net: poll() optimizations
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / sctp / socket.c
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
17 *
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
23 *
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
66 #include <linux/ip.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73
74 #include <net/ip.h>
75 #include <net/icmp.h>
76 #include <net/route.h>
77 #include <net/ipv6.h>
78 #include <net/inet_common.h>
79
80 #include <linux/socket.h> /* for sa_family_t */
81 #include <net/sock.h>
82 #include <net/sctp/sctp.h>
83 #include <net/sctp/sm.h>
84
85 /* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
88 */
89
90 /* Forward declarations for internal helper functions. */
91 static int sctp_writeable(struct sock *sk);
92 static void sctp_wfree(struct sk_buff *skb);
93 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97 static int sctp_wait_for_accept(struct sock *sk, long timeo);
98 static void sctp_wait_for_close(struct sock *sk, long timeo);
99 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108 static int sctp_autobind(struct sock *sk);
109 static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
112
113 extern struct kmem_cache *sctp_bucket_cachep;
114 extern int sysctl_sctp_mem[3];
115 extern int sysctl_sctp_rmem[3];
116 extern int sysctl_sctp_wmem[3];
117
118 static int sctp_memory_pressure;
119 static atomic_t sctp_memory_allocated;
120 struct percpu_counter sctp_sockets_allocated;
121
122 static void sctp_enter_memory_pressure(struct sock *sk)
123 {
124 sctp_memory_pressure = 1;
125 }
126
127
128 /* Get the sndbuf space available at the time on the association. */
129 static inline int sctp_wspace(struct sctp_association *asoc)
130 {
131 int amt;
132
133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
136 amt = sk_wmem_alloc_get(asoc->base.sk);
137
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
145 }
146 } else {
147 amt = asoc->base.sk->sk_sndbuf - amt;
148 }
149 return amt;
150 }
151
152 /* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
155 *
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
160 */
161 static 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
169 skb_set_owner_w(chunk->skb, sk);
170
171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
174
175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
178
179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
182 }
183
184 /* Verify that this is a valid address. */
185 static 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? */
196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
197 return -EINVAL;
198
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
201
202 return 0;
203 }
204
205 /* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
207 */
208 struct 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 */
246 static 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
254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
255 laddr,
256 &transport);
257
258 if (!addr_asoc)
259 return NULL;
260
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
264
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
267
268 return transport;
269 }
270
271 /* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
273 *
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
275 *
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
280 */
281 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
282 {
283 int retval = 0;
284
285 sctp_lock_sock(sk);
286
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
289
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
293 addr_len);
294 else
295 retval = -EINVAL;
296
297 sctp_release_sock(sk);
298
299 return retval;
300 }
301
302 static long sctp_get_port_local(struct sock *, union sctp_addr *);
303
304 /* Verify this is a valid sockaddr. */
305 static 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
314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
323 }
324
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
327
328 if (len < af->sockaddr_len)
329 return NULL;
330
331 return af;
332 }
333
334 /* Bind a local address either to an endpoint or to an association. */
335 SCTP_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
344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
349 return -EINVAL;
350 }
351
352 snum = ntohs(addr->v4.sin_port);
353
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
356 sk,
357 addr,
358 bp->port, snum,
359 len);
360
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
364
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
368 */
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
376 return -EINVAL;
377 }
378 }
379
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
382
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
385 */
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
388
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
392 */
393 addr->v4.sin_port = htons(snum);
394 if ((ret = sctp_get_port_local(sk, addr))) {
395 return -EADDRINUSE;
396 }
397
398 /* Refresh ephemeral port. */
399 if (!bp->port)
400 bp->port = inet_sk(sk)->inet_num;
401
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
404 */
405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
406
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
410 af->to_sk_saddr(addr, sk);
411 }
412
413 return ret;
414 }
415
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 *
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
425 */
426 static 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.
433 */
434 if (asoc->addip_last_asconf) {
435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
436 goto out;
437 }
438
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
446
447 out:
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 */
463 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
464 {
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
470
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
473
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
478 */
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
484 }
485
486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
487 af->sockaddr_len);
488
489 addr_buf += af->sockaddr_len;
490
491 err_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 *
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
510 *
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 */
513 static int sctp_send_asconf_add_ip(struct sock *sk,
514 struct sockaddr *addrs,
515 int addrcnt)
516 {
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
524 union sctp_addr saveaddr;
525 void *addr_buf;
526 struct sctp_af *af;
527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!sctp_addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__, sk, addrs, addrcnt);
539
540 list_for_each_entry(asoc, &ep->asocs, asocs) {
541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
575 */
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
586 retval = sctp_send_asconf(asoc, chunk);
587 if (retval)
588 goto out;
589
590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
592 */
593 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);
598 retval = sctp_add_bind_addr(bp, &saveaddr,
599 SCTP_ADDR_NEW, GFP_ATOMIC);
600 addr_buf += af->sockaddr_len;
601 }
602 }
603
604 out:
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 */
623 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
624 {
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;
630 void *addr_buf;
631 union sctp_addr *sa_addr;
632 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
649 sa_addr = (union sctp_addr *)addr_buf;
650 af = sctp_get_af_specific(sa_addr->sa.sa_family);
651 if (!af) {
652 retval = -EINVAL;
653 goto err_bindx_rem;
654 }
655
656 if (!af->addr_valid(sa_addr, sp, NULL)) {
657 retval = -EADDRNOTAVAIL;
658 goto err_bindx_rem;
659 }
660
661 if (sa_addr->v4.sin_port != htons(bp->port)) {
662 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 */
673 retval = sctp_del_bind_addr(bp, sa_addr);
674
675 addr_buf += af->sockaddr_len;
676 err_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 *
692 * If any of the addresses is already in the bind address list of the
693 * 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 */
698 static 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;
705 struct sctp_transport *transport;
706 struct sctp_bind_addr *bp;
707 struct sctp_chunk *chunk;
708 union sctp_addr *laddr;
709 void *addr_buf;
710 struct sctp_af *af;
711 struct sctp_sockaddr_entry *saddr;
712 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",
722 __func__, sk, addrs, addrcnt);
723
724 list_for_each_entry(asoc, &ep->asocs, asocs) {
725
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
736 * not present in the bind address list of the association.
737 * 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 */
762 bp = &asoc->base.bind_addr;
763 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
764 addrcnt, sp);
765 if (!laddr)
766 continue;
767
768 /* We do not need RCU protection throughout this loop
769 * because this is done under a socket lock from the
770 * setsockopt call.
771 */
772 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
779 /* Reset use_as_src flag for the addresses in the bind address
780 * list that are to be deleted.
781 */
782 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);
786 list_for_each_entry(saddr, &bp->address_list, list) {
787 if (sctp_cmp_addr_exact(&saddr->a, laddr))
788 saddr->state = SCTP_ADDR_DEL;
789 }
790 addr_buf += af->sockaddr_len;
791 }
792
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 */
797 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
798 transports) {
799 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);
805 }
806 out:
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
826 * must be used to distinguish the address length (note that this
827 * 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.
865 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
866 * from userspace.
867 *
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 */
886 SCTP_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. */
909 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
910 if (unlikely(!kaddrs))
911 return -ENOMEM;
912
913 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
914 kfree(kaddrs);
915 return -EFAULT;
916 }
917
918 /* Walk through the addrs buffer and count the number of addresses. */
919 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.
926 */
927 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;
955 }
956
957 out:
958 kfree(kaddrs);
959
960 return err;
961 }
962
963 /* __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 */
968 static int __sctp_connect(struct sock* sk,
969 struct sockaddr *kaddrs,
970 int addrs_size,
971 sctp_assoc_t *assoc_id)
972 {
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;
985 union sctp_addr *sa_addr = NULL;
986 void *addr_buf;
987 unsigned short port;
988 unsigned int f_flags = 0;
989
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) {
1007 sa_addr = (union sctp_addr *)addr_buf;
1008 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1009 port = ntohs(sa_addr->v4.sin_port);
1010
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
1019 /* 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);
1023 if (err)
1024 goto out_free;
1025
1026 /* 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
1032
1033 /* Check if there already is a matching association on the
1034 * endpoint (other than the one created here).
1035 */
1036 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1037 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 */
1049 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1050 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 }
1065 } else {
1066 /*
1067 * 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
1071 * 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 }
1078 }
1079
1080 scope = sctp_scope(&to);
1081 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1082 if (!asoc) {
1083 err = -ENOMEM;
1084 goto out_free;
1085 }
1086
1087 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1088 GFP_KERNEL);
1089 if (err < 0) {
1090 goto out_free;
1091 }
1092
1093 }
1094
1095 /* Prime the peer's transport structures. */
1096 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1097 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
1108 /* 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
1117 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() */
1123 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1124 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1125 af->to_sk_daddr(sa_addr, sk);
1126 sk->sk_err = 0;
1127
1128 /* 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
1136 err = sctp_wait_for_connect(asoc, &timeo);
1137 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1138 *assoc_id = asoc->assoc_id;
1139
1140 /* Don't free association on exit. */
1141 asoc = NULL;
1142
1143 out_free:
1144
1145 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1146 " kaddrs: %p err: %d\n",
1147 asoc, kaddrs, err);
1148 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
1156 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1157 * sctp_assoc_t *asoc);
1158 *
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 *
1173 * 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.
1177 *
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 *
1213 * Returns >=0 if ok, <0 errno code on error.
1214 */
1215 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1216 struct sockaddr __user *addrs,
1217 int addrs_size,
1218 sctp_assoc_t *assoc_id)
1219 {
1220 int err = 0;
1221 struct sockaddr *kaddrs;
1222
1223 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1224 __func__, sk, addrs, addrs_size);
1225
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. */
1234 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1235 if (unlikely(!kaddrs))
1236 return -ENOMEM;
1237
1238 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1239 err = -EFAULT;
1240 } else {
1241 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1242 }
1243
1244 kfree(kaddrs);
1245
1246 return err;
1247 }
1248
1249 /*
1250 * This is an older interface. It's kept for backward compatibility
1251 * to the option that doesn't provide association id.
1252 */
1253 SCTP_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 */
1266 SCTP_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
1281 /*
1282 * 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.
1288 */
1289 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1290 char __user *optval,
1291 int __user *optlen)
1292 {
1293 struct sctp_getaddrs_old param;
1294 sctp_assoc_t assoc_id = 0;
1295 int err = 0;
1296
1297 if (len < sizeof(param))
1298 return -EINVAL;
1299
1300 if (copy_from_user(&param, optval, sizeof(param)))
1301 return -EFAULT;
1302
1303 err = __sctp_setsockopt_connectx(sk,
1304 (struct sockaddr __user *)param.addrs,
1305 param.addr_num, &assoc_id);
1306
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
1317 /* 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 */
1367 SCTP_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;
1377 sk->sk_state = SCTP_SS_CLOSING;
1378
1379 ep = sctp_sk(sk)->ep;
1380
1381 /* Walk all associations on an endpoint. */
1382 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);
1394 continue;
1395 }
1396 }
1397
1398 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
1405 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. */
1440 static 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
1472 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1473
1474 SCTP_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;
1482 union sctp_addr to;
1483 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;
1494 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
1503 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1504
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);
1535 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
1550 /* 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))) {
1552 err = -EINVAL;
1553 goto out_nounlock;
1554 }
1555
1556 /* 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
1559 * the msg_iov set to the user abort reason.
1560 */
1561 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1562 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1563 err = -EINVAL;
1564 goto out_nounlock;
1565 }
1566
1567 /* If SCTP_ADDR_OVER is set, there must be an address
1568 * specified in msg_name.
1569 */
1570 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1571 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. */
1584 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1585 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)) ||
1593 sctp_endpoint_is_peeled_off(ep, &to)) {
1594 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
1619 if (sinfo_flags & SCTP_EOF) {
1620 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1621 asoc);
1622 sctp_primitive_SHUTDOWN(asoc, NULL);
1623 err = 0;
1624 goto out_unlock;
1625 }
1626 if (sinfo_flags & SCTP_ABORT) {
1627
1628 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1629 if (!chunk) {
1630 err = -ENOMEM;
1631 goto out_unlock;
1632 }
1633
1634 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1635 sctp_primitive_ABORT(asoc, chunk);
1636 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
1645 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1646 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 }
1683 } 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 }
1696 }
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;
1705 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1706 if (err < 0) {
1707 err = -ENOMEM;
1708 goto out_free;
1709 }
1710
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) {
1728 asoc->max_init_timeo =
1729 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1730 }
1731 }
1732
1733 /* Prime the peer's transport structures. */
1734 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1735 if (!transport) {
1736 err = -ENOMEM;
1737 goto out_free;
1738 }
1739 }
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
1765 if (asoc->pmtu_pending)
1766 sctp_assoc_pending_pmtu(asoc);
1767
1768 /* 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
1795 * when SCTP_ADDR_OVER flag is set in the UDP model.
1796 */
1797 if ((sctp_style(sk, TCP) && msg_name) ||
1798 (sinfo_flags & SCTP_ADDR_OVER)) {
1799 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1800 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. */
1823 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1824 sctp_chunk_hold(chunk);
1825
1826 /* Do accounting for the write space. */
1827 sctp_set_owner_w(chunk);
1828
1829 chunk->transport = chunk_tp;
1830 }
1831
1832 /* 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
1846 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
1856 out_free:
1857 if (new_asoc)
1858 sctp_association_free(asoc);
1859 out_unlock:
1860 sctp_release_sock(sk);
1861
1862 out_nounlock:
1863 return sctp_error(sk, msg_flags, err);
1864
1865 #if 0
1866 do_sock_err:
1867 if (msg_len)
1868 err = msg_len;
1869 else
1870 err = sock_error(sk);
1871 goto out;
1872
1873 do_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 */
1887 static 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
1900 skb_walk_frags(skb, list) {
1901 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 */
1929 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1930
1931 SCTP_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
1974 sock_recv_ts_and_drops(msg, sk, skb);
1975 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 */
2008 if (!sctp_ulpevent_is_notification(event))
2009 sctp_assoc_rwnd_increase(event->asoc, copied);
2010 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
2017 out_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 }
2030 out:
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 */
2042 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2043 char __user *optval,
2044 unsigned int optlen)
2045 {
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
2059 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2060 unsigned int optlen)
2061 {
2062 if (optlen > sizeof(struct sctp_event_subscribe))
2063 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 */
2080 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2081 unsigned int optlen)
2082 {
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;
2092 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2093 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2094
2095 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 {
2108 * 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 * };
2116 *
2117 * spp_assoc_id - (one-to-many style socket) This is filled in the
2118 * application, and identifies the association for
2119 * this query.
2120 * spp_address - This specifies which address is of interest.
2121 * spp_hbinterval - This contains the value of the heartbeat interval,
2122 * in milliseconds. If a value of zero
2123 * is present in this field then no changes are to
2124 * be made to this parameter.
2125 * spp_pathmaxrxt - This contains the maximum number of
2126 * retransmissions before this address shall be
2127 * 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 *
2165 * 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 *
2169 * 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.
2194 */
2195 static 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)
2202 {
2203 int error;
2204
2205 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2206 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2207 if (error)
2208 return error;
2209 }
2210
2211 /* 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 }
2235 }
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
2251 /* 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) {
2257 if (trans) {
2258 trans->pathmtu = params->spp_pathmtu;
2259 sctp_assoc_sync_pmtu(asoc);
2260 } else if (asoc) {
2261 asoc->pathmtu = params->spp_pathmtu;
2262 sctp_frag_point(asoc, params->spp_pathmtu);
2263 } 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
2287 /* 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) {
2292 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
2319 /* Note that a value of zero indicates the current setting should be
2320 left unchanged.
2321 */
2322 if (params->spp_pathmaxrxt) {
2323 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
2335 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2336 char __user *optval,
2337 unsigned int optlen)
2338 {
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 ||
2361 (params.spp_pathmtu &&
2362 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2363 return -EINVAL;
2364
2365 /* If an address other than INADDR_ANY is specified, and
2366 * no transport is found, then the request is invalid.
2367 */
2368 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2369 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) {
2401 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2402 transports) {
2403 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2404 hb_change, pmtud_change,
2405 sackdelay_change);
2406 }
2407 }
2408
2409 return 0;
2410 }
2411
2412 /*
2413 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2414 *
2415 * 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.
2425 *
2426 * struct sctp_sack_info {
2427 * sctp_assoc_t sack_assoc_id;
2428 * uint32_t sack_delay;
2429 * uint32_t sack_freq;
2430 * };
2431 *
2432 * 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).
2436 *
2437 * 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.
2441 *
2442 * 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.
2446 */
2447
2448 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2449 char __user *optval, unsigned int optlen)
2450 {
2451 struct sctp_sack_info params;
2452 struct sctp_transport *trans = NULL;
2453 struct sctp_association *asoc = NULL;
2454 struct sctp_sock *sp = sctp_sk(sk);
2455
2456 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)) {
2463 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");
2465 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
2473 return - EINVAL;
2474
2475 /* Validate value parameter. */
2476 if (params.sack_delay > 500)
2477 return -EINVAL;
2478
2479 /* Get association, if sack_assoc_id != 0 and the socket is a one
2480 * to many style socket, and an association was not found, then
2481 * the id was invalid.
2482 */
2483 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2484 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2485 return -EINVAL;
2486
2487 if (params.sack_delay) {
2488 if (asoc) {
2489 asoc->sackdelay =
2490 msecs_to_jiffies(params.sack_delay);
2491 asoc->param_flags =
2492 (asoc->param_flags & ~SPP_SACKDELAY) |
2493 SPP_SACKDELAY_ENABLE;
2494 } else {
2495 sp->sackdelay = params.sack_delay;
2496 sp->param_flags =
2497 (sp->param_flags & ~SPP_SACKDELAY) |
2498 SPP_SACKDELAY_ENABLE;
2499 }
2500 }
2501
2502 if (params.sack_freq == 1) {
2503 if (asoc) {
2504 asoc->param_flags =
2505 (asoc->param_flags & ~SPP_SACKDELAY) |
2506 SPP_SACKDELAY_DISABLE;
2507 } else {
2508 sp->param_flags =
2509 (sp->param_flags & ~SPP_SACKDELAY) |
2510 SPP_SACKDELAY_DISABLE;
2511 }
2512 } 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 }
2524 }
2525
2526 /* If change is for association, also apply to each transport. */
2527 if (asoc) {
2528 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2529 transports) {
2530 if (params.sack_delay) {
2531 trans->sackdelay =
2532 msecs_to_jiffies(params.sack_delay);
2533 trans->param_flags =
2534 (trans->param_flags & ~SPP_SACKDELAY) |
2535 SPP_SACKDELAY_ENABLE;
2536 }
2537 if (params.sack_freq == 1) {
2538 trans->param_flags =
2539 (trans->param_flags & ~SPP_SACKDELAY) |
2540 SPP_SACKDELAY_DISABLE;
2541 } 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;
2546 }
2547 }
2548 }
2549
2550 return 0;
2551 }
2552
2553 /* 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 */
2564 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2565 {
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)
2575 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2576 if (sinit.sinit_max_instreams)
2577 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2578 if (sinit.sinit_max_attempts)
2579 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2580 if (sinit.sinit_max_init_timeo)
2581 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2582
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 */
2600 static int sctp_setsockopt_default_send_param(struct sock *sk,
2601 char __user *optval,
2602 unsigned int optlen)
2603 {
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 */
2640 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2641 unsigned int optlen)
2642 {
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 */
2669 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2670 unsigned int optlen)
2671 {
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 */
2695 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2696 {
2697 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)
2714 asoc->rto_initial =
2715 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 *
2741 * This option is used to tune the maximum retransmission attempts
2742 * 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 */
2748 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2749 {
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) {
2766 if (assocparams.sasoc_asocmaxrxt != 0) {
2767 __u32 path_sum = 0;
2768 int paths = 0;
2769 struct sctp_transport *peer_addr;
2770
2771 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2772 transports) {
2773 path_sum += peer_addr->pathmaxrxt;
2774 paths++;
2775 }
2776
2777 /* Only validate asocmaxrxt if we have more than
2778 * 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
2786 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2787 }
2788
2789 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 */
2820 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2821 {
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 /*
2838 * 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
2841 * 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
2844 * 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.
2863 */
2864 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2865 {
2866 struct sctp_assoc_value params;
2867 struct sctp_association *asoc;
2868 struct sctp_sock *sp = sctp_sk(sk);
2869 int val;
2870
2871 if (optlen == sizeof(int)) {
2872 pr_warn("Use of int in maxseg socket option deprecated\n");
2873 pr_warn("Use struct sctp_assoc_value instead\n");
2874 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
2882 return -EINVAL;
2883
2884 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2885 return -EINVAL;
2886
2887 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 }
2898 asoc->user_frag = val;
2899 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2900 } else {
2901 sp->user_frag = val;
2902 }
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 */
2916 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2917 unsigned int optlen)
2918 {
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);
2939 if (!asoc)
2940 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
2967 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2968 unsigned int optlen)
2969 {
2970 struct sctp_setadaptation adaptation;
2971
2972 if (optlen != sizeof(struct sctp_setadaptation))
2973 return -EINVAL;
2974 if (copy_from_user(&adaptation, optval, optlen))
2975 return -EFAULT;
2976
2977 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2978
2979 return 0;
2980 }
2981
2982 /*
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 */
2996 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2997 unsigned int optlen)
2998 {
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
3022 /*
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 */
3046 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3047 char __user *optval,
3048 unsigned int optlen)
3049 {
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
3062 /*
3063 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3064 * (SCTP_PARTIAL_DELIVERY_POINT)
3065 *
3066 * 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
3069 * lower value will cause partial deliveries to happen more often. The
3070 * calls argument is an integer that sets or gets the partial delivery
3071 * 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.
3078 */
3079 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3080 char __user *optval,
3081 unsigned int optlen)
3082 {
3083 u32 val;
3084
3085 if (optlen != sizeof(u32))
3086 return -EINVAL;
3087 if (get_user(val, (int __user *)optval))
3088 return -EFAULT;
3089
3090 /* 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
3096 sctp_sk(sk)->pd_point = val;
3097
3098 return 0; /* is this the right error code? */
3099 }
3100
3101 /*
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 */
3112 static int sctp_setsockopt_maxburst(struct sock *sk,
3113 char __user *optval,
3114 unsigned int optlen)
3115 {
3116 struct sctp_assoc_value params;
3117 struct sctp_sock *sp;
3118 struct sctp_association *asoc;
3119 int val;
3120 int assoc_id = 0;
3121
3122 if (optlen == sizeof(int)) {
3123 pr_warn("Use of int in max_burst socket option deprecated\n");
3124 pr_warn("Use struct sctp_assoc_value instead\n");
3125 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;
3144
3145 return 0;
3146 }
3147
3148 /*
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 */
3155 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3156 char __user *optval,
3157 unsigned int optlen)
3158 {
3159 struct sctp_authchunk val;
3160
3161 if (!sctp_auth_enable)
3162 return -EACCES;
3163
3164 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 */
3187 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3188 char __user *optval,
3189 unsigned int optlen)
3190 {
3191 struct sctp_hmacalgo *hmacs;
3192 u32 idents;
3193 int err;
3194
3195 if (!sctp_auth_enable)
3196 return -EACCES;
3197
3198 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
3210 idents = hmacs->shmac_num_idents;
3211 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3212 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3213 err = -EINVAL;
3214 goto out;
3215 }
3216
3217 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3218 out:
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 */
3229 static int sctp_setsockopt_auth_key(struct sock *sk,
3230 char __user *optval,
3231 unsigned int optlen)
3232 {
3233 struct sctp_authkey *authkey;
3234 struct sctp_association *asoc;
3235 int ret;
3236
3237 if (!sctp_auth_enable)
3238 return -EACCES;
3239
3240 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
3252 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3253 ret = -EINVAL;
3254 goto out;
3255 }
3256
3257 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);
3264 out:
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 */
3275 static int sctp_setsockopt_active_key(struct sock *sk,
3276 char __user *optval,
3277 unsigned int optlen)
3278 {
3279 struct sctp_authkeyid val;
3280 struct sctp_association *asoc;
3281
3282 if (!sctp_auth_enable)
3283 return -EACCES;
3284
3285 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 */
3303 static int sctp_setsockopt_del_key(struct sock *sk,
3304 char __user *optval,
3305 unsigned int optlen)
3306 {
3307 struct sctp_authkeyid val;
3308 struct sctp_association *asoc;
3309
3310 if (!sctp_auth_enable)
3311 return -EACCES;
3312
3313 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
3328 /* 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 */
3347 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3348 char __user *optval, unsigned int optlen)
3349 {
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
3382 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
3389 case SCTP_SOCKOPT_CONNECTX:
3390 /* 'optlen' is the size of the addresses buffer. */
3391 retval = sctp_setsockopt_connectx(sk,
3392 (struct sockaddr __user *)optval,
3393 optlen);
3394 break;
3395
3396 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
3412 case SCTP_DELAYED_ACK:
3413 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3414 break;
3415 case SCTP_PARTIAL_DELIVERY_POINT:
3416 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3417 break;
3418
3419 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;
3447 case SCTP_ADAPTATION_LAYER:
3448 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3449 break;
3450 case SCTP_CONTEXT:
3451 retval = sctp_setsockopt_context(sk, optval, optlen);
3452 break;
3453 case SCTP_FRAGMENT_INTERLEAVE:
3454 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3455 break;
3456 case SCTP_MAX_BURST:
3457 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3458 break;
3459 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;
3474 default:
3475 retval = -ENOPROTOOPT;
3476 break;
3477 }
3478
3479 sctp_release_sock(sk);
3480
3481 out_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 */
3501 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3502 int addr_len)
3503 {
3504 int err = 0;
3505 struct sctp_af *af;
3506
3507 sctp_lock_sock(sk);
3508
3509 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3510 __func__, sk, addr, addr_len);
3511
3512 /* 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 */
3520 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3521 }
3522
3523 sctp_release_sock(sk);
3524 return err;
3525 }
3526
3527 /* FIXME: Write comments. */
3528 SCTP_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 */
3540 SCTP_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
3564 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3565
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
3586 out:
3587 sctp_release_sock(sk);
3588 *err = error;
3589 return newsk;
3590 }
3591
3592 /* The SCTP ioctl handler. */
3593 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3594 {
3595 return -ENOIOCTLCMD;
3596 }
3597
3598 /* This is the function which gets called during socket creation to
3599 * initialized the SCTP-specific portion of the sock.
3600 * The sock structure should already be zero-filled memory.
3601 */
3602 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3603 {
3604 struct sctp_endpoint *ep;
3605 struct sctp_sock *sp;
3606
3607 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3608
3609 sp = sctp_sk(sk);
3610
3611 /* Initialize the SCTP per socket area. */
3612 switch (sk->sk_type) {
3613 case SOCK_SEQPACKET:
3614 sp->type = SCTP_SOCKET_UDP;
3615 break;
3616 case SOCK_STREAM:
3617 sp->type = SCTP_SOCKET_TCP;
3618 break;
3619 default:
3620 return -ESOCKTNOSUPPORT;
3621 }
3622
3623 /* Initialize default send parameters. These parameters can be
3624 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3625 */
3626 sp->default_stream = 0;
3627 sp->default_ppid = 0;
3628 sp->default_flags = 0;
3629 sp->default_context = 0;
3630 sp->default_timetolive = 0;
3631
3632 sp->default_rcv_context = 0;
3633 sp->max_burst = sctp_max_burst;
3634
3635 /* Initialize default setup parameters. These parameters
3636 * can be modified with the SCTP_INITMSG socket option or
3637 * overridden by the SCTP_INIT CMSG.
3638 */
3639 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3640 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3641 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3642 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3643
3644 /* Initialize default RTO related parameters. These parameters can
3645 * be modified for with the SCTP_RTOINFO socket option.
3646 */
3647 sp->rtoinfo.srto_initial = sctp_rto_initial;
3648 sp->rtoinfo.srto_max = sctp_rto_max;
3649 sp->rtoinfo.srto_min = sctp_rto_min;
3650
3651 /* Initialize default association related parameters. These parameters
3652 * can be modified with the SCTP_ASSOCINFO socket option.
3653 */
3654 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3655 sp->assocparams.sasoc_number_peer_destinations = 0;
3656 sp->assocparams.sasoc_peer_rwnd = 0;
3657 sp->assocparams.sasoc_local_rwnd = 0;
3658 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3659
3660 /* Initialize default event subscriptions. By default, all the
3661 * options are off.
3662 */
3663 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3664
3665 /* Default Peer Address Parameters. These defaults can
3666 * be modified via SCTP_PEER_ADDR_PARAMS
3667 */
3668 sp->hbinterval = sctp_hb_interval;
3669 sp->pathmaxrxt = sctp_max_retrans_path;
3670 sp->pathmtu = 0; // allow default discovery
3671 sp->sackdelay = sctp_sack_timeout;
3672 sp->sackfreq = 2;
3673 sp->param_flags = SPP_HB_ENABLE |
3674 SPP_PMTUD_ENABLE |
3675 SPP_SACKDELAY_ENABLE;
3676
3677 /* If enabled no SCTP message fragmentation will be performed.
3678 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3679 */
3680 sp->disable_fragments = 0;
3681
3682 /* Enable Nagle algorithm by default. */
3683 sp->nodelay = 0;
3684
3685 /* Enable by default. */
3686 sp->v4mapped = 1;
3687
3688 /* Auto-close idle associations after the configured
3689 * number of seconds. A value of 0 disables this
3690 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3691 * for UDP-style sockets only.
3692 */
3693 sp->autoclose = 0;
3694
3695 /* User specified fragmentation limit. */
3696 sp->user_frag = 0;
3697
3698 sp->adaptation_ind = 0;
3699
3700 sp->pf = sctp_get_pf_specific(sk->sk_family);
3701
3702 /* Control variables for partial data delivery. */
3703 atomic_set(&sp->pd_mode, 0);
3704 skb_queue_head_init(&sp->pd_lobby);
3705 sp->frag_interleave = 0;
3706
3707 /* Create a per socket endpoint structure. Even if we
3708 * change the data structure relationships, this may still
3709 * be useful for storing pre-connect address information.
3710 */
3711 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3712 if (!ep)
3713 return -ENOMEM;
3714
3715 sp->ep = ep;
3716 sp->hmac = NULL;
3717
3718 SCTP_DBG_OBJCNT_INC(sock);
3719
3720 local_bh_disable();
3721 percpu_counter_inc(&sctp_sockets_allocated);
3722 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3723 local_bh_enable();
3724
3725 return 0;
3726 }
3727
3728 /* Cleanup any SCTP per socket resources. */
3729 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3730 {
3731 struct sctp_endpoint *ep;
3732
3733 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3734
3735 /* Release our hold on the endpoint. */
3736 ep = sctp_sk(sk)->ep;
3737 sctp_endpoint_free(ep);
3738 local_bh_disable();
3739 percpu_counter_dec(&sctp_sockets_allocated);
3740 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3741 local_bh_enable();
3742 }
3743
3744 /* API 4.1.7 shutdown() - TCP Style Syntax
3745 * int shutdown(int socket, int how);
3746 *
3747 * sd - the socket descriptor of the association to be closed.
3748 * how - Specifies the type of shutdown. The values are
3749 * as follows:
3750 * SHUT_RD
3751 * Disables further receive operations. No SCTP
3752 * protocol action is taken.
3753 * SHUT_WR
3754 * Disables further send operations, and initiates
3755 * the SCTP shutdown sequence.
3756 * SHUT_RDWR
3757 * Disables further send and receive operations
3758 * and initiates the SCTP shutdown sequence.
3759 */
3760 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3761 {
3762 struct sctp_endpoint *ep;
3763 struct sctp_association *asoc;
3764
3765 if (!sctp_style(sk, TCP))
3766 return;
3767
3768 if (how & SEND_SHUTDOWN) {
3769 ep = sctp_sk(sk)->ep;
3770 if (!list_empty(&ep->asocs)) {
3771 asoc = list_entry(ep->asocs.next,
3772 struct sctp_association, asocs);
3773 sctp_primitive_SHUTDOWN(asoc, NULL);
3774 }
3775 }
3776 }
3777
3778 /* 7.2.1 Association Status (SCTP_STATUS)
3779
3780 * Applications can retrieve current status information about an
3781 * association, including association state, peer receiver window size,
3782 * number of unacked data chunks, and number of data chunks pending
3783 * receipt. This information is read-only.
3784 */
3785 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3786 char __user *optval,
3787 int __user *optlen)
3788 {
3789 struct sctp_status status;
3790 struct sctp_association *asoc = NULL;
3791 struct sctp_transport *transport;
3792 sctp_assoc_t associd;
3793 int retval = 0;
3794
3795 if (len < sizeof(status)) {
3796 retval = -EINVAL;
3797 goto out;
3798 }
3799
3800 len = sizeof(status);
3801 if (copy_from_user(&status, optval, len)) {
3802 retval = -EFAULT;
3803 goto out;
3804 }
3805
3806 associd = status.sstat_assoc_id;
3807 asoc = sctp_id2assoc(sk, associd);
3808 if (!asoc) {
3809 retval = -EINVAL;
3810 goto out;
3811 }
3812
3813 transport = asoc->peer.primary_path;
3814
3815 status.sstat_assoc_id = sctp_assoc2id(asoc);
3816 status.sstat_state = asoc->state;
3817 status.sstat_rwnd = asoc->peer.rwnd;
3818 status.sstat_unackdata = asoc->unack_data;
3819
3820 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3821 status.sstat_instrms = asoc->c.sinit_max_instreams;
3822 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3823 status.sstat_fragmentation_point = asoc->frag_point;
3824 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3825 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3826 transport->af_specific->sockaddr_len);
3827 /* Map ipv4 address into v4-mapped-on-v6 address. */
3828 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3829 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3830 status.sstat_primary.spinfo_state = transport->state;
3831 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3832 status.sstat_primary.spinfo_srtt = transport->srtt;
3833 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3834 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3835
3836 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3837 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3838
3839 if (put_user(len, optlen)) {
3840 retval = -EFAULT;
3841 goto out;
3842 }
3843
3844 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3845 len, status.sstat_state, status.sstat_rwnd,
3846 status.sstat_assoc_id);
3847
3848 if (copy_to_user(optval, &status, len)) {
3849 retval = -EFAULT;
3850 goto out;
3851 }
3852
3853 out:
3854 return (retval);
3855 }
3856
3857
3858 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3859 *
3860 * Applications can retrieve information about a specific peer address
3861 * of an association, including its reachability state, congestion
3862 * window, and retransmission timer values. This information is
3863 * read-only.
3864 */
3865 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3866 char __user *optval,
3867 int __user *optlen)
3868 {
3869 struct sctp_paddrinfo pinfo;
3870 struct sctp_transport *transport;
3871 int retval = 0;
3872
3873 if (len < sizeof(pinfo)) {
3874 retval = -EINVAL;
3875 goto out;
3876 }
3877
3878 len = sizeof(pinfo);
3879 if (copy_from_user(&pinfo, optval, len)) {
3880 retval = -EFAULT;
3881 goto out;
3882 }
3883
3884 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3885 pinfo.spinfo_assoc_id);
3886 if (!transport)
3887 return -EINVAL;
3888
3889 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3890 pinfo.spinfo_state = transport->state;
3891 pinfo.spinfo_cwnd = transport->cwnd;
3892 pinfo.spinfo_srtt = transport->srtt;
3893 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3894 pinfo.spinfo_mtu = transport->pathmtu;
3895
3896 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3897 pinfo.spinfo_state = SCTP_ACTIVE;
3898
3899 if (put_user(len, optlen)) {
3900 retval = -EFAULT;
3901 goto out;
3902 }
3903
3904 if (copy_to_user(optval, &pinfo, len)) {
3905 retval = -EFAULT;
3906 goto out;
3907 }
3908
3909 out:
3910 return (retval);
3911 }
3912
3913 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3914 *
3915 * This option is a on/off flag. If enabled no SCTP message
3916 * fragmentation will be performed. Instead if a message being sent
3917 * exceeds the current PMTU size, the message will NOT be sent and
3918 * instead a error will be indicated to the user.
3919 */
3920 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3921 char __user *optval, int __user *optlen)
3922 {
3923 int val;
3924
3925 if (len < sizeof(int))
3926 return -EINVAL;
3927
3928 len = sizeof(int);
3929 val = (sctp_sk(sk)->disable_fragments == 1);
3930 if (put_user(len, optlen))
3931 return -EFAULT;
3932 if (copy_to_user(optval, &val, len))
3933 return -EFAULT;
3934 return 0;
3935 }
3936
3937 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3938 *
3939 * This socket option is used to specify various notifications and
3940 * ancillary data the user wishes to receive.
3941 */
3942 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3943 int __user *optlen)
3944 {
3945 if (len < sizeof(struct sctp_event_subscribe))
3946 return -EINVAL;
3947 len = sizeof(struct sctp_event_subscribe);
3948 if (put_user(len, optlen))
3949 return -EFAULT;
3950 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3951 return -EFAULT;
3952 return 0;
3953 }
3954
3955 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3956 *
3957 * This socket option is applicable to the UDP-style socket only. When
3958 * set it will cause associations that are idle for more than the
3959 * specified number of seconds to automatically close. An association
3960 * being idle is defined an association that has NOT sent or received
3961 * user data. The special value of '0' indicates that no automatic
3962 * close of any associations should be performed. The option expects an
3963 * integer defining the number of seconds of idle time before an
3964 * association is closed.
3965 */
3966 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3967 {
3968 /* Applicable to UDP-style socket only */
3969 if (sctp_style(sk, TCP))
3970 return -EOPNOTSUPP;
3971 if (len < sizeof(int))
3972 return -EINVAL;
3973 len = sizeof(int);
3974 if (put_user(len, optlen))
3975 return -EFAULT;
3976 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3977 return -EFAULT;
3978 return 0;
3979 }
3980
3981 /* Helper routine to branch off an association to a new socket. */
3982 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3983 struct socket **sockp)
3984 {
3985 struct sock *sk = asoc->base.sk;
3986 struct socket *sock;
3987 struct sctp_af *af;
3988 int err = 0;
3989
3990 /* An association cannot be branched off from an already peeled-off
3991 * socket, nor is this supported for tcp style sockets.
3992 */
3993 if (!sctp_style(sk, UDP))
3994 return -EINVAL;
3995
3996 /* Create a new socket. */
3997 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3998 if (err < 0)
3999 return err;
4000
4001 sctp_copy_sock(sock->sk, sk, asoc);
4002
4003 /* Make peeled-off sockets more like 1-1 accepted sockets.
4004 * Set the daddr and initialize id to something more random
4005 */
4006 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4007 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4008
4009 /* Populate the fields of the newsk from the oldsk and migrate the
4010 * asoc to the newsk.
4011 */
4012 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4013
4014 *sockp = sock;
4015
4016 return err;
4017 }
4018
4019 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4020 {
4021 sctp_peeloff_arg_t peeloff;
4022 struct socket *newsock;
4023 int retval = 0;
4024 struct sctp_association *asoc;
4025
4026 if (len < sizeof(sctp_peeloff_arg_t))
4027 return -EINVAL;
4028 len = sizeof(sctp_peeloff_arg_t);
4029 if (copy_from_user(&peeloff, optval, len))
4030 return -EFAULT;
4031
4032 asoc = sctp_id2assoc(sk, peeloff.associd);
4033 if (!asoc) {
4034 retval = -EINVAL;
4035 goto out;
4036 }
4037
4038 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4039
4040 retval = sctp_do_peeloff(asoc, &newsock);
4041 if (retval < 0)
4042 goto out;
4043
4044 /* Map the socket to an unused fd that can be returned to the user. */
4045 retval = sock_map_fd(newsock, 0);
4046 if (retval < 0) {
4047 sock_release(newsock);
4048 goto out;
4049 }
4050
4051 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4052 __func__, sk, asoc, newsock->sk, retval);
4053
4054 /* Return the fd mapped to the new socket. */
4055 peeloff.sd = retval;
4056 if (put_user(len, optlen))
4057 return -EFAULT;
4058 if (copy_to_user(optval, &peeloff, len))
4059 retval = -EFAULT;
4060
4061 out:
4062 return retval;
4063 }
4064
4065 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4066 *
4067 * Applications can enable or disable heartbeats for any peer address of
4068 * an association, modify an address's heartbeat interval, force a
4069 * heartbeat to be sent immediately, and adjust the address's maximum
4070 * number of retransmissions sent before an address is considered
4071 * unreachable. The following structure is used to access and modify an
4072 * address's parameters:
4073 *
4074 * struct sctp_paddrparams {
4075 * sctp_assoc_t spp_assoc_id;
4076 * struct sockaddr_storage spp_address;
4077 * uint32_t spp_hbinterval;
4078 * uint16_t spp_pathmaxrxt;
4079 * uint32_t spp_pathmtu;
4080 * uint32_t spp_sackdelay;
4081 * uint32_t spp_flags;
4082 * };
4083 *
4084 * spp_assoc_id - (one-to-many style socket) This is filled in the
4085 * application, and identifies the association for
4086 * this query.
4087 * spp_address - This specifies which address is of interest.
4088 * spp_hbinterval - This contains the value of the heartbeat interval,
4089 * in milliseconds. If a value of zero
4090 * is present in this field then no changes are to
4091 * be made to this parameter.
4092 * spp_pathmaxrxt - This contains the maximum number of
4093 * retransmissions before this address shall be
4094 * considered unreachable. If a value of zero
4095 * is present in this field then no changes are to
4096 * be made to this parameter.
4097 * spp_pathmtu - When Path MTU discovery is disabled the value
4098 * specified here will be the "fixed" path mtu.
4099 * Note that if the spp_address field is empty
4100 * then all associations on this address will
4101 * have this fixed path mtu set upon them.
4102 *
4103 * spp_sackdelay - When delayed sack is enabled, this value specifies
4104 * the number of milliseconds that sacks will be delayed
4105 * for. This value will apply to all addresses of an
4106 * association if the spp_address field is empty. Note
4107 * also, that if delayed sack is enabled and this
4108 * value is set to 0, no change is made to the last
4109 * recorded delayed sack timer value.
4110 *
4111 * spp_flags - These flags are used to control various features
4112 * on an association. The flag field may contain
4113 * zero or more of the following options.
4114 *
4115 * SPP_HB_ENABLE - Enable heartbeats on the
4116 * specified address. Note that if the address
4117 * field is empty all addresses for the association
4118 * have heartbeats enabled upon them.
4119 *
4120 * SPP_HB_DISABLE - Disable heartbeats on the
4121 * speicifed address. Note that if the address
4122 * field is empty all addresses for the association
4123 * will have their heartbeats disabled. Note also
4124 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4125 * mutually exclusive, only one of these two should
4126 * be specified. Enabling both fields will have
4127 * undetermined results.
4128 *
4129 * SPP_HB_DEMAND - Request a user initiated heartbeat
4130 * to be made immediately.
4131 *
4132 * SPP_PMTUD_ENABLE - This field will enable PMTU
4133 * discovery upon the specified address. Note that
4134 * if the address feild is empty then all addresses
4135 * on the association are effected.
4136 *
4137 * SPP_PMTUD_DISABLE - This field will disable PMTU
4138 * discovery upon the specified address. Note that
4139 * if the address feild is empty then all addresses
4140 * on the association are effected. Not also that
4141 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4142 * exclusive. Enabling both will have undetermined
4143 * results.
4144 *
4145 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4146 * on delayed sack. The time specified in spp_sackdelay
4147 * is used to specify the sack delay for this address. Note
4148 * that if spp_address is empty then all addresses will
4149 * enable delayed sack and take on the sack delay
4150 * value specified in spp_sackdelay.
4151 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4152 * off delayed sack. If the spp_address field is blank then
4153 * delayed sack is disabled for the entire association. Note
4154 * also that this field is mutually exclusive to
4155 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4156 * results.
4157 */
4158 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4159 char __user *optval, int __user *optlen)
4160 {
4161 struct sctp_paddrparams params;
4162 struct sctp_transport *trans = NULL;
4163 struct sctp_association *asoc = NULL;
4164 struct sctp_sock *sp = sctp_sk(sk);
4165
4166 if (len < sizeof(struct sctp_paddrparams))
4167 return -EINVAL;
4168 len = sizeof(struct sctp_paddrparams);
4169 if (copy_from_user(&params, optval, len))
4170 return -EFAULT;
4171
4172 /* If an address other than INADDR_ANY is specified, and
4173 * no transport is found, then the request is invalid.
4174 */
4175 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4176 trans = sctp_addr_id2transport(sk, &params.spp_address,
4177 params.spp_assoc_id);
4178 if (!trans) {
4179 SCTP_DEBUG_PRINTK("Failed no transport\n");
4180 return -EINVAL;
4181 }
4182 }
4183
4184 /* Get association, if assoc_id != 0 and the socket is a one
4185 * to many style socket, and an association was not found, then
4186 * the id was invalid.
4187 */
4188 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4189 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4190 SCTP_DEBUG_PRINTK("Failed no association\n");
4191 return -EINVAL;
4192 }
4193
4194 if (trans) {
4195 /* Fetch transport values. */
4196 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4197 params.spp_pathmtu = trans->pathmtu;
4198 params.spp_pathmaxrxt = trans->pathmaxrxt;
4199 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4200
4201 /*draft-11 doesn't say what to return in spp_flags*/
4202 params.spp_flags = trans->param_flags;
4203 } else if (asoc) {
4204 /* Fetch association values. */
4205 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4206 params.spp_pathmtu = asoc->pathmtu;
4207 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4208 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4209
4210 /*draft-11 doesn't say what to return in spp_flags*/
4211 params.spp_flags = asoc->param_flags;
4212 } else {
4213 /* Fetch socket values. */
4214 params.spp_hbinterval = sp->hbinterval;
4215 params.spp_pathmtu = sp->pathmtu;
4216 params.spp_sackdelay = sp->sackdelay;
4217 params.spp_pathmaxrxt = sp->pathmaxrxt;
4218
4219 /*draft-11 doesn't say what to return in spp_flags*/
4220 params.spp_flags = sp->param_flags;
4221 }
4222
4223 if (copy_to_user(optval, &params, len))
4224 return -EFAULT;
4225
4226 if (put_user(len, optlen))
4227 return -EFAULT;
4228
4229 return 0;
4230 }
4231
4232 /*
4233 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4234 *
4235 * This option will effect the way delayed acks are performed. This
4236 * option allows you to get or set the delayed ack time, in
4237 * milliseconds. It also allows changing the delayed ack frequency.
4238 * Changing the frequency to 1 disables the delayed sack algorithm. If
4239 * the assoc_id is 0, then this sets or gets the endpoints default
4240 * values. If the assoc_id field is non-zero, then the set or get
4241 * effects the specified association for the one to many model (the
4242 * assoc_id field is ignored by the one to one model). Note that if
4243 * sack_delay or sack_freq are 0 when setting this option, then the
4244 * current values will remain unchanged.
4245 *
4246 * struct sctp_sack_info {
4247 * sctp_assoc_t sack_assoc_id;
4248 * uint32_t sack_delay;
4249 * uint32_t sack_freq;
4250 * };
4251 *
4252 * sack_assoc_id - This parameter, indicates which association the user
4253 * is performing an action upon. Note that if this field's value is
4254 * zero then the endpoints default value is changed (effecting future
4255 * associations only).
4256 *
4257 * sack_delay - This parameter contains the number of milliseconds that
4258 * the user is requesting the delayed ACK timer be set to. Note that
4259 * this value is defined in the standard to be between 200 and 500
4260 * milliseconds.
4261 *
4262 * sack_freq - This parameter contains the number of packets that must
4263 * be received before a sack is sent without waiting for the delay
4264 * timer to expire. The default value for this is 2, setting this
4265 * value to 1 will disable the delayed sack algorithm.
4266 */
4267 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4268 char __user *optval,
4269 int __user *optlen)
4270 {
4271 struct sctp_sack_info params;
4272 struct sctp_association *asoc = NULL;
4273 struct sctp_sock *sp = sctp_sk(sk);
4274
4275 if (len >= sizeof(struct sctp_sack_info)) {
4276 len = sizeof(struct sctp_sack_info);
4277
4278 if (copy_from_user(&params, optval, len))
4279 return -EFAULT;
4280 } else if (len == sizeof(struct sctp_assoc_value)) {
4281 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4282 pr_warn("Use struct sctp_sack_info instead\n");
4283 if (copy_from_user(&params, optval, len))
4284 return -EFAULT;
4285 } else
4286 return - EINVAL;
4287
4288 /* Get association, if sack_assoc_id != 0 and the socket is a one
4289 * to many style socket, and an association was not found, then
4290 * the id was invalid.
4291 */
4292 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4293 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4294 return -EINVAL;
4295
4296 if (asoc) {
4297 /* Fetch association values. */
4298 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4299 params.sack_delay = jiffies_to_msecs(
4300 asoc->sackdelay);
4301 params.sack_freq = asoc->sackfreq;
4302
4303 } else {
4304 params.sack_delay = 0;
4305 params.sack_freq = 1;
4306 }
4307 } else {
4308 /* Fetch socket values. */
4309 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4310 params.sack_delay = sp->sackdelay;
4311 params.sack_freq = sp->sackfreq;
4312 } else {
4313 params.sack_delay = 0;
4314 params.sack_freq = 1;
4315 }
4316 }
4317
4318 if (copy_to_user(optval, &params, len))
4319 return -EFAULT;
4320
4321 if (put_user(len, optlen))
4322 return -EFAULT;
4323
4324 return 0;
4325 }
4326
4327 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4328 *
4329 * Applications can specify protocol parameters for the default association
4330 * initialization. The option name argument to setsockopt() and getsockopt()
4331 * is SCTP_INITMSG.
4332 *
4333 * Setting initialization parameters is effective only on an unconnected
4334 * socket (for UDP-style sockets only future associations are effected
4335 * by the change). With TCP-style sockets, this option is inherited by
4336 * sockets derived from a listener socket.
4337 */
4338 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4339 {
4340 if (len < sizeof(struct sctp_initmsg))
4341 return -EINVAL;
4342 len = sizeof(struct sctp_initmsg);
4343 if (put_user(len, optlen))
4344 return -EFAULT;
4345 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4346 return -EFAULT;
4347 return 0;
4348 }
4349
4350
4351 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4352 char __user *optval, int __user *optlen)
4353 {
4354 struct sctp_association *asoc;
4355 int cnt = 0;
4356 struct sctp_getaddrs getaddrs;
4357 struct sctp_transport *from;
4358 void __user *to;
4359 union sctp_addr temp;
4360 struct sctp_sock *sp = sctp_sk(sk);
4361 int addrlen;
4362 size_t space_left;
4363 int bytes_copied;
4364
4365 if (len < sizeof(struct sctp_getaddrs))
4366 return -EINVAL;
4367
4368 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4369 return -EFAULT;
4370
4371 /* For UDP-style sockets, id specifies the association to query. */
4372 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4373 if (!asoc)
4374 return -EINVAL;
4375
4376 to = optval + offsetof(struct sctp_getaddrs,addrs);
4377 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4378
4379 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4380 transports) {
4381 memcpy(&temp, &from->ipaddr, sizeof(temp));
4382 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4383 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4384 if (space_left < addrlen)
4385 return -ENOMEM;
4386 if (copy_to_user(to, &temp, addrlen))
4387 return -EFAULT;
4388 to += addrlen;
4389 cnt++;
4390 space_left -= addrlen;
4391 }
4392
4393 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4394 return -EFAULT;
4395 bytes_copied = ((char __user *)to) - optval;
4396 if (put_user(bytes_copied, optlen))
4397 return -EFAULT;
4398
4399 return 0;
4400 }
4401
4402 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4403 size_t space_left, int *bytes_copied)
4404 {
4405 struct sctp_sockaddr_entry *addr;
4406 union sctp_addr temp;
4407 int cnt = 0;
4408 int addrlen;
4409
4410 rcu_read_lock();
4411 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4412 if (!addr->valid)
4413 continue;
4414
4415 if ((PF_INET == sk->sk_family) &&
4416 (AF_INET6 == addr->a.sa.sa_family))
4417 continue;
4418 if ((PF_INET6 == sk->sk_family) &&
4419 inet_v6_ipv6only(sk) &&
4420 (AF_INET == addr->a.sa.sa_family))
4421 continue;
4422 memcpy(&temp, &addr->a, sizeof(temp));
4423 if (!temp.v4.sin_port)
4424 temp.v4.sin_port = htons(port);
4425
4426 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4427 &temp);
4428 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4429 if (space_left < addrlen) {
4430 cnt = -ENOMEM;
4431 break;
4432 }
4433 memcpy(to, &temp, addrlen);
4434
4435 to += addrlen;
4436 cnt ++;
4437 space_left -= addrlen;
4438 *bytes_copied += addrlen;
4439 }
4440 rcu_read_unlock();
4441
4442 return cnt;
4443 }
4444
4445
4446 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4447 char __user *optval, int __user *optlen)
4448 {
4449 struct sctp_bind_addr *bp;
4450 struct sctp_association *asoc;
4451 int cnt = 0;
4452 struct sctp_getaddrs getaddrs;
4453 struct sctp_sockaddr_entry *addr;
4454 void __user *to;
4455 union sctp_addr temp;
4456 struct sctp_sock *sp = sctp_sk(sk);
4457 int addrlen;
4458 int err = 0;
4459 size_t space_left;
4460 int bytes_copied = 0;
4461 void *addrs;
4462 void *buf;
4463
4464 if (len < sizeof(struct sctp_getaddrs))
4465 return -EINVAL;
4466
4467 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4468 return -EFAULT;
4469
4470 /*
4471 * For UDP-style sockets, id specifies the association to query.
4472 * If the id field is set to the value '0' then the locally bound
4473 * addresses are returned without regard to any particular
4474 * association.
4475 */
4476 if (0 == getaddrs.assoc_id) {
4477 bp = &sctp_sk(sk)->ep->base.bind_addr;
4478 } else {
4479 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4480 if (!asoc)
4481 return -EINVAL;
4482 bp = &asoc->base.bind_addr;
4483 }
4484
4485 to = optval + offsetof(struct sctp_getaddrs,addrs);
4486 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4487
4488 addrs = kmalloc(space_left, GFP_KERNEL);
4489 if (!addrs)
4490 return -ENOMEM;
4491
4492 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4493 * addresses from the global local address list.
4494 */
4495 if (sctp_list_single_entry(&bp->address_list)) {
4496 addr = list_entry(bp->address_list.next,
4497 struct sctp_sockaddr_entry, list);
4498 if (sctp_is_any(sk, &addr->a)) {
4499 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4500 space_left, &bytes_copied);
4501 if (cnt < 0) {
4502 err = cnt;
4503 goto out;
4504 }
4505 goto copy_getaddrs;
4506 }
4507 }
4508
4509 buf = addrs;
4510 /* Protection on the bound address list is not needed since
4511 * in the socket option context we hold a socket lock and
4512 * thus the bound address list can't change.
4513 */
4514 list_for_each_entry(addr, &bp->address_list, list) {
4515 memcpy(&temp, &addr->a, sizeof(temp));
4516 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4517 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4518 if (space_left < addrlen) {
4519 err = -ENOMEM; /*fixme: right error?*/
4520 goto out;
4521 }
4522 memcpy(buf, &temp, addrlen);
4523 buf += addrlen;
4524 bytes_copied += addrlen;
4525 cnt ++;
4526 space_left -= addrlen;
4527 }
4528
4529 copy_getaddrs:
4530 if (copy_to_user(to, addrs, bytes_copied)) {
4531 err = -EFAULT;
4532 goto out;
4533 }
4534 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4535 err = -EFAULT;
4536 goto out;
4537 }
4538 if (put_user(bytes_copied, optlen))
4539 err = -EFAULT;
4540 out:
4541 kfree(addrs);
4542 return err;
4543 }
4544
4545 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4546 *
4547 * Requests that the local SCTP stack use the enclosed peer address as
4548 * the association primary. The enclosed address must be one of the
4549 * association peer's addresses.
4550 */
4551 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4552 char __user *optval, int __user *optlen)
4553 {
4554 struct sctp_prim prim;
4555 struct sctp_association *asoc;
4556 struct sctp_sock *sp = sctp_sk(sk);
4557
4558 if (len < sizeof(struct sctp_prim))
4559 return -EINVAL;
4560
4561 len = sizeof(struct sctp_prim);
4562
4563 if (copy_from_user(&prim, optval, len))
4564 return -EFAULT;
4565
4566 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4567 if (!asoc)
4568 return -EINVAL;
4569
4570 if (!asoc->peer.primary_path)
4571 return -ENOTCONN;
4572
4573 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4574 asoc->peer.primary_path->af_specific->sockaddr_len);
4575
4576 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4577 (union sctp_addr *)&prim.ssp_addr);
4578
4579 if (put_user(len, optlen))
4580 return -EFAULT;
4581 if (copy_to_user(optval, &prim, len))
4582 return -EFAULT;
4583
4584 return 0;
4585 }
4586
4587 /*
4588 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4589 *
4590 * Requests that the local endpoint set the specified Adaptation Layer
4591 * Indication parameter for all future INIT and INIT-ACK exchanges.
4592 */
4593 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4594 char __user *optval, int __user *optlen)
4595 {
4596 struct sctp_setadaptation adaptation;
4597
4598 if (len < sizeof(struct sctp_setadaptation))
4599 return -EINVAL;
4600
4601 len = sizeof(struct sctp_setadaptation);
4602
4603 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4604
4605 if (put_user(len, optlen))
4606 return -EFAULT;
4607 if (copy_to_user(optval, &adaptation, len))
4608 return -EFAULT;
4609
4610 return 0;
4611 }
4612
4613 /*
4614 *
4615 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4616 *
4617 * Applications that wish to use the sendto() system call may wish to
4618 * specify a default set of parameters that would normally be supplied
4619 * through the inclusion of ancillary data. This socket option allows
4620 * such an application to set the default sctp_sndrcvinfo structure.
4621
4622
4623 * The application that wishes to use this socket option simply passes
4624 * in to this call the sctp_sndrcvinfo structure defined in Section
4625 * 5.2.2) The input parameters accepted by this call include
4626 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4627 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4628 * to this call if the caller is using the UDP model.
4629 *
4630 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4631 */
4632 static int sctp_getsockopt_default_send_param(struct sock *sk,
4633 int len, char __user *optval,
4634 int __user *optlen)
4635 {
4636 struct sctp_sndrcvinfo info;
4637 struct sctp_association *asoc;
4638 struct sctp_sock *sp = sctp_sk(sk);
4639
4640 if (len < sizeof(struct sctp_sndrcvinfo))
4641 return -EINVAL;
4642
4643 len = sizeof(struct sctp_sndrcvinfo);
4644
4645 if (copy_from_user(&info, optval, len))
4646 return -EFAULT;
4647
4648 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4649 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4650 return -EINVAL;
4651
4652 if (asoc) {
4653 info.sinfo_stream = asoc->default_stream;
4654 info.sinfo_flags = asoc->default_flags;
4655 info.sinfo_ppid = asoc->default_ppid;
4656 info.sinfo_context = asoc->default_context;
4657 info.sinfo_timetolive = asoc->default_timetolive;
4658 } else {
4659 info.sinfo_stream = sp->default_stream;
4660 info.sinfo_flags = sp->default_flags;
4661 info.sinfo_ppid = sp->default_ppid;
4662 info.sinfo_context = sp->default_context;
4663 info.sinfo_timetolive = sp->default_timetolive;
4664 }
4665
4666 if (put_user(len, optlen))
4667 return -EFAULT;
4668 if (copy_to_user(optval, &info, len))
4669 return -EFAULT;
4670
4671 return 0;
4672 }
4673
4674 /*
4675 *
4676 * 7.1.5 SCTP_NODELAY
4677 *
4678 * Turn on/off any Nagle-like algorithm. This means that packets are
4679 * generally sent as soon as possible and no unnecessary delays are
4680 * introduced, at the cost of more packets in the network. Expects an
4681 * integer boolean flag.
4682 */
4683
4684 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4685 char __user *optval, int __user *optlen)
4686 {
4687 int val;
4688
4689 if (len < sizeof(int))
4690 return -EINVAL;
4691
4692 len = sizeof(int);
4693 val = (sctp_sk(sk)->nodelay == 1);
4694 if (put_user(len, optlen))
4695 return -EFAULT;
4696 if (copy_to_user(optval, &val, len))
4697 return -EFAULT;
4698 return 0;
4699 }
4700
4701 /*
4702 *
4703 * 7.1.1 SCTP_RTOINFO
4704 *
4705 * The protocol parameters used to initialize and bound retransmission
4706 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4707 * and modify these parameters.
4708 * All parameters are time values, in milliseconds. A value of 0, when
4709 * modifying the parameters, indicates that the current value should not
4710 * be changed.
4711 *
4712 */
4713 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4714 char __user *optval,
4715 int __user *optlen) {
4716 struct sctp_rtoinfo rtoinfo;
4717 struct sctp_association *asoc;
4718
4719 if (len < sizeof (struct sctp_rtoinfo))
4720 return -EINVAL;
4721
4722 len = sizeof(struct sctp_rtoinfo);
4723
4724 if (copy_from_user(&rtoinfo, optval, len))
4725 return -EFAULT;
4726
4727 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4728
4729 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4730 return -EINVAL;
4731
4732 /* Values corresponding to the specific association. */
4733 if (asoc) {
4734 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4735 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4736 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4737 } else {
4738 /* Values corresponding to the endpoint. */
4739 struct sctp_sock *sp = sctp_sk(sk);
4740
4741 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4742 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4743 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4744 }
4745
4746 if (put_user(len, optlen))
4747 return -EFAULT;
4748
4749 if (copy_to_user(optval, &rtoinfo, len))
4750 return -EFAULT;
4751
4752 return 0;
4753 }
4754
4755 /*
4756 *
4757 * 7.1.2 SCTP_ASSOCINFO
4758 *
4759 * This option is used to tune the maximum retransmission attempts
4760 * of the association.
4761 * Returns an error if the new association retransmission value is
4762 * greater than the sum of the retransmission value of the peer.
4763 * See [SCTP] for more information.
4764 *
4765 */
4766 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4767 char __user *optval,
4768 int __user *optlen)
4769 {
4770
4771 struct sctp_assocparams assocparams;
4772 struct sctp_association *asoc;
4773 struct list_head *pos;
4774 int cnt = 0;
4775
4776 if (len < sizeof (struct sctp_assocparams))
4777 return -EINVAL;
4778
4779 len = sizeof(struct sctp_assocparams);
4780
4781 if (copy_from_user(&assocparams, optval, len))
4782 return -EFAULT;
4783
4784 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4785
4786 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4787 return -EINVAL;
4788
4789 /* Values correspoinding to the specific association */
4790 if (asoc) {
4791 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4792 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4793 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4794 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4795 * 1000) +
4796 (asoc->cookie_life.tv_usec
4797 / 1000);
4798
4799 list_for_each(pos, &asoc->peer.transport_addr_list) {
4800 cnt ++;
4801 }
4802
4803 assocparams.sasoc_number_peer_destinations = cnt;
4804 } else {
4805 /* Values corresponding to the endpoint */
4806 struct sctp_sock *sp = sctp_sk(sk);
4807
4808 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4809 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4810 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4811 assocparams.sasoc_cookie_life =
4812 sp->assocparams.sasoc_cookie_life;
4813 assocparams.sasoc_number_peer_destinations =
4814 sp->assocparams.
4815 sasoc_number_peer_destinations;
4816 }
4817
4818 if (put_user(len, optlen))
4819 return -EFAULT;
4820
4821 if (copy_to_user(optval, &assocparams, len))
4822 return -EFAULT;
4823
4824 return 0;
4825 }
4826
4827 /*
4828 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4829 *
4830 * This socket option is a boolean flag which turns on or off mapped V4
4831 * addresses. If this option is turned on and the socket is type
4832 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4833 * If this option is turned off, then no mapping will be done of V4
4834 * addresses and a user will receive both PF_INET6 and PF_INET type
4835 * addresses on the socket.
4836 */
4837 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4838 char __user *optval, int __user *optlen)
4839 {
4840 int val;
4841 struct sctp_sock *sp = sctp_sk(sk);
4842
4843 if (len < sizeof(int))
4844 return -EINVAL;
4845
4846 len = sizeof(int);
4847 val = sp->v4mapped;
4848 if (put_user(len, optlen))
4849 return -EFAULT;
4850 if (copy_to_user(optval, &val, len))
4851 return -EFAULT;
4852
4853 return 0;
4854 }
4855
4856 /*
4857 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4858 * (chapter and verse is quoted at sctp_setsockopt_context())
4859 */
4860 static int sctp_getsockopt_context(struct sock *sk, int len,
4861 char __user *optval, int __user *optlen)
4862 {
4863 struct sctp_assoc_value params;
4864 struct sctp_sock *sp;
4865 struct sctp_association *asoc;
4866
4867 if (len < sizeof(struct sctp_assoc_value))
4868 return -EINVAL;
4869
4870 len = sizeof(struct sctp_assoc_value);
4871
4872 if (copy_from_user(&params, optval, len))
4873 return -EFAULT;
4874
4875 sp = sctp_sk(sk);
4876
4877 if (params.assoc_id != 0) {
4878 asoc = sctp_id2assoc(sk, params.assoc_id);
4879 if (!asoc)
4880 return -EINVAL;
4881 params.assoc_value = asoc->default_rcv_context;
4882 } else {
4883 params.assoc_value = sp->default_rcv_context;
4884 }
4885
4886 if (put_user(len, optlen))
4887 return -EFAULT;
4888 if (copy_to_user(optval, &params, len))
4889 return -EFAULT;
4890
4891 return 0;
4892 }
4893
4894 /*
4895 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4896 * This option will get or set the maximum size to put in any outgoing
4897 * SCTP DATA chunk. If a message is larger than this size it will be
4898 * fragmented by SCTP into the specified size. Note that the underlying
4899 * SCTP implementation may fragment into smaller sized chunks when the
4900 * PMTU of the underlying association is smaller than the value set by
4901 * the user. The default value for this option is '0' which indicates
4902 * the user is NOT limiting fragmentation and only the PMTU will effect
4903 * SCTP's choice of DATA chunk size. Note also that values set larger
4904 * than the maximum size of an IP datagram will effectively let SCTP
4905 * control fragmentation (i.e. the same as setting this option to 0).
4906 *
4907 * The following structure is used to access and modify this parameter:
4908 *
4909 * struct sctp_assoc_value {
4910 * sctp_assoc_t assoc_id;
4911 * uint32_t assoc_value;
4912 * };
4913 *
4914 * assoc_id: This parameter is ignored for one-to-one style sockets.
4915 * For one-to-many style sockets this parameter indicates which
4916 * association the user is performing an action upon. Note that if
4917 * this field's value is zero then the endpoints default value is
4918 * changed (effecting future associations only).
4919 * assoc_value: This parameter specifies the maximum size in bytes.
4920 */
4921 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4922 char __user *optval, int __user *optlen)
4923 {
4924 struct sctp_assoc_value params;
4925 struct sctp_association *asoc;
4926
4927 if (len == sizeof(int)) {
4928 pr_warn("Use of int in maxseg socket option deprecated\n");
4929 pr_warn("Use struct sctp_assoc_value instead\n");
4930 params.assoc_id = 0;
4931 } else if (len >= sizeof(struct sctp_assoc_value)) {
4932 len = sizeof(struct sctp_assoc_value);
4933 if (copy_from_user(&params, optval, sizeof(params)))
4934 return -EFAULT;
4935 } else
4936 return -EINVAL;
4937
4938 asoc = sctp_id2assoc(sk, params.assoc_id);
4939 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4940 return -EINVAL;
4941
4942 if (asoc)
4943 params.assoc_value = asoc->frag_point;
4944 else
4945 params.assoc_value = sctp_sk(sk)->user_frag;
4946
4947 if (put_user(len, optlen))
4948 return -EFAULT;
4949 if (len == sizeof(int)) {
4950 if (copy_to_user(optval, &params.assoc_value, len))
4951 return -EFAULT;
4952 } else {
4953 if (copy_to_user(optval, &params, len))
4954 return -EFAULT;
4955 }
4956
4957 return 0;
4958 }
4959
4960 /*
4961 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4962 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4963 */
4964 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4965 char __user *optval, int __user *optlen)
4966 {
4967 int val;
4968
4969 if (len < sizeof(int))
4970 return -EINVAL;
4971
4972 len = sizeof(int);
4973
4974 val = sctp_sk(sk)->frag_interleave;
4975 if (put_user(len, optlen))
4976 return -EFAULT;
4977 if (copy_to_user(optval, &val, len))
4978 return -EFAULT;
4979
4980 return 0;
4981 }
4982
4983 /*
4984 * 7.1.25. Set or Get the sctp partial delivery point
4985 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4986 */
4987 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4988 char __user *optval,
4989 int __user *optlen)
4990 {
4991 u32 val;
4992
4993 if (len < sizeof(u32))
4994 return -EINVAL;
4995
4996 len = sizeof(u32);
4997
4998 val = sctp_sk(sk)->pd_point;
4999 if (put_user(len, optlen))
5000 return -EFAULT;
5001 if (copy_to_user(optval, &val, len))
5002 return -EFAULT;
5003
5004 return -ENOTSUPP;
5005 }
5006
5007 /*
5008 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5009 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5010 */
5011 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5012 char __user *optval,
5013 int __user *optlen)
5014 {
5015 struct sctp_assoc_value params;
5016 struct sctp_sock *sp;
5017 struct sctp_association *asoc;
5018
5019 if (len == sizeof(int)) {
5020 pr_warn("Use of int in max_burst socket option deprecated\n");
5021 pr_warn("Use struct sctp_assoc_value instead\n");
5022 params.assoc_id = 0;
5023 } else if (len >= sizeof(struct sctp_assoc_value)) {
5024 len = sizeof(struct sctp_assoc_value);
5025 if (copy_from_user(&params, optval, len))
5026 return -EFAULT;
5027 } else
5028 return -EINVAL;
5029
5030 sp = sctp_sk(sk);
5031
5032 if (params.assoc_id != 0) {
5033 asoc = sctp_id2assoc(sk, params.assoc_id);
5034 if (!asoc)
5035 return -EINVAL;
5036 params.assoc_value = asoc->max_burst;
5037 } else
5038 params.assoc_value = sp->max_burst;
5039
5040 if (len == sizeof(int)) {
5041 if (copy_to_user(optval, &params.assoc_value, len))
5042 return -EFAULT;
5043 } else {
5044 if (copy_to_user(optval, &params, len))
5045 return -EFAULT;
5046 }
5047
5048 return 0;
5049
5050 }
5051
5052 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5053 char __user *optval, int __user *optlen)
5054 {
5055 struct sctp_hmacalgo __user *p = (void __user *)optval;
5056 struct sctp_hmac_algo_param *hmacs;
5057 __u16 data_len = 0;
5058 u32 num_idents;
5059
5060 if (!sctp_auth_enable)
5061 return -EACCES;
5062
5063 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5064 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5065
5066 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5067 return -EINVAL;
5068
5069 len = sizeof(struct sctp_hmacalgo) + data_len;
5070 num_idents = data_len / sizeof(u16);
5071
5072 if (put_user(len, optlen))
5073 return -EFAULT;
5074 if (put_user(num_idents, &p->shmac_num_idents))
5075 return -EFAULT;
5076 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5077 return -EFAULT;
5078 return 0;
5079 }
5080
5081 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5082 char __user *optval, int __user *optlen)
5083 {
5084 struct sctp_authkeyid val;
5085 struct sctp_association *asoc;
5086
5087 if (!sctp_auth_enable)
5088 return -EACCES;
5089
5090 if (len < sizeof(struct sctp_authkeyid))
5091 return -EINVAL;
5092 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5093 return -EFAULT;
5094
5095 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5096 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5097 return -EINVAL;
5098
5099 if (asoc)
5100 val.scact_keynumber = asoc->active_key_id;
5101 else
5102 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5103
5104 len = sizeof(struct sctp_authkeyid);
5105 if (put_user(len, optlen))
5106 return -EFAULT;
5107 if (copy_to_user(optval, &val, len))
5108 return -EFAULT;
5109
5110 return 0;
5111 }
5112
5113 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5115 {
5116 struct sctp_authchunks __user *p = (void __user *)optval;
5117 struct sctp_authchunks val;
5118 struct sctp_association *asoc;
5119 struct sctp_chunks_param *ch;
5120 u32 num_chunks = 0;
5121 char __user *to;
5122
5123 if (!sctp_auth_enable)
5124 return -EACCES;
5125
5126 if (len < sizeof(struct sctp_authchunks))
5127 return -EINVAL;
5128
5129 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5130 return -EFAULT;
5131
5132 to = p->gauth_chunks;
5133 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5134 if (!asoc)
5135 return -EINVAL;
5136
5137 ch = asoc->peer.peer_chunks;
5138 if (!ch)
5139 goto num;
5140
5141 /* See if the user provided enough room for all the data */
5142 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5143 if (len < num_chunks)
5144 return -EINVAL;
5145
5146 if (copy_to_user(to, ch->chunks, num_chunks))
5147 return -EFAULT;
5148 num:
5149 len = sizeof(struct sctp_authchunks) + num_chunks;
5150 if (put_user(len, optlen)) return -EFAULT;
5151 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5152 return -EFAULT;
5153 return 0;
5154 }
5155
5156 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5157 char __user *optval, int __user *optlen)
5158 {
5159 struct sctp_authchunks __user *p = (void __user *)optval;
5160 struct sctp_authchunks val;
5161 struct sctp_association *asoc;
5162 struct sctp_chunks_param *ch;
5163 u32 num_chunks = 0;
5164 char __user *to;
5165
5166 if (!sctp_auth_enable)
5167 return -EACCES;
5168
5169 if (len < sizeof(struct sctp_authchunks))
5170 return -EINVAL;
5171
5172 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5173 return -EFAULT;
5174
5175 to = p->gauth_chunks;
5176 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5177 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5178 return -EINVAL;
5179
5180 if (asoc)
5181 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5182 else
5183 ch = sctp_sk(sk)->ep->auth_chunk_list;
5184
5185 if (!ch)
5186 goto num;
5187
5188 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5189 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5190 return -EINVAL;
5191
5192 if (copy_to_user(to, ch->chunks, num_chunks))
5193 return -EFAULT;
5194 num:
5195 len = sizeof(struct sctp_authchunks) + num_chunks;
5196 if (put_user(len, optlen))
5197 return -EFAULT;
5198 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5199 return -EFAULT;
5200
5201 return 0;
5202 }
5203
5204 /*
5205 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5206 * This option gets the current number of associations that are attached
5207 * to a one-to-many style socket. The option value is an uint32_t.
5208 */
5209 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5210 char __user *optval, int __user *optlen)
5211 {
5212 struct sctp_sock *sp = sctp_sk(sk);
5213 struct sctp_association *asoc;
5214 u32 val = 0;
5215
5216 if (sctp_style(sk, TCP))
5217 return -EOPNOTSUPP;
5218
5219 if (len < sizeof(u32))
5220 return -EINVAL;
5221
5222 len = sizeof(u32);
5223
5224 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5225 val++;
5226 }
5227
5228 if (put_user(len, optlen))
5229 return -EFAULT;
5230 if (copy_to_user(optval, &val, len))
5231 return -EFAULT;
5232
5233 return 0;
5234 }
5235
5236 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5237 char __user *optval, int __user *optlen)
5238 {
5239 int retval = 0;
5240 int len;
5241
5242 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5243 sk, optname);
5244
5245 /* I can hardly begin to describe how wrong this is. This is
5246 * so broken as to be worse than useless. The API draft
5247 * REALLY is NOT helpful here... I am not convinced that the
5248 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5249 * are at all well-founded.
5250 */
5251 if (level != SOL_SCTP) {
5252 struct sctp_af *af = sctp_sk(sk)->pf->af;
5253
5254 retval = af->getsockopt(sk, level, optname, optval, optlen);
5255 return retval;
5256 }
5257
5258 if (get_user(len, optlen))
5259 return -EFAULT;
5260
5261 sctp_lock_sock(sk);
5262
5263 switch (optname) {
5264 case SCTP_STATUS:
5265 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5266 break;
5267 case SCTP_DISABLE_FRAGMENTS:
5268 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5269 optlen);
5270 break;
5271 case SCTP_EVENTS:
5272 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5273 break;
5274 case SCTP_AUTOCLOSE:
5275 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5276 break;
5277 case SCTP_SOCKOPT_PEELOFF:
5278 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5279 break;
5280 case SCTP_PEER_ADDR_PARAMS:
5281 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5282 optlen);
5283 break;
5284 case SCTP_DELAYED_ACK:
5285 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5286 optlen);
5287 break;
5288 case SCTP_INITMSG:
5289 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5290 break;
5291 case SCTP_GET_PEER_ADDRS:
5292 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5293 optlen);
5294 break;
5295 case SCTP_GET_LOCAL_ADDRS:
5296 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5297 optlen);
5298 break;
5299 case SCTP_SOCKOPT_CONNECTX3:
5300 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5301 break;
5302 case SCTP_DEFAULT_SEND_PARAM:
5303 retval = sctp_getsockopt_default_send_param(sk, len,
5304 optval, optlen);
5305 break;
5306 case SCTP_PRIMARY_ADDR:
5307 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5308 break;
5309 case SCTP_NODELAY:
5310 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5311 break;
5312 case SCTP_RTOINFO:
5313 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5314 break;
5315 case SCTP_ASSOCINFO:
5316 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5317 break;
5318 case SCTP_I_WANT_MAPPED_V4_ADDR:
5319 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5320 break;
5321 case SCTP_MAXSEG:
5322 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5323 break;
5324 case SCTP_GET_PEER_ADDR_INFO:
5325 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5326 optlen);
5327 break;
5328 case SCTP_ADAPTATION_LAYER:
5329 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5330 optlen);
5331 break;
5332 case SCTP_CONTEXT:
5333 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5334 break;
5335 case SCTP_FRAGMENT_INTERLEAVE:
5336 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5337 optlen);
5338 break;
5339 case SCTP_PARTIAL_DELIVERY_POINT:
5340 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5341 optlen);
5342 break;
5343 case SCTP_MAX_BURST:
5344 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5345 break;
5346 case SCTP_AUTH_KEY:
5347 case SCTP_AUTH_CHUNK:
5348 case SCTP_AUTH_DELETE_KEY:
5349 retval = -EOPNOTSUPP;
5350 break;
5351 case SCTP_HMAC_IDENT:
5352 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5353 break;
5354 case SCTP_AUTH_ACTIVE_KEY:
5355 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5356 break;
5357 case SCTP_PEER_AUTH_CHUNKS:
5358 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5359 optlen);
5360 break;
5361 case SCTP_LOCAL_AUTH_CHUNKS:
5362 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5363 optlen);
5364 break;
5365 case SCTP_GET_ASSOC_NUMBER:
5366 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5367 break;
5368 default:
5369 retval = -ENOPROTOOPT;
5370 break;
5371 }
5372
5373 sctp_release_sock(sk);
5374 return retval;
5375 }
5376
5377 static void sctp_hash(struct sock *sk)
5378 {
5379 /* STUB */
5380 }
5381
5382 static void sctp_unhash(struct sock *sk)
5383 {
5384 /* STUB */
5385 }
5386
5387 /* Check if port is acceptable. Possibly find first available port.
5388 *
5389 * The port hash table (contained in the 'global' SCTP protocol storage
5390 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5391 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5392 * list (the list number is the port number hashed out, so as you
5393 * would expect from a hash function, all the ports in a given list have
5394 * such a number that hashes out to the same list number; you were
5395 * expecting that, right?); so each list has a set of ports, with a
5396 * link to the socket (struct sock) that uses it, the port number and
5397 * a fastreuse flag (FIXME: NPI ipg).
5398 */
5399 static struct sctp_bind_bucket *sctp_bucket_create(
5400 struct sctp_bind_hashbucket *head, unsigned short snum);
5401
5402 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5403 {
5404 struct sctp_bind_hashbucket *head; /* hash list */
5405 struct sctp_bind_bucket *pp; /* hash list port iterator */
5406 struct hlist_node *node;
5407 unsigned short snum;
5408 int ret;
5409
5410 snum = ntohs(addr->v4.sin_port);
5411
5412 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5413 sctp_local_bh_disable();
5414
5415 if (snum == 0) {
5416 /* Search for an available port. */
5417 int low, high, remaining, index;
5418 unsigned int rover;
5419
5420 inet_get_local_port_range(&low, &high);
5421 remaining = (high - low) + 1;
5422 rover = net_random() % remaining + low;
5423
5424 do {
5425 rover++;
5426 if ((rover < low) || (rover > high))
5427 rover = low;
5428 if (inet_is_reserved_local_port(rover))
5429 continue;
5430 index = sctp_phashfn(rover);
5431 head = &sctp_port_hashtable[index];
5432 sctp_spin_lock(&head->lock);
5433 sctp_for_each_hentry(pp, node, &head->chain)
5434 if (pp->port == rover)
5435 goto next;
5436 break;
5437 next:
5438 sctp_spin_unlock(&head->lock);
5439 } while (--remaining > 0);
5440
5441 /* Exhausted local port range during search? */
5442 ret = 1;
5443 if (remaining <= 0)
5444 goto fail;
5445
5446 /* OK, here is the one we will use. HEAD (the port
5447 * hash table list entry) is non-NULL and we hold it's
5448 * mutex.
5449 */
5450 snum = rover;
5451 } else {
5452 /* We are given an specific port number; we verify
5453 * that it is not being used. If it is used, we will
5454 * exahust the search in the hash list corresponding
5455 * to the port number (snum) - we detect that with the
5456 * port iterator, pp being NULL.
5457 */
5458 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5459 sctp_spin_lock(&head->lock);
5460 sctp_for_each_hentry(pp, node, &head->chain) {
5461 if (pp->port == snum)
5462 goto pp_found;
5463 }
5464 }
5465 pp = NULL;
5466 goto pp_not_found;
5467 pp_found:
5468 if (!hlist_empty(&pp->owner)) {
5469 /* We had a port hash table hit - there is an
5470 * available port (pp != NULL) and it is being
5471 * used by other socket (pp->owner not empty); that other
5472 * socket is going to be sk2.
5473 */
5474 int reuse = sk->sk_reuse;
5475 struct sock *sk2;
5476
5477 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5478 if (pp->fastreuse && sk->sk_reuse &&
5479 sk->sk_state != SCTP_SS_LISTENING)
5480 goto success;
5481
5482 /* Run through the list of sockets bound to the port
5483 * (pp->port) [via the pointers bind_next and
5484 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5485 * we get the endpoint they describe and run through
5486 * the endpoint's list of IP (v4 or v6) addresses,
5487 * comparing each of the addresses with the address of
5488 * the socket sk. If we find a match, then that means
5489 * that this port/socket (sk) combination are already
5490 * in an endpoint.
5491 */
5492 sk_for_each_bound(sk2, node, &pp->owner) {
5493 struct sctp_endpoint *ep2;
5494 ep2 = sctp_sk(sk2)->ep;
5495
5496 if (sk == sk2 ||
5497 (reuse && sk2->sk_reuse &&
5498 sk2->sk_state != SCTP_SS_LISTENING))
5499 continue;
5500
5501 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5502 sctp_sk(sk2), sctp_sk(sk))) {
5503 ret = (long)sk2;
5504 goto fail_unlock;
5505 }
5506 }
5507 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5508 }
5509 pp_not_found:
5510 /* If there was a hash table miss, create a new port. */
5511 ret = 1;
5512 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5513 goto fail_unlock;
5514
5515 /* In either case (hit or miss), make sure fastreuse is 1 only
5516 * if sk->sk_reuse is too (that is, if the caller requested
5517 * SO_REUSEADDR on this socket -sk-).
5518 */
5519 if (hlist_empty(&pp->owner)) {
5520 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5521 pp->fastreuse = 1;
5522 else
5523 pp->fastreuse = 0;
5524 } else if (pp->fastreuse &&
5525 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5526 pp->fastreuse = 0;
5527
5528 /* We are set, so fill up all the data in the hash table
5529 * entry, tie the socket list information with the rest of the
5530 * sockets FIXME: Blurry, NPI (ipg).
5531 */
5532 success:
5533 if (!sctp_sk(sk)->bind_hash) {
5534 inet_sk(sk)->inet_num = snum;
5535 sk_add_bind_node(sk, &pp->owner);
5536 sctp_sk(sk)->bind_hash = pp;
5537 }
5538 ret = 0;
5539
5540 fail_unlock:
5541 sctp_spin_unlock(&head->lock);
5542
5543 fail:
5544 sctp_local_bh_enable();
5545 return ret;
5546 }
5547
5548 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5549 * port is requested.
5550 */
5551 static int sctp_get_port(struct sock *sk, unsigned short snum)
5552 {
5553 long ret;
5554 union sctp_addr addr;
5555 struct sctp_af *af = sctp_sk(sk)->pf->af;
5556
5557 /* Set up a dummy address struct from the sk. */
5558 af->from_sk(&addr, sk);
5559 addr.v4.sin_port = htons(snum);
5560
5561 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5562 ret = sctp_get_port_local(sk, &addr);
5563
5564 return (ret ? 1 : 0);
5565 }
5566
5567 /*
5568 * Move a socket to LISTENING state.
5569 */
5570 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5571 {
5572 struct sctp_sock *sp = sctp_sk(sk);
5573 struct sctp_endpoint *ep = sp->ep;
5574 struct crypto_hash *tfm = NULL;
5575
5576 /* Allocate HMAC for generating cookie. */
5577 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5578 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5579 if (IS_ERR(tfm)) {
5580 if (net_ratelimit()) {
5581 pr_info("failed to load transform for %s: %ld\n",
5582 sctp_hmac_alg, PTR_ERR(tfm));
5583 }
5584 return -ENOSYS;
5585 }
5586 sctp_sk(sk)->hmac = tfm;
5587 }
5588
5589 /*
5590 * If a bind() or sctp_bindx() is not called prior to a listen()
5591 * call that allows new associations to be accepted, the system
5592 * picks an ephemeral port and will choose an address set equivalent
5593 * to binding with a wildcard address.
5594 *
5595 * This is not currently spelled out in the SCTP sockets
5596 * extensions draft, but follows the practice as seen in TCP
5597 * sockets.
5598 *
5599 */
5600 sk->sk_state = SCTP_SS_LISTENING;
5601 if (!ep->base.bind_addr.port) {
5602 if (sctp_autobind(sk))
5603 return -EAGAIN;
5604 } else {
5605 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5606 sk->sk_state = SCTP_SS_CLOSED;
5607 return -EADDRINUSE;
5608 }
5609 }
5610
5611 sk->sk_max_ack_backlog = backlog;
5612 sctp_hash_endpoint(ep);
5613 return 0;
5614 }
5615
5616 /*
5617 * 4.1.3 / 5.1.3 listen()
5618 *
5619 * By default, new associations are not accepted for UDP style sockets.
5620 * An application uses listen() to mark a socket as being able to
5621 * accept new associations.
5622 *
5623 * On TCP style sockets, applications use listen() to ready the SCTP
5624 * endpoint for accepting inbound associations.
5625 *
5626 * On both types of endpoints a backlog of '0' disables listening.
5627 *
5628 * Move a socket to LISTENING state.
5629 */
5630 int sctp_inet_listen(struct socket *sock, int backlog)
5631 {
5632 struct sock *sk = sock->sk;
5633 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5634 int err = -EINVAL;
5635
5636 if (unlikely(backlog < 0))
5637 return err;
5638
5639 sctp_lock_sock(sk);
5640
5641 /* Peeled-off sockets are not allowed to listen(). */
5642 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5643 goto out;
5644
5645 if (sock->state != SS_UNCONNECTED)
5646 goto out;
5647
5648 /* If backlog is zero, disable listening. */
5649 if (!backlog) {
5650 if (sctp_sstate(sk, CLOSED))
5651 goto out;
5652
5653 err = 0;
5654 sctp_unhash_endpoint(ep);
5655 sk->sk_state = SCTP_SS_CLOSED;
5656 if (sk->sk_reuse)
5657 sctp_sk(sk)->bind_hash->fastreuse = 1;
5658 goto out;
5659 }
5660
5661 /* If we are already listening, just update the backlog */
5662 if (sctp_sstate(sk, LISTENING))
5663 sk->sk_max_ack_backlog = backlog;
5664 else {
5665 err = sctp_listen_start(sk, backlog);
5666 if (err)
5667 goto out;
5668 }
5669
5670 err = 0;
5671 out:
5672 sctp_release_sock(sk);
5673 return err;
5674 }
5675
5676 /*
5677 * This function is done by modeling the current datagram_poll() and the
5678 * tcp_poll(). Note that, based on these implementations, we don't
5679 * lock the socket in this function, even though it seems that,
5680 * ideally, locking or some other mechanisms can be used to ensure
5681 * the integrity of the counters (sndbuf and wmem_alloc) used
5682 * in this place. We assume that we don't need locks either until proven
5683 * otherwise.
5684 *
5685 * Another thing to note is that we include the Async I/O support
5686 * here, again, by modeling the current TCP/UDP code. We don't have
5687 * a good way to test with it yet.
5688 */
5689 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5690 {
5691 struct sock *sk = sock->sk;
5692 struct sctp_sock *sp = sctp_sk(sk);
5693 unsigned int mask;
5694
5695 poll_wait(file, sk_sleep(sk), wait);
5696
5697 /* A TCP-style listening socket becomes readable when the accept queue
5698 * is not empty.
5699 */
5700 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5701 return (!list_empty(&sp->ep->asocs)) ?
5702 (POLLIN | POLLRDNORM) : 0;
5703
5704 mask = 0;
5705
5706 /* Is there any exceptional events? */
5707 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5708 mask |= POLLERR;
5709 if (sk->sk_shutdown & RCV_SHUTDOWN)
5710 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5711 if (sk->sk_shutdown == SHUTDOWN_MASK)
5712 mask |= POLLHUP;
5713
5714 /* Is it readable? Reconsider this code with TCP-style support. */
5715 if (!skb_queue_empty(&sk->sk_receive_queue))
5716 mask |= POLLIN | POLLRDNORM;
5717
5718 /* The association is either gone or not ready. */
5719 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5720 return mask;
5721
5722 /* Is it writable? */
5723 if (sctp_writeable(sk)) {
5724 mask |= POLLOUT | POLLWRNORM;
5725 } else {
5726 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5727 /*
5728 * Since the socket is not locked, the buffer
5729 * might be made available after the writeable check and
5730 * before the bit is set. This could cause a lost I/O
5731 * signal. tcp_poll() has a race breaker for this race
5732 * condition. Based on their implementation, we put
5733 * in the following code to cover it as well.
5734 */
5735 if (sctp_writeable(sk))
5736 mask |= POLLOUT | POLLWRNORM;
5737 }
5738 return mask;
5739 }
5740
5741 /********************************************************************
5742 * 2nd Level Abstractions
5743 ********************************************************************/
5744
5745 static struct sctp_bind_bucket *sctp_bucket_create(
5746 struct sctp_bind_hashbucket *head, unsigned short snum)
5747 {
5748 struct sctp_bind_bucket *pp;
5749
5750 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5751 if (pp) {
5752 SCTP_DBG_OBJCNT_INC(bind_bucket);
5753 pp->port = snum;
5754 pp->fastreuse = 0;
5755 INIT_HLIST_HEAD(&pp->owner);
5756 hlist_add_head(&pp->node, &head->chain);
5757 }
5758 return pp;
5759 }
5760
5761 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5762 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5763 {
5764 if (pp && hlist_empty(&pp->owner)) {
5765 __hlist_del(&pp->node);
5766 kmem_cache_free(sctp_bucket_cachep, pp);
5767 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5768 }
5769 }
5770
5771 /* Release this socket's reference to a local port. */
5772 static inline void __sctp_put_port(struct sock *sk)
5773 {
5774 struct sctp_bind_hashbucket *head =
5775 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5776 struct sctp_bind_bucket *pp;
5777
5778 sctp_spin_lock(&head->lock);
5779 pp = sctp_sk(sk)->bind_hash;
5780 __sk_del_bind_node(sk);
5781 sctp_sk(sk)->bind_hash = NULL;
5782 inet_sk(sk)->inet_num = 0;
5783 sctp_bucket_destroy(pp);
5784 sctp_spin_unlock(&head->lock);
5785 }
5786
5787 void sctp_put_port(struct sock *sk)
5788 {
5789 sctp_local_bh_disable();
5790 __sctp_put_port(sk);
5791 sctp_local_bh_enable();
5792 }
5793
5794 /*
5795 * The system picks an ephemeral port and choose an address set equivalent
5796 * to binding with a wildcard address.
5797 * One of those addresses will be the primary address for the association.
5798 * This automatically enables the multihoming capability of SCTP.
5799 */
5800 static int sctp_autobind(struct sock *sk)
5801 {
5802 union sctp_addr autoaddr;
5803 struct sctp_af *af;
5804 __be16 port;
5805
5806 /* Initialize a local sockaddr structure to INADDR_ANY. */
5807 af = sctp_sk(sk)->pf->af;
5808
5809 port = htons(inet_sk(sk)->inet_num);
5810 af->inaddr_any(&autoaddr, port);
5811
5812 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5813 }
5814
5815 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5816 *
5817 * From RFC 2292
5818 * 4.2 The cmsghdr Structure *
5819 *
5820 * When ancillary data is sent or received, any number of ancillary data
5821 * objects can be specified by the msg_control and msg_controllen members of
5822 * the msghdr structure, because each object is preceded by
5823 * a cmsghdr structure defining the object's length (the cmsg_len member).
5824 * Historically Berkeley-derived implementations have passed only one object
5825 * at a time, but this API allows multiple objects to be
5826 * passed in a single call to sendmsg() or recvmsg(). The following example
5827 * shows two ancillary data objects in a control buffer.
5828 *
5829 * |<--------------------------- msg_controllen -------------------------->|
5830 * | |
5831 *
5832 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5833 *
5834 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5835 * | | |
5836 *
5837 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5838 *
5839 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5840 * | | | | |
5841 *
5842 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5843 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5844 *
5845 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5846 *
5847 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5848 * ^
5849 * |
5850 *
5851 * msg_control
5852 * points here
5853 */
5854 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5855 sctp_cmsgs_t *cmsgs)
5856 {
5857 struct cmsghdr *cmsg;
5858 struct msghdr *my_msg = (struct msghdr *)msg;
5859
5860 for (cmsg = CMSG_FIRSTHDR(msg);
5861 cmsg != NULL;
5862 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5863 if (!CMSG_OK(my_msg, cmsg))
5864 return -EINVAL;
5865
5866 /* Should we parse this header or ignore? */
5867 if (cmsg->cmsg_level != IPPROTO_SCTP)
5868 continue;
5869
5870 /* Strictly check lengths following example in SCM code. */
5871 switch (cmsg->cmsg_type) {
5872 case SCTP_INIT:
5873 /* SCTP Socket API Extension
5874 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5875 *
5876 * This cmsghdr structure provides information for
5877 * initializing new SCTP associations with sendmsg().
5878 * The SCTP_INITMSG socket option uses this same data
5879 * structure. This structure is not used for
5880 * recvmsg().
5881 *
5882 * cmsg_level cmsg_type cmsg_data[]
5883 * ------------ ------------ ----------------------
5884 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5885 */
5886 if (cmsg->cmsg_len !=
5887 CMSG_LEN(sizeof(struct sctp_initmsg)))
5888 return -EINVAL;
5889 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5890 break;
5891
5892 case SCTP_SNDRCV:
5893 /* SCTP Socket API Extension
5894 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5895 *
5896 * This cmsghdr structure specifies SCTP options for
5897 * sendmsg() and describes SCTP header information
5898 * about a received message through recvmsg().
5899 *
5900 * cmsg_level cmsg_type cmsg_data[]
5901 * ------------ ------------ ----------------------
5902 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5903 */
5904 if (cmsg->cmsg_len !=
5905 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5906 return -EINVAL;
5907
5908 cmsgs->info =
5909 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5910
5911 /* Minimally, validate the sinfo_flags. */
5912 if (cmsgs->info->sinfo_flags &
5913 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5914 SCTP_ABORT | SCTP_EOF))
5915 return -EINVAL;
5916 break;
5917
5918 default:
5919 return -EINVAL;
5920 }
5921 }
5922 return 0;
5923 }
5924
5925 /*
5926 * Wait for a packet..
5927 * Note: This function is the same function as in core/datagram.c
5928 * with a few modifications to make lksctp work.
5929 */
5930 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5931 {
5932 int error;
5933 DEFINE_WAIT(wait);
5934
5935 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
5936
5937 /* Socket errors? */
5938 error = sock_error(sk);
5939 if (error)
5940 goto out;
5941
5942 if (!skb_queue_empty(&sk->sk_receive_queue))
5943 goto ready;
5944
5945 /* Socket shut down? */
5946 if (sk->sk_shutdown & RCV_SHUTDOWN)
5947 goto out;
5948
5949 /* Sequenced packets can come disconnected. If so we report the
5950 * problem.
5951 */
5952 error = -ENOTCONN;
5953
5954 /* Is there a good reason to think that we may receive some data? */
5955 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5956 goto out;
5957
5958 /* Handle signals. */
5959 if (signal_pending(current))
5960 goto interrupted;
5961
5962 /* Let another process have a go. Since we are going to sleep
5963 * anyway. Note: This may cause odd behaviors if the message
5964 * does not fit in the user's buffer, but this seems to be the
5965 * only way to honor MSG_DONTWAIT realistically.
5966 */
5967 sctp_release_sock(sk);
5968 *timeo_p = schedule_timeout(*timeo_p);
5969 sctp_lock_sock(sk);
5970
5971 ready:
5972 finish_wait(sk_sleep(sk), &wait);
5973 return 0;
5974
5975 interrupted:
5976 error = sock_intr_errno(*timeo_p);
5977
5978 out:
5979 finish_wait(sk_sleep(sk), &wait);
5980 *err = error;
5981 return error;
5982 }
5983
5984 /* Receive a datagram.
5985 * Note: This is pretty much the same routine as in core/datagram.c
5986 * with a few changes to make lksctp work.
5987 */
5988 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5989 int noblock, int *err)
5990 {
5991 int error;
5992 struct sk_buff *skb;
5993 long timeo;
5994
5995 timeo = sock_rcvtimeo(sk, noblock);
5996
5997 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5998 timeo, MAX_SCHEDULE_TIMEOUT);
5999
6000 do {
6001 /* Again only user level code calls this function,
6002 * so nothing interrupt level
6003 * will suddenly eat the receive_queue.
6004 *
6005 * Look at current nfs client by the way...
6006 * However, this function was corrent in any case. 8)
6007 */
6008 if (flags & MSG_PEEK) {
6009 spin_lock_bh(&sk->sk_receive_queue.lock);
6010 skb = skb_peek(&sk->sk_receive_queue);
6011 if (skb)
6012 atomic_inc(&skb->users);
6013 spin_unlock_bh(&sk->sk_receive_queue.lock);
6014 } else {
6015 skb = skb_dequeue(&sk->sk_receive_queue);
6016 }
6017
6018 if (skb)
6019 return skb;
6020
6021 /* Caller is allowed not to check sk->sk_err before calling. */
6022 error = sock_error(sk);
6023 if (error)
6024 goto no_packet;
6025
6026 if (sk->sk_shutdown & RCV_SHUTDOWN)
6027 break;
6028
6029 /* User doesn't want to wait. */
6030 error = -EAGAIN;
6031 if (!timeo)
6032 goto no_packet;
6033 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6034
6035 return NULL;
6036
6037 no_packet:
6038 *err = error;
6039 return NULL;
6040 }
6041
6042 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6043 static void __sctp_write_space(struct sctp_association *asoc)
6044 {
6045 struct sock *sk = asoc->base.sk;
6046 struct socket *sock = sk->sk_socket;
6047
6048 if ((sctp_wspace(asoc) > 0) && sock) {
6049 if (waitqueue_active(&asoc->wait))
6050 wake_up_interruptible(&asoc->wait);
6051
6052 if (sctp_writeable(sk)) {
6053 if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
6054 wake_up_interruptible(sk_sleep(sk));
6055
6056 /* Note that we try to include the Async I/O support
6057 * here by modeling from the current TCP/UDP code.
6058 * We have not tested with it yet.
6059 */
6060 if (sock->wq->fasync_list &&
6061 !(sk->sk_shutdown & SEND_SHUTDOWN))
6062 sock_wake_async(sock,
6063 SOCK_WAKE_SPACE, POLL_OUT);
6064 }
6065 }
6066 }
6067
6068 /* Do accounting for the sndbuf space.
6069 * Decrement the used sndbuf space of the corresponding association by the
6070 * data size which was just transmitted(freed).
6071 */
6072 static void sctp_wfree(struct sk_buff *skb)
6073 {
6074 struct sctp_association *asoc;
6075 struct sctp_chunk *chunk;
6076 struct sock *sk;
6077
6078 /* Get the saved chunk pointer. */
6079 chunk = *((struct sctp_chunk **)(skb->cb));
6080 asoc = chunk->asoc;
6081 sk = asoc->base.sk;
6082 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6083 sizeof(struct sk_buff) +
6084 sizeof(struct sctp_chunk);
6085
6086 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6087
6088 /*
6089 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6090 */
6091 sk->sk_wmem_queued -= skb->truesize;
6092 sk_mem_uncharge(sk, skb->truesize);
6093
6094 sock_wfree(skb);
6095 __sctp_write_space(asoc);
6096
6097 sctp_association_put(asoc);
6098 }
6099
6100 /* Do accounting for the receive space on the socket.
6101 * Accounting for the association is done in ulpevent.c
6102 * We set this as a destructor for the cloned data skbs so that
6103 * accounting is done at the correct time.
6104 */
6105 void sctp_sock_rfree(struct sk_buff *skb)
6106 {
6107 struct sock *sk = skb->sk;
6108 struct sctp_ulpevent *event = sctp_skb2event(skb);
6109
6110 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6111
6112 /*
6113 * Mimic the behavior of sock_rfree
6114 */
6115 sk_mem_uncharge(sk, event->rmem_len);
6116 }
6117
6118
6119 /* Helper function to wait for space in the sndbuf. */
6120 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6121 size_t msg_len)
6122 {
6123 struct sock *sk = asoc->base.sk;
6124 int err = 0;
6125 long current_timeo = *timeo_p;
6126 DEFINE_WAIT(wait);
6127
6128 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6129 asoc, (long)(*timeo_p), msg_len);
6130
6131 /* Increment the association's refcnt. */
6132 sctp_association_hold(asoc);
6133
6134 /* Wait on the association specific sndbuf space. */
6135 for (;;) {
6136 prepare_to_wait_exclusive(&asoc->wait, &wait,
6137 TASK_INTERRUPTIBLE);
6138 if (!*timeo_p)
6139 goto do_nonblock;
6140 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6141 asoc->base.dead)
6142 goto do_error;
6143 if (signal_pending(current))
6144 goto do_interrupted;
6145 if (msg_len <= sctp_wspace(asoc))
6146 break;
6147
6148 /* Let another process have a go. Since we are going
6149 * to sleep anyway.
6150 */
6151 sctp_release_sock(sk);
6152 current_timeo = schedule_timeout(current_timeo);
6153 BUG_ON(sk != asoc->base.sk);
6154 sctp_lock_sock(sk);
6155
6156 *timeo_p = current_timeo;
6157 }
6158
6159 out:
6160 finish_wait(&asoc->wait, &wait);
6161
6162 /* Release the association's refcnt. */
6163 sctp_association_put(asoc);
6164
6165 return err;
6166
6167 do_error:
6168 err = -EPIPE;
6169 goto out;
6170
6171 do_interrupted:
6172 err = sock_intr_errno(*timeo_p);
6173 goto out;
6174
6175 do_nonblock:
6176 err = -EAGAIN;
6177 goto out;
6178 }
6179
6180 void sctp_data_ready(struct sock *sk, int len)
6181 {
6182 struct socket_wq *wq;
6183
6184 rcu_read_lock();
6185 wq = rcu_dereference(sk->sk_wq);
6186 if (wq_has_sleeper(wq))
6187 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6188 POLLRDNORM | POLLRDBAND);
6189 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6190 rcu_read_unlock();
6191 }
6192
6193 /* If socket sndbuf has changed, wake up all per association waiters. */
6194 void sctp_write_space(struct sock *sk)
6195 {
6196 struct sctp_association *asoc;
6197
6198 /* Wake up the tasks in each wait queue. */
6199 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6200 __sctp_write_space(asoc);
6201 }
6202 }
6203
6204 /* Is there any sndbuf space available on the socket?
6205 *
6206 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6207 * associations on the same socket. For a UDP-style socket with
6208 * multiple associations, it is possible for it to be "unwriteable"
6209 * prematurely. I assume that this is acceptable because
6210 * a premature "unwriteable" is better than an accidental "writeable" which
6211 * would cause an unwanted block under certain circumstances. For the 1-1
6212 * UDP-style sockets or TCP-style sockets, this code should work.
6213 * - Daisy
6214 */
6215 static int sctp_writeable(struct sock *sk)
6216 {
6217 int amt = 0;
6218
6219 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6220 if (amt < 0)
6221 amt = 0;
6222 return amt;
6223 }
6224
6225 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6226 * returns immediately with EINPROGRESS.
6227 */
6228 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6229 {
6230 struct sock *sk = asoc->base.sk;
6231 int err = 0;
6232 long current_timeo = *timeo_p;
6233 DEFINE_WAIT(wait);
6234
6235 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6236 (long)(*timeo_p));
6237
6238 /* Increment the association's refcnt. */
6239 sctp_association_hold(asoc);
6240
6241 for (;;) {
6242 prepare_to_wait_exclusive(&asoc->wait, &wait,
6243 TASK_INTERRUPTIBLE);
6244 if (!*timeo_p)
6245 goto do_nonblock;
6246 if (sk->sk_shutdown & RCV_SHUTDOWN)
6247 break;
6248 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6249 asoc->base.dead)
6250 goto do_error;
6251 if (signal_pending(current))
6252 goto do_interrupted;
6253
6254 if (sctp_state(asoc, ESTABLISHED))
6255 break;
6256
6257 /* Let another process have a go. Since we are going
6258 * to sleep anyway.
6259 */
6260 sctp_release_sock(sk);
6261 current_timeo = schedule_timeout(current_timeo);
6262 sctp_lock_sock(sk);
6263
6264 *timeo_p = current_timeo;
6265 }
6266
6267 out:
6268 finish_wait(&asoc->wait, &wait);
6269
6270 /* Release the association's refcnt. */
6271 sctp_association_put(asoc);
6272
6273 return err;
6274
6275 do_error:
6276 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6277 err = -ETIMEDOUT;
6278 else
6279 err = -ECONNREFUSED;
6280 goto out;
6281
6282 do_interrupted:
6283 err = sock_intr_errno(*timeo_p);
6284 goto out;
6285
6286 do_nonblock:
6287 err = -EINPROGRESS;
6288 goto out;
6289 }
6290
6291 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6292 {
6293 struct sctp_endpoint *ep;
6294 int err = 0;
6295 DEFINE_WAIT(wait);
6296
6297 ep = sctp_sk(sk)->ep;
6298
6299
6300 for (;;) {
6301 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6302 TASK_INTERRUPTIBLE);
6303
6304 if (list_empty(&ep->asocs)) {
6305 sctp_release_sock(sk);
6306 timeo = schedule_timeout(timeo);
6307 sctp_lock_sock(sk);
6308 }
6309
6310 err = -EINVAL;
6311 if (!sctp_sstate(sk, LISTENING))
6312 break;
6313
6314 err = 0;
6315 if (!list_empty(&ep->asocs))
6316 break;
6317
6318 err = sock_intr_errno(timeo);
6319 if (signal_pending(current))
6320 break;
6321
6322 err = -EAGAIN;
6323 if (!timeo)
6324 break;
6325 }
6326
6327 finish_wait(sk_sleep(sk), &wait);
6328
6329 return err;
6330 }
6331
6332 static void sctp_wait_for_close(struct sock *sk, long timeout)
6333 {
6334 DEFINE_WAIT(wait);
6335
6336 do {
6337 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6338 if (list_empty(&sctp_sk(sk)->ep->asocs))
6339 break;
6340 sctp_release_sock(sk);
6341 timeout = schedule_timeout(timeout);
6342 sctp_lock_sock(sk);
6343 } while (!signal_pending(current) && timeout);
6344
6345 finish_wait(sk_sleep(sk), &wait);
6346 }
6347
6348 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6349 {
6350 struct sk_buff *frag;
6351
6352 if (!skb->data_len)
6353 goto done;
6354
6355 /* Don't forget the fragments. */
6356 skb_walk_frags(skb, frag)
6357 sctp_skb_set_owner_r_frag(frag, sk);
6358
6359 done:
6360 sctp_skb_set_owner_r(skb, sk);
6361 }
6362
6363 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6364 struct sctp_association *asoc)
6365 {
6366 struct inet_sock *inet = inet_sk(sk);
6367 struct inet_sock *newinet;
6368
6369 newsk->sk_type = sk->sk_type;
6370 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6371 newsk->sk_flags = sk->sk_flags;
6372 newsk->sk_no_check = sk->sk_no_check;
6373 newsk->sk_reuse = sk->sk_reuse;
6374
6375 newsk->sk_shutdown = sk->sk_shutdown;
6376 newsk->sk_destruct = inet_sock_destruct;
6377 newsk->sk_family = sk->sk_family;
6378 newsk->sk_protocol = IPPROTO_SCTP;
6379 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6380 newsk->sk_sndbuf = sk->sk_sndbuf;
6381 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6382 newsk->sk_lingertime = sk->sk_lingertime;
6383 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6384 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6385
6386 newinet = inet_sk(newsk);
6387
6388 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6389 * getsockname() and getpeername()
6390 */
6391 newinet->inet_sport = inet->inet_sport;
6392 newinet->inet_saddr = inet->inet_saddr;
6393 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6394 newinet->inet_dport = htons(asoc->peer.port);
6395 newinet->pmtudisc = inet->pmtudisc;
6396 newinet->inet_id = asoc->next_tsn ^ jiffies;
6397
6398 newinet->uc_ttl = inet->uc_ttl;
6399 newinet->mc_loop = 1;
6400 newinet->mc_ttl = 1;
6401 newinet->mc_index = 0;
6402 newinet->mc_list = NULL;
6403 }
6404
6405 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6406 * and its messages to the newsk.
6407 */
6408 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6409 struct sctp_association *assoc,
6410 sctp_socket_type_t type)
6411 {
6412 struct sctp_sock *oldsp = sctp_sk(oldsk);
6413 struct sctp_sock *newsp = sctp_sk(newsk);
6414 struct sctp_bind_bucket *pp; /* hash list port iterator */
6415 struct sctp_endpoint *newep = newsp->ep;
6416 struct sk_buff *skb, *tmp;
6417 struct sctp_ulpevent *event;
6418 struct sctp_bind_hashbucket *head;
6419
6420 /* Migrate socket buffer sizes and all the socket level options to the
6421 * new socket.
6422 */
6423 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6424 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6425 /* Brute force copy old sctp opt. */
6426 inet_sk_copy_descendant(newsk, oldsk);
6427
6428 /* Restore the ep value that was overwritten with the above structure
6429 * copy.
6430 */
6431 newsp->ep = newep;
6432 newsp->hmac = NULL;
6433
6434 /* Hook this new socket in to the bind_hash list. */
6435 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6436 sctp_local_bh_disable();
6437 sctp_spin_lock(&head->lock);
6438 pp = sctp_sk(oldsk)->bind_hash;
6439 sk_add_bind_node(newsk, &pp->owner);
6440 sctp_sk(newsk)->bind_hash = pp;
6441 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6442 sctp_spin_unlock(&head->lock);
6443 sctp_local_bh_enable();
6444
6445 /* Copy the bind_addr list from the original endpoint to the new
6446 * endpoint so that we can handle restarts properly
6447 */
6448 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6449 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6450
6451 /* Move any messages in the old socket's receive queue that are for the
6452 * peeled off association to the new socket's receive queue.
6453 */
6454 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6455 event = sctp_skb2event(skb);
6456 if (event->asoc == assoc) {
6457 __skb_unlink(skb, &oldsk->sk_receive_queue);
6458 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6459 sctp_skb_set_owner_r_frag(skb, newsk);
6460 }
6461 }
6462
6463 /* Clean up any messages pending delivery due to partial
6464 * delivery. Three cases:
6465 * 1) No partial deliver; no work.
6466 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6467 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6468 */
6469 skb_queue_head_init(&newsp->pd_lobby);
6470 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6471
6472 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6473 struct sk_buff_head *queue;
6474
6475 /* Decide which queue to move pd_lobby skbs to. */
6476 if (assoc->ulpq.pd_mode) {
6477 queue = &newsp->pd_lobby;
6478 } else
6479 queue = &newsk->sk_receive_queue;
6480
6481 /* Walk through the pd_lobby, looking for skbs that
6482 * need moved to the new socket.
6483 */
6484 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6485 event = sctp_skb2event(skb);
6486 if (event->asoc == assoc) {
6487 __skb_unlink(skb, &oldsp->pd_lobby);
6488 __skb_queue_tail(queue, skb);
6489 sctp_skb_set_owner_r_frag(skb, newsk);
6490 }
6491 }
6492
6493 /* Clear up any skbs waiting for the partial
6494 * delivery to finish.
6495 */
6496 if (assoc->ulpq.pd_mode)
6497 sctp_clear_pd(oldsk, NULL);
6498
6499 }
6500
6501 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6502 sctp_skb_set_owner_r_frag(skb, newsk);
6503
6504 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6505 sctp_skb_set_owner_r_frag(skb, newsk);
6506
6507 /* Set the type of socket to indicate that it is peeled off from the
6508 * original UDP-style socket or created with the accept() call on a
6509 * TCP-style socket..
6510 */
6511 newsp->type = type;
6512
6513 /* Mark the new socket "in-use" by the user so that any packets
6514 * that may arrive on the association after we've moved it are
6515 * queued to the backlog. This prevents a potential race between
6516 * backlog processing on the old socket and new-packet processing
6517 * on the new socket.
6518 *
6519 * The caller has just allocated newsk so we can guarantee that other
6520 * paths won't try to lock it and then oldsk.
6521 */
6522 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6523 sctp_assoc_migrate(assoc, newsk);
6524
6525 /* If the association on the newsk is already closed before accept()
6526 * is called, set RCV_SHUTDOWN flag.
6527 */
6528 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6529 newsk->sk_shutdown |= RCV_SHUTDOWN;
6530
6531 newsk->sk_state = SCTP_SS_ESTABLISHED;
6532 sctp_release_sock(newsk);
6533 }
6534
6535
6536 /* This proto struct describes the ULP interface for SCTP. */
6537 struct proto sctp_prot = {
6538 .name = "SCTP",
6539 .owner = THIS_MODULE,
6540 .close = sctp_close,
6541 .connect = sctp_connect,
6542 .disconnect = sctp_disconnect,
6543 .accept = sctp_accept,
6544 .ioctl = sctp_ioctl,
6545 .init = sctp_init_sock,
6546 .destroy = sctp_destroy_sock,
6547 .shutdown = sctp_shutdown,
6548 .setsockopt = sctp_setsockopt,
6549 .getsockopt = sctp_getsockopt,
6550 .sendmsg = sctp_sendmsg,
6551 .recvmsg = sctp_recvmsg,
6552 .bind = sctp_bind,
6553 .backlog_rcv = sctp_backlog_rcv,
6554 .hash = sctp_hash,
6555 .unhash = sctp_unhash,
6556 .get_port = sctp_get_port,
6557 .obj_size = sizeof(struct sctp_sock),
6558 .sysctl_mem = sysctl_sctp_mem,
6559 .sysctl_rmem = sysctl_sctp_rmem,
6560 .sysctl_wmem = sysctl_sctp_wmem,
6561 .memory_pressure = &sctp_memory_pressure,
6562 .enter_memory_pressure = sctp_enter_memory_pressure,
6563 .memory_allocated = &sctp_memory_allocated,
6564 .sockets_allocated = &sctp_sockets_allocated,
6565 };
6566
6567 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6568
6569 struct proto sctpv6_prot = {
6570 .name = "SCTPv6",
6571 .owner = THIS_MODULE,
6572 .close = sctp_close,
6573 .connect = sctp_connect,
6574 .disconnect = sctp_disconnect,
6575 .accept = sctp_accept,
6576 .ioctl = sctp_ioctl,
6577 .init = sctp_init_sock,
6578 .destroy = sctp_destroy_sock,
6579 .shutdown = sctp_shutdown,
6580 .setsockopt = sctp_setsockopt,
6581 .getsockopt = sctp_getsockopt,
6582 .sendmsg = sctp_sendmsg,
6583 .recvmsg = sctp_recvmsg,
6584 .bind = sctp_bind,
6585 .backlog_rcv = sctp_backlog_rcv,
6586 .hash = sctp_hash,
6587 .unhash = sctp_unhash,
6588 .get_port = sctp_get_port,
6589 .obj_size = sizeof(struct sctp6_sock),
6590 .sysctl_mem = sysctl_sctp_mem,
6591 .sysctl_rmem = sysctl_sctp_rmem,
6592 .sysctl_wmem = sysctl_sctp_wmem,
6593 .memory_pressure = &sctp_memory_pressure,
6594 .enter_memory_pressure = sctp_enter_memory_pressure,
6595 .memory_allocated = &sctp_memory_allocated,
6596 .sockets_allocated = &sctp_sockets_allocated,
6597 };
6598 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */