->param_flags =
2521 (sp->param_flags & ~SPP_SACKDELAY) |
2522 SPP_SACKDELAY_ENABLE;
2523 }
Frank Filz77086102005-12-22 11:37:30 -08002524 }
2525
2526 /* If change is for association, also apply to each transport. */
2527 if (asoc) {
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002528 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2529 transports) {
Wei Yongjund364d922008-05-09 15:13:26 -07002530 if (params.sack_delay) {
Frank Filz77086102005-12-22 11:37:30 -08002531 trans->sackdelay =
Wei Yongjund364d922008-05-09 15:13:26 -07002532 msecs_to_jiffies(params.sack_delay);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002533 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002534 (trans->param_flags & ~SPP_SACKDELAY) |
2535 SPP_SACKDELAY_ENABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002536 }
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07002537 if (params.sack_freq == 1) {
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002538 trans->param_flags =
Frank Filz77086102005-12-22 11:37:30 -08002539 (trans->param_flags & ~SPP_SACKDELAY) |
2540 SPP_SACKDELAY_DISABLE;
Wei Yongjund364d922008-05-09 15:13:26 -07002541 } else if (params.sack_freq > 1) {
2542 trans->sackfreq = params.sack_freq;
2543 trans->param_flags =
2544 (trans->param_flags & ~SPP_SACKDELAY) |
2545 SPP_SACKDELAY_ENABLE;
Frank Filz77086102005-12-22 11:37:30 -08002546 }
2547 }
2548 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002549
Frank Filz77086102005-12-22 11:37:30 -08002550 return 0;
2551}
2552
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2554 *
2555 * Applications can specify protocol parameters for the default association
2556 * initialization. The option name argument to setsockopt() and getsockopt()
2557 * is SCTP_INITMSG.
2558 *
2559 * Setting initialization parameters is effective only on an unconnected
2560 * socket (for UDP-style sockets only future associations are effected
2561 * by the change). With TCP-style sockets, this option is inherited by
2562 * sockets derived from a listener socket.
2563 */
David S. Millerb7058842009-09-30 16:12:20 -07002564static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565{
2566 struct sctp_initmsg sinit;
2567 struct sctp_sock *sp = sctp_sk(sk);
2568
2569 if (optlen != sizeof(struct sctp_initmsg))
2570 return -EINVAL;
2571 if (copy_from_user(&sinit, optval, optlen))
2572 return -EFAULT;
2573
2574 if (sinit.sinit_num_ostreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002575 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576 if (sinit.sinit_max_instreams)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002577 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 if (sinit.sinit_max_attempts)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002579 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 if (sinit.sinit_max_init_timeo)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002581 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002582
2583 return 0;
2584}
2585
2586/*
2587 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2588 *
2589 * Applications that wish to use the sendto() system call may wish to
2590 * specify a default set of parameters that would normally be supplied
2591 * through the inclusion of ancillary data. This socket option allows
2592 * such an application to set the default sctp_sndrcvinfo structure.
2593 * The application that wishes to use this socket option simply passes
2594 * in to this call the sctp_sndrcvinfo structure defined in Section
2595 * 5.2.2) The input parameters accepted by this call include
2596 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2597 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2598 * to this call if the caller is using the UDP model.
2599 */
2600static int sctp_setsockopt_default_send_param(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07002601 char __user *optval,
2602 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603{
2604 struct sctp_sndrcvinfo info;
2605 struct sctp_association *asoc;
2606 struct sctp_sock *sp = sctp_sk(sk);
2607
2608 if (optlen != sizeof(struct sctp_sndrcvinfo))
2609 return -EINVAL;
2610 if (copy_from_user(&info, optval, optlen))
2611 return -EFAULT;
2612
2613 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2614 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2615 return -EINVAL;
2616
2617 if (asoc) {
2618 asoc->default_stream = info.sinfo_stream;
2619 asoc->default_flags = info.sinfo_flags;
2620 asoc->default_ppid = info.sinfo_ppid;
2621 asoc->default_context = info.sinfo_context;
2622 asoc->default_timetolive = info.sinfo_timetolive;
2623 } else {
2624 sp->default_stream = info.sinfo_stream;
2625 sp->default_flags = info.sinfo_flags;
2626 sp->default_ppid = info.sinfo_ppid;
2627 sp->default_context = info.sinfo_context;
2628 sp->default_timetolive = info.sinfo_timetolive;
2629 }
2630
2631 return 0;
2632}
2633
2634/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2635 *
2636 * Requests that the local SCTP stack use the enclosed peer address as
2637 * the association primary. The enclosed address must be one of the
2638 * association peer's addresses.
2639 */
2640static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002641 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642{
2643 struct sctp_prim prim;
2644 struct sctp_transport *trans;
2645
2646 if (optlen != sizeof(struct sctp_prim))
2647 return -EINVAL;
2648
2649 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2650 return -EFAULT;
2651
2652 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2653 if (!trans)
2654 return -EINVAL;
2655
2656 sctp_assoc_set_primary(trans->asoc, trans);
2657
2658 return 0;
2659}
2660
2661/*
2662 * 7.1.5 SCTP_NODELAY
2663 *
2664 * Turn on/off any Nagle-like algorithm. This means that packets are
2665 * generally sent as soon as possible and no unnecessary delays are
2666 * introduced, at the cost of more packets in the network. Expects an
2667 * integer boolean flag.
2668 */
2669static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002670 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002671{
2672 int val;
2673
2674 if (optlen < sizeof(int))
2675 return -EINVAL;
2676 if (get_user(val, (int __user *)optval))
2677 return -EFAULT;
2678
2679 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2680 return 0;
2681}
2682
2683/*
2684 *
2685 * 7.1.1 SCTP_RTOINFO
2686 *
2687 * The protocol parameters used to initialize and bound retransmission
2688 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2689 * and modify these parameters.
2690 * All parameters are time values, in milliseconds. A value of 0, when
2691 * modifying the parameters, indicates that the current value should not
2692 * be changed.
2693 *
2694 */
David S. Millerb7058842009-09-30 16:12:20 -07002695static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2696{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697 struct sctp_rtoinfo rtoinfo;
2698 struct sctp_association *asoc;
2699
2700 if (optlen != sizeof (struct sctp_rtoinfo))
2701 return -EINVAL;
2702
2703 if (copy_from_user(&rtoinfo, optval, optlen))
2704 return -EFAULT;
2705
2706 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2707
2708 /* Set the values to the specific association */
2709 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2710 return -EINVAL;
2711
2712 if (asoc) {
2713 if (rtoinfo.srto_initial != 0)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002714 asoc->rto_initial =
Linus Torvalds1da177e2005-04-16 15:20:36 -07002715 msecs_to_jiffies(rtoinfo.srto_initial);
2716 if (rtoinfo.srto_max != 0)
2717 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2718 if (rtoinfo.srto_min != 0)
2719 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2720 } else {
2721 /* If there is no association or the association-id = 0
2722 * set the values to the endpoint.
2723 */
2724 struct sctp_sock *sp = sctp_sk(sk);
2725
2726 if (rtoinfo.srto_initial != 0)
2727 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2728 if (rtoinfo.srto_max != 0)
2729 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2730 if (rtoinfo.srto_min != 0)
2731 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2732 }
2733
2734 return 0;
2735}
2736
2737/*
2738 *
2739 * 7.1.2 SCTP_ASSOCINFO
2740 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02002741 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07002742 * of the association.
2743 * Returns an error if the new association retransmission value is
2744 * greater than the sum of the retransmission value of the peer.
2745 * See [SCTP] for more information.
2746 *
2747 */
David S. Millerb7058842009-09-30 16:12:20 -07002748static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749{
2750
2751 struct sctp_assocparams assocparams;
2752 struct sctp_association *asoc;
2753
2754 if (optlen != sizeof(struct sctp_assocparams))
2755 return -EINVAL;
2756 if (copy_from_user(&assocparams, optval, optlen))
2757 return -EFAULT;
2758
2759 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2760
2761 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2762 return -EINVAL;
2763
2764 /* Set the values to the specific association */
2765 if (asoc) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002766 if (assocparams.sasoc_asocmaxrxt != 0) {
2767 __u32 path_sum = 0;
2768 int paths = 0;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002769 struct sctp_transport *peer_addr;
2770
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07002771 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2772 transports) {
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002773 path_sum += peer_addr->pathmaxrxt;
2774 paths++;
2775 }
2776
Frederik Schwarzer025dfda2008-10-16 19:02:37 +02002777 /* Only validate asocmaxrxt if we have more than
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002778 * one path/transport. We do this because path
2779 * retransmissions are only counted when we have more
2780 * then one path.
2781 */
2782 if (paths > 1 &&
2783 assocparams.sasoc_asocmaxrxt > path_sum)
2784 return -EINVAL;
2785
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
Vlad Yasevich402d68c2006-06-17 22:54:51 -07002787 }
2788
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789 if (assocparams.sasoc_cookie_life != 0) {
2790 asoc->cookie_life.tv_sec =
2791 assocparams.sasoc_cookie_life / 1000;
2792 asoc->cookie_life.tv_usec =
2793 (assocparams.sasoc_cookie_life % 1000)
2794 * 1000;
2795 }
2796 } else {
2797 /* Set the values to the endpoint */
2798 struct sctp_sock *sp = sctp_sk(sk);
2799
2800 if (assocparams.sasoc_asocmaxrxt != 0)
2801 sp->assocparams.sasoc_asocmaxrxt =
2802 assocparams.sasoc_asocmaxrxt;
2803 if (assocparams.sasoc_cookie_life != 0)
2804 sp->assocparams.sasoc_cookie_life =
2805 assocparams.sasoc_cookie_life;
2806 }
2807 return 0;
2808}
2809
2810/*
2811 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2812 *
2813 * This socket option is a boolean flag which turns on or off mapped V4
2814 * addresses. If this option is turned on and the socket is type
2815 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2816 * If this option is turned off, then no mapping will be done of V4
2817 * addresses and a user will receive both PF_INET6 and PF_INET type
2818 * addresses on the socket.
2819 */
David S. Millerb7058842009-09-30 16:12:20 -07002820static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002821{
2822 int val;
2823 struct sctp_sock *sp = sctp_sk(sk);
2824
2825 if (optlen < sizeof(int))
2826 return -EINVAL;
2827 if (get_user(val, (int __user *)optval))
2828 return -EFAULT;
2829 if (val)
2830 sp->v4mapped = 1;
2831 else
2832 sp->v4mapped = 0;
2833
2834 return 0;
2835}
2836
2837/*
Wei Yongjune89c2092008-12-25 16:54:58 -08002838 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2839 * This option will get or set the maximum size to put in any outgoing
2840 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841 * fragmented by SCTP into the specified size. Note that the underlying
2842 * SCTP implementation may fragment into smaller sized chunks when the
2843 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08002844 * the user. The default value for this option is '0' which indicates
2845 * the user is NOT limiting fragmentation and only the PMTU will effect
2846 * SCTP's choice of DATA chunk size. Note also that values set larger
2847 * than the maximum size of an IP datagram will effectively let SCTP
2848 * control fragmentation (i.e. the same as setting this option to 0).
2849 *
2850 * The following structure is used to access and modify this parameter:
2851 *
2852 * struct sctp_assoc_value {
2853 * sctp_assoc_t assoc_id;
2854 * uint32_t assoc_value;
2855 * };
2856 *
2857 * assoc_id: This parameter is ignored for one-to-one style sockets.
2858 * For one-to-many style sockets this parameter indicates which
2859 * association the user is performing an action upon. Note that if
2860 * this field's value is zero then the endpoints default value is
2861 * changed (effecting future associations only).
2862 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863 */
David S. Millerb7058842009-09-30 16:12:20 -07002864static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865{
Wei Yongjune89c2092008-12-25 16:54:58 -08002866 struct sctp_assoc_value params;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868 struct sctp_sock *sp = sctp_sk(sk);
2869 int val;
2870
Wei Yongjune89c2092008-12-25 16:54:58 -08002871 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00002872 pr_warn("Use of int in maxseg socket option deprecated\n");
2873 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08002874 if (copy_from_user(&val, optval, optlen))
2875 return -EFAULT;
2876 params.assoc_id = 0;
2877 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2878 if (copy_from_user(&params, optval, optlen))
2879 return -EFAULT;
2880 val = params.assoc_value;
2881 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882 return -EINVAL;
Wei Yongjune89c2092008-12-25 16:54:58 -08002883
Ivan Skytte Jorgensen96a33992005-10-28 15:36:12 -07002884 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002885 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886
Wei Yongjune89c2092008-12-25 16:54:58 -08002887 asoc = sctp_id2assoc(sk, params.assoc_id);
2888 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2889 return -EINVAL;
2890
2891 if (asoc) {
2892 if (val == 0) {
2893 val = asoc->pathmtu;
2894 val -= sp->pf->af->net_header_len;
2895 val -= sizeof(struct sctphdr) +
2896 sizeof(struct sctp_data_chunk);
2897 }
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04002898 asoc->user_frag = val;
2899 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
Wei Yongjune89c2092008-12-25 16:54:58 -08002900 } else {
2901 sp->user_frag = val;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902 }
2903
2904 return 0;
2905}
2906
2907
2908/*
2909 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2910 *
2911 * Requests that the peer mark the enclosed address as the association
2912 * primary. The enclosed address must be one of the association's
2913 * locally bound addresses. The following structure is used to make a
2914 * set primary request:
2915 */
2916static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002917 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002918{
2919 struct sctp_sock *sp;
2920 struct sctp_endpoint *ep;
2921 struct sctp_association *asoc = NULL;
2922 struct sctp_setpeerprim prim;
2923 struct sctp_chunk *chunk;
2924 int err;
2925
2926 sp = sctp_sk(sk);
2927 ep = sp->ep;
2928
2929 if (!sctp_addip_enable)
2930 return -EPERM;
2931
2932 if (optlen != sizeof(struct sctp_setpeerprim))
2933 return -EINVAL;
2934
2935 if (copy_from_user(&prim, optval, optlen))
2936 return -EFAULT;
2937
2938 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09002939 if (!asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002940 return -EINVAL;
2941
2942 if (!asoc->peer.asconf_capable)
2943 return -EPERM;
2944
2945 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2946 return -EPERM;
2947
2948 if (!sctp_state(asoc, ESTABLISHED))
2949 return -ENOTCONN;
2950
2951 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2952 return -EADDRNOTAVAIL;
2953
2954 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2955 chunk = sctp_make_asconf_set_prim(asoc,
2956 (union sctp_addr *)&prim.sspp_addr);
2957 if (!chunk)
2958 return -ENOMEM;
2959
2960 err = sctp_send_asconf(asoc, chunk);
2961
2962 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2963
2964 return err;
2965}
2966
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002967static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002968 unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002969{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002970 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002971
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002972 if (optlen != sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002973 return -EINVAL;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002974 if (copy_from_user(&adaptation, optval, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975 return -EFAULT;
2976
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08002977 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002978
2979 return 0;
2980}
2981
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08002982/*
2983 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2984 *
2985 * The context field in the sctp_sndrcvinfo structure is normally only
2986 * used when a failed message is retrieved holding the value that was
2987 * sent down on the actual send call. This option allows the setting of
2988 * a default context on an association basis that will be received on
2989 * reading messages from the peer. This is especially helpful in the
2990 * one-2-many model for an application to keep some reference to an
2991 * internal state machine that is processing messages on the
2992 * association. Note that the setting of this value only effects
2993 * received messages from the peer and does not effect the value that is
2994 * saved with outbound messages.
2995 */
2996static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07002997 unsigned int optlen)
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08002998{
2999 struct sctp_assoc_value params;
3000 struct sctp_sock *sp;
3001 struct sctp_association *asoc;
3002
3003 if (optlen != sizeof(struct sctp_assoc_value))
3004 return -EINVAL;
3005 if (copy_from_user(&params, optval, optlen))
3006 return -EFAULT;
3007
3008 sp = sctp_sk(sk);
3009
3010 if (params.assoc_id != 0) {
3011 asoc = sctp_id2assoc(sk, params.assoc_id);
3012 if (!asoc)
3013 return -EINVAL;
3014 asoc->default_rcv_context = params.assoc_value;
3015 } else {
3016 sp->default_rcv_context = params.assoc_value;
3017 }
3018
3019 return 0;
3020}
3021
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003022/*
3023 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3024 *
3025 * This options will at a minimum specify if the implementation is doing
3026 * fragmented interleave. Fragmented interleave, for a one to many
3027 * socket, is when subsequent calls to receive a message may return
3028 * parts of messages from different associations. Some implementations
3029 * may allow you to turn this value on or off. If so, when turned off,
3030 * no fragment interleave will occur (which will cause a head of line
3031 * blocking amongst multiple associations sharing the same one to many
3032 * socket). When this option is turned on, then each receive call may
3033 * come from a different association (thus the user must receive data
3034 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3035 * association each receive belongs to.
3036 *
3037 * This option takes a boolean value. A non-zero value indicates that
3038 * fragmented interleave is on. A value of zero indicates that
3039 * fragmented interleave is off.
3040 *
3041 * Note that it is important that an implementation that allows this
3042 * option to be turned on, have it off by default. Otherwise an unaware
3043 * application using the one to many model may become confused and act
3044 * incorrectly.
3045 */
3046static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3047 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003048 unsigned int optlen)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003049{
3050 int val;
3051
3052 if (optlen != sizeof(int))
3053 return -EINVAL;
3054 if (get_user(val, (int __user *)optval))
3055 return -EFAULT;
3056
3057 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3058
3059 return 0;
3060}
3061
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003062/*
Wei Yongjun8510b932008-12-25 16:59:03 -08003063 * 8.1.21. Set or Get the SCTP Partial Delivery Point
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003064 * (SCTP_PARTIAL_DELIVERY_POINT)
Wei Yongjun8510b932008-12-25 16:59:03 -08003065 *
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003066 * This option will set or get the SCTP partial delivery point. This
3067 * point is the size of a message where the partial delivery API will be
3068 * invoked to help free up rwnd space for the peer. Setting this to a
Wei Yongjun8510b932008-12-25 16:59:03 -08003069 * lower value will cause partial deliveries to happen more often. The
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003070 * calls argument is an integer that sets or gets the partial delivery
Wei Yongjun8510b932008-12-25 16:59:03 -08003071 * point. Note also that the call will fail if the user attempts to set
3072 * this value larger than the socket receive buffer size.
3073 *
3074 * Note that any single message having a length smaller than or equal to
3075 * the SCTP partial delivery point will be delivered in one single read
3076 * call as long as the user provided buffer is large enough to hold the
3077 * message.
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003078 */
3079static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3080 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003081 unsigned int optlen)
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003082{
3083 u32 val;
3084
3085 if (optlen != sizeof(u32))
3086 return -EINVAL;
3087 if (get_user(val, (int __user *)optval))
3088 return -EFAULT;
3089
Wei Yongjun8510b932008-12-25 16:59:03 -08003090 /* Note: We double the receive buffer from what the user sets
3091 * it to be, also initial rwnd is based on rcvbuf/2.
3092 */
3093 if (val > (sk->sk_rcvbuf >> 1))
3094 return -EINVAL;
3095
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003096 sctp_sk(sk)->pd_point = val;
3097
3098 return 0; /* is this the right error code? */
3099}
3100
Vlad Yasevich70331572007-03-23 11:34:36 -07003101/*
3102 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3103 *
3104 * This option will allow a user to change the maximum burst of packets
3105 * that can be emitted by this association. Note that the default value
3106 * is 4, and some implementations may restrict this setting so that it
3107 * can only be lowered.
3108 *
3109 * NOTE: This text doesn't seem right. Do this on a socket basis with
3110 * future associations inheriting the socket value.
3111 */
3112static int sctp_setsockopt_maxburst(struct sock *sk,
3113 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003114 unsigned int optlen)
Vlad Yasevich70331572007-03-23 11:34:36 -07003115{
Neil Horman219b99a2008-03-05 13:44:46 -08003116 struct sctp_assoc_value params;
3117 struct sctp_sock *sp;
3118 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07003119 int val;
Neil Horman219b99a2008-03-05 13:44:46 -08003120 int assoc_id = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003121
Neil Horman219b99a2008-03-05 13:44:46 -08003122 if (optlen == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00003123 pr_warn("Use of int in max_burst socket option deprecated\n");
3124 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08003125 if (copy_from_user(&val, optval, optlen))
3126 return -EFAULT;
3127 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3128 if (copy_from_user(&params, optval, optlen))
3129 return -EFAULT;
3130 val = params.assoc_value;
3131 assoc_id = params.assoc_id;
3132 } else
3133 return -EINVAL;
3134
3135 sp = sctp_sk(sk);
3136
3137 if (assoc_id != 0) {
3138 asoc = sctp_id2assoc(sk, assoc_id);
3139 if (!asoc)
3140 return -EINVAL;
3141 asoc->max_burst = val;
3142 } else
3143 sp->max_burst = val;
Vlad Yasevich70331572007-03-23 11:34:36 -07003144
3145 return 0;
3146}
3147
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003148/*
3149 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3150 *
3151 * This set option adds a chunk type that the user is requesting to be
3152 * received only in an authenticated way. Changes to the list of chunks
3153 * will only effect future associations on the socket.
3154 */
3155static int sctp_setsockopt_auth_chunk(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003156 char __user *optval,
3157 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003158{
3159 struct sctp_authchunk val;
3160
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003161 if (!sctp_auth_enable)
3162 return -EACCES;
3163
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003164 if (optlen != sizeof(struct sctp_authchunk))
3165 return -EINVAL;
3166 if (copy_from_user(&val, optval, optlen))
3167 return -EFAULT;
3168
3169 switch (val.sauth_chunk) {
3170 case SCTP_CID_INIT:
3171 case SCTP_CID_INIT_ACK:
3172 case SCTP_CID_SHUTDOWN_COMPLETE:
3173 case SCTP_CID_AUTH:
3174 return -EINVAL;
3175 }
3176
3177 /* add this chunk id to the endpoint */
3178 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3179}
3180
3181/*
3182 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3183 *
3184 * This option gets or sets the list of HMAC algorithms that the local
3185 * endpoint requires the peer to use.
3186 */
3187static int sctp_setsockopt_hmac_ident(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003188 char __user *optval,
3189 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003190{
3191 struct sctp_hmacalgo *hmacs;
Vlad Yasevichd9724052008-08-27 16:09:49 -07003192 u32 idents;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003193 int err;
3194
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003195 if (!sctp_auth_enable)
3196 return -EACCES;
3197
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003198 if (optlen < sizeof(struct sctp_hmacalgo))
3199 return -EINVAL;
3200
3201 hmacs = kmalloc(optlen, GFP_KERNEL);
3202 if (!hmacs)
3203 return -ENOMEM;
3204
3205 if (copy_from_user(hmacs, optval, optlen)) {
3206 err = -EFAULT;
3207 goto out;
3208 }
3209
Vlad Yasevichd9724052008-08-27 16:09:49 -07003210 idents = hmacs->shmac_num_idents;
3211 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3212 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003213 err = -EINVAL;
3214 goto out;
3215 }
3216
3217 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3218out:
3219 kfree(hmacs);
3220 return err;
3221}
3222
3223/*
3224 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3225 *
3226 * This option will set a shared secret key which is used to build an
3227 * association shared key.
3228 */
3229static int sctp_setsockopt_auth_key(struct sock *sk,
3230 char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07003231 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003232{
3233 struct sctp_authkey *authkey;
3234 struct sctp_association *asoc;
3235 int ret;
3236
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003237 if (!sctp_auth_enable)
3238 return -EACCES;
3239
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003240 if (optlen <= sizeof(struct sctp_authkey))
3241 return -EINVAL;
3242
3243 authkey = kmalloc(optlen, GFP_KERNEL);
3244 if (!authkey)
3245 return -ENOMEM;
3246
3247 if (copy_from_user(authkey, optval, optlen)) {
3248 ret = -EFAULT;
3249 goto out;
3250 }
3251
Vlad Yasevich328fc472008-08-27 16:08:54 -07003252 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
Vlad Yasevich30c22352008-08-25 15:16:19 -07003253 ret = -EINVAL;
3254 goto out;
3255 }
3256
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003257 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3258 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3259 ret = -EINVAL;
3260 goto out;
3261 }
3262
3263 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3264out:
3265 kfree(authkey);
3266 return ret;
3267}
3268
3269/*
3270 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3271 *
3272 * This option will get or set the active shared key to be used to build
3273 * the association shared key.
3274 */
3275static int sctp_setsockopt_active_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003276 char __user *optval,
3277 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003278{
3279 struct sctp_authkeyid val;
3280 struct sctp_association *asoc;
3281
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003282 if (!sctp_auth_enable)
3283 return -EACCES;
3284
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003285 if (optlen != sizeof(struct sctp_authkeyid))
3286 return -EINVAL;
3287 if (copy_from_user(&val, optval, optlen))
3288 return -EFAULT;
3289
3290 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3291 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3292 return -EINVAL;
3293
3294 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3295 val.scact_keynumber);
3296}
3297
3298/*
3299 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3300 *
3301 * This set option will delete a shared secret key from use.
3302 */
3303static int sctp_setsockopt_del_key(struct sock *sk,
David S. Millerb7058842009-09-30 16:12:20 -07003304 char __user *optval,
3305 unsigned int optlen)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003306{
3307 struct sctp_authkeyid val;
3308 struct sctp_association *asoc;
3309
Vlad Yasevich5e739d12008-08-21 03:34:25 -07003310 if (!sctp_auth_enable)
3311 return -EACCES;
3312
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003313 if (optlen != sizeof(struct sctp_authkeyid))
3314 return -EINVAL;
3315 if (copy_from_user(&val, optval, optlen))
3316 return -EFAULT;
3317
3318 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3319 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3320 return -EINVAL;
3321
3322 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3323 val.scact_keynumber);
3324
3325}
3326
3327
Linus Torvalds1da177e2005-04-16 15:20:36 -07003328/* API 6.2 setsockopt(), getsockopt()
3329 *
3330 * Applications use setsockopt() and getsockopt() to set or retrieve
3331 * socket options. Socket options are used to change the default
3332 * behavior of sockets calls. They are described in Section 7.
3333 *
3334 * The syntax is:
3335 *
3336 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3337 * int __user *optlen);
3338 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3339 * int optlen);
3340 *
3341 * sd - the socket descript.
3342 * level - set to IPPROTO_SCTP for all SCTP options.
3343 * optname - the option name.
3344 * optval - the buffer to store the value of the option.
3345 * optlen - the size of the buffer.
3346 */
3347SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003348 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003349{
3350 int retval = 0;
3351
3352 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3353 sk, optname);
3354
3355 /* I can hardly begin to describe how wrong this is. This is
3356 * so broken as to be worse than useless. The API draft
3357 * REALLY is NOT helpful here... I am not convinced that the
3358 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3359 * are at all well-founded.
3360 */
3361 if (level != SOL_SCTP) {
3362 struct sctp_af *af = sctp_sk(sk)->pf->af;
3363 retval = af->setsockopt(sk, level, optname, optval, optlen);
3364 goto out_nounlock;
3365 }
3366
3367 sctp_lock_sock(sk);
3368
3369 switch (optname) {
3370 case SCTP_SOCKOPT_BINDX_ADD:
3371 /* 'optlen' is the size of the addresses buffer. */
3372 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3373 optlen, SCTP_BINDX_ADD_ADDR);
3374 break;
3375
3376 case SCTP_SOCKOPT_BINDX_REM:
3377 /* 'optlen' is the size of the addresses buffer. */
3378 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3379 optlen, SCTP_BINDX_REM_ADDR);
3380 break;
3381
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003382 case SCTP_SOCKOPT_CONNECTX_OLD:
3383 /* 'optlen' is the size of the addresses buffer. */
3384 retval = sctp_setsockopt_connectx_old(sk,
3385 (struct sockaddr __user *)optval,
3386 optlen);
3387 break;
3388
Frank Filz3f7a87d2005-06-20 13:14:57 -07003389 case SCTP_SOCKOPT_CONNECTX:
3390 /* 'optlen' is the size of the addresses buffer. */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003391 retval = sctp_setsockopt_connectx(sk,
3392 (struct sockaddr __user *)optval,
3393 optlen);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003394 break;
3395
Linus Torvalds1da177e2005-04-16 15:20:36 -07003396 case SCTP_DISABLE_FRAGMENTS:
3397 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3398 break;
3399
3400 case SCTP_EVENTS:
3401 retval = sctp_setsockopt_events(sk, optval, optlen);
3402 break;
3403
3404 case SCTP_AUTOCLOSE:
3405 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3406 break;
3407
3408 case SCTP_PEER_ADDR_PARAMS:
3409 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3410 break;
3411
Wei Yongjund364d922008-05-09 15:13:26 -07003412 case SCTP_DELAYED_ACK:
3413 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
Frank Filz77086102005-12-22 11:37:30 -08003414 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07003415 case SCTP_PARTIAL_DELIVERY_POINT:
3416 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3417 break;
Frank Filz77086102005-12-22 11:37:30 -08003418
Linus Torvalds1da177e2005-04-16 15:20:36 -07003419 case SCTP_INITMSG:
3420 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3421 break;
3422 case SCTP_DEFAULT_SEND_PARAM:
3423 retval = sctp_setsockopt_default_send_param(sk, optval,
3424 optlen);
3425 break;
3426 case SCTP_PRIMARY_ADDR:
3427 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3428 break;
3429 case SCTP_SET_PEER_PRIMARY_ADDR:
3430 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3431 break;
3432 case SCTP_NODELAY:
3433 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3434 break;
3435 case SCTP_RTOINFO:
3436 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3437 break;
3438 case SCTP_ASSOCINFO:
3439 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3440 break;
3441 case SCTP_I_WANT_MAPPED_V4_ADDR:
3442 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3443 break;
3444 case SCTP_MAXSEG:
3445 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3446 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003447 case SCTP_ADAPTATION_LAYER:
3448 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003449 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003450 case SCTP_CONTEXT:
3451 retval = sctp_setsockopt_context(sk, optval, optlen);
3452 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003453 case SCTP_FRAGMENT_INTERLEAVE:
3454 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3455 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07003456 case SCTP_MAX_BURST:
3457 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3458 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07003459 case SCTP_AUTH_CHUNK:
3460 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3461 break;
3462 case SCTP_HMAC_IDENT:
3463 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3464 break;
3465 case SCTP_AUTH_KEY:
3466 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3467 break;
3468 case SCTP_AUTH_ACTIVE_KEY:
3469 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3470 break;
3471 case SCTP_AUTH_DELETE_KEY:
3472 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3473 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003474 default:
3475 retval = -ENOPROTOOPT;
3476 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07003477 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003478
3479 sctp_release_sock(sk);
3480
3481out_nounlock:
3482 return retval;
3483}
3484
3485/* API 3.1.6 connect() - UDP Style Syntax
3486 *
3487 * An application may use the connect() call in the UDP model to initiate an
3488 * association without sending data.
3489 *
3490 * The syntax is:
3491 *
3492 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3493 *
3494 * sd: the socket descriptor to have a new association added to.
3495 *
3496 * nam: the address structure (either struct sockaddr_in or struct
3497 * sockaddr_in6 defined in RFC2553 [7]).
3498 *
3499 * len: the size of the address.
3500 */
Frank Filz3f7a87d2005-06-20 13:14:57 -07003501SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003502 int addr_len)
3503{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003504 int err = 0;
Frank Filz3f7a87d2005-06-20 13:14:57 -07003505 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003506
3507 sctp_lock_sock(sk);
3508
Frank Filz3f7a87d2005-06-20 13:14:57 -07003509 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08003510 __func__, sk, addr, addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003511
Frank Filz3f7a87d2005-06-20 13:14:57 -07003512 /* Validate addr_len before calling common connect/connectx routine. */
3513 af = sctp_get_af_specific(addr->sa_family);
3514 if (!af || addr_len < af->sockaddr_len) {
3515 err = -EINVAL;
3516 } else {
3517 /* Pass correct addr len to common routine (so it knows there
3518 * is only one address being passed.
3519 */
Vlad Yasevich88a0a942008-05-09 15:14:11 -07003520 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003521 }
3522
Linus Torvalds1da177e2005-04-16 15:20:36 -07003523 sctp_release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003524 return err;
3525}
3526
3527/* FIXME: Write comments. */
3528SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3529{
3530 return -EOPNOTSUPP; /* STUB */
3531}
3532
3533/* 4.1.4 accept() - TCP Style Syntax
3534 *
3535 * Applications use accept() call to remove an established SCTP
3536 * association from the accept queue of the endpoint. A new socket
3537 * descriptor will be returned from accept() to represent the newly
3538 * formed association.
3539 */
3540SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3541{
3542 struct sctp_sock *sp;
3543 struct sctp_endpoint *ep;
3544 struct sock *newsk = NULL;
3545 struct sctp_association *asoc;
3546 long timeo;
3547 int error = 0;
3548
3549 sctp_lock_sock(sk);
3550
3551 sp = sctp_sk(sk);
3552 ep = sp->ep;
3553
3554 if (!sctp_style(sk, TCP)) {
3555 error = -EOPNOTSUPP;
3556 goto out;
3557 }
3558
3559 if (!sctp_sstate(sk, LISTENING)) {
3560 error = -EINVAL;
3561 goto out;
3562 }
3563
Sridhar Samudrala8abfedd2006-08-22 00:24:09 -07003564 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003565
3566 error = sctp_wait_for_accept(sk, timeo);
3567 if (error)
3568 goto out;
3569
3570 /* We treat the list of associations on the endpoint as the accept
3571 * queue and pick the first association on the list.
3572 */
3573 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3574
3575 newsk = sp->pf->create_accept_sk(sk, asoc);
3576 if (!newsk) {
3577 error = -ENOMEM;
3578 goto out;
3579 }
3580
3581 /* Populate the fields of the newsk from the oldsk and migrate the
3582 * asoc to the newsk.
3583 */
3584 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3585
3586out:
3587 sctp_release_sock(sk);
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003588 *err = error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003589 return newsk;
3590}
3591
3592/* The SCTP ioctl handler. */
3593SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3594{
Diego Elio 'Flameeyes' Pettenò65040c32010-09-03 03:47:03 +00003595 int rc = -ENOTCONN;
3596
3597 sctp_lock_sock(sk);
3598
3599 /*
3600 * SEQPACKET-style sockets in LISTENING state are valid, for
3601 * SCTP, so only discard TCP-style sockets in LISTENING state.
3602 */
3603 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3604 goto out;
3605
3606 switch (cmd) {
3607 case SIOCINQ: {
3608 struct sk_buff *skb;
3609 unsigned int amount = 0;
3610
3611 skb = skb_peek(&sk->sk_receive_queue);
3612 if (skb != NULL) {
3613 /*
3614 * We will only return the amount of this packet since
3615 * that is all that will be read.
3616 */
3617 amount = skb->len;
3618 }
3619 rc = put_user(amount, (int __user *)arg);
3620 }
3621 break;
3622 default:
3623 rc = -ENOIOCTLCMD;
3624 break;
3625 }
3626out:
3627 sctp_release_sock(sk);
3628 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003629}
3630
3631/* This is the function which gets called during socket creation to
3632 * initialized the SCTP-specific portion of the sock.
3633 * The sock structure should already be zero-filled memory.
3634 */
3635SCTP_STATIC int sctp_init_sock(struct sock *sk)
3636{
3637 struct sctp_endpoint *ep;
3638 struct sctp_sock *sp;
3639
3640 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3641
3642 sp = sctp_sk(sk);
3643
3644 /* Initialize the SCTP per socket area. */
3645 switch (sk->sk_type) {
3646 case SOCK_SEQPACKET:
3647 sp->type = SCTP_SOCKET_UDP;
3648 break;
3649 case SOCK_STREAM:
3650 sp->type = SCTP_SOCKET_TCP;
3651 break;
3652 default:
3653 return -ESOCKTNOSUPPORT;
3654 }
3655
3656 /* Initialize default send parameters. These parameters can be
3657 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3658 */
3659 sp->default_stream = 0;
3660 sp->default_ppid = 0;
3661 sp->default_flags = 0;
3662 sp->default_context = 0;
3663 sp->default_timetolive = 0;
3664
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003665 sp->default_rcv_context = 0;
Vlad Yasevich70331572007-03-23 11:34:36 -07003666 sp->max_burst = sctp_max_burst;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08003667
Linus Torvalds1da177e2005-04-16 15:20:36 -07003668 /* Initialize default setup parameters. These parameters
3669 * can be modified with the SCTP_INITMSG socket option or
3670 * overridden by the SCTP_INIT CMSG.
3671 */
3672 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3673 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3674 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003675 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003676
3677 /* Initialize default RTO related parameters. These parameters can
3678 * be modified for with the SCTP_RTOINFO socket option.
3679 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003680 sp->rtoinfo.srto_initial = sctp_rto_initial;
3681 sp->rtoinfo.srto_max = sctp_rto_max;
3682 sp->rtoinfo.srto_min = sctp_rto_min;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003683
3684 /* Initialize default association related parameters. These parameters
3685 * can be modified with the SCTP_ASSOCINFO socket option.
3686 */
3687 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3688 sp->assocparams.sasoc_number_peer_destinations = 0;
3689 sp->assocparams.sasoc_peer_rwnd = 0;
3690 sp->assocparams.sasoc_local_rwnd = 0;
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003691 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003692
3693 /* Initialize default event subscriptions. By default, all the
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003694 * options are off.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003695 */
3696 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3697
3698 /* Default Peer Address Parameters. These defaults can
3699 * be modified via SCTP_PEER_ADDR_PARAMS
3700 */
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003701 sp->hbinterval = sctp_hb_interval;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003702 sp->pathmaxrxt = sctp_max_retrans_path;
3703 sp->pathmtu = 0; // allow default discovery
Vladislav Yasevich3fd091e2006-08-22 13:29:17 -07003704 sp->sackdelay = sctp_sack_timeout;
Vlad Yasevich7bfe8bd2008-06-09 15:45:05 -07003705 sp->sackfreq = 2;
Frank Filz52ccb8e2005-12-22 11:36:46 -08003706 sp->param_flags = SPP_HB_ENABLE |
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09003707 SPP_PMTUD_ENABLE |
3708 SPP_SACKDELAY_ENABLE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003709
3710 /* If enabled no SCTP message fragmentation will be performed.
3711 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3712 */
3713 sp->disable_fragments = 0;
3714
Sridhar Samudrala208edef2006-09-29 17:08:01 -07003715 /* Enable Nagle algorithm by default. */
3716 sp->nodelay = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003717
3718 /* Enable by default. */
3719 sp->v4mapped = 1;
3720
3721 /* Auto-close idle associations after the configured
3722 * number of seconds. A value of 0 disables this
3723 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3724 * for UDP-style sockets only.
3725 */
3726 sp->autoclose = 0;
3727
3728 /* User specified fragmentation limit. */
3729 sp->user_frag = 0;
3730
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08003731 sp->adaptation_ind = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003732
3733 sp->pf = sctp_get_pf_specific(sk->sk_family);
3734
3735 /* Control variables for partial data delivery. */
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003736 atomic_set(&sp->pd_mode, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003737 skb_queue_head_init(&sp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07003738 sp->frag_interleave = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003739
3740 /* Create a per socket endpoint structure. Even if we
3741 * change the data structure relationships, this may still
3742 * be useful for storing pre-connect address information.
3743 */
3744 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3745 if (!ep)
3746 return -ENOMEM;
3747
3748 sp->ep = ep;
3749 sp->hmac = NULL;
3750
3751 SCTP_DBG_OBJCNT_INC(sock);
David S. Miller6f756a82008-11-23 17:34:03 -08003752
3753 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003754 percpu_counter_inc(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003755 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
David S. Miller6f756a82008-11-23 17:34:03 -08003756 local_bh_enable();
3757
Linus Torvalds1da177e2005-04-16 15:20:36 -07003758 return 0;
3759}
3760
3761/* Cleanup any SCTP per socket resources. */
Brian Haley7d06b2e2008-06-14 17:04:49 -07003762SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003763{
3764 struct sctp_endpoint *ep;
3765
3766 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3767
3768 /* Release our hold on the endpoint. */
3769 ep = sctp_sk(sk)->ep;
3770 sctp_endpoint_free(ep);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003771 local_bh_disable();
Vlad Yasevich81419d82010-04-28 08:47:20 +00003772 percpu_counter_dec(&sctp_sockets_allocated);
Eric Dumazet9a57f7f2008-11-17 02:41:00 -08003773 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet5bc0b3b2008-11-25 13:53:27 -08003774 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003775}
3776
3777/* API 4.1.7 shutdown() - TCP Style Syntax
3778 * int shutdown(int socket, int how);
3779 *
3780 * sd - the socket descriptor of the association to be closed.
3781 * how - Specifies the type of shutdown. The values are
3782 * as follows:
3783 * SHUT_RD
3784 * Disables further receive operations. No SCTP
3785 * protocol action is taken.
3786 * SHUT_WR
3787 * Disables further send operations, and initiates
3788 * the SCTP shutdown sequence.
3789 * SHUT_RDWR
3790 * Disables further send and receive operations
3791 * and initiates the SCTP shutdown sequence.
3792 */
3793SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3794{
3795 struct sctp_endpoint *ep;
3796 struct sctp_association *asoc;
3797
3798 if (!sctp_style(sk, TCP))
3799 return;
3800
3801 if (how & SEND_SHUTDOWN) {
3802 ep = sctp_sk(sk)->ep;
3803 if (!list_empty(&ep->asocs)) {
3804 asoc = list_entry(ep->asocs.next,
3805 struct sctp_association, asocs);
3806 sctp_primitive_SHUTDOWN(asoc, NULL);
3807 }
3808 }
3809}
3810
3811/* 7.2.1 Association Status (SCTP_STATUS)
3812
3813 * Applications can retrieve current status information about an
3814 * association, including association state, peer receiver window size,
3815 * number of unacked data chunks, and number of data chunks pending
3816 * receipt. This information is read-only.
3817 */
3818static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3819 char __user *optval,
3820 int __user *optlen)
3821{
3822 struct sctp_status status;
3823 struct sctp_association *asoc = NULL;
3824 struct sctp_transport *transport;
3825 sctp_assoc_t associd;
3826 int retval = 0;
3827
Neil Horman408f22e2007-06-16 14:03:45 -04003828 if (len < sizeof(status)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003829 retval = -EINVAL;
3830 goto out;
3831 }
3832
Neil Horman408f22e2007-06-16 14:03:45 -04003833 len = sizeof(status);
3834 if (copy_from_user(&status, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003835 retval = -EFAULT;
3836 goto out;
3837 }
3838
3839 associd = status.sstat_assoc_id;
3840 asoc = sctp_id2assoc(sk, associd);
3841 if (!asoc) {
3842 retval = -EINVAL;
3843 goto out;
3844 }
3845
3846 transport = asoc->peer.primary_path;
3847
3848 status.sstat_assoc_id = sctp_assoc2id(asoc);
3849 status.sstat_state = asoc->state;
3850 status.sstat_rwnd = asoc->peer.rwnd;
3851 status.sstat_unackdata = asoc->unack_data;
3852
3853 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3854 status.sstat_instrms = asoc->c.sinit_max_instreams;
3855 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3856 status.sstat_fragmentation_point = asoc->frag_point;
3857 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Al Viro8cec6b82006-11-20 17:23:01 -08003858 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3859 transport->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003860 /* Map ipv4 address into v4-mapped-on-v6 address. */
3861 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3862 (union sctp_addr *)&status.sstat_primary.spinfo_address);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003863 status.sstat_primary.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003864 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3865 status.sstat_primary.spinfo_srtt = transport->srtt;
3866 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003867 status.sstat_primary.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003868
Frank Filz3f7a87d2005-06-20 13:14:57 -07003869 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3870 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3871
Linus Torvalds1da177e2005-04-16 15:20:36 -07003872 if (put_user(len, optlen)) {
3873 retval = -EFAULT;
3874 goto out;
3875 }
3876
3877 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3878 len, status.sstat_state, status.sstat_rwnd,
3879 status.sstat_assoc_id);
3880
3881 if (copy_to_user(optval, &status, len)) {
3882 retval = -EFAULT;
3883 goto out;
3884 }
3885
3886out:
3887 return (retval);
3888}
3889
3890
3891/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3892 *
3893 * Applications can retrieve information about a specific peer address
3894 * of an association, including its reachability state, congestion
3895 * window, and retransmission timer values. This information is
3896 * read-only.
3897 */
3898static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3899 char __user *optval,
3900 int __user *optlen)
3901{
3902 struct sctp_paddrinfo pinfo;
3903 struct sctp_transport *transport;
3904 int retval = 0;
3905
Neil Horman408f22e2007-06-16 14:03:45 -04003906 if (len < sizeof(pinfo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003907 retval = -EINVAL;
3908 goto out;
3909 }
3910
Neil Horman408f22e2007-06-16 14:03:45 -04003911 len = sizeof(pinfo);
3912 if (copy_from_user(&pinfo, optval, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003913 retval = -EFAULT;
3914 goto out;
3915 }
3916
3917 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3918 pinfo.spinfo_assoc_id);
3919 if (!transport)
3920 return -EINVAL;
3921
3922 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
Frank Filz3f7a87d2005-06-20 13:14:57 -07003923 pinfo.spinfo_state = transport->state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003924 pinfo.spinfo_cwnd = transport->cwnd;
3925 pinfo.spinfo_srtt = transport->srtt;
3926 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
Frank Filz52ccb8e2005-12-22 11:36:46 -08003927 pinfo.spinfo_mtu = transport->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003928
Frank Filz3f7a87d2005-06-20 13:14:57 -07003929 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3930 pinfo.spinfo_state = SCTP_ACTIVE;
3931
Linus Torvalds1da177e2005-04-16 15:20:36 -07003932 if (put_user(len, optlen)) {
3933 retval = -EFAULT;
3934 goto out;
3935 }
3936
3937 if (copy_to_user(optval, &pinfo, len)) {
3938 retval = -EFAULT;
3939 goto out;
3940 }
3941
3942out:
3943 return (retval);
3944}
3945
3946/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3947 *
3948 * This option is a on/off flag. If enabled no SCTP message
3949 * fragmentation will be performed. Instead if a message being sent
3950 * exceeds the current PMTU size, the message will NOT be sent and
3951 * instead a error will be indicated to the user.
3952 */
3953static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3954 char __user *optval, int __user *optlen)
3955{
3956 int val;
3957
3958 if (len < sizeof(int))
3959 return -EINVAL;
3960
3961 len = sizeof(int);
3962 val = (sctp_sk(sk)->disable_fragments == 1);
3963 if (put_user(len, optlen))
3964 return -EFAULT;
3965 if (copy_to_user(optval, &val, len))
3966 return -EFAULT;
3967 return 0;
3968}
3969
3970/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3971 *
3972 * This socket option is used to specify various notifications and
3973 * ancillary data the user wishes to receive.
3974 */
3975static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3976 int __user *optlen)
3977{
Neil Horman408f22e2007-06-16 14:03:45 -04003978 if (len < sizeof(struct sctp_event_subscribe))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003979 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04003980 len = sizeof(struct sctp_event_subscribe);
3981 if (put_user(len, optlen))
3982 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003983 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3984 return -EFAULT;
3985 return 0;
3986}
3987
3988/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3989 *
3990 * This socket option is applicable to the UDP-style socket only. When
3991 * set it will cause associations that are idle for more than the
3992 * specified number of seconds to automatically close. An association
3993 * being idle is defined an association that has NOT sent or received
3994 * user data. The special value of '0' indicates that no automatic
3995 * close of any associations should be performed. The option expects an
3996 * integer defining the number of seconds of idle time before an
3997 * association is closed.
3998 */
3999static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4000{
4001 /* Applicable to UDP-style socket only */
4002 if (sctp_style(sk, TCP))
4003 return -EOPNOTSUPP;
Neil Horman408f22e2007-06-16 14:03:45 -04004004 if (len < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004005 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004006 len = sizeof(int);
4007 if (put_user(len, optlen))
4008 return -EFAULT;
4009 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004010 return -EFAULT;
4011 return 0;
4012}
4013
4014/* Helper routine to branch off an association to a new socket. */
4015SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4016 struct socket **sockp)
4017{
4018 struct sock *sk = asoc->base.sk;
4019 struct socket *sock;
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004020 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004021 int err = 0;
4022
4023 /* An association cannot be branched off from an already peeled-off
4024 * socket, nor is this supported for tcp style sockets.
4025 */
4026 if (!sctp_style(sk, UDP))
4027 return -EINVAL;
4028
4029 /* Create a new socket. */
4030 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4031 if (err < 0)
4032 return err;
4033
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004034 sctp_copy_sock(sock->sk, sk, asoc);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004035
4036 /* Make peeled-off sockets more like 1-1 accepted sockets.
4037 * Set the daddr and initialize id to something more random
4038 */
Vlad Yasevichd570ee42007-05-15 16:32:39 -04004039 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4040 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00004041
4042 /* Populate the fields of the newsk from the oldsk and migrate the
4043 * asoc to the newsk.
4044 */
4045 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
Vlad Yasevich4f444302006-10-30 18:54:32 -08004046
Linus Torvalds1da177e2005-04-16 15:20:36 -07004047 *sockp = sock;
4048
4049 return err;
4050}
4051
4052static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4053{
4054 sctp_peeloff_arg_t peeloff;
4055 struct socket *newsock;
4056 int retval = 0;
4057 struct sctp_association *asoc;
4058
Neil Horman408f22e2007-06-16 14:03:45 -04004059 if (len < sizeof(sctp_peeloff_arg_t))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004060 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004061 len = sizeof(sctp_peeloff_arg_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004062 if (copy_from_user(&peeloff, optval, len))
4063 return -EFAULT;
4064
4065 asoc = sctp_id2assoc(sk, peeloff.associd);
4066 if (!asoc) {
4067 retval = -EINVAL;
4068 goto out;
4069 }
4070
Harvey Harrison0dc47872008-03-05 20:47:47 -08004071 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004072
4073 retval = sctp_do_peeloff(asoc, &newsock);
4074 if (retval < 0)
4075 goto out;
4076
4077 /* Map the socket to an unused fd that can be returned to the user. */
Ulrich Dreppera677a032008-07-23 21:29:17 -07004078 retval = sock_map_fd(newsock, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004079 if (retval < 0) {
4080 sock_release(newsock);
4081 goto out;
4082 }
4083
4084 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08004085 __func__, sk, asoc, newsock->sk, retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004086
4087 /* Return the fd mapped to the new socket. */
4088 peeloff.sd = retval;
Neil Horman408f22e2007-06-16 14:03:45 -04004089 if (put_user(len, optlen))
4090 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004091 if (copy_to_user(optval, &peeloff, len))
4092 retval = -EFAULT;
4093
4094out:
4095 return retval;
4096}
4097
4098/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4099 *
4100 * Applications can enable or disable heartbeats for any peer address of
4101 * an association, modify an address's heartbeat interval, force a
4102 * heartbeat to be sent immediately, and adjust the address's maximum
4103 * number of retransmissions sent before an address is considered
4104 * unreachable. The following structure is used to access and modify an
4105 * address's parameters:
4106 *
4107 * struct sctp_paddrparams {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004108 * sctp_assoc_t spp_assoc_id;
4109 * struct sockaddr_storage spp_address;
4110 * uint32_t spp_hbinterval;
4111 * uint16_t spp_pathmaxrxt;
4112 * uint32_t spp_pathmtu;
4113 * uint32_t spp_sackdelay;
4114 * uint32_t spp_flags;
4115 * };
Linus Torvalds1da177e2005-04-16 15:20:36 -07004116 *
Frank Filz52ccb8e2005-12-22 11:36:46 -08004117 * spp_assoc_id - (one-to-many style socket) This is filled in the
4118 * application, and identifies the association for
4119 * this query.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004120 * spp_address - This specifies which address is of interest.
4121 * spp_hbinterval - This contains the value of the heartbeat interval,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004122 * in milliseconds. If a value of zero
4123 * is present in this field then no changes are to
4124 * be made to this parameter.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004125 * spp_pathmaxrxt - This contains the maximum number of
4126 * retransmissions before this address shall be
Frank Filz52ccb8e2005-12-22 11:36:46 -08004127 * considered unreachable. If a value of zero
4128 * is present in this field then no changes are to
4129 * be made to this parameter.
4130 * spp_pathmtu - When Path MTU discovery is disabled the value
4131 * specified here will be the "fixed" path mtu.
4132 * Note that if the spp_address field is empty
4133 * then all associations on this address will
4134 * have this fixed path mtu set upon them.
4135 *
4136 * spp_sackdelay - When delayed sack is enabled, this value specifies
4137 * the number of milliseconds that sacks will be delayed
4138 * for. This value will apply to all addresses of an
4139 * association if the spp_address field is empty. Note
4140 * also, that if delayed sack is enabled and this
4141 * value is set to 0, no change is made to the last
4142 * recorded delayed sack timer value.
4143 *
4144 * spp_flags - These flags are used to control various features
4145 * on an association. The flag field may contain
4146 * zero or more of the following options.
4147 *
4148 * SPP_HB_ENABLE - Enable heartbeats on the
4149 * specified address. Note that if the address
4150 * field is empty all addresses for the association
4151 * have heartbeats enabled upon them.
4152 *
4153 * SPP_HB_DISABLE - Disable heartbeats on the
4154 * speicifed address. Note that if the address
4155 * field is empty all addresses for the association
4156 * will have their heartbeats disabled. Note also
4157 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4158 * mutually exclusive, only one of these two should
4159 * be specified. Enabling both fields will have
4160 * undetermined results.
4161 *
4162 * SPP_HB_DEMAND - Request a user initiated heartbeat
4163 * to be made immediately.
4164 *
4165 * SPP_PMTUD_ENABLE - This field will enable PMTU
4166 * discovery upon the specified address. Note that
4167 * if the address feild is empty then all addresses
4168 * on the association are effected.
4169 *
4170 * SPP_PMTUD_DISABLE - This field will disable PMTU
4171 * discovery upon the specified address. Note that
4172 * if the address feild is empty then all addresses
4173 * on the association are effected. Not also that
4174 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4175 * exclusive. Enabling both will have undetermined
4176 * results.
4177 *
4178 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4179 * on delayed sack. The time specified in spp_sackdelay
4180 * is used to specify the sack delay for this address. Note
4181 * that if spp_address is empty then all addresses will
4182 * enable delayed sack and take on the sack delay
4183 * value specified in spp_sackdelay.
4184 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4185 * off delayed sack. If the spp_address field is blank then
4186 * delayed sack is disabled for the entire association. Note
4187 * also that this field is mutually exclusive to
4188 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4189 * results.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004190 */
4191static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
Frank Filz52ccb8e2005-12-22 11:36:46 -08004192 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004193{
Frank Filz52ccb8e2005-12-22 11:36:46 -08004194 struct sctp_paddrparams params;
4195 struct sctp_transport *trans = NULL;
4196 struct sctp_association *asoc = NULL;
4197 struct sctp_sock *sp = sctp_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004198
Neil Horman408f22e2007-06-16 14:03:45 -04004199 if (len < sizeof(struct sctp_paddrparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004200 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004201 len = sizeof(struct sctp_paddrparams);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004202 if (copy_from_user(&params, optval, len))
4203 return -EFAULT;
4204
Frank Filz52ccb8e2005-12-22 11:36:46 -08004205 /* If an address other than INADDR_ANY is specified, and
4206 * no transport is found, then the request is invalid.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004207 */
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004208 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08004209 trans = sctp_addr_id2transport(sk, &params.spp_address,
4210 params.spp_assoc_id);
4211 if (!trans) {
4212 SCTP_DEBUG_PRINTK("Failed no transport\n");
4213 return -EINVAL;
4214 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004215 }
4216
Frank Filz52ccb8e2005-12-22 11:36:46 -08004217 /* Get association, if assoc_id != 0 and the socket is a one
4218 * to many style socket, and an association was not found, then
4219 * the id was invalid.
4220 */
4221 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4222 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4223 SCTP_DEBUG_PRINTK("Failed no association\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07004224 return -EINVAL;
Frank Filz52ccb8e2005-12-22 11:36:46 -08004225 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004226
Frank Filz52ccb8e2005-12-22 11:36:46 -08004227 if (trans) {
4228 /* Fetch transport values. */
4229 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4230 params.spp_pathmtu = trans->pathmtu;
4231 params.spp_pathmaxrxt = trans->pathmaxrxt;
4232 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004233
Frank Filz52ccb8e2005-12-22 11:36:46 -08004234 /*draft-11 doesn't say what to return in spp_flags*/
4235 params.spp_flags = trans->param_flags;
4236 } else if (asoc) {
4237 /* Fetch association values. */
4238 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4239 params.spp_pathmtu = asoc->pathmtu;
4240 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4241 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004242
Frank Filz52ccb8e2005-12-22 11:36:46 -08004243 /*draft-11 doesn't say what to return in spp_flags*/
4244 params.spp_flags = asoc->param_flags;
4245 } else {
4246 /* Fetch socket values. */
4247 params.spp_hbinterval = sp->hbinterval;
4248 params.spp_pathmtu = sp->pathmtu;
4249 params.spp_sackdelay = sp->sackdelay;
4250 params.spp_pathmaxrxt = sp->pathmaxrxt;
4251
4252 /*draft-11 doesn't say what to return in spp_flags*/
4253 params.spp_flags = sp->param_flags;
4254 }
4255
Linus Torvalds1da177e2005-04-16 15:20:36 -07004256 if (copy_to_user(optval, &params, len))
4257 return -EFAULT;
4258
4259 if (put_user(len, optlen))
4260 return -EFAULT;
4261
4262 return 0;
4263}
4264
Wei Yongjund364d922008-05-09 15:13:26 -07004265/*
4266 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
Frank Filz77086102005-12-22 11:37:30 -08004267 *
Wei Yongjund364d922008-05-09 15:13:26 -07004268 * This option will effect the way delayed acks are performed. This
4269 * option allows you to get or set the delayed ack time, in
4270 * milliseconds. It also allows changing the delayed ack frequency.
4271 * Changing the frequency to 1 disables the delayed sack algorithm. If
4272 * the assoc_id is 0, then this sets or gets the endpoints default
4273 * values. If the assoc_id field is non-zero, then the set or get
4274 * effects the specified association for the one to many model (the
4275 * assoc_id field is ignored by the one to one model). Note that if
4276 * sack_delay or sack_freq are 0 when setting this option, then the
4277 * current values will remain unchanged.
Frank Filz77086102005-12-22 11:37:30 -08004278 *
Wei Yongjund364d922008-05-09 15:13:26 -07004279 * struct sctp_sack_info {
4280 * sctp_assoc_t sack_assoc_id;
4281 * uint32_t sack_delay;
4282 * uint32_t sack_freq;
4283 * };
Frank Filz77086102005-12-22 11:37:30 -08004284 *
Wei Yongjund364d922008-05-09 15:13:26 -07004285 * sack_assoc_id - This parameter, indicates which association the user
4286 * is performing an action upon. Note that if this field's value is
4287 * zero then the endpoints default value is changed (effecting future
4288 * associations only).
Frank Filz77086102005-12-22 11:37:30 -08004289 *
Wei Yongjund364d922008-05-09 15:13:26 -07004290 * sack_delay - This parameter contains the number of milliseconds that
4291 * the user is requesting the delayed ACK timer be set to. Note that
4292 * this value is defined in the standard to be between 200 and 500
4293 * milliseconds.
Frank Filz77086102005-12-22 11:37:30 -08004294 *
Wei Yongjund364d922008-05-09 15:13:26 -07004295 * sack_freq - This parameter contains the number of packets that must
4296 * be received before a sack is sent without waiting for the delay
4297 * timer to expire. The default value for this is 2, setting this
4298 * value to 1 will disable the delayed sack algorithm.
Frank Filz77086102005-12-22 11:37:30 -08004299 */
Wei Yongjund364d922008-05-09 15:13:26 -07004300static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
Frank Filz77086102005-12-22 11:37:30 -08004301 char __user *optval,
4302 int __user *optlen)
4303{
Wei Yongjund364d922008-05-09 15:13:26 -07004304 struct sctp_sack_info params;
Frank Filz77086102005-12-22 11:37:30 -08004305 struct sctp_association *asoc = NULL;
4306 struct sctp_sock *sp = sctp_sk(sk);
4307
Wei Yongjund364d922008-05-09 15:13:26 -07004308 if (len >= sizeof(struct sctp_sack_info)) {
4309 len = sizeof(struct sctp_sack_info);
4310
4311 if (copy_from_user(&params, optval, len))
4312 return -EFAULT;
4313 } else if (len == sizeof(struct sctp_assoc_value)) {
Joe Perches145ce502010-08-24 13:21:08 +00004314 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4315 pr_warn("Use struct sctp_sack_info instead\n");
Wei Yongjund364d922008-05-09 15:13:26 -07004316 if (copy_from_user(&params, optval, len))
4317 return -EFAULT;
4318 } else
Frank Filz77086102005-12-22 11:37:30 -08004319 return - EINVAL;
4320
Wei Yongjund364d922008-05-09 15:13:26 -07004321 /* Get association, if sack_assoc_id != 0 and the socket is a one
Frank Filz77086102005-12-22 11:37:30 -08004322 * to many style socket, and an association was not found, then
4323 * the id was invalid.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004324 */
Wei Yongjund364d922008-05-09 15:13:26 -07004325 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4326 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
Frank Filz77086102005-12-22 11:37:30 -08004327 return -EINVAL;
4328
4329 if (asoc) {
4330 /* Fetch association values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004331 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4332 params.sack_delay = jiffies_to_msecs(
Frank Filz77086102005-12-22 11:37:30 -08004333 asoc->sackdelay);
Wei Yongjund364d922008-05-09 15:13:26 -07004334 params.sack_freq = asoc->sackfreq;
4335
4336 } else {
4337 params.sack_delay = 0;
4338 params.sack_freq = 1;
4339 }
Frank Filz77086102005-12-22 11:37:30 -08004340 } else {
4341 /* Fetch socket values. */
Wei Yongjund364d922008-05-09 15:13:26 -07004342 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4343 params.sack_delay = sp->sackdelay;
4344 params.sack_freq = sp->sackfreq;
4345 } else {
4346 params.sack_delay = 0;
4347 params.sack_freq = 1;
4348 }
Frank Filz77086102005-12-22 11:37:30 -08004349 }
4350
4351 if (copy_to_user(optval, &params, len))
4352 return -EFAULT;
4353
4354 if (put_user(len, optlen))
4355 return -EFAULT;
4356
4357 return 0;
4358}
4359
Linus Torvalds1da177e2005-04-16 15:20:36 -07004360/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4361 *
4362 * Applications can specify protocol parameters for the default association
4363 * initialization. The option name argument to setsockopt() and getsockopt()
4364 * is SCTP_INITMSG.
4365 *
4366 * Setting initialization parameters is effective only on an unconnected
4367 * socket (for UDP-style sockets only future associations are effected
4368 * by the change). With TCP-style sockets, this option is inherited by
4369 * sockets derived from a listener socket.
4370 */
4371static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4372{
Neil Horman408f22e2007-06-16 14:03:45 -04004373 if (len < sizeof(struct sctp_initmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004374 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004375 len = sizeof(struct sctp_initmsg);
4376 if (put_user(len, optlen))
4377 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004378 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4379 return -EFAULT;
4380 return 0;
4381}
4382
Linus Torvalds1da177e2005-04-16 15:20:36 -07004383
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004384static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4385 char __user *optval, int __user *optlen)
4386{
4387 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004388 int cnt = 0;
4389 struct sctp_getaddrs getaddrs;
4390 struct sctp_transport *from;
4391 void __user *to;
4392 union sctp_addr temp;
4393 struct sctp_sock *sp = sctp_sk(sk);
4394 int addrlen;
4395 size_t space_left;
4396 int bytes_copied;
4397
4398 if (len < sizeof(struct sctp_getaddrs))
4399 return -EINVAL;
4400
4401 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4402 return -EFAULT;
4403
4404 /* For UDP-style sockets, id specifies the association to query. */
4405 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4406 if (!asoc)
4407 return -EINVAL;
4408
4409 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004410 space_left = len - offsetof(struct sctp_getaddrs,addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004411
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07004412 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4413 transports) {
Al Virob3f5b3b2006-11-20 17:22:43 -08004414 memcpy(&temp, &from->ipaddr, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004415 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
Vlad Yasevicha5f4cea2010-04-30 21:42:42 -04004416 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004417 if (space_left < addrlen)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004418 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004419 if (copy_to_user(to, &temp, addrlen))
4420 return -EFAULT;
4421 to += addrlen;
4422 cnt++;
4423 space_left -= addrlen;
4424 }
4425
4426 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4427 return -EFAULT;
4428 bytes_copied = ((char __user *)to) - optval;
4429 if (put_user(bytes_copied, optlen))
4430 return -EFAULT;
4431
4432 return 0;
4433}
4434
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004435static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4436 size_t space_left, int *bytes_copied)
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004437{
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004438 struct sctp_sockaddr_entry *addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004439 union sctp_addr temp;
4440 int cnt = 0;
4441 int addrlen;
4442
Vlad Yasevich29303542007-09-16 16:02:12 -07004443 rcu_read_lock();
4444 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4445 if (!addr->valid)
4446 continue;
4447
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004448 if ((PF_INET == sk->sk_family) &&
Al Viro6244be42006-11-20 17:21:44 -08004449 (AF_INET6 == addr->a.sa.sa_family))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004450 continue;
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07004451 if ((PF_INET6 == sk->sk_family) &&
4452 inet_v6_ipv6only(sk) &&
4453 (AF_INET == addr->a.sa.sa_family))
4454 continue;
Al Viro6244be42006-11-20 17:21:44 -08004455 memcpy(&temp, &addr->a, sizeof(temp));
Vlad Yasevichb46ae362008-01-28 14:25:36 -05004456 if (!temp.v4.sin_port)
4457 temp.v4.sin_port = htons(port);
4458
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004459 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4460 &temp);
4461 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevich29303542007-09-16 16:02:12 -07004462 if (space_left < addrlen) {
4463 cnt = -ENOMEM;
4464 break;
4465 }
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004466 memcpy(to, &temp, addrlen);
Sridhar Samudrala29c7cf92006-12-13 16:26:26 -08004467
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004468 to += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004469 cnt ++;
4470 space_left -= addrlen;
Vlad Yasevich3663c302007-07-03 12:43:12 -04004471 *bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004472 }
Vlad Yasevich29303542007-09-16 16:02:12 -07004473 rcu_read_unlock();
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004474
4475 return cnt;
4476}
4477
Linus Torvalds1da177e2005-04-16 15:20:36 -07004478
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004479static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4480 char __user *optval, int __user *optlen)
4481{
4482 struct sctp_bind_addr *bp;
4483 struct sctp_association *asoc;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004484 int cnt = 0;
4485 struct sctp_getaddrs getaddrs;
4486 struct sctp_sockaddr_entry *addr;
4487 void __user *to;
4488 union sctp_addr temp;
4489 struct sctp_sock *sp = sctp_sk(sk);
4490 int addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004491 int err = 0;
4492 size_t space_left;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004493 int bytes_copied = 0;
4494 void *addrs;
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004495 void *buf;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004496
Neil Horman408f22e2007-06-16 14:03:45 -04004497 if (len < sizeof(struct sctp_getaddrs))
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004498 return -EINVAL;
4499
4500 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4501 return -EFAULT;
4502
4503 /*
4504 * For UDP-style sockets, id specifies the association to query.
4505 * If the id field is set to the value '0' then the locally bound
4506 * addresses are returned without regard to any particular
4507 * association.
4508 */
4509 if (0 == getaddrs.assoc_id) {
4510 bp = &sctp_sk(sk)->ep->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004511 } else {
4512 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4513 if (!asoc)
4514 return -EINVAL;
4515 bp = &asoc->base.bind_addr;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004516 }
4517
4518 to = optval + offsetof(struct sctp_getaddrs,addrs);
Neil Horman186e2342007-06-18 19:59:16 -04004519 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4520
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004521 addrs = kmalloc(space_left, GFP_KERNEL);
4522 if (!addrs)
4523 return -ENOMEM;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004524
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004525 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4526 * addresses from the global local address list.
4527 */
4528 if (sctp_list_single_entry(&bp->address_list)) {
4529 addr = list_entry(bp->address_list.next,
4530 struct sctp_sockaddr_entry, list);
Vlad Yasevich52cae8f2008-08-18 10:34:34 -04004531 if (sctp_is_any(sk, &addr->a)) {
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004532 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4533 space_left, &bytes_copied);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004534 if (cnt < 0) {
4535 err = cnt;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004536 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004537 }
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004538 goto copy_getaddrs;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004539 }
4540 }
4541
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004542 buf = addrs;
Vlad Yasevich559cf712007-09-16 16:03:28 -07004543 /* Protection on the bound address list is not needed since
4544 * in the socket option context we hold a socket lock and
4545 * thus the bound address list can't change.
4546 */
4547 list_for_each_entry(addr, &bp->address_list, list) {
Al Viro6244be42006-11-20 17:21:44 -08004548 memcpy(&temp, &addr->a, sizeof(temp));
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004549 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4550 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004551 if (space_left < addrlen) {
4552 err = -ENOMEM; /*fixme: right error?*/
Vlad Yasevich559cf712007-09-16 16:03:28 -07004553 goto out;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004554 }
Vlad Yasevich70b57b82007-05-09 13:51:31 -07004555 memcpy(buf, &temp, addrlen);
4556 buf += addrlen;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004557 bytes_copied += addrlen;
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004558 cnt ++;
4559 space_left -= addrlen;
4560 }
4561
4562copy_getaddrs:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004563 if (copy_to_user(to, addrs, bytes_copied)) {
4564 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004565 goto out;
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004566 }
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004567 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4568 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004569 goto out;
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004570 }
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004571 if (put_user(bytes_copied, optlen))
Vlad Yasevichfe979ac2007-05-23 11:11:37 -04004572 err = -EFAULT;
Sebastian Siewiord6f9fda2007-07-27 22:55:59 +02004573out:
Vlad Yasevichaad97f32007-04-28 21:09:04 -07004574 kfree(addrs);
Ivan Skytte Jørgensen5fe467e2005-10-06 21:36:17 -07004575 return err;
4576}
4577
Linus Torvalds1da177e2005-04-16 15:20:36 -07004578/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4579 *
4580 * Requests that the local SCTP stack use the enclosed peer address as
4581 * the association primary. The enclosed address must be one of the
4582 * association peer's addresses.
4583 */
4584static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4585 char __user *optval, int __user *optlen)
4586{
4587 struct sctp_prim prim;
4588 struct sctp_association *asoc;
4589 struct sctp_sock *sp = sctp_sk(sk);
4590
Neil Horman408f22e2007-06-16 14:03:45 -04004591 if (len < sizeof(struct sctp_prim))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004592 return -EINVAL;
4593
Neil Horman408f22e2007-06-16 14:03:45 -04004594 len = sizeof(struct sctp_prim);
4595
4596 if (copy_from_user(&prim, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004597 return -EFAULT;
4598
4599 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4600 if (!asoc)
4601 return -EINVAL;
4602
4603 if (!asoc->peer.primary_path)
4604 return -ENOTCONN;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09004605
Al Viro8cec6b82006-11-20 17:23:01 -08004606 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4607 asoc->peer.primary_path->af_specific->sockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07004608
4609 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4610 (union sctp_addr *)&prim.ssp_addr);
4611
Neil Horman408f22e2007-06-16 14:03:45 -04004612 if (put_user(len, optlen))
4613 return -EFAULT;
4614 if (copy_to_user(optval, &prim, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004615 return -EFAULT;
4616
4617 return 0;
4618}
4619
4620/*
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004621 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004622 *
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004623 * Requests that the local endpoint set the specified Adaptation Layer
Linus Torvalds1da177e2005-04-16 15:20:36 -07004624 * Indication parameter for all future INIT and INIT-ACK exchanges.
4625 */
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004626static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07004627 char __user *optval, int __user *optlen)
4628{
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004629 struct sctp_setadaptation adaptation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004630
Neil Horman408f22e2007-06-16 14:03:45 -04004631 if (len < sizeof(struct sctp_setadaptation))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004632 return -EINVAL;
4633
Neil Horman408f22e2007-06-16 14:03:45 -04004634 len = sizeof(struct sctp_setadaptation);
4635
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004636 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
Neil Horman408f22e2007-06-16 14:03:45 -04004637
4638 if (put_user(len, optlen))
4639 return -EFAULT;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08004640 if (copy_to_user(optval, &adaptation, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004641 return -EFAULT;
Ivan Skytte Jorgensena1ab3582005-10-28 15:33:24 -07004642
Linus Torvalds1da177e2005-04-16 15:20:36 -07004643 return 0;
4644}
4645
4646/*
4647 *
4648 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4649 *
4650 * Applications that wish to use the sendto() system call may wish to
4651 * specify a default set of parameters that would normally be supplied
4652 * through the inclusion of ancillary data. This socket option allows
4653 * such an application to set the default sctp_sndrcvinfo structure.
4654
4655
4656 * The application that wishes to use this socket option simply passes
4657 * in to this call the sctp_sndrcvinfo structure defined in Section
4658 * 5.2.2) The input parameters accepted by this call include
4659 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4660 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4661 * to this call if the caller is using the UDP model.
4662 *
4663 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4664 */
4665static int sctp_getsockopt_default_send_param(struct sock *sk,
4666 int len, char __user *optval,
4667 int __user *optlen)
4668{
4669 struct sctp_sndrcvinfo info;
4670 struct sctp_association *asoc;
4671 struct sctp_sock *sp = sctp_sk(sk);
4672
Neil Horman408f22e2007-06-16 14:03:45 -04004673 if (len < sizeof(struct sctp_sndrcvinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004674 return -EINVAL;
Neil Horman408f22e2007-06-16 14:03:45 -04004675
4676 len = sizeof(struct sctp_sndrcvinfo);
4677
4678 if (copy_from_user(&info, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004679 return -EFAULT;
4680
4681 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4682 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4683 return -EINVAL;
4684
4685 if (asoc) {
4686 info.sinfo_stream = asoc->default_stream;
4687 info.sinfo_flags = asoc->default_flags;
4688 info.sinfo_ppid = asoc->default_ppid;
4689 info.sinfo_context = asoc->default_context;
4690 info.sinfo_timetolive = asoc->default_timetolive;
4691 } else {
4692 info.sinfo_stream = sp->default_stream;
4693 info.sinfo_flags = sp->default_flags;
4694 info.sinfo_ppid = sp->default_ppid;
4695 info.sinfo_context = sp->default_context;
4696 info.sinfo_timetolive = sp->default_timetolive;
4697 }
4698
Neil Horman408f22e2007-06-16 14:03:45 -04004699 if (put_user(len, optlen))
4700 return -EFAULT;
4701 if (copy_to_user(optval, &info, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004702 return -EFAULT;
4703
4704 return 0;
4705}
4706
4707/*
4708 *
4709 * 7.1.5 SCTP_NODELAY
4710 *
4711 * Turn on/off any Nagle-like algorithm. This means that packets are
4712 * generally sent as soon as possible and no unnecessary delays are
4713 * introduced, at the cost of more packets in the network. Expects an
4714 * integer boolean flag.
4715 */
4716
4717static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4718 char __user *optval, int __user *optlen)
4719{
4720 int val;
4721
4722 if (len < sizeof(int))
4723 return -EINVAL;
4724
4725 len = sizeof(int);
4726 val = (sctp_sk(sk)->nodelay == 1);
4727 if (put_user(len, optlen))
4728 return -EFAULT;
4729 if (copy_to_user(optval, &val, len))
4730 return -EFAULT;
4731 return 0;
4732}
4733
4734/*
4735 *
4736 * 7.1.1 SCTP_RTOINFO
4737 *
4738 * The protocol parameters used to initialize and bound retransmission
4739 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4740 * and modify these parameters.
4741 * All parameters are time values, in milliseconds. A value of 0, when
4742 * modifying the parameters, indicates that the current value should not
4743 * be changed.
4744 *
4745 */
4746static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4747 char __user *optval,
4748 int __user *optlen) {
4749 struct sctp_rtoinfo rtoinfo;
4750 struct sctp_association *asoc;
4751
Neil Horman408f22e2007-06-16 14:03:45 -04004752 if (len < sizeof (struct sctp_rtoinfo))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004753 return -EINVAL;
4754
Neil Horman408f22e2007-06-16 14:03:45 -04004755 len = sizeof(struct sctp_rtoinfo);
4756
4757 if (copy_from_user(&rtoinfo, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004758 return -EFAULT;
4759
4760 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4761
4762 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4763 return -EINVAL;
4764
4765 /* Values corresponding to the specific association. */
4766 if (asoc) {
4767 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4768 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4769 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4770 } else {
4771 /* Values corresponding to the endpoint. */
4772 struct sctp_sock *sp = sctp_sk(sk);
4773
4774 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4775 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4776 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4777 }
4778
4779 if (put_user(len, optlen))
4780 return -EFAULT;
4781
4782 if (copy_to_user(optval, &rtoinfo, len))
4783 return -EFAULT;
4784
4785 return 0;
4786}
4787
4788/*
4789 *
4790 * 7.1.2 SCTP_ASSOCINFO
4791 *
Michael Opdenacker59c51592007-05-09 08:57:56 +02004792 * This option is used to tune the maximum retransmission attempts
Linus Torvalds1da177e2005-04-16 15:20:36 -07004793 * of the association.
4794 * Returns an error if the new association retransmission value is
4795 * greater than the sum of the retransmission value of the peer.
4796 * See [SCTP] for more information.
4797 *
4798 */
4799static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4800 char __user *optval,
4801 int __user *optlen)
4802{
4803
4804 struct sctp_assocparams assocparams;
4805 struct sctp_association *asoc;
4806 struct list_head *pos;
4807 int cnt = 0;
4808
Neil Horman408f22e2007-06-16 14:03:45 -04004809 if (len < sizeof (struct sctp_assocparams))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004810 return -EINVAL;
4811
Neil Horman408f22e2007-06-16 14:03:45 -04004812 len = sizeof(struct sctp_assocparams);
4813
4814 if (copy_from_user(&assocparams, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07004815 return -EFAULT;
4816
4817 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4818
4819 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4820 return -EINVAL;
4821
4822 /* Values correspoinding to the specific association */
Vladislav Yasevich17337212005-04-28 11:57:54 -07004823 if (asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07004824 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4825 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4826 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4827 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4828 * 1000) +
4829 (asoc->cookie_life.tv_usec
4830 / 1000);
4831
4832 list_for_each(pos, &asoc->peer.transport_addr_list) {
4833 cnt ++;
4834 }
4835
4836 assocparams.sasoc_number_peer_destinations = cnt;
4837 } else {
4838 /* Values corresponding to the endpoint */
4839 struct sctp_sock *sp = sctp_sk(sk);
4840
4841 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4842 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4843 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4844 assocparams.sasoc_cookie_life =
4845 sp->assocparams.sasoc_cookie_life;
4846 assocparams.sasoc_number_peer_destinations =
4847 sp->assocparams.
4848 sasoc_number_peer_destinations;
4849 }
4850
4851 if (put_user(len, optlen))
4852 return -EFAULT;
4853
4854 if (copy_to_user(optval, &assocparams, len))
4855 return -EFAULT;
4856
4857 return 0;
4858}
4859
4860/*
4861 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4862 *
4863 * This socket option is a boolean flag which turns on or off mapped V4
4864 * addresses. If this option is turned on and the socket is type
4865 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4866 * If this option is turned off, then no mapping will be done of V4
4867 * addresses and a user will receive both PF_INET6 and PF_INET type
4868 * addresses on the socket.
4869 */
4870static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4871 char __user *optval, int __user *optlen)
4872{
4873 int val;
4874 struct sctp_sock *sp = sctp_sk(sk);
4875
4876 if (len < sizeof(int))
4877 return -EINVAL;
4878
4879 len = sizeof(int);
4880 val = sp->v4mapped;
4881 if (put_user(len, optlen))
4882 return -EFAULT;
4883 if (copy_to_user(optval, &val, len))
4884 return -EFAULT;
4885
4886 return 0;
4887}
4888
4889/*
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004890 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4891 * (chapter and verse is quoted at sctp_setsockopt_context())
4892 */
4893static int sctp_getsockopt_context(struct sock *sk, int len,
4894 char __user *optval, int __user *optlen)
4895{
4896 struct sctp_assoc_value params;
4897 struct sctp_sock *sp;
4898 struct sctp_association *asoc;
4899
Neil Horman408f22e2007-06-16 14:03:45 -04004900 if (len < sizeof(struct sctp_assoc_value))
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004901 return -EINVAL;
4902
Neil Horman408f22e2007-06-16 14:03:45 -04004903 len = sizeof(struct sctp_assoc_value);
4904
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08004905 if (copy_from_user(&params, optval, len))
4906 return -EFAULT;
4907
4908 sp = sctp_sk(sk);
4909
4910 if (params.assoc_id != 0) {
4911 asoc = sctp_id2assoc(sk, params.assoc_id);
4912 if (!asoc)
4913 return -EINVAL;
4914 params.assoc_value = asoc->default_rcv_context;
4915 } else {
4916 params.assoc_value = sp->default_rcv_context;
4917 }
4918
4919 if (put_user(len, optlen))
4920 return -EFAULT;
4921 if (copy_to_user(optval, &params, len))
4922 return -EFAULT;
4923
4924 return 0;
4925}
4926
4927/*
Wei Yongjune89c2092008-12-25 16:54:58 -08004928 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4929 * This option will get or set the maximum size to put in any outgoing
4930 * SCTP DATA chunk. If a message is larger than this size it will be
Linus Torvalds1da177e2005-04-16 15:20:36 -07004931 * fragmented by SCTP into the specified size. Note that the underlying
4932 * SCTP implementation may fragment into smaller sized chunks when the
4933 * PMTU of the underlying association is smaller than the value set by
Wei Yongjune89c2092008-12-25 16:54:58 -08004934 * the user. The default value for this option is '0' which indicates
4935 * the user is NOT limiting fragmentation and only the PMTU will effect
4936 * SCTP's choice of DATA chunk size. Note also that values set larger
4937 * than the maximum size of an IP datagram will effectively let SCTP
4938 * control fragmentation (i.e. the same as setting this option to 0).
4939 *
4940 * The following structure is used to access and modify this parameter:
4941 *
4942 * struct sctp_assoc_value {
4943 * sctp_assoc_t assoc_id;
4944 * uint32_t assoc_value;
4945 * };
4946 *
4947 * assoc_id: This parameter is ignored for one-to-one style sockets.
4948 * For one-to-many style sockets this parameter indicates which
4949 * association the user is performing an action upon. Note that if
4950 * this field's value is zero then the endpoints default value is
4951 * changed (effecting future associations only).
4952 * assoc_value: This parameter specifies the maximum size in bytes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004953 */
4954static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4955 char __user *optval, int __user *optlen)
4956{
Wei Yongjune89c2092008-12-25 16:54:58 -08004957 struct sctp_assoc_value params;
4958 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004959
Wei Yongjune89c2092008-12-25 16:54:58 -08004960 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00004961 pr_warn("Use of int in maxseg socket option deprecated\n");
4962 pr_warn("Use struct sctp_assoc_value instead\n");
Wei Yongjune89c2092008-12-25 16:54:58 -08004963 params.assoc_id = 0;
4964 } else if (len >= sizeof(struct sctp_assoc_value)) {
4965 len = sizeof(struct sctp_assoc_value);
4966 if (copy_from_user(&params, optval, sizeof(params)))
4967 return -EFAULT;
4968 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -07004969 return -EINVAL;
4970
Wei Yongjune89c2092008-12-25 16:54:58 -08004971 asoc = sctp_id2assoc(sk, params.assoc_id);
4972 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4973 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07004974
Wei Yongjune89c2092008-12-25 16:54:58 -08004975 if (asoc)
4976 params.assoc_value = asoc->frag_point;
4977 else
4978 params.assoc_value = sctp_sk(sk)->user_frag;
4979
Linus Torvalds1da177e2005-04-16 15:20:36 -07004980 if (put_user(len, optlen))
4981 return -EFAULT;
Wei Yongjune89c2092008-12-25 16:54:58 -08004982 if (len == sizeof(int)) {
4983 if (copy_to_user(optval, &params.assoc_value, len))
4984 return -EFAULT;
4985 } else {
4986 if (copy_to_user(optval, &params, len))
4987 return -EFAULT;
4988 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07004989
4990 return 0;
4991}
4992
Vlad Yasevichb6e13312007-04-20 12:23:15 -07004993/*
4994 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4995 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4996 */
4997static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4998 char __user *optval, int __user *optlen)
4999{
5000 int val;
5001
5002 if (len < sizeof(int))
5003 return -EINVAL;
5004
5005 len = sizeof(int);
5006
5007 val = sctp_sk(sk)->frag_interleave;
5008 if (put_user(len, optlen))
5009 return -EFAULT;
5010 if (copy_to_user(optval, &val, len))
5011 return -EFAULT;
5012
5013 return 0;
5014}
5015
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005016/*
5017 * 7.1.25. Set or Get the sctp partial delivery point
5018 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5019 */
5020static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5021 char __user *optval,
5022 int __user *optlen)
5023{
YOSHIFUJI Hideaki9cbcbf42007-07-19 10:44:50 +09005024 u32 val;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005025
5026 if (len < sizeof(u32))
5027 return -EINVAL;
5028
5029 len = sizeof(u32);
5030
5031 val = sctp_sk(sk)->pd_point;
5032 if (put_user(len, optlen))
5033 return -EFAULT;
5034 if (copy_to_user(optval, &val, len))
5035 return -EFAULT;
5036
5037 return -ENOTSUPP;
5038}
5039
Vlad Yasevich70331572007-03-23 11:34:36 -07005040/*
5041 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5042 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5043 */
5044static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5045 char __user *optval,
5046 int __user *optlen)
5047{
Neil Horman219b99a2008-03-05 13:44:46 -08005048 struct sctp_assoc_value params;
5049 struct sctp_sock *sp;
5050 struct sctp_association *asoc;
Vlad Yasevich70331572007-03-23 11:34:36 -07005051
Neil Horman219b99a2008-03-05 13:44:46 -08005052 if (len == sizeof(int)) {
Joe Perches145ce502010-08-24 13:21:08 +00005053 pr_warn("Use of int in max_burst socket option deprecated\n");
5054 pr_warn("Use struct sctp_assoc_value instead\n");
Neil Horman219b99a2008-03-05 13:44:46 -08005055 params.assoc_id = 0;
Wei Yongjunc6db93a2009-03-02 09:46:12 +00005056 } else if (len >= sizeof(struct sctp_assoc_value)) {
5057 len = sizeof(struct sctp_assoc_value);
Neil Horman219b99a2008-03-05 13:44:46 -08005058 if (copy_from_user(&params, optval, len))
5059 return -EFAULT;
5060 } else
5061 return -EINVAL;
Vlad Yasevich70331572007-03-23 11:34:36 -07005062
Neil Horman219b99a2008-03-05 13:44:46 -08005063 sp = sctp_sk(sk);
Vlad Yasevich70331572007-03-23 11:34:36 -07005064
Neil Horman219b99a2008-03-05 13:44:46 -08005065 if (params.assoc_id != 0) {
5066 asoc = sctp_id2assoc(sk, params.assoc_id);
5067 if (!asoc)
5068 return -EINVAL;
5069 params.assoc_value = asoc->max_burst;
5070 } else
5071 params.assoc_value = sp->max_burst;
5072
5073 if (len == sizeof(int)) {
5074 if (copy_to_user(optval, &params.assoc_value, len))
5075 return -EFAULT;
5076 } else {
5077 if (copy_to_user(optval, &params, len))
5078 return -EFAULT;
5079 }
5080
5081 return 0;
5082
Vlad Yasevich70331572007-03-23 11:34:36 -07005083}
5084
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005085static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5086 char __user *optval, int __user *optlen)
5087{
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005088 struct sctp_hmacalgo __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005089 struct sctp_hmac_algo_param *hmacs;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005090 __u16 data_len = 0;
5091 u32 num_idents;
5092
5093 if (!sctp_auth_enable)
5094 return -EACCES;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005095
5096 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005097 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005098
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005099 if (len < sizeof(struct sctp_hmacalgo) + data_len)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005100 return -EINVAL;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005101
5102 len = sizeof(struct sctp_hmacalgo) + data_len;
5103 num_idents = data_len / sizeof(u16);
5104
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005105 if (put_user(len, optlen))
5106 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005107 if (put_user(num_idents, &p->shmac_num_idents))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005108 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005109 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5110 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005111 return 0;
5112}
5113
5114static int sctp_getsockopt_active_key(struct sock *sk, int len,
5115 char __user *optval, int __user *optlen)
5116{
5117 struct sctp_authkeyid val;
5118 struct sctp_association *asoc;
5119
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005120 if (!sctp_auth_enable)
5121 return -EACCES;
5122
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005123 if (len < sizeof(struct sctp_authkeyid))
5124 return -EINVAL;
5125 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5126 return -EFAULT;
5127
5128 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5129 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5130 return -EINVAL;
5131
5132 if (asoc)
5133 val.scact_keynumber = asoc->active_key_id;
5134 else
5135 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5136
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005137 len = sizeof(struct sctp_authkeyid);
5138 if (put_user(len, optlen))
5139 return -EFAULT;
5140 if (copy_to_user(optval, &val, len))
5141 return -EFAULT;
5142
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005143 return 0;
5144}
5145
5146static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5147 char __user *optval, int __user *optlen)
5148{
Al Viro411223c2007-10-14 19:21:20 +01005149 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005150 struct sctp_authchunks val;
5151 struct sctp_association *asoc;
5152 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005153 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005154 char __user *to;
5155
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005156 if (!sctp_auth_enable)
5157 return -EACCES;
5158
5159 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005160 return -EINVAL;
5161
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005162 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005163 return -EFAULT;
5164
Al Viro411223c2007-10-14 19:21:20 +01005165 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005166 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5167 if (!asoc)
5168 return -EINVAL;
5169
5170 ch = asoc->peer.peer_chunks;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005171 if (!ch)
5172 goto num;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005173
5174 /* See if the user provided enough room for all the data */
Vlad Yasevichb40db682008-02-27 14:40:37 -05005175 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5176 if (len < num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005177 return -EINVAL;
5178
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005179 if (copy_to_user(to, ch->chunks, num_chunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005180 return -EFAULT;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005181num:
5182 len = sizeof(struct sctp_authchunks) + num_chunks;
5183 if (put_user(len, optlen)) return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005184 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5185 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005186 return 0;
5187}
5188
5189static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5190 char __user *optval, int __user *optlen)
5191{
Al Viro411223c2007-10-14 19:21:20 +01005192 struct sctp_authchunks __user *p = (void __user *)optval;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005193 struct sctp_authchunks val;
5194 struct sctp_association *asoc;
5195 struct sctp_chunks_param *ch;
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005196 u32 num_chunks = 0;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005197 char __user *to;
5198
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005199 if (!sctp_auth_enable)
5200 return -EACCES;
5201
5202 if (len < sizeof(struct sctp_authchunks))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005203 return -EINVAL;
5204
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005205 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005206 return -EFAULT;
5207
Al Viro411223c2007-10-14 19:21:20 +01005208 to = p->gauth_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005209 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5210 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5211 return -EINVAL;
5212
5213 if (asoc)
5214 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5215 else
5216 ch = sctp_sk(sk)->ep->auth_chunk_list;
5217
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005218 if (!ch)
5219 goto num;
5220
Vlad Yasevichb40db682008-02-27 14:40:37 -05005221 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005222 if (len < sizeof(struct sctp_authchunks) + num_chunks)
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005223 return -EINVAL;
5224
Vlad Yasevich5e739d12008-08-21 03:34:25 -07005225 if (copy_to_user(to, ch->chunks, num_chunks))
5226 return -EFAULT;
5227num:
5228 len = sizeof(struct sctp_authchunks) + num_chunks;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005229 if (put_user(len, optlen))
5230 return -EFAULT;
Vlad Yasevich7e8616d2008-02-27 16:04:52 -05005231 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5232 return -EFAULT;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005233
5234 return 0;
5235}
5236
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005237/*
5238 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5239 * This option gets the current number of associations that are attached
5240 * to a one-to-many style socket. The option value is an uint32_t.
5241 */
5242static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5243 char __user *optval, int __user *optlen)
5244{
5245 struct sctp_sock *sp = sctp_sk(sk);
5246 struct sctp_association *asoc;
5247 u32 val = 0;
5248
5249 if (sctp_style(sk, TCP))
5250 return -EOPNOTSUPP;
5251
5252 if (len < sizeof(u32))
5253 return -EINVAL;
5254
5255 len = sizeof(u32);
5256
5257 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5258 val++;
5259 }
5260
5261 if (put_user(len, optlen))
5262 return -EFAULT;
5263 if (copy_to_user(optval, &val, len))
5264 return -EFAULT;
5265
5266 return 0;
5267}
5268
Linus Torvalds1da177e2005-04-16 15:20:36 -07005269SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5270 char __user *optval, int __user *optlen)
5271{
5272 int retval = 0;
5273 int len;
5274
Frank Filz3f7a87d2005-06-20 13:14:57 -07005275 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5276 sk, optname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005277
5278 /* I can hardly begin to describe how wrong this is. This is
5279 * so broken as to be worse than useless. The API draft
5280 * REALLY is NOT helpful here... I am not convinced that the
5281 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5282 * are at all well-founded.
5283 */
5284 if (level != SOL_SCTP) {
5285 struct sctp_af *af = sctp_sk(sk)->pf->af;
5286
5287 retval = af->getsockopt(sk, level, optname, optval, optlen);
5288 return retval;
5289 }
5290
5291 if (get_user(len, optlen))
5292 return -EFAULT;
5293
5294 sctp_lock_sock(sk);
5295
5296 switch (optname) {
5297 case SCTP_STATUS:
5298 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5299 break;
5300 case SCTP_DISABLE_FRAGMENTS:
5301 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5302 optlen);
5303 break;
5304 case SCTP_EVENTS:
5305 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5306 break;
5307 case SCTP_AUTOCLOSE:
5308 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5309 break;
5310 case SCTP_SOCKOPT_PEELOFF:
5311 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5312 break;
5313 case SCTP_PEER_ADDR_PARAMS:
5314 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5315 optlen);
5316 break;
Wei Yongjund364d922008-05-09 15:13:26 -07005317 case SCTP_DELAYED_ACK:
5318 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
Frank Filz77086102005-12-22 11:37:30 -08005319 optlen);
5320 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005321 case SCTP_INITMSG:
5322 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5323 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005324 case SCTP_GET_PEER_ADDRS:
5325 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5326 optlen);
5327 break;
5328 case SCTP_GET_LOCAL_ADDRS:
5329 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5330 optlen);
5331 break;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04005332 case SCTP_SOCKOPT_CONNECTX3:
5333 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5334 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005335 case SCTP_DEFAULT_SEND_PARAM:
5336 retval = sctp_getsockopt_default_send_param(sk, len,
5337 optval, optlen);
5338 break;
5339 case SCTP_PRIMARY_ADDR:
5340 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5341 break;
5342 case SCTP_NODELAY:
5343 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5344 break;
5345 case SCTP_RTOINFO:
5346 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5347 break;
5348 case SCTP_ASSOCINFO:
5349 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5350 break;
5351 case SCTP_I_WANT_MAPPED_V4_ADDR:
5352 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5353 break;
5354 case SCTP_MAXSEG:
5355 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5356 break;
5357 case SCTP_GET_PEER_ADDR_INFO:
5358 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5359 optlen);
5360 break;
Ivan Skytte Jorgensen0f3fffd2006-12-20 16:07:04 -08005361 case SCTP_ADAPTATION_LAYER:
5362 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07005363 optlen);
5364 break;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -08005365 case SCTP_CONTEXT:
5366 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5367 break;
Vlad Yasevichb6e13312007-04-20 12:23:15 -07005368 case SCTP_FRAGMENT_INTERLEAVE:
5369 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5370 optlen);
5371 break;
Vlad Yasevichd49d91d2007-03-23 11:32:00 -07005372 case SCTP_PARTIAL_DELIVERY_POINT:
5373 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5374 optlen);
5375 break;
Vlad Yasevich70331572007-03-23 11:34:36 -07005376 case SCTP_MAX_BURST:
5377 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5378 break;
Vlad Yasevich65b07e52007-09-16 19:34:00 -07005379 case SCTP_AUTH_KEY:
5380 case SCTP_AUTH_CHUNK:
5381 case SCTP_AUTH_DELETE_KEY:
5382 retval = -EOPNOTSUPP;
5383 break;
5384 case SCTP_HMAC_IDENT:
5385 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5386 break;
5387 case SCTP_AUTH_ACTIVE_KEY:
5388 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5389 break;
5390 case SCTP_PEER_AUTH_CHUNKS:
5391 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5392 optlen);
5393 break;
5394 case SCTP_LOCAL_AUTH_CHUNKS:
5395 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5396 optlen);
5397 break;
Wei Yongjunaea3c5c2008-12-25 16:57:24 -08005398 case SCTP_GET_ASSOC_NUMBER:
5399 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5400 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005401 default:
5402 retval = -ENOPROTOOPT;
5403 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005404 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005405
5406 sctp_release_sock(sk);
5407 return retval;
5408}
5409
5410static void sctp_hash(struct sock *sk)
5411{
5412 /* STUB */
5413}
5414
5415static void sctp_unhash(struct sock *sk)
5416{
5417 /* STUB */
5418}
5419
5420/* Check if port is acceptable. Possibly find first available port.
5421 *
5422 * The port hash table (contained in the 'global' SCTP protocol storage
5423 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5424 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5425 * list (the list number is the port number hashed out, so as you
5426 * would expect from a hash function, all the ports in a given list have
5427 * such a number that hashes out to the same list number; you were
5428 * expecting that, right?); so each list has a set of ports, with a
5429 * link to the socket (struct sock) that uses it, the port number and
5430 * a fastreuse flag (FIXME: NPI ipg).
5431 */
5432static struct sctp_bind_bucket *sctp_bucket_create(
5433 struct sctp_bind_hashbucket *head, unsigned short snum);
5434
5435static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5436{
5437 struct sctp_bind_hashbucket *head; /* hash list */
5438 struct sctp_bind_bucket *pp; /* hash list port iterator */
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005439 struct hlist_node *node;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005440 unsigned short snum;
5441 int ret;
5442
Al Viro04afd8b2006-11-20 17:02:01 -08005443 snum = ntohs(addr->v4.sin_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005444
5445 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5446 sctp_local_bh_disable();
5447
5448 if (snum == 0) {
Stephen Hemminger06393002007-10-10 17:30:18 -07005449 /* Search for an available port. */
Stephen Hemminger227b60f2007-10-10 17:30:46 -07005450 int low, high, remaining, index;
5451 unsigned int rover;
5452
5453 inet_get_local_port_range(&low, &high);
5454 remaining = (high - low) + 1;
5455 rover = net_random() % remaining + low;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005456
Linus Torvalds1da177e2005-04-16 15:20:36 -07005457 do {
5458 rover++;
5459 if ((rover < low) || (rover > high))
5460 rover = low;
Amerigo Wange3826f12010-05-05 00:27:06 +00005461 if (inet_is_reserved_local_port(rover))
5462 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005463 index = sctp_phashfn(rover);
5464 head = &sctp_port_hashtable[index];
5465 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005466 sctp_for_each_hentry(pp, node, &head->chain)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005467 if (pp->port == rover)
5468 goto next;
5469 break;
5470 next:
5471 sctp_spin_unlock(&head->lock);
5472 } while (--remaining > 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005473
5474 /* Exhausted local port range during search? */
5475 ret = 1;
5476 if (remaining <= 0)
5477 goto fail;
5478
5479 /* OK, here is the one we will use. HEAD (the port
5480 * hash table list entry) is non-NULL and we hold it's
5481 * mutex.
5482 */
5483 snum = rover;
5484 } else {
5485 /* We are given an specific port number; we verify
5486 * that it is not being used. If it is used, we will
5487 * exahust the search in the hash list corresponding
5488 * to the port number (snum) - we detect that with the
5489 * port iterator, pp being NULL.
5490 */
5491 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5492 sctp_spin_lock(&head->lock);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005493 sctp_for_each_hentry(pp, node, &head->chain) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005494 if (pp->port == snum)
5495 goto pp_found;
5496 }
5497 }
5498 pp = NULL;
5499 goto pp_not_found;
5500pp_found:
5501 if (!hlist_empty(&pp->owner)) {
5502 /* We had a port hash table hit - there is an
5503 * available port (pp != NULL) and it is being
5504 * used by other socket (pp->owner not empty); that other
5505 * socket is going to be sk2.
5506 */
5507 int reuse = sk->sk_reuse;
5508 struct sock *sk2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005509
5510 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005511 if (pp->fastreuse && sk->sk_reuse &&
5512 sk->sk_state != SCTP_SS_LISTENING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005513 goto success;
5514
5515 /* Run through the list of sockets bound to the port
5516 * (pp->port) [via the pointers bind_next and
5517 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5518 * we get the endpoint they describe and run through
5519 * the endpoint's list of IP (v4 or v6) addresses,
5520 * comparing each of the addresses with the address of
5521 * the socket sk. If we find a match, then that means
5522 * that this port/socket (sk) combination are already
5523 * in an endpoint.
5524 */
5525 sk_for_each_bound(sk2, node, &pp->owner) {
5526 struct sctp_endpoint *ep2;
5527 ep2 = sctp_sk(sk2)->ep;
5528
Vlad Yasevich4e540642008-07-18 23:06:32 -07005529 if (sk == sk2 ||
5530 (reuse && sk2->sk_reuse &&
5531 sk2->sk_state != SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005532 continue;
5533
Vlad Yasevich7dab83d2008-07-18 23:05:40 -07005534 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5535 sctp_sk(sk2), sctp_sk(sk))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07005536 ret = (long)sk2;
5537 goto fail_unlock;
5538 }
5539 }
5540 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5541 }
5542pp_not_found:
5543 /* If there was a hash table miss, create a new port. */
5544 ret = 1;
5545 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5546 goto fail_unlock;
5547
5548 /* In either case (hit or miss), make sure fastreuse is 1 only
5549 * if sk->sk_reuse is too (that is, if the caller requested
5550 * SO_REUSEADDR on this socket -sk-).
5551 */
Vlad Yasevichce5325c2007-05-04 13:34:49 -07005552 if (hlist_empty(&pp->owner)) {
5553 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5554 pp->fastreuse = 1;
5555 else
5556 pp->fastreuse = 0;
5557 } else if (pp->fastreuse &&
5558 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005559 pp->fastreuse = 0;
5560
5561 /* We are set, so fill up all the data in the hash table
5562 * entry, tie the socket list information with the rest of the
5563 * sockets FIXME: Blurry, NPI (ipg).
5564 */
5565success:
Linus Torvalds1da177e2005-04-16 15:20:36 -07005566 if (!sctp_sk(sk)->bind_hash) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005567 inet_sk(sk)->inet_num = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005568 sk_add_bind_node(sk, &pp->owner);
5569 sctp_sk(sk)->bind_hash = pp;
5570 }
5571 ret = 0;
5572
5573fail_unlock:
5574 sctp_spin_unlock(&head->lock);
5575
5576fail:
5577 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07005578 return ret;
5579}
5580
5581/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5582 * port is requested.
5583 */
5584static int sctp_get_port(struct sock *sk, unsigned short snum)
5585{
5586 long ret;
5587 union sctp_addr addr;
5588 struct sctp_af *af = sctp_sk(sk)->pf->af;
5589
5590 /* Set up a dummy address struct from the sk. */
5591 af->from_sk(&addr, sk);
5592 addr.v4.sin_port = htons(snum);
5593
5594 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5595 ret = sctp_get_port_local(sk, &addr);
5596
5597 return (ret ? 1 : 0);
5598}
5599
5600/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07005601 * Move a socket to LISTENING state.
5602 */
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005603SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07005604{
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005605 struct sctp_sock *sp = sctp_sk(sk);
5606 struct sctp_endpoint *ep = sp->ep;
Herbert Xu1b489e12006-08-20 15:07:14 +10005607 struct crypto_hash *tfm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005608
5609 /* Allocate HMAC for generating cookie. */
Vlad Yasevich23b29ed2008-07-18 23:06:07 -07005610 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
Herbert Xu1b489e12006-08-20 15:07:14 +10005611 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
Vlad Yasevich8dc49842007-05-09 13:50:20 -07005612 if (IS_ERR(tfm)) {
5613 if (net_ratelimit()) {
Joe Perches145ce502010-08-24 13:21:08 +00005614 pr_info("failed to load transform for %s: %ld\n",
Vlad Yasevich8dc49842007-05-09 13:50:20 -07005615 sctp_hmac_alg, PTR_ERR(tfm));
5616 }
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005617 return -ENOSYS;
5618 }
5619 sctp_sk(sk)->hmac = tfm;
5620 }
5621
5622 /*
5623 * If a bind() or sctp_bindx() is not called prior to a listen()
5624 * call that allows new associations to be accepted, the system
5625 * picks an ephemeral port and will choose an address set equivalent
5626 * to binding with a wildcard address.
5627 *
5628 * This is not currently spelled out in the SCTP sockets
5629 * extensions draft, but follows the practice as seen in TCP
5630 * sockets.
5631 *
5632 */
5633 sk->sk_state = SCTP_SS_LISTENING;
5634 if (!ep->base.bind_addr.port) {
5635 if (sctp_autobind(sk))
5636 return -EAGAIN;
5637 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005638 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005639 sk->sk_state = SCTP_SS_CLOSED;
5640 return -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005641 }
5642 }
5643
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005644 sk->sk_max_ack_backlog = backlog;
5645 sctp_hash_endpoint(ep);
5646 return 0;
5647}
5648
5649/*
5650 * 4.1.3 / 5.1.3 listen()
5651 *
5652 * By default, new associations are not accepted for UDP style sockets.
5653 * An application uses listen() to mark a socket as being able to
5654 * accept new associations.
5655 *
5656 * On TCP style sockets, applications use listen() to ready the SCTP
5657 * endpoint for accepting inbound associations.
5658 *
5659 * On both types of endpoints a backlog of '0' disables listening.
5660 *
5661 * Move a socket to LISTENING state.
5662 */
5663int sctp_inet_listen(struct socket *sock, int backlog)
5664{
5665 struct sock *sk = sock->sk;
5666 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5667 int err = -EINVAL;
5668
5669 if (unlikely(backlog < 0))
5670 return err;
5671
5672 sctp_lock_sock(sk);
5673
5674 /* Peeled-off sockets are not allowed to listen(). */
5675 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5676 goto out;
5677
5678 if (sock->state != SS_UNCONNECTED)
5679 goto out;
5680
5681 /* If backlog is zero, disable listening. */
5682 if (!backlog) {
5683 if (sctp_sstate(sk, CLOSED))
5684 goto out;
5685
5686 err = 0;
5687 sctp_unhash_endpoint(ep);
5688 sk->sk_state = SCTP_SS_CLOSED;
5689 if (sk->sk_reuse)
5690 sctp_sk(sk)->bind_hash->fastreuse = 1;
5691 goto out;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005692 }
5693
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005694 /* If we are already listening, just update the backlog */
5695 if (sctp_sstate(sk, LISTENING))
5696 sk->sk_max_ack_backlog = backlog;
5697 else {
5698 err = sctp_listen_start(sk, backlog);
5699 if (err)
5700 goto out;
5701 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005702
Vlad Yasevich5e8f3f72009-03-12 09:49:17 +00005703 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005704out:
5705 sctp_release_sock(sk);
5706 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005707}
5708
5709/*
5710 * This function is done by modeling the current datagram_poll() and the
5711 * tcp_poll(). Note that, based on these implementations, we don't
5712 * lock the socket in this function, even though it seems that,
5713 * ideally, locking or some other mechanisms can be used to ensure
Neil Horman9bffc4a2005-12-19 14:24:40 -08005714 * the integrity of the counters (sndbuf and wmem_alloc) used
Linus Torvalds1da177e2005-04-16 15:20:36 -07005715 * in this place. We assume that we don't need locks either until proven
5716 * otherwise.
5717 *
5718 * Another thing to note is that we include the Async I/O support
5719 * here, again, by modeling the current TCP/UDP code. We don't have
5720 * a good way to test with it yet.
5721 */
5722unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5723{
5724 struct sock *sk = sock->sk;
5725 struct sctp_sock *sp = sctp_sk(sk);
5726 unsigned int mask;
5727
Eric Dumazetaa395142010-04-20 13:03:51 +00005728 poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005729
5730 /* A TCP-style listening socket becomes readable when the accept queue
5731 * is not empty.
5732 */
5733 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5734 return (!list_empty(&sp->ep->asocs)) ?
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09005735 (POLLIN | POLLRDNORM) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005736
5737 mask = 0;
5738
5739 /* Is there any exceptional events? */
5740 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5741 mask |= POLLERR;
Davide Libenzif348d702006-03-25 03:07:39 -08005742 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00005743 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005744 if (sk->sk_shutdown == SHUTDOWN_MASK)
5745 mask |= POLLHUP;
5746
5747 /* Is it readable? Reconsider this code with TCP-style support. */
Eric Dumazetdb409802010-09-06 11:13:50 +00005748 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005749 mask |= POLLIN | POLLRDNORM;
5750
5751 /* The association is either gone or not ready. */
5752 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5753 return mask;
5754
5755 /* Is it writable? */
5756 if (sctp_writeable(sk)) {
5757 mask |= POLLOUT | POLLWRNORM;
5758 } else {
5759 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5760 /*
5761 * Since the socket is not locked, the buffer
5762 * might be made available after the writeable check and
5763 * before the bit is set. This could cause a lost I/O
5764 * signal. tcp_poll() has a race breaker for this race
5765 * condition. Based on their implementation, we put
5766 * in the following code to cover it as well.
5767 */
5768 if (sctp_writeable(sk))
5769 mask |= POLLOUT | POLLWRNORM;
5770 }
5771 return mask;
5772}
5773
5774/********************************************************************
5775 * 2nd Level Abstractions
5776 ********************************************************************/
5777
5778static struct sctp_bind_bucket *sctp_bucket_create(
5779 struct sctp_bind_hashbucket *head, unsigned short snum)
5780{
5781 struct sctp_bind_bucket *pp;
5782
Christoph Lameter54e6ecb2006-12-06 20:33:16 -08005783 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005784 if (pp) {
Li Zefan935a7f62008-04-10 01:58:06 -07005785 SCTP_DBG_OBJCNT_INC(bind_bucket);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005786 pp->port = snum;
5787 pp->fastreuse = 0;
5788 INIT_HLIST_HEAD(&pp->owner);
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005789 hlist_add_head(&pp->node, &head->chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005790 }
5791 return pp;
5792}
5793
5794/* Caller must hold hashbucket lock for this tb with local BH disabled */
5795static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5796{
Sridhar Samudrala37fa6872006-07-21 14:45:47 -07005797 if (pp && hlist_empty(&pp->owner)) {
Vlad Yasevichd970dbf2007-11-09 11:43:40 -05005798 __hlist_del(&pp->node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005799 kmem_cache_free(sctp_bucket_cachep, pp);
5800 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5801 }
5802}
5803
5804/* Release this socket's reference to a local port. */
5805static inline void __sctp_put_port(struct sock *sk)
5806{
5807 struct sctp_bind_hashbucket *head =
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005808 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
Linus Torvalds1da177e2005-04-16 15:20:36 -07005809 struct sctp_bind_bucket *pp;
5810
5811 sctp_spin_lock(&head->lock);
5812 pp = sctp_sk(sk)->bind_hash;
5813 __sk_del_bind_node(sk);
5814 sctp_sk(sk)->bind_hash = NULL;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005815 inet_sk(sk)->inet_num = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005816 sctp_bucket_destroy(pp);
5817 sctp_spin_unlock(&head->lock);
5818}
5819
5820void sctp_put_port(struct sock *sk)
5821{
5822 sctp_local_bh_disable();
5823 __sctp_put_port(sk);
5824 sctp_local_bh_enable();
5825}
5826
5827/*
5828 * The system picks an ephemeral port and choose an address set equivalent
5829 * to binding with a wildcard address.
5830 * One of those addresses will be the primary address for the association.
5831 * This automatically enables the multihoming capability of SCTP.
5832 */
5833static int sctp_autobind(struct sock *sk)
5834{
5835 union sctp_addr autoaddr;
5836 struct sctp_af *af;
Al Viro6fbfa9f2006-11-20 17:24:53 -08005837 __be16 port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005838
5839 /* Initialize a local sockaddr structure to INADDR_ANY. */
5840 af = sctp_sk(sk)->pf->af;
5841
Eric Dumazetc720c7e2009-10-15 06:30:45 +00005842 port = htons(inet_sk(sk)->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005843 af->inaddr_any(&autoaddr, port);
5844
5845 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5846}
5847
5848/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5849 *
5850 * From RFC 2292
5851 * 4.2 The cmsghdr Structure *
5852 *
5853 * When ancillary data is sent or received, any number of ancillary data
5854 * objects can be specified by the msg_control and msg_controllen members of
5855 * the msghdr structure, because each object is preceded by
5856 * a cmsghdr structure defining the object's length (the cmsg_len member).
5857 * Historically Berkeley-derived implementations have passed only one object
5858 * at a time, but this API allows multiple objects to be
5859 * passed in a single call to sendmsg() or recvmsg(). The following example
5860 * shows two ancillary data objects in a control buffer.
5861 *
5862 * |<--------------------------- msg_controllen -------------------------->|
5863 * | |
5864 *
5865 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5866 *
5867 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5868 * | | |
5869 *
5870 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5871 *
5872 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5873 * | | | | |
5874 *
5875 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5876 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5877 *
5878 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5879 *
5880 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5881 * ^
5882 * |
5883 *
5884 * msg_control
5885 * points here
5886 */
5887SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5888 sctp_cmsgs_t *cmsgs)
5889{
5890 struct cmsghdr *cmsg;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07005891 struct msghdr *my_msg = (struct msghdr *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07005892
5893 for (cmsg = CMSG_FIRSTHDR(msg);
5894 cmsg != NULL;
Vlad Yasevichab38fb02008-04-12 18:40:06 -07005895 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5896 if (!CMSG_OK(my_msg, cmsg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005897 return -EINVAL;
5898
5899 /* Should we parse this header or ignore? */
5900 if (cmsg->cmsg_level != IPPROTO_SCTP)
5901 continue;
5902
5903 /* Strictly check lengths following example in SCM code. */
5904 switch (cmsg->cmsg_type) {
5905 case SCTP_INIT:
5906 /* SCTP Socket API Extension
5907 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5908 *
5909 * This cmsghdr structure provides information for
5910 * initializing new SCTP associations with sendmsg().
5911 * The SCTP_INITMSG socket option uses this same data
5912 * structure. This structure is not used for
5913 * recvmsg().
5914 *
5915 * cmsg_level cmsg_type cmsg_data[]
5916 * ------------ ------------ ----------------------
5917 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5918 */
5919 if (cmsg->cmsg_len !=
5920 CMSG_LEN(sizeof(struct sctp_initmsg)))
5921 return -EINVAL;
5922 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5923 break;
5924
5925 case SCTP_SNDRCV:
5926 /* SCTP Socket API Extension
5927 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5928 *
5929 * This cmsghdr structure specifies SCTP options for
5930 * sendmsg() and describes SCTP header information
5931 * about a received message through recvmsg().
5932 *
5933 * cmsg_level cmsg_type cmsg_data[]
5934 * ------------ ------------ ----------------------
5935 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5936 */
5937 if (cmsg->cmsg_len !=
5938 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5939 return -EINVAL;
5940
5941 cmsgs->info =
5942 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5943
5944 /* Minimally, validate the sinfo_flags. */
5945 if (cmsgs->info->sinfo_flags &
Ivan Skytte Jorgenseneaa5c542005-10-28 15:10:00 -07005946 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5947 SCTP_ABORT | SCTP_EOF))
Linus Torvalds1da177e2005-04-16 15:20:36 -07005948 return -EINVAL;
5949 break;
5950
5951 default:
5952 return -EINVAL;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -07005953 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07005954 }
5955 return 0;
5956}
5957
5958/*
5959 * Wait for a packet..
5960 * Note: This function is the same function as in core/datagram.c
5961 * with a few modifications to make lksctp work.
5962 */
5963static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5964{
5965 int error;
5966 DEFINE_WAIT(wait);
5967
Eric Dumazetaa395142010-04-20 13:03:51 +00005968 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07005969
5970 /* Socket errors? */
5971 error = sock_error(sk);
5972 if (error)
5973 goto out;
5974
5975 if (!skb_queue_empty(&sk->sk_receive_queue))
5976 goto ready;
5977
5978 /* Socket shut down? */
5979 if (sk->sk_shutdown & RCV_SHUTDOWN)
5980 goto out;
5981
5982 /* Sequenced packets can come disconnected. If so we report the
5983 * problem.
5984 */
5985 error = -ENOTCONN;
5986
5987 /* Is there a good reason to think that we may receive some data? */
5988 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5989 goto out;
5990
5991 /* Handle signals. */
5992 if (signal_pending(current))
5993 goto interrupted;
5994
5995 /* Let another process have a go. Since we are going to sleep
5996 * anyway. Note: This may cause odd behaviors if the message
5997 * does not fit in the user's buffer, but this seems to be the
5998 * only way to honor MSG_DONTWAIT realistically.
5999 */
6000 sctp_release_sock(sk);
6001 *timeo_p = schedule_timeout(*timeo_p);
6002 sctp_lock_sock(sk);
6003
6004ready:
Eric Dumazetaa395142010-04-20 13:03:51 +00006005 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006006 return 0;
6007
6008interrupted:
6009 error = sock_intr_errno(*timeo_p);
6010
6011out:
Eric Dumazetaa395142010-04-20 13:03:51 +00006012 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006013 *err = error;
6014 return error;
6015}
6016
6017/* Receive a datagram.
6018 * Note: This is pretty much the same routine as in core/datagram.c
6019 * with a few changes to make lksctp work.
6020 */
6021static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6022 int noblock, int *err)
6023{
6024 int error;
6025 struct sk_buff *skb;
6026 long timeo;
6027
Linus Torvalds1da177e2005-04-16 15:20:36 -07006028 timeo = sock_rcvtimeo(sk, noblock);
6029
6030 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6031 timeo, MAX_SCHEDULE_TIMEOUT);
6032
6033 do {
6034 /* Again only user level code calls this function,
6035 * so nothing interrupt level
6036 * will suddenly eat the receive_queue.
6037 *
6038 * Look at current nfs client by the way...
6039 * However, this function was corrent in any case. 8)
6040 */
6041 if (flags & MSG_PEEK) {
Herbert Xu1e061ab2005-06-18 22:56:42 -07006042 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006043 skb = skb_peek(&sk->sk_receive_queue);
6044 if (skb)
6045 atomic_inc(&skb->users);
Herbert Xu1e061ab2005-06-18 22:56:42 -07006046 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006047 } else {
6048 skb = skb_dequeue(&sk->sk_receive_queue);
6049 }
6050
6051 if (skb)
6052 return skb;
6053
Neil Horman6736dc32005-12-02 20:30:06 -08006054 /* Caller is allowed not to check sk->sk_err before calling. */
6055 error = sock_error(sk);
6056 if (error)
6057 goto no_packet;
6058
Linus Torvalds1da177e2005-04-16 15:20:36 -07006059 if (sk->sk_shutdown & RCV_SHUTDOWN)
6060 break;
6061
6062 /* User doesn't want to wait. */
6063 error = -EAGAIN;
6064 if (!timeo)
6065 goto no_packet;
6066 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6067
6068 return NULL;
6069
6070no_packet:
6071 *err = error;
6072 return NULL;
6073}
6074
6075/* If sndbuf has changed, wake up per association sndbuf waiters. */
6076static void __sctp_write_space(struct sctp_association *asoc)
6077{
6078 struct sock *sk = asoc->base.sk;
6079 struct socket *sock = sk->sk_socket;
6080
6081 if ((sctp_wspace(asoc) > 0) && sock) {
6082 if (waitqueue_active(&asoc->wait))
6083 wake_up_interruptible(&asoc->wait);
6084
6085 if (sctp_writeable(sk)) {
Eric Dumazetaa395142010-04-20 13:03:51 +00006086 if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
6087 wake_up_interruptible(sk_sleep(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07006088
6089 /* Note that we try to include the Async I/O support
6090 * here by modeling from the current TCP/UDP code.
6091 * We have not tested with it yet.
6092 */
Eric Dumazet43815482010-04-29 11:01:49 +00006093 if (sock->wq->fasync_list &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07006094 !(sk->sk_shutdown & SEND_SHUTDOWN))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08006095 sock_wake_async(sock,
6096 SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006097 }
6098 }
6099}
6100
6101/* Do accounting for the sndbuf space.
6102 * Decrement the used sndbuf space of the corresponding association by the
6103 * data size which was just transmitted(freed).
6104 */
6105static void sctp_wfree(struct sk_buff *skb)
6106{
6107 struct sctp_association *asoc;
6108 struct sctp_chunk *chunk;
6109 struct sock *sk;
6110
6111 /* Get the saved chunk pointer. */
6112 chunk = *((struct sctp_chunk **)(skb->cb));
6113 asoc = chunk->asoc;
6114 sk = asoc->base.sk;
Neil Horman4eb701d2005-04-28 12:02:04 -07006115 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6116 sizeof(struct sk_buff) +
6117 sizeof(struct sctp_chunk);
6118
Neil Horman4eb701d2005-04-28 12:02:04 -07006119 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6120
Neil Horman4d93df02007-08-15 16:07:44 -07006121 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006122 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
Neil Horman4d93df02007-08-15 16:07:44 -07006123 */
6124 sk->sk_wmem_queued -= skb->truesize;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006125 sk_mem_uncharge(sk, skb->truesize);
Neil Horman4d93df02007-08-15 16:07:44 -07006126
Neil Horman4eb701d2005-04-28 12:02:04 -07006127 sock_wfree(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006128 __sctp_write_space(asoc);
6129
6130 sctp_association_put(asoc);
6131}
6132
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006133/* Do accounting for the receive space on the socket.
6134 * Accounting for the association is done in ulpevent.c
6135 * We set this as a destructor for the cloned data skbs so that
6136 * accounting is done at the correct time.
6137 */
6138void sctp_sock_rfree(struct sk_buff *skb)
6139{
6140 struct sock *sk = skb->sk;
6141 struct sctp_ulpevent *event = sctp_skb2event(skb);
6142
6143 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
Neil Horman4d93df02007-08-15 16:07:44 -07006144
6145 /*
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006146 * Mimic the behavior of sock_rfree
Neil Horman4d93df02007-08-15 16:07:44 -07006147 */
Hideo Aoki3ab224b2007-12-31 00:11:19 -08006148 sk_mem_uncharge(sk, event->rmem_len);
Vlad Yasevich331c4ee2006-10-09 21:34:04 -07006149}
6150
6151
Linus Torvalds1da177e2005-04-16 15:20:36 -07006152/* Helper function to wait for space in the sndbuf. */
6153static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6154 size_t msg_len)
6155{
6156 struct sock *sk = asoc->base.sk;
6157 int err = 0;
6158 long current_timeo = *timeo_p;
6159 DEFINE_WAIT(wait);
6160
6161 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09006162 asoc, (long)(*timeo_p), msg_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006163
6164 /* Increment the association's refcnt. */
6165 sctp_association_hold(asoc);
6166
6167 /* Wait on the association specific sndbuf space. */
6168 for (;;) {
6169 prepare_to_wait_exclusive(&asoc->wait, &wait,
6170 TASK_INTERRUPTIBLE);
6171 if (!*timeo_p)
6172 goto do_nonblock;
6173 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6174 asoc->base.dead)
6175 goto do_error;
6176 if (signal_pending(current))
6177 goto do_interrupted;
6178 if (msg_len <= sctp_wspace(asoc))
6179 break;
6180
6181 /* Let another process have a go. Since we are going
6182 * to sleep anyway.
6183 */
6184 sctp_release_sock(sk);
6185 current_timeo = schedule_timeout(current_timeo);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006186 BUG_ON(sk != asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006187 sctp_lock_sock(sk);
6188
6189 *timeo_p = current_timeo;
6190 }
6191
6192out:
6193 finish_wait(&asoc->wait, &wait);
6194
6195 /* Release the association's refcnt. */
6196 sctp_association_put(asoc);
6197
6198 return err;
6199
6200do_error:
6201 err = -EPIPE;
6202 goto out;
6203
6204do_interrupted:
6205 err = sock_intr_errno(*timeo_p);
6206 goto out;
6207
6208do_nonblock:
6209 err = -EAGAIN;
6210 goto out;
6211}
6212
Wei Yongjun561b1732010-04-28 08:47:18 +00006213void sctp_data_ready(struct sock *sk, int len)
6214{
David S. Miller7ef52732010-05-02 21:43:40 -07006215 struct socket_wq *wq;
6216
6217 rcu_read_lock();
6218 wq = rcu_dereference(sk->sk_wq);
6219 if (wq_has_sleeper(wq))
6220 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
Wei Yongjun561b1732010-04-28 08:47:18 +00006221 POLLRDNORM | POLLRDBAND);
6222 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
David S. Miller7ef52732010-05-02 21:43:40 -07006223 rcu_read_unlock();
Wei Yongjun561b1732010-04-28 08:47:18 +00006224}
6225
Linus Torvalds1da177e2005-04-16 15:20:36 -07006226/* If socket sndbuf has changed, wake up all per association waiters. */
6227void sctp_write_space(struct sock *sk)
6228{
6229 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006230
6231 /* Wake up the tasks in each wait queue. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07006232 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006233 __sctp_write_space(asoc);
6234 }
6235}
6236
6237/* Is there any sndbuf space available on the socket?
6238 *
Neil Horman9bffc4a2005-12-19 14:24:40 -08006239 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07006240 * associations on the same socket. For a UDP-style socket with
6241 * multiple associations, it is possible for it to be "unwriteable"
6242 * prematurely. I assume that this is acceptable because
6243 * a premature "unwriteable" is better than an accidental "writeable" which
6244 * would cause an unwanted block under certain circumstances. For the 1-1
6245 * UDP-style sockets or TCP-style sockets, this code should work.
6246 * - Daisy
6247 */
6248static int sctp_writeable(struct sock *sk)
6249{
6250 int amt = 0;
6251
Eric Dumazet31e6d362009-06-17 19:05:41 -07006252 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006253 if (amt < 0)
6254 amt = 0;
6255 return amt;
6256}
6257
6258/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6259 * returns immediately with EINPROGRESS.
6260 */
6261static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6262{
6263 struct sock *sk = asoc->base.sk;
6264 int err = 0;
6265 long current_timeo = *timeo_p;
6266 DEFINE_WAIT(wait);
6267
Harvey Harrison0dc47872008-03-05 20:47:47 -08006268 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006269 (long)(*timeo_p));
6270
6271 /* Increment the association's refcnt. */
6272 sctp_association_hold(asoc);
6273
6274 for (;;) {
6275 prepare_to_wait_exclusive(&asoc->wait, &wait,
6276 TASK_INTERRUPTIBLE);
6277 if (!*timeo_p)
6278 goto do_nonblock;
6279 if (sk->sk_shutdown & RCV_SHUTDOWN)
6280 break;
6281 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6282 asoc->base.dead)
6283 goto do_error;
6284 if (signal_pending(current))
6285 goto do_interrupted;
6286
6287 if (sctp_state(asoc, ESTABLISHED))
6288 break;
6289
6290 /* Let another process have a go. Since we are going
6291 * to sleep anyway.
6292 */
6293 sctp_release_sock(sk);
6294 current_timeo = schedule_timeout(current_timeo);
6295 sctp_lock_sock(sk);
6296
6297 *timeo_p = current_timeo;
6298 }
6299
6300out:
6301 finish_wait(&asoc->wait, &wait);
6302
6303 /* Release the association's refcnt. */
6304 sctp_association_put(asoc);
6305
6306 return err;
6307
6308do_error:
Vlad Yasevich81845c22006-01-30 15:59:54 -08006309 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006310 err = -ETIMEDOUT;
6311 else
6312 err = -ECONNREFUSED;
6313 goto out;
6314
6315do_interrupted:
6316 err = sock_intr_errno(*timeo_p);
6317 goto out;
6318
6319do_nonblock:
6320 err = -EINPROGRESS;
6321 goto out;
6322}
6323
6324static int sctp_wait_for_accept(struct sock *sk, long timeo)
6325{
6326 struct sctp_endpoint *ep;
6327 int err = 0;
6328 DEFINE_WAIT(wait);
6329
6330 ep = sctp_sk(sk)->ep;
6331
6332
6333 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00006334 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006335 TASK_INTERRUPTIBLE);
6336
6337 if (list_empty(&ep->asocs)) {
6338 sctp_release_sock(sk);
6339 timeo = schedule_timeout(timeo);
6340 sctp_lock_sock(sk);
6341 }
6342
6343 err = -EINVAL;
6344 if (!sctp_sstate(sk, LISTENING))
6345 break;
6346
6347 err = 0;
6348 if (!list_empty(&ep->asocs))
6349 break;
6350
6351 err = sock_intr_errno(timeo);
6352 if (signal_pending(current))
6353 break;
6354
6355 err = -EAGAIN;
6356 if (!timeo)
6357 break;
6358 }
6359
Eric Dumazetaa395142010-04-20 13:03:51 +00006360 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006361
6362 return err;
6363}
6364
sebastian@breakpoint.cc04675212007-07-26 23:21:31 +02006365static void sctp_wait_for_close(struct sock *sk, long timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006366{
6367 DEFINE_WAIT(wait);
6368
6369 do {
Eric Dumazetaa395142010-04-20 13:03:51 +00006370 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006371 if (list_empty(&sctp_sk(sk)->ep->asocs))
6372 break;
6373 sctp_release_sock(sk);
6374 timeout = schedule_timeout(timeout);
6375 sctp_lock_sock(sk);
6376 } while (!signal_pending(current) && timeout);
6377
Eric Dumazetaa395142010-04-20 13:03:51 +00006378 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006379}
6380
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006381static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6382{
6383 struct sk_buff *frag;
6384
6385 if (!skb->data_len)
6386 goto done;
6387
6388 /* Don't forget the fragments. */
David S. Miller1b003be2009-06-09 00:22:35 -07006389 skb_walk_frags(skb, frag)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006390 sctp_skb_set_owner_r_frag(frag, sk);
6391
6392done:
6393 sctp_skb_set_owner_r(skb, sk);
6394}
6395
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006396void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6397 struct sctp_association *asoc)
6398{
6399 struct inet_sock *inet = inet_sk(sk);
Julia Lawall09cb47a2010-01-21 02:43:20 -08006400 struct inet_sock *newinet;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006401
6402 newsk->sk_type = sk->sk_type;
6403 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6404 newsk->sk_flags = sk->sk_flags;
6405 newsk->sk_no_check = sk->sk_no_check;
6406 newsk->sk_reuse = sk->sk_reuse;
6407
6408 newsk->sk_shutdown = sk->sk_shutdown;
6409 newsk->sk_destruct = inet_sock_destruct;
6410 newsk->sk_family = sk->sk_family;
6411 newsk->sk_protocol = IPPROTO_SCTP;
6412 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6413 newsk->sk_sndbuf = sk->sk_sndbuf;
6414 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6415 newsk->sk_lingertime = sk->sk_lingertime;
6416 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6417 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6418
6419 newinet = inet_sk(newsk);
6420
6421 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6422 * getsockname() and getpeername()
6423 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006424 newinet->inet_sport = inet->inet_sport;
6425 newinet->inet_saddr = inet->inet_saddr;
6426 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6427 newinet->inet_dport = htons(asoc->peer.port);
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006428 newinet->pmtudisc = inet->pmtudisc;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006429 newinet->inet_id = asoc->next_tsn ^ jiffies;
Vlad Yasevich914e1c82009-02-13 08:33:44 +00006430
6431 newinet->uc_ttl = inet->uc_ttl;
6432 newinet->mc_loop = 1;
6433 newinet->mc_ttl = 1;
6434 newinet->mc_index = 0;
6435 newinet->mc_list = NULL;
6436}
6437
Linus Torvalds1da177e2005-04-16 15:20:36 -07006438/* Populate the fields of the newsk from the oldsk and migrate the assoc
6439 * and its messages to the newsk.
6440 */
6441static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6442 struct sctp_association *assoc,
6443 sctp_socket_type_t type)
6444{
6445 struct sctp_sock *oldsp = sctp_sk(oldsk);
6446 struct sctp_sock *newsp = sctp_sk(newsk);
6447 struct sctp_bind_bucket *pp; /* hash list port iterator */
6448 struct sctp_endpoint *newep = newsp->ep;
6449 struct sk_buff *skb, *tmp;
6450 struct sctp_ulpevent *event;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006451 struct sctp_bind_hashbucket *head;
Linus Torvalds1da177e2005-04-16 15:20:36 -07006452
6453 /* Migrate socket buffer sizes and all the socket level options to the
6454 * new socket.
6455 */
6456 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6457 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6458 /* Brute force copy old sctp opt. */
6459 inet_sk_copy_descendant(newsk, oldsk);
6460
6461 /* Restore the ep value that was overwritten with the above structure
6462 * copy.
6463 */
6464 newsp->ep = newep;
6465 newsp->hmac = NULL;
6466
6467 /* Hook this new socket in to the bind_hash list. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006468 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006469 sctp_local_bh_disable();
6470 sctp_spin_lock(&head->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006471 pp = sctp_sk(oldsk)->bind_hash;
6472 sk_add_bind_node(newsk, &pp->owner);
6473 sctp_sk(newsk)->bind_hash = pp;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00006474 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
Vlad Yasevichf26f7c42007-12-06 22:50:27 -08006475 sctp_spin_unlock(&head->lock);
6476 sctp_local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07006477
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006478 /* Copy the bind_addr list from the original endpoint to the new
6479 * endpoint so that we can handle restarts properly
6480 */
Vlad Yasevich8e71a112007-12-06 22:50:54 -08006481 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6482 &oldsp->ep->base.bind_addr, GFP_KERNEL);
Vladislav Yasevich4243cac2005-06-13 15:10:49 -07006483
Linus Torvalds1da177e2005-04-16 15:20:36 -07006484 /* Move any messages in the old socket's receive queue that are for the
6485 * peeled off association to the new socket's receive queue.
6486 */
6487 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6488 event = sctp_skb2event(skb);
6489 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006490 __skb_unlink(skb, &oldsk->sk_receive_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006491 __skb_queue_tail(&newsk->sk_receive_queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006492 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006493 }
6494 }
6495
6496 /* Clean up any messages pending delivery due to partial
6497 * delivery. Three cases:
6498 * 1) No partial deliver; no work.
6499 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6500 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6501 */
6502 skb_queue_head_init(&newsp->pd_lobby);
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006503 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006504
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006505 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07006506 struct sk_buff_head *queue;
6507
6508 /* Decide which queue to move pd_lobby skbs to. */
6509 if (assoc->ulpq.pd_mode) {
6510 queue = &newsp->pd_lobby;
6511 } else
6512 queue = &newsk->sk_receive_queue;
6513
6514 /* Walk through the pd_lobby, looking for skbs that
6515 * need moved to the new socket.
6516 */
6517 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6518 event = sctp_skb2event(skb);
6519 if (event->asoc == assoc) {
David S. Miller8728b832005-08-09 19:25:21 -07006520 __skb_unlink(skb, &oldsp->pd_lobby);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006521 __skb_queue_tail(queue, skb);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006522 sctp_skb_set_owner_r_frag(skb, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006523 }
6524 }
6525
6526 /* Clear up any skbs waiting for the partial
6527 * delivery to finish.
6528 */
6529 if (assoc->ulpq.pd_mode)
Vlad Yasevichb6e13312007-04-20 12:23:15 -07006530 sctp_clear_pd(oldsk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006531
6532 }
6533
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006534 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006535 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006536
Wei Yongjun1bc4ee42009-07-05 19:45:48 +00006537 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006538 sctp_skb_set_owner_r_frag(skb, newsk);
Tsutomu Fujiiea2bc482007-04-17 12:49:53 -07006539
Linus Torvalds1da177e2005-04-16 15:20:36 -07006540 /* Set the type of socket to indicate that it is peeled off from the
6541 * original UDP-style socket or created with the accept() call on a
6542 * TCP-style socket..
6543 */
6544 newsp->type = type;
6545
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006546 /* Mark the new socket "in-use" by the user so that any packets
6547 * that may arrive on the association after we've moved it are
6548 * queued to the backlog. This prevents a potential race between
6549 * backlog processing on the old socket and new-packet processing
6550 * on the new socket.
Zach Brown5131a182007-06-22 15:14:46 -07006551 *
6552 * The caller has just allocated newsk so we can guarantee that other
6553 * paths won't try to lock it and then oldsk.
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006554 */
Zach Brown5131a182007-06-22 15:14:46 -07006555 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006556 sctp_assoc_migrate(assoc, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006557
6558 /* If the association on the newsk is already closed before accept()
6559 * is called, set RCV_SHUTDOWN flag.
6560 */
6561 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6562 newsk->sk_shutdown |= RCV_SHUTDOWN;
6563
6564 newsk->sk_state = SCTP_SS_ESTABLISHED;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -07006565 sctp_release_sock(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07006566}
6567
Neil Horman4d93df02007-08-15 16:07:44 -07006568
Linus Torvalds1da177e2005-04-16 15:20:36 -07006569/* This proto struct describes the ULP interface for SCTP. */
6570struct proto sctp_prot = {
6571 .name = "SCTP",
6572 .owner = THIS_MODULE,
6573 .close = sctp_close,
6574 .connect = sctp_connect,
6575 .disconnect = sctp_disconnect,
6576 .accept = sctp_accept,
6577 .ioctl = sctp_ioctl,
6578 .init = sctp_init_sock,
6579 .destroy = sctp_destroy_sock,
6580 .shutdown = sctp_shutdown,
6581 .setsockopt = sctp_setsockopt,
6582 .getsockopt = sctp_getsockopt,
6583 .sendmsg = sctp_sendmsg,
6584 .recvmsg = sctp_recvmsg,
6585 .bind = sctp_bind,
6586 .backlog_rcv = sctp_backlog_rcv,
6587 .hash = sctp_hash,
6588 .unhash = sctp_unhash,
6589 .get_port = sctp_get_port,
6590 .obj_size = sizeof(struct sctp_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006591 .sysctl_mem = sysctl_sctp_mem,
6592 .sysctl_rmem = sysctl_sctp_rmem,
6593 .sysctl_wmem = sysctl_sctp_wmem,
6594 .memory_pressure = &sctp_memory_pressure,
6595 .enter_memory_pressure = sctp_enter_memory_pressure,
6596 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006597 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006598};
6599
6600#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
Eric Dumazet8295b6d2007-11-05 23:40:28 -08006601
Linus Torvalds1da177e2005-04-16 15:20:36 -07006602struct proto sctpv6_prot = {
6603 .name = "SCTPv6",
6604 .owner = THIS_MODULE,
6605 .close = sctp_close,
6606 .connect = sctp_connect,
6607 .disconnect = sctp_disconnect,
6608 .accept = sctp_accept,
6609 .ioctl = sctp_ioctl,
6610 .init = sctp_init_sock,
6611 .destroy = sctp_destroy_sock,
6612 .shutdown = sctp_shutdown,
6613 .setsockopt = sctp_setsockopt,
6614 .getsockopt = sctp_getsockopt,
6615 .sendmsg = sctp_sendmsg,
6616 .recvmsg = sctp_recvmsg,
6617 .bind = sctp_bind,
6618 .backlog_rcv = sctp_backlog_rcv,
6619 .hash = sctp_hash,
6620 .unhash = sctp_unhash,
6621 .get_port = sctp_get_port,
6622 .obj_size = sizeof(struct sctp6_sock),
Neil Horman4d93df02007-08-15 16:07:44 -07006623 .sysctl_mem = sysctl_sctp_mem,
6624 .sysctl_rmem = sysctl_sctp_rmem,
6625 .sysctl_wmem = sysctl_sctp_wmem,
6626 .memory_pressure = &sctp_memory_pressure,
6627 .enter_memory_pressure = sctp_enter_memory_pressure,
6628 .memory_allocated = &sctp_memory_allocated,
Pavel Emelyanov5f318862008-02-20 00:23:01 -08006629 .sockets_allocated = &sctp_sockets_allocated,
Linus Torvalds1da177e2005-04-16 15:20:36 -07006630};
6631#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */