ia64/linux-2.6.18-xen.hg

view net/sctp/socket.c @ 871:9cbcc9008446

xen/x86: don't initialize cpu_data[]'s apicid field on generic code

Afaict, this is not only redundant with the intialization done in
drivers/xen/core/smpboot.c, but actually results - at least for
secondary CPUs - in the Xen-specific value written to be later
overwritten with whatever the generic code determines (with no
guarantee that the two values are identical).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
author Keir Fraser <keir.fraser@citrix.com>
date Thu May 14 10:09:15 2009 +0100 (2009-05-14)
parents 831230e53067
children
line source
1 /* SCTP kernel reference 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 reference 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 * The SCTP reference 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 * The SCTP reference 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 */
60 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
64 #include <linux/ip.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
71 #include <net/ip.h>
72 #include <net/icmp.h>
73 #include <net/route.h>
74 #include <net/ipv6.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
85 */
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110 extern kmem_cache_t *sctp_bucket_cachep;
112 /* Get the sndbuf space available at the time on the association. */
113 static inline int sctp_wspace(struct sctp_association *asoc)
114 {
115 struct sock *sk = asoc->base.sk;
116 int amt = 0;
118 if (asoc->ep->sndbuf_policy) {
119 /* make sure that no association uses more than sk_sndbuf */
120 amt = sk->sk_sndbuf - asoc->sndbuf_used;
121 } else {
122 /* do socket level accounting */
123 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
124 }
126 if (amt < 0)
127 amt = 0;
129 return amt;
130 }
132 /* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
135 *
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
139 * tracking.
140 */
141 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
142 {
143 struct sctp_association *asoc = chunk->asoc;
144 struct sock *sk = asoc->base.sk;
146 /* The sndbuf space is tracked per association. */
147 sctp_association_hold(asoc);
149 skb_set_owner_w(chunk->skb, sk);
151 chunk->skb->destructor = sctp_wfree;
152 /* Save the chunk pointer in skb for sctp_wfree to use later. */
153 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
155 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
156 sizeof(struct sk_buff) +
157 sizeof(struct sctp_chunk);
159 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
160 }
162 /* Verify that this is a valid address. */
163 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
164 int len)
165 {
166 struct sctp_af *af;
168 /* Verify basic sockaddr. */
169 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
170 if (!af)
171 return -EINVAL;
173 /* Is this a valid SCTP address? */
174 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
175 return -EINVAL;
177 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
178 return -EINVAL;
180 return 0;
181 }
183 /* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
185 */
186 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
187 {
188 struct sctp_association *asoc = NULL;
190 /* If this is not a UDP-style socket, assoc id should be ignored. */
191 if (!sctp_style(sk, UDP)) {
192 /* Return NULL if the socket state is not ESTABLISHED. It
193 * could be a TCP-style listening socket or a socket which
194 * hasn't yet called connect() to establish an association.
195 */
196 if (!sctp_sstate(sk, ESTABLISHED))
197 return NULL;
199 /* Get the first and the only association from the list. */
200 if (!list_empty(&sctp_sk(sk)->ep->asocs))
201 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
202 struct sctp_association, asocs);
203 return asoc;
204 }
206 /* Otherwise this is a UDP-style socket. */
207 if (!id || (id == (sctp_assoc_t)-1))
208 return NULL;
210 spin_lock_bh(&sctp_assocs_id_lock);
211 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
212 spin_unlock_bh(&sctp_assocs_id_lock);
214 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
215 return NULL;
217 return asoc;
218 }
220 /* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
222 * the same.
223 */
224 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
226 sctp_assoc_t id)
227 {
228 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
229 struct sctp_transport *transport;
230 union sctp_addr *laddr = (union sctp_addr *)addr;
232 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
233 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
234 (union sctp_addr *)addr,
235 &transport);
236 laddr->v4.sin_port = htons(laddr->v4.sin_port);
238 if (!addr_asoc)
239 return NULL;
241 id_asoc = sctp_id2assoc(sk, id);
242 if (id_asoc && (id_asoc != addr_asoc))
243 return NULL;
245 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
246 (union sctp_addr *)addr);
248 return transport;
249 }
251 /* API 3.1.2 bind() - UDP Style Syntax
252 * The syntax of bind() is,
253 *
254 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
255 *
256 * sd - the socket descriptor returned by socket().
257 * addr - the address structure (struct sockaddr_in or struct
258 * sockaddr_in6 [RFC 2553]),
259 * addr_len - the size of the address structure.
260 */
261 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
262 {
263 int retval = 0;
265 sctp_lock_sock(sk);
267 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
268 sk, addr, addr_len);
270 /* Disallow binding twice. */
271 if (!sctp_sk(sk)->ep->base.bind_addr.port)
272 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
273 addr_len);
274 else
275 retval = -EINVAL;
277 sctp_release_sock(sk);
279 return retval;
280 }
282 static long sctp_get_port_local(struct sock *, union sctp_addr *);
284 /* Verify this is a valid sockaddr. */
285 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
286 union sctp_addr *addr, int len)
287 {
288 struct sctp_af *af;
290 /* Check minimum size. */
291 if (len < sizeof (struct sockaddr))
292 return NULL;
294 /* Does this PF support this AF? */
295 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
296 return NULL;
298 /* If we get this far, af is valid. */
299 af = sctp_get_af_specific(addr->sa.sa_family);
301 if (len < af->sockaddr_len)
302 return NULL;
304 return af;
305 }
307 /* Bind a local address either to an endpoint or to an association. */
308 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
309 {
310 struct sctp_sock *sp = sctp_sk(sk);
311 struct sctp_endpoint *ep = sp->ep;
312 struct sctp_bind_addr *bp = &ep->base.bind_addr;
313 struct sctp_af *af;
314 unsigned short snum;
315 int ret = 0;
317 /* Common sockaddr verification. */
318 af = sctp_sockaddr_af(sp, addr, len);
319 if (!af) {
320 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
321 sk, addr, len);
322 return -EINVAL;
323 }
325 snum = ntohs(addr->v4.sin_port);
327 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
328 ", port: %d, new port: %d, len: %d)\n",
329 sk,
330 addr,
331 bp->port, snum,
332 len);
334 /* PF specific bind() address verification. */
335 if (!sp->pf->bind_verify(sp, addr))
336 return -EADDRNOTAVAIL;
338 /* We must either be unbound, or bind to the same port. */
339 if (bp->port && (snum != bp->port)) {
340 SCTP_DEBUG_PRINTK("sctp_do_bind:"
341 " New port %d does not match existing port "
342 "%d.\n", snum, bp->port);
343 return -EINVAL;
344 }
346 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
347 return -EACCES;
349 /* Make sure we are allowed to bind here.
350 * The function sctp_get_port_local() does duplicate address
351 * detection.
352 */
353 if ((ret = sctp_get_port_local(sk, addr))) {
354 if (ret == (long) sk) {
355 /* This endpoint has a conflicting address. */
356 return -EINVAL;
357 } else {
358 return -EADDRINUSE;
359 }
360 }
362 /* Refresh ephemeral port. */
363 if (!bp->port)
364 bp->port = inet_sk(sk)->num;
366 /* Add the address to the bind address list. */
367 sctp_local_bh_disable();
368 sctp_write_lock(&ep->base.addr_lock);
370 /* Use GFP_ATOMIC since BHs are disabled. */
371 addr->v4.sin_port = ntohs(addr->v4.sin_port);
372 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
373 addr->v4.sin_port = htons(addr->v4.sin_port);
374 sctp_write_unlock(&ep->base.addr_lock);
375 sctp_local_bh_enable();
377 /* Copy back into socket for getsockname() use. */
378 if (!ret) {
379 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
380 af->to_sk_saddr(addr, sk);
381 }
383 return ret;
384 }
386 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
387 *
388 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
389 * at any one time. If a sender, after sending an ASCONF chunk, decides
390 * it needs to transfer another ASCONF Chunk, it MUST wait until the
391 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
392 * subsequent ASCONF. Note this restriction binds each side, so at any
393 * time two ASCONF may be in-transit on any given association (one sent
394 * from each endpoint).
395 */
396 static int sctp_send_asconf(struct sctp_association *asoc,
397 struct sctp_chunk *chunk)
398 {
399 int retval = 0;
401 /* If there is an outstanding ASCONF chunk, queue it for later
402 * transmission.
403 */
404 if (asoc->addip_last_asconf) {
405 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
406 goto out;
407 }
409 /* Hold the chunk until an ASCONF_ACK is received. */
410 sctp_chunk_hold(chunk);
411 retval = sctp_primitive_ASCONF(asoc, chunk);
412 if (retval)
413 sctp_chunk_free(chunk);
414 else
415 asoc->addip_last_asconf = chunk;
417 out:
418 return retval;
419 }
421 /* Add a list of addresses as bind addresses to local endpoint or
422 * association.
423 *
424 * Basically run through each address specified in the addrs/addrcnt
425 * array/length pair, determine if it is IPv6 or IPv4 and call
426 * sctp_do_bind() on it.
427 *
428 * If any of them fails, then the operation will be reversed and the
429 * ones that were added will be removed.
430 *
431 * Only sctp_setsockopt_bindx() is supposed to call this function.
432 */
433 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
434 {
435 int cnt;
436 int retval = 0;
437 void *addr_buf;
438 struct sockaddr *sa_addr;
439 struct sctp_af *af;
441 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
442 sk, addrs, addrcnt);
444 addr_buf = addrs;
445 for (cnt = 0; cnt < addrcnt; cnt++) {
446 /* The list may contain either IPv4 or IPv6 address;
447 * determine the address length for walking thru the list.
448 */
449 sa_addr = (struct sockaddr *)addr_buf;
450 af = sctp_get_af_specific(sa_addr->sa_family);
451 if (!af) {
452 retval = -EINVAL;
453 goto err_bindx_add;
454 }
456 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
457 af->sockaddr_len);
459 addr_buf += af->sockaddr_len;
461 err_bindx_add:
462 if (retval < 0) {
463 /* Failed. Cleanup the ones that have been added */
464 if (cnt > 0)
465 sctp_bindx_rem(sk, addrs, cnt);
466 return retval;
467 }
468 }
470 return retval;
471 }
473 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
474 * associations that are part of the endpoint indicating that a list of local
475 * addresses are added to the endpoint.
476 *
477 * If any of the addresses is already in the bind address list of the
478 * association, we do not send the chunk for that association. But it will not
479 * affect other associations.
480 *
481 * Only sctp_setsockopt_bindx() is supposed to call this function.
482 */
483 static int sctp_send_asconf_add_ip(struct sock *sk,
484 struct sockaddr *addrs,
485 int addrcnt)
486 {
487 struct sctp_sock *sp;
488 struct sctp_endpoint *ep;
489 struct sctp_association *asoc;
490 struct sctp_bind_addr *bp;
491 struct sctp_chunk *chunk;
492 struct sctp_sockaddr_entry *laddr;
493 union sctp_addr *addr;
494 union sctp_addr saveaddr;
495 void *addr_buf;
496 struct sctp_af *af;
497 struct list_head *pos;
498 struct list_head *p;
499 int i;
500 int retval = 0;
502 if (!sctp_addip_enable)
503 return retval;
505 sp = sctp_sk(sk);
506 ep = sp->ep;
508 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
509 __FUNCTION__, sk, addrs, addrcnt);
511 list_for_each(pos, &ep->asocs) {
512 asoc = list_entry(pos, struct sctp_association, asocs);
514 if (!asoc->peer.asconf_capable)
515 continue;
517 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
518 continue;
520 if (!sctp_state(asoc, ESTABLISHED))
521 continue;
523 /* Check if any address in the packed array of addresses is
524 * in the bind address list of the association. If so,
525 * do not send the asconf chunk to its peer, but continue with
526 * other associations.
527 */
528 addr_buf = addrs;
529 for (i = 0; i < addrcnt; i++) {
530 addr = (union sctp_addr *)addr_buf;
531 af = sctp_get_af_specific(addr->v4.sin_family);
532 if (!af) {
533 retval = -EINVAL;
534 goto out;
535 }
537 if (sctp_assoc_lookup_laddr(asoc, addr))
538 break;
540 addr_buf += af->sockaddr_len;
541 }
542 if (i < addrcnt)
543 continue;
545 /* Use the first address in bind addr list of association as
546 * Address Parameter of ASCONF CHUNK.
547 */
548 sctp_read_lock(&asoc->base.addr_lock);
549 bp = &asoc->base.bind_addr;
550 p = bp->address_list.next;
551 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
552 sctp_read_unlock(&asoc->base.addr_lock);
554 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
555 addrcnt, SCTP_PARAM_ADD_IP);
556 if (!chunk) {
557 retval = -ENOMEM;
558 goto out;
559 }
561 retval = sctp_send_asconf(asoc, chunk);
562 if (retval)
563 goto out;
565 /* Add the new addresses to the bind address list with
566 * use_as_src set to 0.
567 */
568 sctp_local_bh_disable();
569 sctp_write_lock(&asoc->base.addr_lock);
570 addr_buf = addrs;
571 for (i = 0; i < addrcnt; i++) {
572 addr = (union sctp_addr *)addr_buf;
573 af = sctp_get_af_specific(addr->v4.sin_family);
574 memcpy(&saveaddr, addr, af->sockaddr_len);
575 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
576 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
577 GFP_ATOMIC);
578 addr_buf += af->sockaddr_len;
579 }
580 sctp_write_unlock(&asoc->base.addr_lock);
581 sctp_local_bh_enable();
582 }
584 out:
585 return retval;
586 }
588 /* Remove a list of addresses from bind addresses list. Do not remove the
589 * last address.
590 *
591 * Basically run through each address specified in the addrs/addrcnt
592 * array/length pair, determine if it is IPv6 or IPv4 and call
593 * sctp_del_bind() on it.
594 *
595 * If any of them fails, then the operation will be reversed and the
596 * ones that were removed will be added back.
597 *
598 * At least one address has to be left; if only one address is
599 * available, the operation will return -EBUSY.
600 *
601 * Only sctp_setsockopt_bindx() is supposed to call this function.
602 */
603 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
604 {
605 struct sctp_sock *sp = sctp_sk(sk);
606 struct sctp_endpoint *ep = sp->ep;
607 int cnt;
608 struct sctp_bind_addr *bp = &ep->base.bind_addr;
609 int retval = 0;
610 union sctp_addr saveaddr;
611 void *addr_buf;
612 struct sockaddr *sa_addr;
613 struct sctp_af *af;
615 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
616 sk, addrs, addrcnt);
618 addr_buf = addrs;
619 for (cnt = 0; cnt < addrcnt; cnt++) {
620 /* If the bind address list is empty or if there is only one
621 * bind address, there is nothing more to be removed (we need
622 * at least one address here).
623 */
624 if (list_empty(&bp->address_list) ||
625 (sctp_list_single_entry(&bp->address_list))) {
626 retval = -EBUSY;
627 goto err_bindx_rem;
628 }
630 /* The list may contain either IPv4 or IPv6 address;
631 * determine the address length to copy the address to
632 * saveaddr.
633 */
634 sa_addr = (struct sockaddr *)addr_buf;
635 af = sctp_get_af_specific(sa_addr->sa_family);
636 if (!af) {
637 retval = -EINVAL;
638 goto err_bindx_rem;
639 }
640 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
641 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
642 if (saveaddr.v4.sin_port != bp->port) {
643 retval = -EINVAL;
644 goto err_bindx_rem;
645 }
647 /* FIXME - There is probably a need to check if sk->sk_saddr and
648 * sk->sk_rcv_addr are currently set to one of the addresses to
649 * be removed. This is something which needs to be looked into
650 * when we are fixing the outstanding issues with multi-homing
651 * socket routing and failover schemes. Refer to comments in
652 * sctp_do_bind(). -daisy
653 */
654 sctp_local_bh_disable();
655 sctp_write_lock(&ep->base.addr_lock);
657 retval = sctp_del_bind_addr(bp, &saveaddr);
659 sctp_write_unlock(&ep->base.addr_lock);
660 sctp_local_bh_enable();
662 addr_buf += af->sockaddr_len;
663 err_bindx_rem:
664 if (retval < 0) {
665 /* Failed. Add the ones that has been removed back */
666 if (cnt > 0)
667 sctp_bindx_add(sk, addrs, cnt);
668 return retval;
669 }
670 }
672 return retval;
673 }
675 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
676 * the associations that are part of the endpoint indicating that a list of
677 * local addresses are removed from the endpoint.
678 *
679 * If any of the addresses is already in the bind address list of the
680 * association, we do not send the chunk for that association. But it will not
681 * affect other associations.
682 *
683 * Only sctp_setsockopt_bindx() is supposed to call this function.
684 */
685 static int sctp_send_asconf_del_ip(struct sock *sk,
686 struct sockaddr *addrs,
687 int addrcnt)
688 {
689 struct sctp_sock *sp;
690 struct sctp_endpoint *ep;
691 struct sctp_association *asoc;
692 struct sctp_transport *transport;
693 struct sctp_bind_addr *bp;
694 struct sctp_chunk *chunk;
695 union sctp_addr *laddr;
696 union sctp_addr saveaddr;
697 void *addr_buf;
698 struct sctp_af *af;
699 struct list_head *pos, *pos1;
700 struct sctp_sockaddr_entry *saddr;
701 int i;
702 int retval = 0;
704 if (!sctp_addip_enable)
705 return retval;
707 sp = sctp_sk(sk);
708 ep = sp->ep;
710 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 __FUNCTION__, sk, addrs, addrcnt);
713 list_for_each(pos, &ep->asocs) {
714 asoc = list_entry(pos, struct sctp_association, asocs);
716 if (!asoc->peer.asconf_capable)
717 continue;
719 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
720 continue;
722 if (!sctp_state(asoc, ESTABLISHED))
723 continue;
725 /* Check if any address in the packed array of addresses is
726 * not present in the bind address list of the association.
727 * If so, do not send the asconf chunk to its peer, but
728 * continue with other associations.
729 */
730 addr_buf = addrs;
731 for (i = 0; i < addrcnt; i++) {
732 laddr = (union sctp_addr *)addr_buf;
733 af = sctp_get_af_specific(laddr->v4.sin_family);
734 if (!af) {
735 retval = -EINVAL;
736 goto out;
737 }
739 if (!sctp_assoc_lookup_laddr(asoc, laddr))
740 break;
742 addr_buf += af->sockaddr_len;
743 }
744 if (i < addrcnt)
745 continue;
747 /* Find one address in the association's bind address list
748 * that is not in the packed array of addresses. This is to
749 * make sure that we do not delete all the addresses in the
750 * association.
751 */
752 sctp_read_lock(&asoc->base.addr_lock);
753 bp = &asoc->base.bind_addr;
754 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
755 addrcnt, sp);
756 sctp_read_unlock(&asoc->base.addr_lock);
757 if (!laddr)
758 continue;
760 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
761 SCTP_PARAM_DEL_IP);
762 if (!chunk) {
763 retval = -ENOMEM;
764 goto out;
765 }
767 /* Reset use_as_src flag for the addresses in the bind address
768 * list that are to be deleted.
769 */
770 sctp_local_bh_disable();
771 sctp_write_lock(&asoc->base.addr_lock);
772 addr_buf = addrs;
773 for (i = 0; i < addrcnt; i++) {
774 laddr = (union sctp_addr *)addr_buf;
775 af = sctp_get_af_specific(laddr->v4.sin_family);
776 memcpy(&saveaddr, laddr, af->sockaddr_len);
777 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
778 list_for_each(pos1, &bp->address_list) {
779 saddr = list_entry(pos1,
780 struct sctp_sockaddr_entry,
781 list);
782 if (sctp_cmp_addr_exact(&saddr->a, &saveaddr))
783 saddr->use_as_src = 0;
784 }
785 addr_buf += af->sockaddr_len;
786 }
787 sctp_write_unlock(&asoc->base.addr_lock);
788 sctp_local_bh_enable();
790 /* Update the route and saddr entries for all the transports
791 * as some of the addresses in the bind address list are
792 * about to be deleted and cannot be used as source addresses.
793 */
794 list_for_each(pos1, &asoc->peer.transport_addr_list) {
795 transport = list_entry(pos1, struct sctp_transport,
796 transports);
797 dst_release(transport->dst);
798 sctp_transport_route(transport, NULL,
799 sctp_sk(asoc->base.sk));
800 }
802 retval = sctp_send_asconf(asoc, chunk);
803 }
804 out:
805 return retval;
806 }
808 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809 *
810 * API 8.1
811 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
812 * int flags);
813 *
814 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
815 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
816 * or IPv6 addresses.
817 *
818 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
819 * Section 3.1.2 for this usage.
820 *
821 * addrs is a pointer to an array of one or more socket addresses. Each
822 * address is contained in its appropriate structure (i.e. struct
823 * sockaddr_in or struct sockaddr_in6) the family of the address type
824 * must be used to distengish the address length (note that this
825 * representation is termed a "packed array" of addresses). The caller
826 * specifies the number of addresses in the array with addrcnt.
827 *
828 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
829 * -1, and sets errno to the appropriate error code.
830 *
831 * For SCTP, the port given in each socket address must be the same, or
832 * sctp_bindx() will fail, setting errno to EINVAL.
833 *
834 * The flags parameter is formed from the bitwise OR of zero or more of
835 * the following currently defined flags:
836 *
837 * SCTP_BINDX_ADD_ADDR
838 *
839 * SCTP_BINDX_REM_ADDR
840 *
841 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
842 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
843 * addresses from the association. The two flags are mutually exclusive;
844 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
845 * not remove all addresses from an association; sctp_bindx() will
846 * reject such an attempt with EINVAL.
847 *
848 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
849 * additional addresses with an endpoint after calling bind(). Or use
850 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
851 * socket is associated with so that no new association accepted will be
852 * associated with those addresses. If the endpoint supports dynamic
853 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
854 * endpoint to send the appropriate message to the peer to change the
855 * peers address lists.
856 *
857 * Adding and removing addresses from a connected association is
858 * optional functionality. Implementations that do not support this
859 * functionality should return EOPNOTSUPP.
860 *
861 * Basically do nothing but copying the addresses from user to kernel
862 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
863 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
864 * from userspace.
865 *
866 * We don't use copy_from_user() for optimization: we first do the
867 * sanity checks (buffer size -fast- and access check-healthy
868 * pointer); if all of those succeed, then we can alloc the memory
869 * (expensive operation) needed to copy the data to kernel. Then we do
870 * the copying without checking the user space area
871 * (__copy_from_user()).
872 *
873 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
874 * it.
875 *
876 * sk The sk of the socket
877 * addrs The pointer to the addresses in user land
878 * addrssize Size of the addrs buffer
879 * op Operation to perform (add or remove, see the flags of
880 * sctp_bindx)
881 *
882 * Returns 0 if ok, <0 errno code on error.
883 */
884 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
885 struct sockaddr __user *addrs,
886 int addrs_size, int op)
887 {
888 struct sockaddr *kaddrs;
889 int err;
890 int addrcnt = 0;
891 int walk_size = 0;
892 struct sockaddr *sa_addr;
893 void *addr_buf;
894 struct sctp_af *af;
896 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
897 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
899 if (unlikely(addrs_size <= 0))
900 return -EINVAL;
902 /* Check the user passed a healthy pointer. */
903 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
904 return -EFAULT;
906 /* Alloc space for the address array in kernel memory. */
907 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
908 if (unlikely(!kaddrs))
909 return -ENOMEM;
911 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
912 kfree(kaddrs);
913 return -EFAULT;
914 }
916 /* Walk through the addrs buffer and count the number of addresses. */
917 addr_buf = kaddrs;
918 while (walk_size < addrs_size) {
919 sa_addr = (struct sockaddr *)addr_buf;
920 af = sctp_get_af_specific(sa_addr->sa_family);
922 /* If the address family is not supported or if this address
923 * causes the address buffer to overflow return EINVAL.
924 */
925 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
926 kfree(kaddrs);
927 return -EINVAL;
928 }
929 addrcnt++;
930 addr_buf += af->sockaddr_len;
931 walk_size += af->sockaddr_len;
932 }
934 /* Do the work. */
935 switch (op) {
936 case SCTP_BINDX_ADD_ADDR:
937 err = sctp_bindx_add(sk, kaddrs, addrcnt);
938 if (err)
939 goto out;
940 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
941 break;
943 case SCTP_BINDX_REM_ADDR:
944 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
945 if (err)
946 goto out;
947 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
948 break;
950 default:
951 err = -EINVAL;
952 break;
953 };
955 out:
956 kfree(kaddrs);
958 return err;
959 }
961 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962 *
963 * Common routine for handling connect() and sctp_connectx().
964 * Connect will come in with just a single address.
965 */
966 static int __sctp_connect(struct sock* sk,
967 struct sockaddr *kaddrs,
968 int addrs_size)
969 {
970 struct sctp_sock *sp;
971 struct sctp_endpoint *ep;
972 struct sctp_association *asoc = NULL;
973 struct sctp_association *asoc2;
974 struct sctp_transport *transport;
975 union sctp_addr to;
976 struct sctp_af *af;
977 sctp_scope_t scope;
978 long timeo;
979 int err = 0;
980 int addrcnt = 0;
981 int walk_size = 0;
982 struct sockaddr *sa_addr;
983 void *addr_buf;
985 sp = sctp_sk(sk);
986 ep = sp->ep;
988 /* connect() cannot be done on a socket that is already in ESTABLISHED
989 * state - UDP-style peeled off socket or a TCP-style socket that
990 * is already connected.
991 * It cannot be done even on a TCP-style listening socket.
992 */
993 if (sctp_sstate(sk, ESTABLISHED) ||
994 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
995 err = -EISCONN;
996 goto out_free;
997 }
999 /* Walk through the addrs buffer and count the number of addresses. */
1000 addr_buf = kaddrs;
1001 while (walk_size < addrs_size) {
1002 sa_addr = (struct sockaddr *)addr_buf;
1003 af = sctp_get_af_specific(sa_addr->sa_family);
1005 /* If the address family is not supported or if this address
1006 * causes the address buffer to overflow return EINVAL.
1007 */
1008 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1009 err = -EINVAL;
1010 goto out_free;
1013 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
1014 af->sockaddr_len);
1015 if (err)
1016 goto out_free;
1018 memcpy(&to, sa_addr, af->sockaddr_len);
1019 to.v4.sin_port = ntohs(to.v4.sin_port);
1021 /* Check if there already is a matching association on the
1022 * endpoint (other than the one created here).
1023 */
1024 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1025 if (asoc2 && asoc2 != asoc) {
1026 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1027 err = -EISCONN;
1028 else
1029 err = -EALREADY;
1030 goto out_free;
1033 /* If we could not find a matching association on the endpoint,
1034 * make sure that there is no peeled-off association matching
1035 * the peer address even on another socket.
1036 */
1037 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1038 err = -EADDRNOTAVAIL;
1039 goto out_free;
1042 if (!asoc) {
1043 /* If a bind() or sctp_bindx() is not called prior to
1044 * an sctp_connectx() call, the system picks an
1045 * ephemeral port and will choose an address set
1046 * equivalent to binding with a wildcard address.
1047 */
1048 if (!ep->base.bind_addr.port) {
1049 if (sctp_autobind(sk)) {
1050 err = -EAGAIN;
1051 goto out_free;
1053 } else {
1054 /*
1055 * If an unprivileged user inherits a 1-many
1056 * style socket with open associations on a
1057 * privileged port, it MAY be permitted to
1058 * accept new associations, but it SHOULD NOT
1059 * be permitted to open new associations.
1060 */
1061 if (ep->base.bind_addr.port < PROT_SOCK &&
1062 !capable(CAP_NET_BIND_SERVICE)) {
1063 err = -EACCES;
1064 goto out_free;
1068 scope = sctp_scope(&to);
1069 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1070 if (!asoc) {
1071 err = -ENOMEM;
1072 goto out_free;
1076 /* Prime the peer's transport structures. */
1077 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1078 SCTP_UNKNOWN);
1079 if (!transport) {
1080 err = -ENOMEM;
1081 goto out_free;
1084 addrcnt++;
1085 addr_buf += af->sockaddr_len;
1086 walk_size += af->sockaddr_len;
1089 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1090 if (err < 0) {
1091 goto out_free;
1094 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1095 if (err < 0) {
1096 goto out_free;
1099 /* Initialize sk's dport and daddr for getpeername() */
1100 inet_sk(sk)->dport = htons(asoc->peer.port);
1101 af = sctp_get_af_specific(to.sa.sa_family);
1102 af->to_sk_daddr(&to, sk);
1103 sk->sk_err = 0;
1105 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1106 err = sctp_wait_for_connect(asoc, &timeo);
1108 /* Don't free association on exit. */
1109 asoc = NULL;
1111 out_free:
1113 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1114 " kaddrs: %p err: %d\n",
1115 asoc, kaddrs, err);
1116 if (asoc)
1117 sctp_association_free(asoc);
1118 return err;
1121 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1123 * API 8.9
1124 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1126 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1127 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1128 * or IPv6 addresses.
1130 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1131 * Section 3.1.2 for this usage.
1133 * addrs is a pointer to an array of one or more socket addresses. Each
1134 * address is contained in its appropriate structure (i.e. struct
1135 * sockaddr_in or struct sockaddr_in6) the family of the address type
1136 * must be used to distengish the address length (note that this
1137 * representation is termed a "packed array" of addresses). The caller
1138 * specifies the number of addresses in the array with addrcnt.
1140 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1141 * -1, and sets errno to the appropriate error code.
1143 * For SCTP, the port given in each socket address must be the same, or
1144 * sctp_connectx() will fail, setting errno to EINVAL.
1146 * An application can use sctp_connectx to initiate an association with
1147 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1148 * allows a caller to specify multiple addresses at which a peer can be
1149 * reached. The way the SCTP stack uses the list of addresses to set up
1150 * the association is implementation dependant. This function only
1151 * specifies that the stack will try to make use of all the addresses in
1152 * the list when needed.
1154 * Note that the list of addresses passed in is only used for setting up
1155 * the association. It does not necessarily equal the set of addresses
1156 * the peer uses for the resulting association. If the caller wants to
1157 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1158 * retrieve them after the association has been set up.
1160 * Basically do nothing but copying the addresses from user to kernel
1161 * land and invoking either sctp_connectx(). This is used for tunneling
1162 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1164 * We don't use copy_from_user() for optimization: we first do the
1165 * sanity checks (buffer size -fast- and access check-healthy
1166 * pointer); if all of those succeed, then we can alloc the memory
1167 * (expensive operation) needed to copy the data to kernel. Then we do
1168 * the copying without checking the user space area
1169 * (__copy_from_user()).
1171 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1172 * it.
1174 * sk The sk of the socket
1175 * addrs The pointer to the addresses in user land
1176 * addrssize Size of the addrs buffer
1178 * Returns 0 if ok, <0 errno code on error.
1179 */
1180 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1181 struct sockaddr __user *addrs,
1182 int addrs_size)
1184 int err = 0;
1185 struct sockaddr *kaddrs;
1187 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1188 __FUNCTION__, sk, addrs, addrs_size);
1190 if (unlikely(addrs_size <= 0))
1191 return -EINVAL;
1193 /* Check the user passed a healthy pointer. */
1194 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1195 return -EFAULT;
1197 /* Alloc space for the address array in kernel memory. */
1198 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1199 if (unlikely(!kaddrs))
1200 return -ENOMEM;
1202 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1203 err = -EFAULT;
1204 } else {
1205 err = __sctp_connect(sk, kaddrs, addrs_size);
1208 kfree(kaddrs);
1209 return err;
1212 /* API 3.1.4 close() - UDP Style Syntax
1213 * Applications use close() to perform graceful shutdown (as described in
1214 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1215 * by a UDP-style socket.
1217 * The syntax is
1219 * ret = close(int sd);
1221 * sd - the socket descriptor of the associations to be closed.
1223 * To gracefully shutdown a specific association represented by the
1224 * UDP-style socket, an application should use the sendmsg() call,
1225 * passing no user data, but including the appropriate flag in the
1226 * ancillary data (see Section xxxx).
1228 * If sd in the close() call is a branched-off socket representing only
1229 * one association, the shutdown is performed on that association only.
1231 * 4.1.6 close() - TCP Style Syntax
1233 * Applications use close() to gracefully close down an association.
1235 * The syntax is:
1237 * int close(int sd);
1239 * sd - the socket descriptor of the association to be closed.
1241 * After an application calls close() on a socket descriptor, no further
1242 * socket operations will succeed on that descriptor.
1244 * API 7.1.4 SO_LINGER
1246 * An application using the TCP-style socket can use this option to
1247 * perform the SCTP ABORT primitive. The linger option structure is:
1249 * struct linger {
1250 * int l_onoff; // option on/off
1251 * int l_linger; // linger time
1252 * };
1254 * To enable the option, set l_onoff to 1. If the l_linger value is set
1255 * to 0, calling close() is the same as the ABORT primitive. If the
1256 * value is set to a negative value, the setsockopt() call will return
1257 * an error. If the value is set to a positive value linger_time, the
1258 * close() can be blocked for at most linger_time ms. If the graceful
1259 * shutdown phase does not finish during this period, close() will
1260 * return but the graceful shutdown phase continues in the system.
1261 */
1262 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1264 struct sctp_endpoint *ep;
1265 struct sctp_association *asoc;
1266 struct list_head *pos, *temp;
1268 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1270 sctp_lock_sock(sk);
1271 sk->sk_shutdown = SHUTDOWN_MASK;
1273 ep = sctp_sk(sk)->ep;
1275 /* Walk all associations on an endpoint. */
1276 list_for_each_safe(pos, temp, &ep->asocs) {
1277 asoc = list_entry(pos, struct sctp_association, asocs);
1279 if (sctp_style(sk, TCP)) {
1280 /* A closed association can still be in the list if
1281 * it belongs to a TCP-style listening socket that is
1282 * not yet accepted. If so, free it. If not, send an
1283 * ABORT or SHUTDOWN based on the linger options.
1284 */
1285 if (sctp_state(asoc, CLOSED)) {
1286 sctp_unhash_established(asoc);
1287 sctp_association_free(asoc);
1288 continue;
1292 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1293 struct sctp_chunk *chunk;
1295 chunk = sctp_make_abort_user(asoc, NULL, 0);
1296 if (chunk)
1297 sctp_primitive_ABORT(asoc, chunk);
1298 } else
1299 sctp_primitive_SHUTDOWN(asoc, NULL);
1302 /* Clean up any skbs sitting on the receive queue. */
1303 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1304 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1306 /* On a TCP-style socket, block for at most linger_time if set. */
1307 if (sctp_style(sk, TCP) && timeout)
1308 sctp_wait_for_close(sk, timeout);
1310 /* This will run the backlog queue. */
1311 sctp_release_sock(sk);
1313 /* Supposedly, no process has access to the socket, but
1314 * the net layers still may.
1315 */
1316 sctp_local_bh_disable();
1317 sctp_bh_lock_sock(sk);
1319 /* Hold the sock, since sk_common_release() will put sock_put()
1320 * and we have just a little more cleanup.
1321 */
1322 sock_hold(sk);
1323 sk_common_release(sk);
1325 sctp_bh_unlock_sock(sk);
1326 sctp_local_bh_enable();
1328 sock_put(sk);
1330 SCTP_DBG_OBJCNT_DEC(sock);
1333 /* Handle EPIPE error. */
1334 static int sctp_error(struct sock *sk, int flags, int err)
1336 if (err == -EPIPE)
1337 err = sock_error(sk) ? : -EPIPE;
1338 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1339 send_sig(SIGPIPE, current, 0);
1340 return err;
1343 /* API 3.1.3 sendmsg() - UDP Style Syntax
1345 * An application uses sendmsg() and recvmsg() calls to transmit data to
1346 * and receive data from its peer.
1348 * ssize_t sendmsg(int socket, const struct msghdr *message,
1349 * int flags);
1351 * socket - the socket descriptor of the endpoint.
1352 * message - pointer to the msghdr structure which contains a single
1353 * user message and possibly some ancillary data.
1355 * See Section 5 for complete description of the data
1356 * structures.
1358 * flags - flags sent or received with the user message, see Section
1359 * 5 for complete description of the flags.
1361 * Note: This function could use a rewrite especially when explicit
1362 * connect support comes in.
1363 */
1364 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1366 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1368 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1369 struct msghdr *msg, size_t msg_len)
1371 struct sctp_sock *sp;
1372 struct sctp_endpoint *ep;
1373 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1374 struct sctp_transport *transport, *chunk_tp;
1375 struct sctp_chunk *chunk;
1376 union sctp_addr to;
1377 struct sockaddr *msg_name = NULL;
1378 struct sctp_sndrcvinfo default_sinfo = { 0 };
1379 struct sctp_sndrcvinfo *sinfo;
1380 struct sctp_initmsg *sinit;
1381 sctp_assoc_t associd = 0;
1382 sctp_cmsgs_t cmsgs = { NULL };
1383 int err;
1384 sctp_scope_t scope;
1385 long timeo;
1386 __u16 sinfo_flags = 0;
1387 struct sctp_datamsg *datamsg;
1388 struct list_head *pos;
1389 int msg_flags = msg->msg_flags;
1391 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1392 sk, msg, msg_len);
1394 err = 0;
1395 sp = sctp_sk(sk);
1396 ep = sp->ep;
1398 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1400 /* We cannot send a message over a TCP-style listening socket. */
1401 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1402 err = -EPIPE;
1403 goto out_nounlock;
1406 /* Parse out the SCTP CMSGs. */
1407 err = sctp_msghdr_parse(msg, &cmsgs);
1409 if (err) {
1410 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1411 goto out_nounlock;
1414 /* Fetch the destination address for this packet. This
1415 * address only selects the association--it is not necessarily
1416 * the address we will send to.
1417 * For a peeled-off socket, msg_name is ignored.
1418 */
1419 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1420 int msg_namelen = msg->msg_namelen;
1422 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1423 msg_namelen);
1424 if (err)
1425 return err;
1427 if (msg_namelen > sizeof(to))
1428 msg_namelen = sizeof(to);
1429 memcpy(&to, msg->msg_name, msg_namelen);
1430 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1431 "0x%x:%u.\n",
1432 to.v4.sin_addr.s_addr, to.v4.sin_port);
1434 to.v4.sin_port = ntohs(to.v4.sin_port);
1435 msg_name = msg->msg_name;
1438 sinfo = cmsgs.info;
1439 sinit = cmsgs.init;
1441 /* Did the user specify SNDRCVINFO? */
1442 if (sinfo) {
1443 sinfo_flags = sinfo->sinfo_flags;
1444 associd = sinfo->sinfo_assoc_id;
1447 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1448 msg_len, sinfo_flags);
1450 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1451 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1452 err = -EINVAL;
1453 goto out_nounlock;
1456 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1457 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1458 * If SCTP_ABORT is set, the message length could be non zero with
1459 * the msg_iov set to the user abort reason.
1460 */
1461 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1462 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1463 err = -EINVAL;
1464 goto out_nounlock;
1467 /* If SCTP_ADDR_OVER is set, there must be an address
1468 * specified in msg_name.
1469 */
1470 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1471 err = -EINVAL;
1472 goto out_nounlock;
1475 transport = NULL;
1477 SCTP_DEBUG_PRINTK("About to look up association.\n");
1479 sctp_lock_sock(sk);
1481 /* If a msg_name has been specified, assume this is to be used. */
1482 if (msg_name) {
1483 /* Look for a matching association on the endpoint. */
1484 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1485 if (!asoc) {
1486 /* If we could not find a matching association on the
1487 * endpoint, make sure that it is not a TCP-style
1488 * socket that already has an association or there is
1489 * no peeled-off association on another socket.
1490 */
1491 if ((sctp_style(sk, TCP) &&
1492 sctp_sstate(sk, ESTABLISHED)) ||
1493 sctp_endpoint_is_peeled_off(ep, &to)) {
1494 err = -EADDRNOTAVAIL;
1495 goto out_unlock;
1498 } else {
1499 asoc = sctp_id2assoc(sk, associd);
1500 if (!asoc) {
1501 err = -EPIPE;
1502 goto out_unlock;
1506 if (asoc) {
1507 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1509 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1510 * socket that has an association in CLOSED state. This can
1511 * happen when an accepted socket has an association that is
1512 * already CLOSED.
1513 */
1514 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1515 err = -EPIPE;
1516 goto out_unlock;
1519 if (sinfo_flags & SCTP_EOF) {
1520 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1521 asoc);
1522 sctp_primitive_SHUTDOWN(asoc, NULL);
1523 err = 0;
1524 goto out_unlock;
1526 if (sinfo_flags & SCTP_ABORT) {
1527 struct sctp_chunk *chunk;
1529 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1530 if (!chunk) {
1531 err = -ENOMEM;
1532 goto out_unlock;
1535 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1536 sctp_primitive_ABORT(asoc, chunk);
1537 err = 0;
1538 goto out_unlock;
1542 /* Do we need to create the association? */
1543 if (!asoc) {
1544 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1546 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1547 err = -EINVAL;
1548 goto out_unlock;
1551 /* Check for invalid stream against the stream counts,
1552 * either the default or the user specified stream counts.
1553 */
1554 if (sinfo) {
1555 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1556 /* Check against the defaults. */
1557 if (sinfo->sinfo_stream >=
1558 sp->initmsg.sinit_num_ostreams) {
1559 err = -EINVAL;
1560 goto out_unlock;
1562 } else {
1563 /* Check against the requested. */
1564 if (sinfo->sinfo_stream >=
1565 sinit->sinit_num_ostreams) {
1566 err = -EINVAL;
1567 goto out_unlock;
1572 /*
1573 * API 3.1.2 bind() - UDP Style Syntax
1574 * If a bind() or sctp_bindx() is not called prior to a
1575 * sendmsg() call that initiates a new association, the
1576 * system picks an ephemeral port and will choose an address
1577 * set equivalent to binding with a wildcard address.
1578 */
1579 if (!ep->base.bind_addr.port) {
1580 if (sctp_autobind(sk)) {
1581 err = -EAGAIN;
1582 goto out_unlock;
1584 } else {
1585 /*
1586 * If an unprivileged user inherits a one-to-many
1587 * style socket with open associations on a privileged
1588 * port, it MAY be permitted to accept new associations,
1589 * but it SHOULD NOT be permitted to open new
1590 * associations.
1591 */
1592 if (ep->base.bind_addr.port < PROT_SOCK &&
1593 !capable(CAP_NET_BIND_SERVICE)) {
1594 err = -EACCES;
1595 goto out_unlock;
1599 scope = sctp_scope(&to);
1600 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1601 if (!new_asoc) {
1602 err = -ENOMEM;
1603 goto out_unlock;
1605 asoc = new_asoc;
1607 /* If the SCTP_INIT ancillary data is specified, set all
1608 * the association init values accordingly.
1609 */
1610 if (sinit) {
1611 if (sinit->sinit_num_ostreams) {
1612 asoc->c.sinit_num_ostreams =
1613 sinit->sinit_num_ostreams;
1615 if (sinit->sinit_max_instreams) {
1616 asoc->c.sinit_max_instreams =
1617 sinit->sinit_max_instreams;
1619 if (sinit->sinit_max_attempts) {
1620 asoc->max_init_attempts
1621 = sinit->sinit_max_attempts;
1623 if (sinit->sinit_max_init_timeo) {
1624 asoc->max_init_timeo =
1625 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1629 /* Prime the peer's transport structures. */
1630 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1631 if (!transport) {
1632 err = -ENOMEM;
1633 goto out_free;
1635 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1636 if (err < 0) {
1637 err = -ENOMEM;
1638 goto out_free;
1642 /* ASSERT: we have a valid association at this point. */
1643 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1645 if (!sinfo) {
1646 /* If the user didn't specify SNDRCVINFO, make up one with
1647 * some defaults.
1648 */
1649 default_sinfo.sinfo_stream = asoc->default_stream;
1650 default_sinfo.sinfo_flags = asoc->default_flags;
1651 default_sinfo.sinfo_ppid = asoc->default_ppid;
1652 default_sinfo.sinfo_context = asoc->default_context;
1653 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1654 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1655 sinfo = &default_sinfo;
1658 /* API 7.1.7, the sndbuf size per association bounds the
1659 * maximum size of data that can be sent in a single send call.
1660 */
1661 if (msg_len > sk->sk_sndbuf) {
1662 err = -EMSGSIZE;
1663 goto out_free;
1666 /* If fragmentation is disabled and the message length exceeds the
1667 * association fragmentation point, return EMSGSIZE. The I-D
1668 * does not specify what this error is, but this looks like
1669 * a great fit.
1670 */
1671 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1672 err = -EMSGSIZE;
1673 goto out_free;
1676 if (sinfo) {
1677 /* Check for invalid stream. */
1678 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1679 err = -EINVAL;
1680 goto out_free;
1684 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1685 if (!sctp_wspace(asoc)) {
1686 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1687 if (err)
1688 goto out_free;
1691 /* If an address is passed with the sendto/sendmsg call, it is used
1692 * to override the primary destination address in the TCP model, or
1693 * when SCTP_ADDR_OVER flag is set in the UDP model.
1694 */
1695 if ((sctp_style(sk, TCP) && msg_name) ||
1696 (sinfo_flags & SCTP_ADDR_OVER)) {
1697 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1698 if (!chunk_tp) {
1699 err = -EINVAL;
1700 goto out_free;
1702 } else
1703 chunk_tp = NULL;
1705 /* Auto-connect, if we aren't connected already. */
1706 if (sctp_state(asoc, CLOSED)) {
1707 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1708 if (err < 0)
1709 goto out_free;
1710 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1713 /* Break the message into multiple chunks of maximum size. */
1714 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1715 if (!datamsg) {
1716 err = -ENOMEM;
1717 goto out_free;
1720 /* Now send the (possibly) fragmented message. */
1721 list_for_each(pos, &datamsg->chunks) {
1722 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1723 sctp_datamsg_track(chunk);
1725 /* Do accounting for the write space. */
1726 sctp_set_owner_w(chunk);
1728 chunk->transport = chunk_tp;
1730 /* Send it to the lower layers. Note: all chunks
1731 * must either fail or succeed. The lower layer
1732 * works that way today. Keep it that way or this
1733 * breaks.
1734 */
1735 err = sctp_primitive_SEND(asoc, chunk);
1736 /* Did the lower layer accept the chunk? */
1737 if (err)
1738 sctp_chunk_free(chunk);
1739 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1742 sctp_datamsg_free(datamsg);
1743 if (err)
1744 goto out_free;
1745 else
1746 err = msg_len;
1748 /* If we are already past ASSOCIATE, the lower
1749 * layers are responsible for association cleanup.
1750 */
1751 goto out_unlock;
1753 out_free:
1754 if (new_asoc)
1755 sctp_association_free(asoc);
1756 out_unlock:
1757 sctp_release_sock(sk);
1759 out_nounlock:
1760 return sctp_error(sk, msg_flags, err);
1762 #if 0
1763 do_sock_err:
1764 if (msg_len)
1765 err = msg_len;
1766 else
1767 err = sock_error(sk);
1768 goto out;
1770 do_interrupted:
1771 if (msg_len)
1772 err = msg_len;
1773 goto out;
1774 #endif /* 0 */
1777 /* This is an extended version of skb_pull() that removes the data from the
1778 * start of a skb even when data is spread across the list of skb's in the
1779 * frag_list. len specifies the total amount of data that needs to be removed.
1780 * when 'len' bytes could be removed from the skb, it returns 0.
1781 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1782 * could not be removed.
1783 */
1784 static int sctp_skb_pull(struct sk_buff *skb, int len)
1786 struct sk_buff *list;
1787 int skb_len = skb_headlen(skb);
1788 int rlen;
1790 if (len <= skb_len) {
1791 __skb_pull(skb, len);
1792 return 0;
1794 len -= skb_len;
1795 __skb_pull(skb, skb_len);
1797 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1798 rlen = sctp_skb_pull(list, len);
1799 skb->len -= (len-rlen);
1800 skb->data_len -= (len-rlen);
1802 if (!rlen)
1803 return 0;
1805 len = rlen;
1808 return len;
1811 /* API 3.1.3 recvmsg() - UDP Style Syntax
1813 * ssize_t recvmsg(int socket, struct msghdr *message,
1814 * int flags);
1816 * socket - the socket descriptor of the endpoint.
1817 * message - pointer to the msghdr structure which contains a single
1818 * user message and possibly some ancillary data.
1820 * See Section 5 for complete description of the data
1821 * structures.
1823 * flags - flags sent or received with the user message, see Section
1824 * 5 for complete description of the flags.
1825 */
1826 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1828 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1829 struct msghdr *msg, size_t len, int noblock,
1830 int flags, int *addr_len)
1832 struct sctp_ulpevent *event = NULL;
1833 struct sctp_sock *sp = sctp_sk(sk);
1834 struct sk_buff *skb;
1835 int copied;
1836 int err = 0;
1837 int skb_len;
1839 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1840 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1841 "len", len, "knoblauch", noblock,
1842 "flags", flags, "addr_len", addr_len);
1844 sctp_lock_sock(sk);
1846 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1847 err = -ENOTCONN;
1848 goto out;
1851 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1852 if (!skb)
1853 goto out;
1855 /* Get the total length of the skb including any skb's in the
1856 * frag_list.
1857 */
1858 skb_len = skb->len;
1860 copied = skb_len;
1861 if (copied > len)
1862 copied = len;
1864 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1866 event = sctp_skb2event(skb);
1868 if (err)
1869 goto out_free;
1871 sock_recv_timestamp(msg, sk, skb);
1872 if (sctp_ulpevent_is_notification(event)) {
1873 msg->msg_flags |= MSG_NOTIFICATION;
1874 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1875 } else {
1876 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1879 /* Check if we allow SCTP_SNDRCVINFO. */
1880 if (sp->subscribe.sctp_data_io_event)
1881 sctp_ulpevent_read_sndrcvinfo(event, msg);
1882 #if 0
1883 /* FIXME: we should be calling IP/IPv6 layers. */
1884 if (sk->sk_protinfo.af_inet.cmsg_flags)
1885 ip_cmsg_recv(msg, skb);
1886 #endif
1888 err = copied;
1890 /* If skb's length exceeds the user's buffer, update the skb and
1891 * push it back to the receive_queue so that the next call to
1892 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1893 */
1894 if (skb_len > copied) {
1895 msg->msg_flags &= ~MSG_EOR;
1896 if (flags & MSG_PEEK)
1897 goto out_free;
1898 sctp_skb_pull(skb, copied);
1899 skb_queue_head(&sk->sk_receive_queue, skb);
1901 /* When only partial message is copied to the user, increase
1902 * rwnd by that amount. If all the data in the skb is read,
1903 * rwnd is updated when the event is freed.
1904 */
1905 sctp_assoc_rwnd_increase(event->asoc, copied);
1906 goto out;
1907 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1908 (event->msg_flags & MSG_EOR))
1909 msg->msg_flags |= MSG_EOR;
1910 else
1911 msg->msg_flags &= ~MSG_EOR;
1913 out_free:
1914 if (flags & MSG_PEEK) {
1915 /* Release the skb reference acquired after peeking the skb in
1916 * sctp_skb_recv_datagram().
1917 */
1918 kfree_skb(skb);
1919 } else {
1920 /* Free the event which includes releasing the reference to
1921 * the owner of the skb, freeing the skb and updating the
1922 * rwnd.
1923 */
1924 sctp_ulpevent_free(event);
1926 out:
1927 sctp_release_sock(sk);
1928 return err;
1931 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1933 * This option is a on/off flag. If enabled no SCTP message
1934 * fragmentation will be performed. Instead if a message being sent
1935 * exceeds the current PMTU size, the message will NOT be sent and
1936 * instead a error will be indicated to the user.
1937 */
1938 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1939 char __user *optval, int optlen)
1941 int val;
1943 if (optlen < sizeof(int))
1944 return -EINVAL;
1946 if (get_user(val, (int __user *)optval))
1947 return -EFAULT;
1949 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1951 return 0;
1954 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1955 int optlen)
1957 if (optlen != sizeof(struct sctp_event_subscribe))
1958 return -EINVAL;
1959 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1960 return -EFAULT;
1961 return 0;
1964 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1966 * This socket option is applicable to the UDP-style socket only. When
1967 * set it will cause associations that are idle for more than the
1968 * specified number of seconds to automatically close. An association
1969 * being idle is defined an association that has NOT sent or received
1970 * user data. The special value of '0' indicates that no automatic
1971 * close of any associations should be performed. The option expects an
1972 * integer defining the number of seconds of idle time before an
1973 * association is closed.
1974 */
1975 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1976 int optlen)
1978 struct sctp_sock *sp = sctp_sk(sk);
1980 /* Applicable to UDP-style socket only */
1981 if (sctp_style(sk, TCP))
1982 return -EOPNOTSUPP;
1983 if (optlen != sizeof(int))
1984 return -EINVAL;
1985 if (copy_from_user(&sp->autoclose, optval, optlen))
1986 return -EFAULT;
1988 return 0;
1991 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1993 * Applications can enable or disable heartbeats for any peer address of
1994 * an association, modify an address's heartbeat interval, force a
1995 * heartbeat to be sent immediately, and adjust the address's maximum
1996 * number of retransmissions sent before an address is considered
1997 * unreachable. The following structure is used to access and modify an
1998 * address's parameters:
2000 * struct sctp_paddrparams {
2001 * sctp_assoc_t spp_assoc_id;
2002 * struct sockaddr_storage spp_address;
2003 * uint32_t spp_hbinterval;
2004 * uint16_t spp_pathmaxrxt;
2005 * uint32_t spp_pathmtu;
2006 * uint32_t spp_sackdelay;
2007 * uint32_t spp_flags;
2008 * };
2010 * spp_assoc_id - (one-to-many style socket) This is filled in the
2011 * application, and identifies the association for
2012 * this query.
2013 * spp_address - This specifies which address is of interest.
2014 * spp_hbinterval - This contains the value of the heartbeat interval,
2015 * in milliseconds. If a value of zero
2016 * is present in this field then no changes are to
2017 * be made to this parameter.
2018 * spp_pathmaxrxt - This contains the maximum number of
2019 * retransmissions before this address shall be
2020 * considered unreachable. If a value of zero
2021 * is present in this field then no changes are to
2022 * be made to this parameter.
2023 * spp_pathmtu - When Path MTU discovery is disabled the value
2024 * specified here will be the "fixed" path mtu.
2025 * Note that if the spp_address field is empty
2026 * then all associations on this address will
2027 * have this fixed path mtu set upon them.
2029 * spp_sackdelay - When delayed sack is enabled, this value specifies
2030 * the number of milliseconds that sacks will be delayed
2031 * for. This value will apply to all addresses of an
2032 * association if the spp_address field is empty. Note
2033 * also, that if delayed sack is enabled and this
2034 * value is set to 0, no change is made to the last
2035 * recorded delayed sack timer value.
2037 * spp_flags - These flags are used to control various features
2038 * on an association. The flag field may contain
2039 * zero or more of the following options.
2041 * SPP_HB_ENABLE - Enable heartbeats on the
2042 * specified address. Note that if the address
2043 * field is empty all addresses for the association
2044 * have heartbeats enabled upon them.
2046 * SPP_HB_DISABLE - Disable heartbeats on the
2047 * speicifed address. Note that if the address
2048 * field is empty all addresses for the association
2049 * will have their heartbeats disabled. Note also
2050 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2051 * mutually exclusive, only one of these two should
2052 * be specified. Enabling both fields will have
2053 * undetermined results.
2055 * SPP_HB_DEMAND - Request a user initiated heartbeat
2056 * to be made immediately.
2058 * SPP_PMTUD_ENABLE - This field will enable PMTU
2059 * discovery upon the specified address. Note that
2060 * if the address feild is empty then all addresses
2061 * on the association are effected.
2063 * SPP_PMTUD_DISABLE - This field will disable PMTU
2064 * discovery upon the specified address. Note that
2065 * if the address feild is empty then all addresses
2066 * on the association are effected. Not also that
2067 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2068 * exclusive. Enabling both will have undetermined
2069 * results.
2071 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2072 * on delayed sack. The time specified in spp_sackdelay
2073 * is used to specify the sack delay for this address. Note
2074 * that if spp_address is empty then all addresses will
2075 * enable delayed sack and take on the sack delay
2076 * value specified in spp_sackdelay.
2077 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2078 * off delayed sack. If the spp_address field is blank then
2079 * delayed sack is disabled for the entire association. Note
2080 * also that this field is mutually exclusive to
2081 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2082 * results.
2083 */
2084 int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2085 struct sctp_transport *trans,
2086 struct sctp_association *asoc,
2087 struct sctp_sock *sp,
2088 int hb_change,
2089 int pmtud_change,
2090 int sackdelay_change)
2092 int error;
2094 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2095 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2096 if (error)
2097 return error;
2100 if (params->spp_hbinterval) {
2101 if (trans) {
2102 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2103 } else if (asoc) {
2104 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2105 } else {
2106 sp->hbinterval = params->spp_hbinterval;
2110 if (hb_change) {
2111 if (trans) {
2112 trans->param_flags =
2113 (trans->param_flags & ~SPP_HB) | hb_change;
2114 } else if (asoc) {
2115 asoc->param_flags =
2116 (asoc->param_flags & ~SPP_HB) | hb_change;
2117 } else {
2118 sp->param_flags =
2119 (sp->param_flags & ~SPP_HB) | hb_change;
2123 if (params->spp_pathmtu) {
2124 if (trans) {
2125 trans->pathmtu = params->spp_pathmtu;
2126 sctp_assoc_sync_pmtu(asoc);
2127 } else if (asoc) {
2128 asoc->pathmtu = params->spp_pathmtu;
2129 sctp_frag_point(sp, params->spp_pathmtu);
2130 } else {
2131 sp->pathmtu = params->spp_pathmtu;
2135 if (pmtud_change) {
2136 if (trans) {
2137 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2138 (params->spp_flags & SPP_PMTUD_ENABLE);
2139 trans->param_flags =
2140 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2141 if (update) {
2142 sctp_transport_pmtu(trans);
2143 sctp_assoc_sync_pmtu(asoc);
2145 } else if (asoc) {
2146 asoc->param_flags =
2147 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2148 } else {
2149 sp->param_flags =
2150 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2154 if (params->spp_sackdelay) {
2155 if (trans) {
2156 trans->sackdelay =
2157 msecs_to_jiffies(params->spp_sackdelay);
2158 } else if (asoc) {
2159 asoc->sackdelay =
2160 msecs_to_jiffies(params->spp_sackdelay);
2161 } else {
2162 sp->sackdelay = params->spp_sackdelay;
2166 if (sackdelay_change) {
2167 if (trans) {
2168 trans->param_flags =
2169 (trans->param_flags & ~SPP_SACKDELAY) |
2170 sackdelay_change;
2171 } else if (asoc) {
2172 asoc->param_flags =
2173 (asoc->param_flags & ~SPP_SACKDELAY) |
2174 sackdelay_change;
2175 } else {
2176 sp->param_flags =
2177 (sp->param_flags & ~SPP_SACKDELAY) |
2178 sackdelay_change;
2182 if (params->spp_pathmaxrxt) {
2183 if (trans) {
2184 trans->pathmaxrxt = params->spp_pathmaxrxt;
2185 } else if (asoc) {
2186 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2187 } else {
2188 sp->pathmaxrxt = params->spp_pathmaxrxt;
2192 return 0;
2195 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2196 char __user *optval, int optlen)
2198 struct sctp_paddrparams params;
2199 struct sctp_transport *trans = NULL;
2200 struct sctp_association *asoc = NULL;
2201 struct sctp_sock *sp = sctp_sk(sk);
2202 int error;
2203 int hb_change, pmtud_change, sackdelay_change;
2205 if (optlen != sizeof(struct sctp_paddrparams))
2206 return - EINVAL;
2208 if (copy_from_user(&params, optval, optlen))
2209 return -EFAULT;
2211 /* Validate flags and value parameters. */
2212 hb_change = params.spp_flags & SPP_HB;
2213 pmtud_change = params.spp_flags & SPP_PMTUD;
2214 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2216 if (hb_change == SPP_HB ||
2217 pmtud_change == SPP_PMTUD ||
2218 sackdelay_change == SPP_SACKDELAY ||
2219 params.spp_sackdelay > 500 ||
2220 (params.spp_pathmtu
2221 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2222 return -EINVAL;
2224 /* If an address other than INADDR_ANY is specified, and
2225 * no transport is found, then the request is invalid.
2226 */
2227 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2228 trans = sctp_addr_id2transport(sk, &params.spp_address,
2229 params.spp_assoc_id);
2230 if (!trans)
2231 return -EINVAL;
2234 /* Get association, if assoc_id != 0 and the socket is a one
2235 * to many style socket, and an association was not found, then
2236 * the id was invalid.
2237 */
2238 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2239 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2240 return -EINVAL;
2242 /* Heartbeat demand can only be sent on a transport or
2243 * association, but not a socket.
2244 */
2245 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2246 return -EINVAL;
2248 /* Process parameters. */
2249 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2250 hb_change, pmtud_change,
2251 sackdelay_change);
2253 if (error)
2254 return error;
2256 /* If changes are for association, also apply parameters to each
2257 * transport.
2258 */
2259 if (!trans && asoc) {
2260 struct list_head *pos;
2262 list_for_each(pos, &asoc->peer.transport_addr_list) {
2263 trans = list_entry(pos, struct sctp_transport,
2264 transports);
2265 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2266 hb_change, pmtud_change,
2267 sackdelay_change);
2271 return 0;
2274 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2276 * This options will get or set the delayed ack timer. The time is set
2277 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2278 * endpoints default delayed ack timer value. If the assoc_id field is
2279 * non-zero, then the set or get effects the specified association.
2281 * struct sctp_assoc_value {
2282 * sctp_assoc_t assoc_id;
2283 * uint32_t assoc_value;
2284 * };
2286 * assoc_id - This parameter, indicates which association the
2287 * user is preforming an action upon. Note that if
2288 * this field's value is zero then the endpoints
2289 * default value is changed (effecting future
2290 * associations only).
2292 * assoc_value - This parameter contains the number of milliseconds
2293 * that the user is requesting the delayed ACK timer
2294 * be set to. Note that this value is defined in
2295 * the standard to be between 200 and 500 milliseconds.
2297 * Note: a value of zero will leave the value alone,
2298 * but disable SACK delay. A non-zero value will also
2299 * enable SACK delay.
2300 */
2302 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2303 char __user *optval, int optlen)
2305 struct sctp_assoc_value params;
2306 struct sctp_transport *trans = NULL;
2307 struct sctp_association *asoc = NULL;
2308 struct sctp_sock *sp = sctp_sk(sk);
2310 if (optlen != sizeof(struct sctp_assoc_value))
2311 return - EINVAL;
2313 if (copy_from_user(&params, optval, optlen))
2314 return -EFAULT;
2316 /* Validate value parameter. */
2317 if (params.assoc_value > 500)
2318 return -EINVAL;
2320 /* Get association, if assoc_id != 0 and the socket is a one
2321 * to many style socket, and an association was not found, then
2322 * the id was invalid.
2323 */
2324 asoc = sctp_id2assoc(sk, params.assoc_id);
2325 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2326 return -EINVAL;
2328 if (params.assoc_value) {
2329 if (asoc) {
2330 asoc->sackdelay =
2331 msecs_to_jiffies(params.assoc_value);
2332 asoc->param_flags =
2333 (asoc->param_flags & ~SPP_SACKDELAY) |
2334 SPP_SACKDELAY_ENABLE;
2335 } else {
2336 sp->sackdelay = params.assoc_value;
2337 sp->param_flags =
2338 (sp->param_flags & ~SPP_SACKDELAY) |
2339 SPP_SACKDELAY_ENABLE;
2341 } else {
2342 if (asoc) {
2343 asoc->param_flags =
2344 (asoc->param_flags & ~SPP_SACKDELAY) |
2345 SPP_SACKDELAY_DISABLE;
2346 } else {
2347 sp->param_flags =
2348 (sp->param_flags & ~SPP_SACKDELAY) |
2349 SPP_SACKDELAY_DISABLE;
2353 /* If change is for association, also apply to each transport. */
2354 if (asoc) {
2355 struct list_head *pos;
2357 list_for_each(pos, &asoc->peer.transport_addr_list) {
2358 trans = list_entry(pos, struct sctp_transport,
2359 transports);
2360 if (params.assoc_value) {
2361 trans->sackdelay =
2362 msecs_to_jiffies(params.assoc_value);
2363 trans->param_flags =
2364 (trans->param_flags & ~SPP_SACKDELAY) |
2365 SPP_SACKDELAY_ENABLE;
2366 } else {
2367 trans->param_flags =
2368 (trans->param_flags & ~SPP_SACKDELAY) |
2369 SPP_SACKDELAY_DISABLE;
2374 return 0;
2377 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2379 * Applications can specify protocol parameters for the default association
2380 * initialization. The option name argument to setsockopt() and getsockopt()
2381 * is SCTP_INITMSG.
2383 * Setting initialization parameters is effective only on an unconnected
2384 * socket (for UDP-style sockets only future associations are effected
2385 * by the change). With TCP-style sockets, this option is inherited by
2386 * sockets derived from a listener socket.
2387 */
2388 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2390 struct sctp_initmsg sinit;
2391 struct sctp_sock *sp = sctp_sk(sk);
2393 if (optlen != sizeof(struct sctp_initmsg))
2394 return -EINVAL;
2395 if (copy_from_user(&sinit, optval, optlen))
2396 return -EFAULT;
2398 if (sinit.sinit_num_ostreams)
2399 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2400 if (sinit.sinit_max_instreams)
2401 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2402 if (sinit.sinit_max_attempts)
2403 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2404 if (sinit.sinit_max_init_timeo)
2405 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2407 return 0;
2410 /*
2411 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2413 * Applications that wish to use the sendto() system call may wish to
2414 * specify a default set of parameters that would normally be supplied
2415 * through the inclusion of ancillary data. This socket option allows
2416 * such an application to set the default sctp_sndrcvinfo structure.
2417 * The application that wishes to use this socket option simply passes
2418 * in to this call the sctp_sndrcvinfo structure defined in Section
2419 * 5.2.2) The input parameters accepted by this call include
2420 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2421 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2422 * to this call if the caller is using the UDP model.
2423 */
2424 static int sctp_setsockopt_default_send_param(struct sock *sk,
2425 char __user *optval, int optlen)
2427 struct sctp_sndrcvinfo info;
2428 struct sctp_association *asoc;
2429 struct sctp_sock *sp = sctp_sk(sk);
2431 if (optlen != sizeof(struct sctp_sndrcvinfo))
2432 return -EINVAL;
2433 if (copy_from_user(&info, optval, optlen))
2434 return -EFAULT;
2436 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2437 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2438 return -EINVAL;
2440 if (asoc) {
2441 asoc->default_stream = info.sinfo_stream;
2442 asoc->default_flags = info.sinfo_flags;
2443 asoc->default_ppid = info.sinfo_ppid;
2444 asoc->default_context = info.sinfo_context;
2445 asoc->default_timetolive = info.sinfo_timetolive;
2446 } else {
2447 sp->default_stream = info.sinfo_stream;
2448 sp->default_flags = info.sinfo_flags;
2449 sp->default_ppid = info.sinfo_ppid;
2450 sp->default_context = info.sinfo_context;
2451 sp->default_timetolive = info.sinfo_timetolive;
2454 return 0;
2457 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2459 * Requests that the local SCTP stack use the enclosed peer address as
2460 * the association primary. The enclosed address must be one of the
2461 * association peer's addresses.
2462 */
2463 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2464 int optlen)
2466 struct sctp_prim prim;
2467 struct sctp_transport *trans;
2469 if (optlen != sizeof(struct sctp_prim))
2470 return -EINVAL;
2472 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2473 return -EFAULT;
2475 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2476 if (!trans)
2477 return -EINVAL;
2479 sctp_assoc_set_primary(trans->asoc, trans);
2481 return 0;
2484 /*
2485 * 7.1.5 SCTP_NODELAY
2487 * Turn on/off any Nagle-like algorithm. This means that packets are
2488 * generally sent as soon as possible and no unnecessary delays are
2489 * introduced, at the cost of more packets in the network. Expects an
2490 * integer boolean flag.
2491 */
2492 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2493 int optlen)
2495 int val;
2497 if (optlen < sizeof(int))
2498 return -EINVAL;
2499 if (get_user(val, (int __user *)optval))
2500 return -EFAULT;
2502 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2503 return 0;
2506 /*
2508 * 7.1.1 SCTP_RTOINFO
2510 * The protocol parameters used to initialize and bound retransmission
2511 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2512 * and modify these parameters.
2513 * All parameters are time values, in milliseconds. A value of 0, when
2514 * modifying the parameters, indicates that the current value should not
2515 * be changed.
2517 */
2518 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2519 struct sctp_rtoinfo rtoinfo;
2520 struct sctp_association *asoc;
2522 if (optlen != sizeof (struct sctp_rtoinfo))
2523 return -EINVAL;
2525 if (copy_from_user(&rtoinfo, optval, optlen))
2526 return -EFAULT;
2528 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2530 /* Set the values to the specific association */
2531 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2532 return -EINVAL;
2534 if (asoc) {
2535 if (rtoinfo.srto_initial != 0)
2536 asoc->rto_initial =
2537 msecs_to_jiffies(rtoinfo.srto_initial);
2538 if (rtoinfo.srto_max != 0)
2539 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2540 if (rtoinfo.srto_min != 0)
2541 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2542 } else {
2543 /* If there is no association or the association-id = 0
2544 * set the values to the endpoint.
2545 */
2546 struct sctp_sock *sp = sctp_sk(sk);
2548 if (rtoinfo.srto_initial != 0)
2549 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2550 if (rtoinfo.srto_max != 0)
2551 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2552 if (rtoinfo.srto_min != 0)
2553 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2556 return 0;
2559 /*
2561 * 7.1.2 SCTP_ASSOCINFO
2563 * This option is used to tune the the maximum retransmission attempts
2564 * of the association.
2565 * Returns an error if the new association retransmission value is
2566 * greater than the sum of the retransmission value of the peer.
2567 * See [SCTP] for more information.
2569 */
2570 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2573 struct sctp_assocparams assocparams;
2574 struct sctp_association *asoc;
2576 if (optlen != sizeof(struct sctp_assocparams))
2577 return -EINVAL;
2578 if (copy_from_user(&assocparams, optval, optlen))
2579 return -EFAULT;
2581 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2583 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2584 return -EINVAL;
2586 /* Set the values to the specific association */
2587 if (asoc) {
2588 if (assocparams.sasoc_asocmaxrxt != 0) {
2589 __u32 path_sum = 0;
2590 int paths = 0;
2591 struct list_head *pos;
2592 struct sctp_transport *peer_addr;
2594 list_for_each(pos, &asoc->peer.transport_addr_list) {
2595 peer_addr = list_entry(pos,
2596 struct sctp_transport,
2597 transports);
2598 path_sum += peer_addr->pathmaxrxt;
2599 paths++;
2602 /* Only validate asocmaxrxt if we have more then
2603 * one path/transport. We do this because path
2604 * retransmissions are only counted when we have more
2605 * then one path.
2606 */
2607 if (paths > 1 &&
2608 assocparams.sasoc_asocmaxrxt > path_sum)
2609 return -EINVAL;
2611 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2614 if (assocparams.sasoc_cookie_life != 0) {
2615 asoc->cookie_life.tv_sec =
2616 assocparams.sasoc_cookie_life / 1000;
2617 asoc->cookie_life.tv_usec =
2618 (assocparams.sasoc_cookie_life % 1000)
2619 * 1000;
2621 } else {
2622 /* Set the values to the endpoint */
2623 struct sctp_sock *sp = sctp_sk(sk);
2625 if (assocparams.sasoc_asocmaxrxt != 0)
2626 sp->assocparams.sasoc_asocmaxrxt =
2627 assocparams.sasoc_asocmaxrxt;
2628 if (assocparams.sasoc_cookie_life != 0)
2629 sp->assocparams.sasoc_cookie_life =
2630 assocparams.sasoc_cookie_life;
2632 return 0;
2635 /*
2636 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2638 * This socket option is a boolean flag which turns on or off mapped V4
2639 * addresses. If this option is turned on and the socket is type
2640 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2641 * If this option is turned off, then no mapping will be done of V4
2642 * addresses and a user will receive both PF_INET6 and PF_INET type
2643 * addresses on the socket.
2644 */
2645 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2647 int val;
2648 struct sctp_sock *sp = sctp_sk(sk);
2650 if (optlen < sizeof(int))
2651 return -EINVAL;
2652 if (get_user(val, (int __user *)optval))
2653 return -EFAULT;
2654 if (val)
2655 sp->v4mapped = 1;
2656 else
2657 sp->v4mapped = 0;
2659 return 0;
2662 /*
2663 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2665 * This socket option specifies the maximum size to put in any outgoing
2666 * SCTP chunk. If a message is larger than this size it will be
2667 * fragmented by SCTP into the specified size. Note that the underlying
2668 * SCTP implementation may fragment into smaller sized chunks when the
2669 * PMTU of the underlying association is smaller than the value set by
2670 * the user.
2671 */
2672 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2674 struct sctp_association *asoc;
2675 struct list_head *pos;
2676 struct sctp_sock *sp = sctp_sk(sk);
2677 int val;
2679 if (optlen < sizeof(int))
2680 return -EINVAL;
2681 if (get_user(val, (int __user *)optval))
2682 return -EFAULT;
2683 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2684 return -EINVAL;
2685 sp->user_frag = val;
2687 /* Update the frag_point of the existing associations. */
2688 list_for_each(pos, &(sp->ep->asocs)) {
2689 asoc = list_entry(pos, struct sctp_association, asocs);
2690 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2693 return 0;
2697 /*
2698 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2700 * Requests that the peer mark the enclosed address as the association
2701 * primary. The enclosed address must be one of the association's
2702 * locally bound addresses. The following structure is used to make a
2703 * set primary request:
2704 */
2705 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2706 int optlen)
2708 struct sctp_sock *sp;
2709 struct sctp_endpoint *ep;
2710 struct sctp_association *asoc = NULL;
2711 struct sctp_setpeerprim prim;
2712 struct sctp_chunk *chunk;
2713 int err;
2715 sp = sctp_sk(sk);
2716 ep = sp->ep;
2718 if (!sctp_addip_enable)
2719 return -EPERM;
2721 if (optlen != sizeof(struct sctp_setpeerprim))
2722 return -EINVAL;
2724 if (copy_from_user(&prim, optval, optlen))
2725 return -EFAULT;
2727 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2728 if (!asoc)
2729 return -EINVAL;
2731 if (!asoc->peer.asconf_capable)
2732 return -EPERM;
2734 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2735 return -EPERM;
2737 if (!sctp_state(asoc, ESTABLISHED))
2738 return -ENOTCONN;
2740 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2741 return -EADDRNOTAVAIL;
2743 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2744 chunk = sctp_make_asconf_set_prim(asoc,
2745 (union sctp_addr *)&prim.sspp_addr);
2746 if (!chunk)
2747 return -ENOMEM;
2749 err = sctp_send_asconf(asoc, chunk);
2751 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2753 return err;
2756 static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2757 int optlen)
2759 struct sctp_setadaption adaption;
2761 if (optlen != sizeof(struct sctp_setadaption))
2762 return -EINVAL;
2763 if (copy_from_user(&adaption, optval, optlen))
2764 return -EFAULT;
2766 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
2768 return 0;
2771 /* API 6.2 setsockopt(), getsockopt()
2773 * Applications use setsockopt() and getsockopt() to set or retrieve
2774 * socket options. Socket options are used to change the default
2775 * behavior of sockets calls. They are described in Section 7.
2777 * The syntax is:
2779 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2780 * int __user *optlen);
2781 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2782 * int optlen);
2784 * sd - the socket descript.
2785 * level - set to IPPROTO_SCTP for all SCTP options.
2786 * optname - the option name.
2787 * optval - the buffer to store the value of the option.
2788 * optlen - the size of the buffer.
2789 */
2790 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2791 char __user *optval, int optlen)
2793 int retval = 0;
2795 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2796 sk, optname);
2798 /* I can hardly begin to describe how wrong this is. This is
2799 * so broken as to be worse than useless. The API draft
2800 * REALLY is NOT helpful here... I am not convinced that the
2801 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2802 * are at all well-founded.
2803 */
2804 if (level != SOL_SCTP) {
2805 struct sctp_af *af = sctp_sk(sk)->pf->af;
2806 retval = af->setsockopt(sk, level, optname, optval, optlen);
2807 goto out_nounlock;
2810 sctp_lock_sock(sk);
2812 switch (optname) {
2813 case SCTP_SOCKOPT_BINDX_ADD:
2814 /* 'optlen' is the size of the addresses buffer. */
2815 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2816 optlen, SCTP_BINDX_ADD_ADDR);
2817 break;
2819 case SCTP_SOCKOPT_BINDX_REM:
2820 /* 'optlen' is the size of the addresses buffer. */
2821 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2822 optlen, SCTP_BINDX_REM_ADDR);
2823 break;
2825 case SCTP_SOCKOPT_CONNECTX:
2826 /* 'optlen' is the size of the addresses buffer. */
2827 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2828 optlen);
2829 break;
2831 case SCTP_DISABLE_FRAGMENTS:
2832 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2833 break;
2835 case SCTP_EVENTS:
2836 retval = sctp_setsockopt_events(sk, optval, optlen);
2837 break;
2839 case SCTP_AUTOCLOSE:
2840 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2841 break;
2843 case SCTP_PEER_ADDR_PARAMS:
2844 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2845 break;
2847 case SCTP_DELAYED_ACK_TIME:
2848 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2849 break;
2851 case SCTP_INITMSG:
2852 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2853 break;
2854 case SCTP_DEFAULT_SEND_PARAM:
2855 retval = sctp_setsockopt_default_send_param(sk, optval,
2856 optlen);
2857 break;
2858 case SCTP_PRIMARY_ADDR:
2859 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2860 break;
2861 case SCTP_SET_PEER_PRIMARY_ADDR:
2862 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2863 break;
2864 case SCTP_NODELAY:
2865 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2866 break;
2867 case SCTP_RTOINFO:
2868 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2869 break;
2870 case SCTP_ASSOCINFO:
2871 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2872 break;
2873 case SCTP_I_WANT_MAPPED_V4_ADDR:
2874 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2875 break;
2876 case SCTP_MAXSEG:
2877 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2878 break;
2879 case SCTP_ADAPTION_LAYER:
2880 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2881 break;
2883 default:
2884 retval = -ENOPROTOOPT;
2885 break;
2886 };
2888 sctp_release_sock(sk);
2890 out_nounlock:
2891 return retval;
2894 /* API 3.1.6 connect() - UDP Style Syntax
2896 * An application may use the connect() call in the UDP model to initiate an
2897 * association without sending data.
2899 * The syntax is:
2901 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2903 * sd: the socket descriptor to have a new association added to.
2905 * nam: the address structure (either struct sockaddr_in or struct
2906 * sockaddr_in6 defined in RFC2553 [7]).
2908 * len: the size of the address.
2909 */
2910 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2911 int addr_len)
2913 int err = 0;
2914 struct sctp_af *af;
2916 sctp_lock_sock(sk);
2918 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2919 __FUNCTION__, sk, addr, addr_len);
2921 /* Validate addr_len before calling common connect/connectx routine. */
2922 af = sctp_get_af_specific(addr->sa_family);
2923 if (!af || addr_len < af->sockaddr_len) {
2924 err = -EINVAL;
2925 } else {
2926 /* Pass correct addr len to common routine (so it knows there
2927 * is only one address being passed.
2928 */
2929 err = __sctp_connect(sk, addr, af->sockaddr_len);
2932 sctp_release_sock(sk);
2933 return err;
2936 /* FIXME: Write comments. */
2937 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2939 return -EOPNOTSUPP; /* STUB */
2942 /* 4.1.4 accept() - TCP Style Syntax
2944 * Applications use accept() call to remove an established SCTP
2945 * association from the accept queue of the endpoint. A new socket
2946 * descriptor will be returned from accept() to represent the newly
2947 * formed association.
2948 */
2949 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2951 struct sctp_sock *sp;
2952 struct sctp_endpoint *ep;
2953 struct sock *newsk = NULL;
2954 struct sctp_association *asoc;
2955 long timeo;
2956 int error = 0;
2958 sctp_lock_sock(sk);
2960 sp = sctp_sk(sk);
2961 ep = sp->ep;
2963 if (!sctp_style(sk, TCP)) {
2964 error = -EOPNOTSUPP;
2965 goto out;
2968 if (!sctp_sstate(sk, LISTENING)) {
2969 error = -EINVAL;
2970 goto out;
2973 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2975 error = sctp_wait_for_accept(sk, timeo);
2976 if (error)
2977 goto out;
2979 /* We treat the list of associations on the endpoint as the accept
2980 * queue and pick the first association on the list.
2981 */
2982 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2984 newsk = sp->pf->create_accept_sk(sk, asoc);
2985 if (!newsk) {
2986 error = -ENOMEM;
2987 goto out;
2990 /* Populate the fields of the newsk from the oldsk and migrate the
2991 * asoc to the newsk.
2992 */
2993 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2995 out:
2996 sctp_release_sock(sk);
2997 *err = error;
2998 return newsk;
3001 /* The SCTP ioctl handler. */
3002 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3004 return -ENOIOCTLCMD;
3007 /* This is the function which gets called during socket creation to
3008 * initialized the SCTP-specific portion of the sock.
3009 * The sock structure should already be zero-filled memory.
3010 */
3011 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3013 struct sctp_endpoint *ep;
3014 struct sctp_sock *sp;
3016 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3018 sp = sctp_sk(sk);
3020 /* Initialize the SCTP per socket area. */
3021 switch (sk->sk_type) {
3022 case SOCK_SEQPACKET:
3023 sp->type = SCTP_SOCKET_UDP;
3024 break;
3025 case SOCK_STREAM:
3026 sp->type = SCTP_SOCKET_TCP;
3027 break;
3028 default:
3029 return -ESOCKTNOSUPPORT;
3032 /* Initialize default send parameters. These parameters can be
3033 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3034 */
3035 sp->default_stream = 0;
3036 sp->default_ppid = 0;
3037 sp->default_flags = 0;
3038 sp->default_context = 0;
3039 sp->default_timetolive = 0;
3041 /* Initialize default setup parameters. These parameters
3042 * can be modified with the SCTP_INITMSG socket option or
3043 * overridden by the SCTP_INIT CMSG.
3044 */
3045 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3046 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3047 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3048 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
3050 /* Initialize default RTO related parameters. These parameters can
3051 * be modified for with the SCTP_RTOINFO socket option.
3052 */
3053 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
3054 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
3055 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
3057 /* Initialize default association related parameters. These parameters
3058 * can be modified with the SCTP_ASSOCINFO socket option.
3059 */
3060 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3061 sp->assocparams.sasoc_number_peer_destinations = 0;
3062 sp->assocparams.sasoc_peer_rwnd = 0;
3063 sp->assocparams.sasoc_local_rwnd = 0;
3064 sp->assocparams.sasoc_cookie_life =
3065 jiffies_to_msecs(sctp_valid_cookie_life);
3067 /* Initialize default event subscriptions. By default, all the
3068 * options are off.
3069 */
3070 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3072 /* Default Peer Address Parameters. These defaults can
3073 * be modified via SCTP_PEER_ADDR_PARAMS
3074 */
3075 sp->hbinterval = jiffies_to_msecs(sctp_hb_interval);
3076 sp->pathmaxrxt = sctp_max_retrans_path;
3077 sp->pathmtu = 0; // allow default discovery
3078 sp->sackdelay = jiffies_to_msecs(sctp_sack_timeout);
3079 sp->param_flags = SPP_HB_ENABLE |
3080 SPP_PMTUD_ENABLE |
3081 SPP_SACKDELAY_ENABLE;
3083 /* If enabled no SCTP message fragmentation will be performed.
3084 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3085 */
3086 sp->disable_fragments = 0;
3088 /* Turn on/off any Nagle-like algorithm. */
3089 sp->nodelay = 1;
3091 /* Enable by default. */
3092 sp->v4mapped = 1;
3094 /* Auto-close idle associations after the configured
3095 * number of seconds. A value of 0 disables this
3096 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3097 * for UDP-style sockets only.
3098 */
3099 sp->autoclose = 0;
3101 /* User specified fragmentation limit. */
3102 sp->user_frag = 0;
3104 sp->adaption_ind = 0;
3106 sp->pf = sctp_get_pf_specific(sk->sk_family);
3108 /* Control variables for partial data delivery. */
3109 sp->pd_mode = 0;
3110 skb_queue_head_init(&sp->pd_lobby);
3112 /* Create a per socket endpoint structure. Even if we
3113 * change the data structure relationships, this may still
3114 * be useful for storing pre-connect address information.
3115 */
3116 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3117 if (!ep)
3118 return -ENOMEM;
3120 sp->ep = ep;
3121 sp->hmac = NULL;
3123 SCTP_DBG_OBJCNT_INC(sock);
3124 return 0;
3127 /* Cleanup any SCTP per socket resources. */
3128 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3130 struct sctp_endpoint *ep;
3132 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3134 /* Release our hold on the endpoint. */
3135 ep = sctp_sk(sk)->ep;
3136 sctp_endpoint_free(ep);
3138 return 0;
3141 /* API 4.1.7 shutdown() - TCP Style Syntax
3142 * int shutdown(int socket, int how);
3144 * sd - the socket descriptor of the association to be closed.
3145 * how - Specifies the type of shutdown. The values are
3146 * as follows:
3147 * SHUT_RD
3148 * Disables further receive operations. No SCTP
3149 * protocol action is taken.
3150 * SHUT_WR
3151 * Disables further send operations, and initiates
3152 * the SCTP shutdown sequence.
3153 * SHUT_RDWR
3154 * Disables further send and receive operations
3155 * and initiates the SCTP shutdown sequence.
3156 */
3157 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3159 struct sctp_endpoint *ep;
3160 struct sctp_association *asoc;
3162 if (!sctp_style(sk, TCP))
3163 return;
3165 if (how & SEND_SHUTDOWN) {
3166 ep = sctp_sk(sk)->ep;
3167 if (!list_empty(&ep->asocs)) {
3168 asoc = list_entry(ep->asocs.next,
3169 struct sctp_association, asocs);
3170 sctp_primitive_SHUTDOWN(asoc, NULL);
3175 /* 7.2.1 Association Status (SCTP_STATUS)
3177 * Applications can retrieve current status information about an
3178 * association, including association state, peer receiver window size,
3179 * number of unacked data chunks, and number of data chunks pending
3180 * receipt. This information is read-only.
3181 */
3182 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3183 char __user *optval,
3184 int __user *optlen)
3186 struct sctp_status status;
3187 struct sctp_association *asoc = NULL;
3188 struct sctp_transport *transport;
3189 sctp_assoc_t associd;
3190 int retval = 0;
3192 if (len != sizeof(status)) {
3193 retval = -EINVAL;
3194 goto out;
3197 if (copy_from_user(&status, optval, sizeof(status))) {
3198 retval = -EFAULT;
3199 goto out;
3202 associd = status.sstat_assoc_id;
3203 asoc = sctp_id2assoc(sk, associd);
3204 if (!asoc) {
3205 retval = -EINVAL;
3206 goto out;
3209 transport = asoc->peer.primary_path;
3211 status.sstat_assoc_id = sctp_assoc2id(asoc);
3212 status.sstat_state = asoc->state;
3213 status.sstat_rwnd = asoc->peer.rwnd;
3214 status.sstat_unackdata = asoc->unack_data;
3216 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3217 status.sstat_instrms = asoc->c.sinit_max_instreams;
3218 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3219 status.sstat_fragmentation_point = asoc->frag_point;
3220 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3221 memcpy(&status.sstat_primary.spinfo_address,
3222 &(transport->ipaddr), sizeof(union sctp_addr));
3223 /* Map ipv4 address into v4-mapped-on-v6 address. */
3224 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3225 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3226 status.sstat_primary.spinfo_state = transport->state;
3227 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3228 status.sstat_primary.spinfo_srtt = transport->srtt;
3229 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3230 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3232 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3233 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3235 if (put_user(len, optlen)) {
3236 retval = -EFAULT;
3237 goto out;
3240 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3241 len, status.sstat_state, status.sstat_rwnd,
3242 status.sstat_assoc_id);
3244 if (copy_to_user(optval, &status, len)) {
3245 retval = -EFAULT;
3246 goto out;
3249 out:
3250 return (retval);
3254 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3256 * Applications can retrieve information about a specific peer address
3257 * of an association, including its reachability state, congestion
3258 * window, and retransmission timer values. This information is
3259 * read-only.
3260 */
3261 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3262 char __user *optval,
3263 int __user *optlen)
3265 struct sctp_paddrinfo pinfo;
3266 struct sctp_transport *transport;
3267 int retval = 0;
3269 if (len != sizeof(pinfo)) {
3270 retval = -EINVAL;
3271 goto out;
3274 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3275 retval = -EFAULT;
3276 goto out;
3279 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3280 pinfo.spinfo_assoc_id);
3281 if (!transport)
3282 return -EINVAL;
3284 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3285 pinfo.spinfo_state = transport->state;
3286 pinfo.spinfo_cwnd = transport->cwnd;
3287 pinfo.spinfo_srtt = transport->srtt;
3288 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3289 pinfo.spinfo_mtu = transport->pathmtu;
3291 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3292 pinfo.spinfo_state = SCTP_ACTIVE;
3294 if (put_user(len, optlen)) {
3295 retval = -EFAULT;
3296 goto out;
3299 if (copy_to_user(optval, &pinfo, len)) {
3300 retval = -EFAULT;
3301 goto out;
3304 out:
3305 return (retval);
3308 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3310 * This option is a on/off flag. If enabled no SCTP message
3311 * fragmentation will be performed. Instead if a message being sent
3312 * exceeds the current PMTU size, the message will NOT be sent and
3313 * instead a error will be indicated to the user.
3314 */
3315 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3316 char __user *optval, int __user *optlen)
3318 int val;
3320 if (len < sizeof(int))
3321 return -EINVAL;
3323 len = sizeof(int);
3324 val = (sctp_sk(sk)->disable_fragments == 1);
3325 if (put_user(len, optlen))
3326 return -EFAULT;
3327 if (copy_to_user(optval, &val, len))
3328 return -EFAULT;
3329 return 0;
3332 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3334 * This socket option is used to specify various notifications and
3335 * ancillary data the user wishes to receive.
3336 */
3337 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3338 int __user *optlen)
3340 if (len != sizeof(struct sctp_event_subscribe))
3341 return -EINVAL;
3342 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3343 return -EFAULT;
3344 return 0;
3347 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3349 * This socket option is applicable to the UDP-style socket only. When
3350 * set it will cause associations that are idle for more than the
3351 * specified number of seconds to automatically close. An association
3352 * being idle is defined an association that has NOT sent or received
3353 * user data. The special value of '0' indicates that no automatic
3354 * close of any associations should be performed. The option expects an
3355 * integer defining the number of seconds of idle time before an
3356 * association is closed.
3357 */
3358 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3360 /* Applicable to UDP-style socket only */
3361 if (sctp_style(sk, TCP))
3362 return -EOPNOTSUPP;
3363 if (len != sizeof(int))
3364 return -EINVAL;
3365 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3366 return -EFAULT;
3367 return 0;
3370 /* Helper routine to branch off an association to a new socket. */
3371 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3372 struct socket **sockp)
3374 struct sock *sk = asoc->base.sk;
3375 struct socket *sock;
3376 int err = 0;
3378 /* An association cannot be branched off from an already peeled-off
3379 * socket, nor is this supported for tcp style sockets.
3380 */
3381 if (!sctp_style(sk, UDP))
3382 return -EINVAL;
3384 /* Create a new socket. */
3385 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3386 if (err < 0)
3387 return err;
3389 /* Populate the fields of the newsk from the oldsk and migrate the
3390 * asoc to the newsk.
3391 */
3392 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3393 *sockp = sock;
3395 return err;
3398 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3400 sctp_peeloff_arg_t peeloff;
3401 struct socket *newsock;
3402 int retval = 0;
3403 struct sctp_association *asoc;
3405 if (len != sizeof(sctp_peeloff_arg_t))
3406 return -EINVAL;
3407 if (copy_from_user(&peeloff, optval, len))
3408 return -EFAULT;
3410 asoc = sctp_id2assoc(sk, peeloff.associd);
3411 if (!asoc) {
3412 retval = -EINVAL;
3413 goto out;
3416 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3418 retval = sctp_do_peeloff(asoc, &newsock);
3419 if (retval < 0)
3420 goto out;
3422 /* Map the socket to an unused fd that can be returned to the user. */
3423 retval = sock_map_fd(newsock);
3424 if (retval < 0) {
3425 sock_release(newsock);
3426 goto out;
3429 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3430 __FUNCTION__, sk, asoc, newsock->sk, retval);
3432 /* Return the fd mapped to the new socket. */
3433 peeloff.sd = retval;
3434 if (copy_to_user(optval, &peeloff, len))
3435 retval = -EFAULT;
3437 out:
3438 return retval;
3441 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3443 * Applications can enable or disable heartbeats for any peer address of
3444 * an association, modify an address's heartbeat interval, force a
3445 * heartbeat to be sent immediately, and adjust the address's maximum
3446 * number of retransmissions sent before an address is considered
3447 * unreachable. The following structure is used to access and modify an
3448 * address's parameters:
3450 * struct sctp_paddrparams {
3451 * sctp_assoc_t spp_assoc_id;
3452 * struct sockaddr_storage spp_address;
3453 * uint32_t spp_hbinterval;
3454 * uint16_t spp_pathmaxrxt;
3455 * uint32_t spp_pathmtu;
3456 * uint32_t spp_sackdelay;
3457 * uint32_t spp_flags;
3458 * };
3460 * spp_assoc_id - (one-to-many style socket) This is filled in the
3461 * application, and identifies the association for
3462 * this query.
3463 * spp_address - This specifies which address is of interest.
3464 * spp_hbinterval - This contains the value of the heartbeat interval,
3465 * in milliseconds. If a value of zero
3466 * is present in this field then no changes are to
3467 * be made to this parameter.
3468 * spp_pathmaxrxt - This contains the maximum number of
3469 * retransmissions before this address shall be
3470 * considered unreachable. If a value of zero
3471 * is present in this field then no changes are to
3472 * be made to this parameter.
3473 * spp_pathmtu - When Path MTU discovery is disabled the value
3474 * specified here will be the "fixed" path mtu.
3475 * Note that if the spp_address field is empty
3476 * then all associations on this address will
3477 * have this fixed path mtu set upon them.
3479 * spp_sackdelay - When delayed sack is enabled, this value specifies
3480 * the number of milliseconds that sacks will be delayed
3481 * for. This value will apply to all addresses of an
3482 * association if the spp_address field is empty. Note
3483 * also, that if delayed sack is enabled and this
3484 * value is set to 0, no change is made to the last
3485 * recorded delayed sack timer value.
3487 * spp_flags - These flags are used to control various features
3488 * on an association. The flag field may contain
3489 * zero or more of the following options.
3491 * SPP_HB_ENABLE - Enable heartbeats on the
3492 * specified address. Note that if the address
3493 * field is empty all addresses for the association
3494 * have heartbeats enabled upon them.
3496 * SPP_HB_DISABLE - Disable heartbeats on the
3497 * speicifed address. Note that if the address
3498 * field is empty all addresses for the association
3499 * will have their heartbeats disabled. Note also
3500 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3501 * mutually exclusive, only one of these two should
3502 * be specified. Enabling both fields will have
3503 * undetermined results.
3505 * SPP_HB_DEMAND - Request a user initiated heartbeat
3506 * to be made immediately.
3508 * SPP_PMTUD_ENABLE - This field will enable PMTU
3509 * discovery upon the specified address. Note that
3510 * if the address feild is empty then all addresses
3511 * on the association are effected.
3513 * SPP_PMTUD_DISABLE - This field will disable PMTU
3514 * discovery upon the specified address. Note that
3515 * if the address feild is empty then all addresses
3516 * on the association are effected. Not also that
3517 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3518 * exclusive. Enabling both will have undetermined
3519 * results.
3521 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3522 * on delayed sack. The time specified in spp_sackdelay
3523 * is used to specify the sack delay for this address. Note
3524 * that if spp_address is empty then all addresses will
3525 * enable delayed sack and take on the sack delay
3526 * value specified in spp_sackdelay.
3527 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3528 * off delayed sack. If the spp_address field is blank then
3529 * delayed sack is disabled for the entire association. Note
3530 * also that this field is mutually exclusive to
3531 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3532 * results.
3533 */
3534 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3535 char __user *optval, int __user *optlen)
3537 struct sctp_paddrparams params;
3538 struct sctp_transport *trans = NULL;
3539 struct sctp_association *asoc = NULL;
3540 struct sctp_sock *sp = sctp_sk(sk);
3542 if (len != sizeof(struct sctp_paddrparams))
3543 return -EINVAL;
3545 if (copy_from_user(&params, optval, len))
3546 return -EFAULT;
3548 /* If an address other than INADDR_ANY is specified, and
3549 * no transport is found, then the request is invalid.
3550 */
3551 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3552 trans = sctp_addr_id2transport(sk, &params.spp_address,
3553 params.spp_assoc_id);
3554 if (!trans) {
3555 SCTP_DEBUG_PRINTK("Failed no transport\n");
3556 return -EINVAL;
3560 /* Get association, if assoc_id != 0 and the socket is a one
3561 * to many style socket, and an association was not found, then
3562 * the id was invalid.
3563 */
3564 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3565 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3566 SCTP_DEBUG_PRINTK("Failed no association\n");
3567 return -EINVAL;
3570 if (trans) {
3571 /* Fetch transport values. */
3572 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3573 params.spp_pathmtu = trans->pathmtu;
3574 params.spp_pathmaxrxt = trans->pathmaxrxt;
3575 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3577 /*draft-11 doesn't say what to return in spp_flags*/
3578 params.spp_flags = trans->param_flags;
3579 } else if (asoc) {
3580 /* Fetch association values. */
3581 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3582 params.spp_pathmtu = asoc->pathmtu;
3583 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3584 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3586 /*draft-11 doesn't say what to return in spp_flags*/
3587 params.spp_flags = asoc->param_flags;
3588 } else {
3589 /* Fetch socket values. */
3590 params.spp_hbinterval = sp->hbinterval;
3591 params.spp_pathmtu = sp->pathmtu;
3592 params.spp_sackdelay = sp->sackdelay;
3593 params.spp_pathmaxrxt = sp->pathmaxrxt;
3595 /*draft-11 doesn't say what to return in spp_flags*/
3596 params.spp_flags = sp->param_flags;
3599 if (copy_to_user(optval, &params, len))
3600 return -EFAULT;
3602 if (put_user(len, optlen))
3603 return -EFAULT;
3605 return 0;
3608 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3610 * This options will get or set the delayed ack timer. The time is set
3611 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3612 * endpoints default delayed ack timer value. If the assoc_id field is
3613 * non-zero, then the set or get effects the specified association.
3615 * struct sctp_assoc_value {
3616 * sctp_assoc_t assoc_id;
3617 * uint32_t assoc_value;
3618 * };
3620 * assoc_id - This parameter, indicates which association the
3621 * user is preforming an action upon. Note that if
3622 * this field's value is zero then the endpoints
3623 * default value is changed (effecting future
3624 * associations only).
3626 * assoc_value - This parameter contains the number of milliseconds
3627 * that the user is requesting the delayed ACK timer
3628 * be set to. Note that this value is defined in
3629 * the standard to be between 200 and 500 milliseconds.
3631 * Note: a value of zero will leave the value alone,
3632 * but disable SACK delay. A non-zero value will also
3633 * enable SACK delay.
3634 */
3635 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3636 char __user *optval,
3637 int __user *optlen)
3639 struct sctp_assoc_value params;
3640 struct sctp_association *asoc = NULL;
3641 struct sctp_sock *sp = sctp_sk(sk);
3643 if (len != sizeof(struct sctp_assoc_value))
3644 return - EINVAL;
3646 if (copy_from_user(&params, optval, len))
3647 return -EFAULT;
3649 /* Get association, if assoc_id != 0 and the socket is a one
3650 * to many style socket, and an association was not found, then
3651 * the id was invalid.
3652 */
3653 asoc = sctp_id2assoc(sk, params.assoc_id);
3654 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3655 return -EINVAL;
3657 if (asoc) {
3658 /* Fetch association values. */
3659 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3660 params.assoc_value = jiffies_to_msecs(
3661 asoc->sackdelay);
3662 else
3663 params.assoc_value = 0;
3664 } else {
3665 /* Fetch socket values. */
3666 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3667 params.assoc_value = sp->sackdelay;
3668 else
3669 params.assoc_value = 0;
3672 if (copy_to_user(optval, &params, len))
3673 return -EFAULT;
3675 if (put_user(len, optlen))
3676 return -EFAULT;
3678 return 0;
3681 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3683 * Applications can specify protocol parameters for the default association
3684 * initialization. The option name argument to setsockopt() and getsockopt()
3685 * is SCTP_INITMSG.
3687 * Setting initialization parameters is effective only on an unconnected
3688 * socket (for UDP-style sockets only future associations are effected
3689 * by the change). With TCP-style sockets, this option is inherited by
3690 * sockets derived from a listener socket.
3691 */
3692 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3694 if (len != sizeof(struct sctp_initmsg))
3695 return -EINVAL;
3696 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3697 return -EFAULT;
3698 return 0;
3701 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3702 char __user *optval,
3703 int __user *optlen)
3705 sctp_assoc_t id;
3706 struct sctp_association *asoc;
3707 struct list_head *pos;
3708 int cnt = 0;
3710 if (len != sizeof(sctp_assoc_t))
3711 return -EINVAL;
3713 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3714 return -EFAULT;
3716 /* For UDP-style sockets, id specifies the association to query. */
3717 asoc = sctp_id2assoc(sk, id);
3718 if (!asoc)
3719 return -EINVAL;
3721 list_for_each(pos, &asoc->peer.transport_addr_list) {
3722 cnt ++;
3725 return cnt;
3728 /*
3729 * Old API for getting list of peer addresses. Does not work for 32-bit
3730 * programs running on a 64-bit kernel
3731 */
3732 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3733 char __user *optval,
3734 int __user *optlen)
3736 struct sctp_association *asoc;
3737 struct list_head *pos;
3738 int cnt = 0;
3739 struct sctp_getaddrs_old getaddrs;
3740 struct sctp_transport *from;
3741 void __user *to;
3742 union sctp_addr temp;
3743 struct sctp_sock *sp = sctp_sk(sk);
3744 int addrlen;
3746 if (len != sizeof(struct sctp_getaddrs_old))
3747 return -EINVAL;
3749 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3750 return -EFAULT;
3752 if (getaddrs.addr_num <= 0) return -EINVAL;
3754 /* For UDP-style sockets, id specifies the association to query. */
3755 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3756 if (!asoc)
3757 return -EINVAL;
3759 to = (void __user *)getaddrs.addrs;
3760 list_for_each(pos, &asoc->peer.transport_addr_list) {
3761 from = list_entry(pos, struct sctp_transport, transports);
3762 memcpy(&temp, &from->ipaddr, sizeof(temp));
3763 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3764 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3765 temp.v4.sin_port = htons(temp.v4.sin_port);
3766 if (copy_to_user(to, &temp, addrlen))
3767 return -EFAULT;
3768 to += addrlen ;
3769 cnt ++;
3770 if (cnt >= getaddrs.addr_num) break;
3772 getaddrs.addr_num = cnt;
3773 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3774 return -EFAULT;
3776 return 0;
3779 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3780 char __user *optval, int __user *optlen)
3782 struct sctp_association *asoc;
3783 struct list_head *pos;
3784 int cnt = 0;
3785 struct sctp_getaddrs getaddrs;
3786 struct sctp_transport *from;
3787 void __user *to;
3788 union sctp_addr temp;
3789 struct sctp_sock *sp = sctp_sk(sk);
3790 int addrlen;
3791 size_t space_left;
3792 int bytes_copied;
3794 if (len < sizeof(struct sctp_getaddrs))
3795 return -EINVAL;
3797 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3798 return -EFAULT;
3800 /* For UDP-style sockets, id specifies the association to query. */
3801 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3802 if (!asoc)
3803 return -EINVAL;
3805 to = optval + offsetof(struct sctp_getaddrs,addrs);
3806 space_left = len - sizeof(struct sctp_getaddrs) -
3807 offsetof(struct sctp_getaddrs,addrs);
3809 list_for_each(pos, &asoc->peer.transport_addr_list) {
3810 from = list_entry(pos, struct sctp_transport, transports);
3811 memcpy(&temp, &from->ipaddr, sizeof(temp));
3812 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3813 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3814 if(space_left < addrlen)
3815 return -ENOMEM;
3816 temp.v4.sin_port = htons(temp.v4.sin_port);
3817 if (copy_to_user(to, &temp, addrlen))
3818 return -EFAULT;
3819 to += addrlen;
3820 cnt++;
3821 space_left -= addrlen;
3824 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3825 return -EFAULT;
3826 bytes_copied = ((char __user *)to) - optval;
3827 if (put_user(bytes_copied, optlen))
3828 return -EFAULT;
3830 return 0;
3833 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3834 char __user *optval,
3835 int __user *optlen)
3837 sctp_assoc_t id;
3838 struct sctp_bind_addr *bp;
3839 struct sctp_association *asoc;
3840 struct list_head *pos;
3841 struct sctp_sockaddr_entry *addr;
3842 rwlock_t *addr_lock;
3843 unsigned long flags;
3844 int cnt = 0;
3846 if (len != sizeof(sctp_assoc_t))
3847 return -EINVAL;
3849 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3850 return -EFAULT;
3852 /*
3853 * For UDP-style sockets, id specifies the association to query.
3854 * If the id field is set to the value '0' then the locally bound
3855 * addresses are returned without regard to any particular
3856 * association.
3857 */
3858 if (0 == id) {
3859 bp = &sctp_sk(sk)->ep->base.bind_addr;
3860 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3861 } else {
3862 asoc = sctp_id2assoc(sk, id);
3863 if (!asoc)
3864 return -EINVAL;
3865 bp = &asoc->base.bind_addr;
3866 addr_lock = &asoc->base.addr_lock;
3869 sctp_read_lock(addr_lock);
3871 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3872 * addresses from the global local address list.
3873 */
3874 if (sctp_list_single_entry(&bp->address_list)) {
3875 addr = list_entry(bp->address_list.next,
3876 struct sctp_sockaddr_entry, list);
3877 if (sctp_is_any(&addr->a)) {
3878 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3879 list_for_each(pos, &sctp_local_addr_list) {
3880 addr = list_entry(pos,
3881 struct sctp_sockaddr_entry,
3882 list);
3883 if ((PF_INET == sk->sk_family) &&
3884 (AF_INET6 == addr->a.sa.sa_family))
3885 continue;
3886 cnt++;
3888 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3889 flags);
3890 } else {
3891 cnt = 1;
3893 goto done;
3896 list_for_each(pos, &bp->address_list) {
3897 cnt ++;
3900 done:
3901 sctp_read_unlock(addr_lock);
3902 return cnt;
3905 /* Helper function that copies local addresses to user and returns the number
3906 * of addresses copied.
3907 */
3908 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3909 void __user *to)
3911 struct list_head *pos;
3912 struct sctp_sockaddr_entry *addr;
3913 unsigned long flags;
3914 union sctp_addr temp;
3915 int cnt = 0;
3916 int addrlen;
3918 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3919 list_for_each(pos, &sctp_local_addr_list) {
3920 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3921 if ((PF_INET == sk->sk_family) &&
3922 (AF_INET6 == addr->a.sa.sa_family))
3923 continue;
3924 memcpy(&temp, &addr->a, sizeof(temp));
3925 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3926 &temp);
3927 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3928 temp.v4.sin_port = htons(port);
3929 if (copy_to_user(to, &temp, addrlen)) {
3930 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3931 flags);
3932 return -EFAULT;
3934 to += addrlen;
3935 cnt ++;
3936 if (cnt >= max_addrs) break;
3938 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3940 return cnt;
3943 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
3944 void __user **to, size_t space_left)
3946 struct list_head *pos;
3947 struct sctp_sockaddr_entry *addr;
3948 unsigned long flags;
3949 union sctp_addr temp;
3950 int cnt = 0;
3951 int addrlen;
3953 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3954 list_for_each(pos, &sctp_local_addr_list) {
3955 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3956 if ((PF_INET == sk->sk_family) &&
3957 (AF_INET6 == addr->a.sa.sa_family))
3958 continue;
3959 memcpy(&temp, &addr->a, sizeof(temp));
3960 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3961 &temp);
3962 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3963 if(space_left<addrlen)
3964 return -ENOMEM;
3965 temp.v4.sin_port = htons(port);
3966 if (copy_to_user(*to, &temp, addrlen)) {
3967 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3968 flags);
3969 return -EFAULT;
3971 *to += addrlen;
3972 cnt ++;
3973 space_left -= addrlen;
3975 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3977 return cnt;
3980 /* Old API for getting list of local addresses. Does not work for 32-bit
3981 * programs running on a 64-bit kernel
3982 */
3983 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3984 char __user *optval, int __user *optlen)
3986 struct sctp_bind_addr *bp;
3987 struct sctp_association *asoc;
3988 struct list_head *pos;
3989 int cnt = 0;
3990 struct sctp_getaddrs_old getaddrs;
3991 struct sctp_sockaddr_entry *addr;
3992 void __user *to;
3993 union sctp_addr temp;
3994 struct sctp_sock *sp = sctp_sk(sk);
3995 int addrlen;
3996 rwlock_t *addr_lock;
3997 int err = 0;
3999 if (len != sizeof(struct sctp_getaddrs_old))
4000 return -EINVAL;
4002 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
4003 return -EFAULT;
4005 if (getaddrs.addr_num <= 0) return -EINVAL;
4006 /*
4007 * For UDP-style sockets, id specifies the association to query.
4008 * If the id field is set to the value '0' then the locally bound
4009 * addresses are returned without regard to any particular
4010 * association.
4011 */
4012 if (0 == getaddrs.assoc_id) {
4013 bp = &sctp_sk(sk)->ep->base.bind_addr;
4014 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4015 } else {
4016 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4017 if (!asoc)
4018 return -EINVAL;
4019 bp = &asoc->base.bind_addr;
4020 addr_lock = &asoc->base.addr_lock;
4023 to = getaddrs.addrs;
4025 sctp_read_lock(addr_lock);
4027 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4028 * addresses from the global local address list.
4029 */
4030 if (sctp_list_single_entry(&bp->address_list)) {
4031 addr = list_entry(bp->address_list.next,
4032 struct sctp_sockaddr_entry, list);
4033 if (sctp_is_any(&addr->a)) {
4034 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4035 getaddrs.addr_num,
4036 to);
4037 if (cnt < 0) {
4038 err = cnt;
4039 goto unlock;
4041 goto copy_getaddrs;
4045 list_for_each(pos, &bp->address_list) {
4046 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4047 memcpy(&temp, &addr->a, sizeof(temp));
4048 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4049 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4050 temp.v4.sin_port = htons(temp.v4.sin_port);
4051 if (copy_to_user(to, &temp, addrlen)) {
4052 err = -EFAULT;
4053 goto unlock;
4055 to += addrlen;
4056 cnt ++;
4057 if (cnt >= getaddrs.addr_num) break;
4060 copy_getaddrs:
4061 getaddrs.addr_num = cnt;
4062 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4063 err = -EFAULT;
4065 unlock:
4066 sctp_read_unlock(addr_lock);
4067 return err;
4070 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4071 char __user *optval, int __user *optlen)
4073 struct sctp_bind_addr *bp;
4074 struct sctp_association *asoc;
4075 struct list_head *pos;
4076 int cnt = 0;
4077 struct sctp_getaddrs getaddrs;
4078 struct sctp_sockaddr_entry *addr;
4079 void __user *to;
4080 union sctp_addr temp;
4081 struct sctp_sock *sp = sctp_sk(sk);
4082 int addrlen;
4083 rwlock_t *addr_lock;
4084 int err = 0;
4085 size_t space_left;
4086 int bytes_copied;
4088 if (len <= sizeof(struct sctp_getaddrs))
4089 return -EINVAL;
4091 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4092 return -EFAULT;
4094 /*
4095 * For UDP-style sockets, id specifies the association to query.
4096 * If the id field is set to the value '0' then the locally bound
4097 * addresses are returned without regard to any particular
4098 * association.
4099 */
4100 if (0 == getaddrs.assoc_id) {
4101 bp = &sctp_sk(sk)->ep->base.bind_addr;
4102 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4103 } else {
4104 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4105 if (!asoc)
4106 return -EINVAL;
4107 bp = &asoc->base.bind_addr;
4108 addr_lock = &asoc->base.addr_lock;
4111 to = optval + offsetof(struct sctp_getaddrs,addrs);
4112 space_left = len - sizeof(struct sctp_getaddrs) -
4113 offsetof(struct sctp_getaddrs,addrs);
4115 sctp_read_lock(addr_lock);
4117 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4118 * addresses from the global local address list.
4119 */
4120 if (sctp_list_single_entry(&bp->address_list)) {
4121 addr = list_entry(bp->address_list.next,
4122 struct sctp_sockaddr_entry, list);
4123 if (sctp_is_any(&addr->a)) {
4124 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4125 &to, space_left);
4126 if (cnt < 0) {
4127 err = cnt;
4128 goto unlock;
4130 goto copy_getaddrs;
4134 list_for_each(pos, &bp->address_list) {
4135 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4136 memcpy(&temp, &addr->a, sizeof(temp));
4137 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4138 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4139 if(space_left < addrlen)
4140 return -ENOMEM; /*fixme: right error?*/
4141 temp.v4.sin_port = htons(temp.v4.sin_port);
4142 if (copy_to_user(to, &temp, addrlen)) {
4143 err = -EFAULT;
4144 goto unlock;
4146 to += addrlen;
4147 cnt ++;
4148 space_left -= addrlen;
4151 copy_getaddrs:
4152 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4153 return -EFAULT;
4154 bytes_copied = ((char __user *)to) - optval;
4155 if (put_user(bytes_copied, optlen))
4156 return -EFAULT;
4158 unlock:
4159 sctp_read_unlock(addr_lock);
4160 return err;
4163 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4165 * Requests that the local SCTP stack use the enclosed peer address as
4166 * the association primary. The enclosed address must be one of the
4167 * association peer's addresses.
4168 */
4169 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4170 char __user *optval, int __user *optlen)
4172 struct sctp_prim prim;
4173 struct sctp_association *asoc;
4174 struct sctp_sock *sp = sctp_sk(sk);
4176 if (len != sizeof(struct sctp_prim))
4177 return -EINVAL;
4179 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4180 return -EFAULT;
4182 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4183 if (!asoc)
4184 return -EINVAL;
4186 if (!asoc->peer.primary_path)
4187 return -ENOTCONN;
4189 asoc->peer.primary_path->ipaddr.v4.sin_port =
4190 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
4191 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4192 sizeof(union sctp_addr));
4193 asoc->peer.primary_path->ipaddr.v4.sin_port =
4194 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
4196 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4197 (union sctp_addr *)&prim.ssp_addr);
4199 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4200 return -EFAULT;
4202 return 0;
4205 /*
4206 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4208 * Requests that the local endpoint set the specified Adaption Layer
4209 * Indication parameter for all future INIT and INIT-ACK exchanges.
4210 */
4211 static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4212 char __user *optval, int __user *optlen)
4214 struct sctp_setadaption adaption;
4216 if (len != sizeof(struct sctp_setadaption))
4217 return -EINVAL;
4219 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4220 if (copy_to_user(optval, &adaption, len))
4221 return -EFAULT;
4223 return 0;
4226 /*
4228 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4230 * Applications that wish to use the sendto() system call may wish to
4231 * specify a default set of parameters that would normally be supplied
4232 * through the inclusion of ancillary data. This socket option allows
4233 * such an application to set the default sctp_sndrcvinfo structure.
4236 * The application that wishes to use this socket option simply passes
4237 * in to this call the sctp_sndrcvinfo structure defined in Section
4238 * 5.2.2) The input parameters accepted by this call include
4239 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4240 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4241 * to this call if the caller is using the UDP model.
4243 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4244 */
4245 static int sctp_getsockopt_default_send_param(struct sock *sk,
4246 int len, char __user *optval,
4247 int __user *optlen)
4249 struct sctp_sndrcvinfo info;
4250 struct sctp_association *asoc;
4251 struct sctp_sock *sp = sctp_sk(sk);
4253 if (len != sizeof(struct sctp_sndrcvinfo))
4254 return -EINVAL;
4255 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4256 return -EFAULT;
4258 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4259 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4260 return -EINVAL;
4262 if (asoc) {
4263 info.sinfo_stream = asoc->default_stream;
4264 info.sinfo_flags = asoc->default_flags;
4265 info.sinfo_ppid = asoc->default_ppid;
4266 info.sinfo_context = asoc->default_context;
4267 info.sinfo_timetolive = asoc->default_timetolive;
4268 } else {
4269 info.sinfo_stream = sp->default_stream;
4270 info.sinfo_flags = sp->default_flags;
4271 info.sinfo_ppid = sp->default_ppid;
4272 info.sinfo_context = sp->default_context;
4273 info.sinfo_timetolive = sp->default_timetolive;
4276 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4277 return -EFAULT;
4279 return 0;
4282 /*
4284 * 7.1.5 SCTP_NODELAY
4286 * Turn on/off any Nagle-like algorithm. This means that packets are
4287 * generally sent as soon as possible and no unnecessary delays are
4288 * introduced, at the cost of more packets in the network. Expects an
4289 * integer boolean flag.
4290 */
4292 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4293 char __user *optval, int __user *optlen)
4295 int val;
4297 if (len < sizeof(int))
4298 return -EINVAL;
4300 len = sizeof(int);
4301 val = (sctp_sk(sk)->nodelay == 1);
4302 if (put_user(len, optlen))
4303 return -EFAULT;
4304 if (copy_to_user(optval, &val, len))
4305 return -EFAULT;
4306 return 0;
4309 /*
4311 * 7.1.1 SCTP_RTOINFO
4313 * The protocol parameters used to initialize and bound retransmission
4314 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4315 * and modify these parameters.
4316 * All parameters are time values, in milliseconds. A value of 0, when
4317 * modifying the parameters, indicates that the current value should not
4318 * be changed.
4320 */
4321 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4322 char __user *optval,
4323 int __user *optlen) {
4324 struct sctp_rtoinfo rtoinfo;
4325 struct sctp_association *asoc;
4327 if (len != sizeof (struct sctp_rtoinfo))
4328 return -EINVAL;
4330 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4331 return -EFAULT;
4333 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4335 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4336 return -EINVAL;
4338 /* Values corresponding to the specific association. */
4339 if (asoc) {
4340 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4341 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4342 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4343 } else {
4344 /* Values corresponding to the endpoint. */
4345 struct sctp_sock *sp = sctp_sk(sk);
4347 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4348 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4349 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4352 if (put_user(len, optlen))
4353 return -EFAULT;
4355 if (copy_to_user(optval, &rtoinfo, len))
4356 return -EFAULT;
4358 return 0;
4361 /*
4363 * 7.1.2 SCTP_ASSOCINFO
4365 * This option is used to tune the the maximum retransmission attempts
4366 * of the association.
4367 * Returns an error if the new association retransmission value is
4368 * greater than the sum of the retransmission value of the peer.
4369 * See [SCTP] for more information.
4371 */
4372 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4373 char __user *optval,
4374 int __user *optlen)
4377 struct sctp_assocparams assocparams;
4378 struct sctp_association *asoc;
4379 struct list_head *pos;
4380 int cnt = 0;
4382 if (len != sizeof (struct sctp_assocparams))
4383 return -EINVAL;
4385 if (copy_from_user(&assocparams, optval,
4386 sizeof (struct sctp_assocparams)))
4387 return -EFAULT;
4389 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4391 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4392 return -EINVAL;
4394 /* Values correspoinding to the specific association */
4395 if (asoc) {
4396 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4397 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4398 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4399 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4400 * 1000) +
4401 (asoc->cookie_life.tv_usec
4402 / 1000);
4404 list_for_each(pos, &asoc->peer.transport_addr_list) {
4405 cnt ++;
4408 assocparams.sasoc_number_peer_destinations = cnt;
4409 } else {
4410 /* Values corresponding to the endpoint */
4411 struct sctp_sock *sp = sctp_sk(sk);
4413 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4414 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4415 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4416 assocparams.sasoc_cookie_life =
4417 sp->assocparams.sasoc_cookie_life;
4418 assocparams.sasoc_number_peer_destinations =
4419 sp->assocparams.
4420 sasoc_number_peer_destinations;
4423 if (put_user(len, optlen))
4424 return -EFAULT;
4426 if (copy_to_user(optval, &assocparams, len))
4427 return -EFAULT;
4429 return 0;
4432 /*
4433 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4435 * This socket option is a boolean flag which turns on or off mapped V4
4436 * addresses. If this option is turned on and the socket is type
4437 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4438 * If this option is turned off, then no mapping will be done of V4
4439 * addresses and a user will receive both PF_INET6 and PF_INET type
4440 * addresses on the socket.
4441 */
4442 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4443 char __user *optval, int __user *optlen)
4445 int val;
4446 struct sctp_sock *sp = sctp_sk(sk);
4448 if (len < sizeof(int))
4449 return -EINVAL;
4451 len = sizeof(int);
4452 val = sp->v4mapped;
4453 if (put_user(len, optlen))
4454 return -EFAULT;
4455 if (copy_to_user(optval, &val, len))
4456 return -EFAULT;
4458 return 0;
4461 /*
4462 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4464 * This socket option specifies the maximum size to put in any outgoing
4465 * SCTP chunk. If a message is larger than this size it will be
4466 * fragmented by SCTP into the specified size. Note that the underlying
4467 * SCTP implementation may fragment into smaller sized chunks when the
4468 * PMTU of the underlying association is smaller than the value set by
4469 * the user.
4470 */
4471 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4472 char __user *optval, int __user *optlen)
4474 int val;
4476 if (len < sizeof(int))
4477 return -EINVAL;
4479 len = sizeof(int);
4481 val = sctp_sk(sk)->user_frag;
4482 if (put_user(len, optlen))
4483 return -EFAULT;
4484 if (copy_to_user(optval, &val, len))
4485 return -EFAULT;
4487 return 0;
4490 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4491 char __user *optval, int __user *optlen)
4493 int retval = 0;
4494 int len;
4496 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4497 sk, optname);
4499 /* I can hardly begin to describe how wrong this is. This is
4500 * so broken as to be worse than useless. The API draft
4501 * REALLY is NOT helpful here... I am not convinced that the
4502 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4503 * are at all well-founded.
4504 */
4505 if (level != SOL_SCTP) {
4506 struct sctp_af *af = sctp_sk(sk)->pf->af;
4508 retval = af->getsockopt(sk, level, optname, optval, optlen);
4509 return retval;
4512 if (get_user(len, optlen))
4513 return -EFAULT;
4515 sctp_lock_sock(sk);
4517 switch (optname) {
4518 case SCTP_STATUS:
4519 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4520 break;
4521 case SCTP_DISABLE_FRAGMENTS:
4522 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4523 optlen);
4524 break;
4525 case SCTP_EVENTS:
4526 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4527 break;
4528 case SCTP_AUTOCLOSE:
4529 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4530 break;
4531 case SCTP_SOCKOPT_PEELOFF:
4532 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4533 break;
4534 case SCTP_PEER_ADDR_PARAMS:
4535 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4536 optlen);
4537 break;
4538 case SCTP_DELAYED_ACK_TIME:
4539 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4540 optlen);
4541 break;
4542 case SCTP_INITMSG:
4543 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4544 break;
4545 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4546 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4547 optlen);
4548 break;
4549 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4550 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4551 optlen);
4552 break;
4553 case SCTP_GET_PEER_ADDRS_OLD:
4554 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4555 optlen);
4556 break;
4557 case SCTP_GET_LOCAL_ADDRS_OLD:
4558 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4559 optlen);
4560 break;
4561 case SCTP_GET_PEER_ADDRS:
4562 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4563 optlen);
4564 break;
4565 case SCTP_GET_LOCAL_ADDRS:
4566 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4567 optlen);
4568 break;
4569 case SCTP_DEFAULT_SEND_PARAM:
4570 retval = sctp_getsockopt_default_send_param(sk, len,
4571 optval, optlen);
4572 break;
4573 case SCTP_PRIMARY_ADDR:
4574 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4575 break;
4576 case SCTP_NODELAY:
4577 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4578 break;
4579 case SCTP_RTOINFO:
4580 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4581 break;
4582 case SCTP_ASSOCINFO:
4583 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4584 break;
4585 case SCTP_I_WANT_MAPPED_V4_ADDR:
4586 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4587 break;
4588 case SCTP_MAXSEG:
4589 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4590 break;
4591 case SCTP_GET_PEER_ADDR_INFO:
4592 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4593 optlen);
4594 break;
4595 case SCTP_ADAPTION_LAYER:
4596 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4597 optlen);
4598 break;
4599 default:
4600 retval = -ENOPROTOOPT;
4601 break;
4602 };
4604 sctp_release_sock(sk);
4605 return retval;
4608 static void sctp_hash(struct sock *sk)
4610 /* STUB */
4613 static void sctp_unhash(struct sock *sk)
4615 /* STUB */
4618 /* Check if port is acceptable. Possibly find first available port.
4620 * The port hash table (contained in the 'global' SCTP protocol storage
4621 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4622 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4623 * list (the list number is the port number hashed out, so as you
4624 * would expect from a hash function, all the ports in a given list have
4625 * such a number that hashes out to the same list number; you were
4626 * expecting that, right?); so each list has a set of ports, with a
4627 * link to the socket (struct sock) that uses it, the port number and
4628 * a fastreuse flag (FIXME: NPI ipg).
4629 */
4630 static struct sctp_bind_bucket *sctp_bucket_create(
4631 struct sctp_bind_hashbucket *head, unsigned short snum);
4633 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4635 struct sctp_bind_hashbucket *head; /* hash list */
4636 struct sctp_bind_bucket *pp; /* hash list port iterator */
4637 unsigned short snum;
4638 int ret;
4640 /* NOTE: Remember to put this back to net order. */
4641 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4642 snum = addr->v4.sin_port;
4644 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4645 sctp_local_bh_disable();
4647 if (snum == 0) {
4648 /* Search for an available port.
4650 * 'sctp_port_rover' was the last port assigned, so
4651 * we start to search from 'sctp_port_rover +
4652 * 1'. What we do is first check if port 'rover' is
4653 * already in the hash table; if not, we use that; if
4654 * it is, we try next.
4655 */
4656 int low = sysctl_local_port_range[0];
4657 int high = sysctl_local_port_range[1];
4658 int remaining = (high - low) + 1;
4659 int rover;
4660 int index;
4662 sctp_spin_lock(&sctp_port_alloc_lock);
4663 rover = sctp_port_rover;
4664 do {
4665 rover++;
4666 if ((rover < low) || (rover > high))
4667 rover = low;
4668 index = sctp_phashfn(rover);
4669 head = &sctp_port_hashtable[index];
4670 sctp_spin_lock(&head->lock);
4671 for (pp = head->chain; pp; pp = pp->next)
4672 if (pp->port == rover)
4673 goto next;
4674 break;
4675 next:
4676 sctp_spin_unlock(&head->lock);
4677 } while (--remaining > 0);
4678 sctp_port_rover = rover;
4679 sctp_spin_unlock(&sctp_port_alloc_lock);
4681 /* Exhausted local port range during search? */
4682 ret = 1;
4683 if (remaining <= 0)
4684 goto fail;
4686 /* OK, here is the one we will use. HEAD (the port
4687 * hash table list entry) is non-NULL and we hold it's
4688 * mutex.
4689 */
4690 snum = rover;
4691 } else {
4692 /* We are given an specific port number; we verify
4693 * that it is not being used. If it is used, we will
4694 * exahust the search in the hash list corresponding
4695 * to the port number (snum) - we detect that with the
4696 * port iterator, pp being NULL.
4697 */
4698 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4699 sctp_spin_lock(&head->lock);
4700 for (pp = head->chain; pp; pp = pp->next) {
4701 if (pp->port == snum)
4702 goto pp_found;
4705 pp = NULL;
4706 goto pp_not_found;
4707 pp_found:
4708 if (!hlist_empty(&pp->owner)) {
4709 /* We had a port hash table hit - there is an
4710 * available port (pp != NULL) and it is being
4711 * used by other socket (pp->owner not empty); that other
4712 * socket is going to be sk2.
4713 */
4714 int reuse = sk->sk_reuse;
4715 struct sock *sk2;
4716 struct hlist_node *node;
4718 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4719 if (pp->fastreuse && sk->sk_reuse)
4720 goto success;
4722 /* Run through the list of sockets bound to the port
4723 * (pp->port) [via the pointers bind_next and
4724 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4725 * we get the endpoint they describe and run through
4726 * the endpoint's list of IP (v4 or v6) addresses,
4727 * comparing each of the addresses with the address of
4728 * the socket sk. If we find a match, then that means
4729 * that this port/socket (sk) combination are already
4730 * in an endpoint.
4731 */
4732 sk_for_each_bound(sk2, node, &pp->owner) {
4733 struct sctp_endpoint *ep2;
4734 ep2 = sctp_sk(sk2)->ep;
4736 if (reuse && sk2->sk_reuse)
4737 continue;
4739 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4740 sctp_sk(sk))) {
4741 ret = (long)sk2;
4742 goto fail_unlock;
4745 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4747 pp_not_found:
4748 /* If there was a hash table miss, create a new port. */
4749 ret = 1;
4750 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4751 goto fail_unlock;
4753 /* In either case (hit or miss), make sure fastreuse is 1 only
4754 * if sk->sk_reuse is too (that is, if the caller requested
4755 * SO_REUSEADDR on this socket -sk-).
4756 */
4757 if (hlist_empty(&pp->owner))
4758 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4759 else if (pp->fastreuse && !sk->sk_reuse)
4760 pp->fastreuse = 0;
4762 /* We are set, so fill up all the data in the hash table
4763 * entry, tie the socket list information with the rest of the
4764 * sockets FIXME: Blurry, NPI (ipg).
4765 */
4766 success:
4767 inet_sk(sk)->num = snum;
4768 if (!sctp_sk(sk)->bind_hash) {
4769 sk_add_bind_node(sk, &pp->owner);
4770 sctp_sk(sk)->bind_hash = pp;
4772 ret = 0;
4774 fail_unlock:
4775 sctp_spin_unlock(&head->lock);
4777 fail:
4778 sctp_local_bh_enable();
4779 addr->v4.sin_port = htons(addr->v4.sin_port);
4780 return ret;
4783 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4784 * port is requested.
4785 */
4786 static int sctp_get_port(struct sock *sk, unsigned short snum)
4788 long ret;
4789 union sctp_addr addr;
4790 struct sctp_af *af = sctp_sk(sk)->pf->af;
4792 /* Set up a dummy address struct from the sk. */
4793 af->from_sk(&addr, sk);
4794 addr.v4.sin_port = htons(snum);
4796 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4797 ret = sctp_get_port_local(sk, &addr);
4799 return (ret ? 1 : 0);
4802 /*
4803 * 3.1.3 listen() - UDP Style Syntax
4805 * By default, new associations are not accepted for UDP style sockets.
4806 * An application uses listen() to mark a socket as being able to
4807 * accept new associations.
4808 */
4809 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4811 struct sctp_sock *sp = sctp_sk(sk);
4812 struct sctp_endpoint *ep = sp->ep;
4814 /* Only UDP style sockets that are not peeled off are allowed to
4815 * listen().
4816 */
4817 if (!sctp_style(sk, UDP))
4818 return -EINVAL;
4820 /* If backlog is zero, disable listening. */
4821 if (!backlog) {
4822 if (sctp_sstate(sk, CLOSED))
4823 return 0;
4825 sctp_unhash_endpoint(ep);
4826 sk->sk_state = SCTP_SS_CLOSED;
4829 /* Return if we are already listening. */
4830 if (sctp_sstate(sk, LISTENING))
4831 return 0;
4833 /*
4834 * If a bind() or sctp_bindx() is not called prior to a listen()
4835 * call that allows new associations to be accepted, the system
4836 * picks an ephemeral port and will choose an address set equivalent
4837 * to binding with a wildcard address.
4839 * This is not currently spelled out in the SCTP sockets
4840 * extensions draft, but follows the practice as seen in TCP
4841 * sockets.
4842 */
4843 if (!ep->base.bind_addr.port) {
4844 if (sctp_autobind(sk))
4845 return -EAGAIN;
4847 sk->sk_state = SCTP_SS_LISTENING;
4848 sctp_hash_endpoint(ep);
4849 return 0;
4852 /*
4853 * 4.1.3 listen() - TCP Style Syntax
4855 * Applications uses listen() to ready the SCTP endpoint for accepting
4856 * inbound associations.
4857 */
4858 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4860 struct sctp_sock *sp = sctp_sk(sk);
4861 struct sctp_endpoint *ep = sp->ep;
4863 /* If backlog is zero, disable listening. */
4864 if (!backlog) {
4865 if (sctp_sstate(sk, CLOSED))
4866 return 0;
4868 sctp_unhash_endpoint(ep);
4869 sk->sk_state = SCTP_SS_CLOSED;
4872 if (sctp_sstate(sk, LISTENING))
4873 return 0;
4875 /*
4876 * If a bind() or sctp_bindx() is not called prior to a listen()
4877 * call that allows new associations to be accepted, the system
4878 * picks an ephemeral port and will choose an address set equivalent
4879 * to binding with a wildcard address.
4881 * This is not currently spelled out in the SCTP sockets
4882 * extensions draft, but follows the practice as seen in TCP
4883 * sockets.
4884 */
4885 if (!ep->base.bind_addr.port) {
4886 if (sctp_autobind(sk))
4887 return -EAGAIN;
4889 sk->sk_state = SCTP_SS_LISTENING;
4890 sk->sk_max_ack_backlog = backlog;
4891 sctp_hash_endpoint(ep);
4892 return 0;
4895 /*
4896 * Move a socket to LISTENING state.
4897 */
4898 int sctp_inet_listen(struct socket *sock, int backlog)
4900 struct sock *sk = sock->sk;
4901 struct crypto_tfm *tfm=NULL;
4902 int err = -EINVAL;
4904 if (unlikely(backlog < 0))
4905 goto out;
4907 sctp_lock_sock(sk);
4909 if (sock->state != SS_UNCONNECTED)
4910 goto out;
4912 /* Allocate HMAC for generating cookie. */
4913 if (sctp_hmac_alg) {
4914 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4915 if (!tfm) {
4916 err = -ENOSYS;
4917 goto out;
4921 switch (sock->type) {
4922 case SOCK_SEQPACKET:
4923 err = sctp_seqpacket_listen(sk, backlog);
4924 break;
4925 case SOCK_STREAM:
4926 err = sctp_stream_listen(sk, backlog);
4927 break;
4928 default:
4929 break;
4930 };
4931 if (err)
4932 goto cleanup;
4934 /* Store away the transform reference. */
4935 sctp_sk(sk)->hmac = tfm;
4936 out:
4937 sctp_release_sock(sk);
4938 return err;
4939 cleanup:
4940 sctp_crypto_free_tfm(tfm);
4941 goto out;
4944 /*
4945 * This function is done by modeling the current datagram_poll() and the
4946 * tcp_poll(). Note that, based on these implementations, we don't
4947 * lock the socket in this function, even though it seems that,
4948 * ideally, locking or some other mechanisms can be used to ensure
4949 * the integrity of the counters (sndbuf and wmem_alloc) used
4950 * in this place. We assume that we don't need locks either until proven
4951 * otherwise.
4953 * Another thing to note is that we include the Async I/O support
4954 * here, again, by modeling the current TCP/UDP code. We don't have
4955 * a good way to test with it yet.
4956 */
4957 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4959 struct sock *sk = sock->sk;
4960 struct sctp_sock *sp = sctp_sk(sk);
4961 unsigned int mask;
4963 poll_wait(file, sk->sk_sleep, wait);
4965 /* A TCP-style listening socket becomes readable when the accept queue
4966 * is not empty.
4967 */
4968 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4969 return (!list_empty(&sp->ep->asocs)) ?
4970 (POLLIN | POLLRDNORM) : 0;
4972 mask = 0;
4974 /* Is there any exceptional events? */
4975 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4976 mask |= POLLERR;
4977 if (sk->sk_shutdown & RCV_SHUTDOWN)
4978 mask |= POLLRDHUP;
4979 if (sk->sk_shutdown == SHUTDOWN_MASK)
4980 mask |= POLLHUP;
4982 /* Is it readable? Reconsider this code with TCP-style support. */
4983 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4984 (sk->sk_shutdown & RCV_SHUTDOWN))
4985 mask |= POLLIN | POLLRDNORM;
4987 /* The association is either gone or not ready. */
4988 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4989 return mask;
4991 /* Is it writable? */
4992 if (sctp_writeable(sk)) {
4993 mask |= POLLOUT | POLLWRNORM;
4994 } else {
4995 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4996 /*
4997 * Since the socket is not locked, the buffer
4998 * might be made available after the writeable check and
4999 * before the bit is set. This could cause a lost I/O
5000 * signal. tcp_poll() has a race breaker for this race
5001 * condition. Based on their implementation, we put
5002 * in the following code to cover it as well.
5003 */
5004 if (sctp_writeable(sk))
5005 mask |= POLLOUT | POLLWRNORM;
5007 return mask;
5010 /********************************************************************
5011 * 2nd Level Abstractions
5012 ********************************************************************/
5014 static struct sctp_bind_bucket *sctp_bucket_create(
5015 struct sctp_bind_hashbucket *head, unsigned short snum)
5017 struct sctp_bind_bucket *pp;
5019 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
5020 SCTP_DBG_OBJCNT_INC(bind_bucket);
5021 if (pp) {
5022 pp->port = snum;
5023 pp->fastreuse = 0;
5024 INIT_HLIST_HEAD(&pp->owner);
5025 if ((pp->next = head->chain) != NULL)
5026 pp->next->pprev = &pp->next;
5027 head->chain = pp;
5028 pp->pprev = &head->chain;
5030 return pp;
5033 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5034 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5036 if (pp && hlist_empty(&pp->owner)) {
5037 if (pp->next)
5038 pp->next->pprev = pp->pprev;
5039 *(pp->pprev) = pp->next;
5040 kmem_cache_free(sctp_bucket_cachep, pp);
5041 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5045 /* Release this socket's reference to a local port. */
5046 static inline void __sctp_put_port(struct sock *sk)
5048 struct sctp_bind_hashbucket *head =
5049 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5050 struct sctp_bind_bucket *pp;
5052 sctp_spin_lock(&head->lock);
5053 pp = sctp_sk(sk)->bind_hash;
5054 __sk_del_bind_node(sk);
5055 sctp_sk(sk)->bind_hash = NULL;
5056 inet_sk(sk)->num = 0;
5057 sctp_bucket_destroy(pp);
5058 sctp_spin_unlock(&head->lock);
5061 void sctp_put_port(struct sock *sk)
5063 sctp_local_bh_disable();
5064 __sctp_put_port(sk);
5065 sctp_local_bh_enable();
5068 /*
5069 * The system picks an ephemeral port and choose an address set equivalent
5070 * to binding with a wildcard address.
5071 * One of those addresses will be the primary address for the association.
5072 * This automatically enables the multihoming capability of SCTP.
5073 */
5074 static int sctp_autobind(struct sock *sk)
5076 union sctp_addr autoaddr;
5077 struct sctp_af *af;
5078 unsigned short port;
5080 /* Initialize a local sockaddr structure to INADDR_ANY. */
5081 af = sctp_sk(sk)->pf->af;
5083 port = htons(inet_sk(sk)->num);
5084 af->inaddr_any(&autoaddr, port);
5086 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5089 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5091 * From RFC 2292
5092 * 4.2 The cmsghdr Structure *
5094 * When ancillary data is sent or received, any number of ancillary data
5095 * objects can be specified by the msg_control and msg_controllen members of
5096 * the msghdr structure, because each object is preceded by
5097 * a cmsghdr structure defining the object's length (the cmsg_len member).
5098 * Historically Berkeley-derived implementations have passed only one object
5099 * at a time, but this API allows multiple objects to be
5100 * passed in a single call to sendmsg() or recvmsg(). The following example
5101 * shows two ancillary data objects in a control buffer.
5103 * |<--------------------------- msg_controllen -------------------------->|
5104 * | |
5106 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5108 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5109 * | | |
5111 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5113 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5114 * | | | | |
5116 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5117 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5119 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5121 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5122 * ^
5123 * |
5125 * msg_control
5126 * points here
5127 */
5128 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5129 sctp_cmsgs_t *cmsgs)
5131 struct cmsghdr *cmsg;
5133 for (cmsg = CMSG_FIRSTHDR(msg);
5134 cmsg != NULL;
5135 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5136 if (!CMSG_OK(msg, cmsg))
5137 return -EINVAL;
5139 /* Should we parse this header or ignore? */
5140 if (cmsg->cmsg_level != IPPROTO_SCTP)
5141 continue;
5143 /* Strictly check lengths following example in SCM code. */
5144 switch (cmsg->cmsg_type) {
5145 case SCTP_INIT:
5146 /* SCTP Socket API Extension
5147 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5149 * This cmsghdr structure provides information for
5150 * initializing new SCTP associations with sendmsg().
5151 * The SCTP_INITMSG socket option uses this same data
5152 * structure. This structure is not used for
5153 * recvmsg().
5155 * cmsg_level cmsg_type cmsg_data[]
5156 * ------------ ------------ ----------------------
5157 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5158 */
5159 if (cmsg->cmsg_len !=
5160 CMSG_LEN(sizeof(struct sctp_initmsg)))
5161 return -EINVAL;
5162 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5163 break;
5165 case SCTP_SNDRCV:
5166 /* SCTP Socket API Extension
5167 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5169 * This cmsghdr structure specifies SCTP options for
5170 * sendmsg() and describes SCTP header information
5171 * about a received message through recvmsg().
5173 * cmsg_level cmsg_type cmsg_data[]
5174 * ------------ ------------ ----------------------
5175 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5176 */
5177 if (cmsg->cmsg_len !=
5178 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5179 return -EINVAL;
5181 cmsgs->info =
5182 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5184 /* Minimally, validate the sinfo_flags. */
5185 if (cmsgs->info->sinfo_flags &
5186 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5187 SCTP_ABORT | SCTP_EOF))
5188 return -EINVAL;
5189 break;
5191 default:
5192 return -EINVAL;
5193 };
5195 return 0;
5198 /*
5199 * Wait for a packet..
5200 * Note: This function is the same function as in core/datagram.c
5201 * with a few modifications to make lksctp work.
5202 */
5203 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5205 int error;
5206 DEFINE_WAIT(wait);
5208 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5210 /* Socket errors? */
5211 error = sock_error(sk);
5212 if (error)
5213 goto out;
5215 if (!skb_queue_empty(&sk->sk_receive_queue))
5216 goto ready;
5218 /* Socket shut down? */
5219 if (sk->sk_shutdown & RCV_SHUTDOWN)
5220 goto out;
5222 /* Sequenced packets can come disconnected. If so we report the
5223 * problem.
5224 */
5225 error = -ENOTCONN;
5227 /* Is there a good reason to think that we may receive some data? */
5228 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5229 goto out;
5231 /* Handle signals. */
5232 if (signal_pending(current))
5233 goto interrupted;
5235 /* Let another process have a go. Since we are going to sleep
5236 * anyway. Note: This may cause odd behaviors if the message
5237 * does not fit in the user's buffer, but this seems to be the
5238 * only way to honor MSG_DONTWAIT realistically.
5239 */
5240 sctp_release_sock(sk);
5241 *timeo_p = schedule_timeout(*timeo_p);
5242 sctp_lock_sock(sk);
5244 ready:
5245 finish_wait(sk->sk_sleep, &wait);
5246 return 0;
5248 interrupted:
5249 error = sock_intr_errno(*timeo_p);
5251 out:
5252 finish_wait(sk->sk_sleep, &wait);
5253 *err = error;
5254 return error;
5257 /* Receive a datagram.
5258 * Note: This is pretty much the same routine as in core/datagram.c
5259 * with a few changes to make lksctp work.
5260 */
5261 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5262 int noblock, int *err)
5264 int error;
5265 struct sk_buff *skb;
5266 long timeo;
5268 timeo = sock_rcvtimeo(sk, noblock);
5270 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5271 timeo, MAX_SCHEDULE_TIMEOUT);
5273 do {
5274 /* Again only user level code calls this function,
5275 * so nothing interrupt level
5276 * will suddenly eat the receive_queue.
5278 * Look at current nfs client by the way...
5279 * However, this function was corrent in any case. 8)
5280 */
5281 if (flags & MSG_PEEK) {
5282 spin_lock_bh(&sk->sk_receive_queue.lock);
5283 skb = skb_peek(&sk->sk_receive_queue);
5284 if (skb)
5285 atomic_inc(&skb->users);
5286 spin_unlock_bh(&sk->sk_receive_queue.lock);
5287 } else {
5288 skb = skb_dequeue(&sk->sk_receive_queue);
5291 if (skb)
5292 return skb;
5294 /* Caller is allowed not to check sk->sk_err before calling. */
5295 error = sock_error(sk);
5296 if (error)
5297 goto no_packet;
5299 if (sk->sk_shutdown & RCV_SHUTDOWN)
5300 break;
5302 /* User doesn't want to wait. */
5303 error = -EAGAIN;
5304 if (!timeo)
5305 goto no_packet;
5306 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5308 return NULL;
5310 no_packet:
5311 *err = error;
5312 return NULL;
5315 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5316 static void __sctp_write_space(struct sctp_association *asoc)
5318 struct sock *sk = asoc->base.sk;
5319 struct socket *sock = sk->sk_socket;
5321 if ((sctp_wspace(asoc) > 0) && sock) {
5322 if (waitqueue_active(&asoc->wait))
5323 wake_up_interruptible(&asoc->wait);
5325 if (sctp_writeable(sk)) {
5326 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5327 wake_up_interruptible(sk->sk_sleep);
5329 /* Note that we try to include the Async I/O support
5330 * here by modeling from the current TCP/UDP code.
5331 * We have not tested with it yet.
5332 */
5333 if (sock->fasync_list &&
5334 !(sk->sk_shutdown & SEND_SHUTDOWN))
5335 sock_wake_async(sock, 2, POLL_OUT);
5340 /* Do accounting for the sndbuf space.
5341 * Decrement the used sndbuf space of the corresponding association by the
5342 * data size which was just transmitted(freed).
5343 */
5344 static void sctp_wfree(struct sk_buff *skb)
5346 struct sctp_association *asoc;
5347 struct sctp_chunk *chunk;
5348 struct sock *sk;
5350 /* Get the saved chunk pointer. */
5351 chunk = *((struct sctp_chunk **)(skb->cb));
5352 asoc = chunk->asoc;
5353 sk = asoc->base.sk;
5354 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5355 sizeof(struct sk_buff) +
5356 sizeof(struct sctp_chunk);
5358 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5360 sock_wfree(skb);
5361 __sctp_write_space(asoc);
5363 sctp_association_put(asoc);
5366 /* Helper function to wait for space in the sndbuf. */
5367 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5368 size_t msg_len)
5370 struct sock *sk = asoc->base.sk;
5371 int err = 0;
5372 long current_timeo = *timeo_p;
5373 DEFINE_WAIT(wait);
5375 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5376 asoc, (long)(*timeo_p), msg_len);
5378 /* Increment the association's refcnt. */
5379 sctp_association_hold(asoc);
5381 /* Wait on the association specific sndbuf space. */
5382 for (;;) {
5383 prepare_to_wait_exclusive(&asoc->wait, &wait,
5384 TASK_INTERRUPTIBLE);
5385 if (!*timeo_p)
5386 goto do_nonblock;
5387 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5388 asoc->base.dead)
5389 goto do_error;
5390 if (signal_pending(current))
5391 goto do_interrupted;
5392 if (msg_len <= sctp_wspace(asoc))
5393 break;
5395 /* Let another process have a go. Since we are going
5396 * to sleep anyway.
5397 */
5398 sctp_release_sock(sk);
5399 current_timeo = schedule_timeout(current_timeo);
5400 BUG_ON(sk != asoc->base.sk);
5401 sctp_lock_sock(sk);
5403 *timeo_p = current_timeo;
5406 out:
5407 finish_wait(&asoc->wait, &wait);
5409 /* Release the association's refcnt. */
5410 sctp_association_put(asoc);
5412 return err;
5414 do_error:
5415 err = -EPIPE;
5416 goto out;
5418 do_interrupted:
5419 err = sock_intr_errno(*timeo_p);
5420 goto out;
5422 do_nonblock:
5423 err = -EAGAIN;
5424 goto out;
5427 /* If socket sndbuf has changed, wake up all per association waiters. */
5428 void sctp_write_space(struct sock *sk)
5430 struct sctp_association *asoc;
5431 struct list_head *pos;
5433 /* Wake up the tasks in each wait queue. */
5434 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5435 asoc = list_entry(pos, struct sctp_association, asocs);
5436 __sctp_write_space(asoc);
5440 /* Is there any sndbuf space available on the socket?
5442 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5443 * associations on the same socket. For a UDP-style socket with
5444 * multiple associations, it is possible for it to be "unwriteable"
5445 * prematurely. I assume that this is acceptable because
5446 * a premature "unwriteable" is better than an accidental "writeable" which
5447 * would cause an unwanted block under certain circumstances. For the 1-1
5448 * UDP-style sockets or TCP-style sockets, this code should work.
5449 * - Daisy
5450 */
5451 static int sctp_writeable(struct sock *sk)
5453 int amt = 0;
5455 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5456 if (amt < 0)
5457 amt = 0;
5458 return amt;
5461 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5462 * returns immediately with EINPROGRESS.
5463 */
5464 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5466 struct sock *sk = asoc->base.sk;
5467 int err = 0;
5468 long current_timeo = *timeo_p;
5469 DEFINE_WAIT(wait);
5471 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5472 (long)(*timeo_p));
5474 /* Increment the association's refcnt. */
5475 sctp_association_hold(asoc);
5477 for (;;) {
5478 prepare_to_wait_exclusive(&asoc->wait, &wait,
5479 TASK_INTERRUPTIBLE);
5480 if (!*timeo_p)
5481 goto do_nonblock;
5482 if (sk->sk_shutdown & RCV_SHUTDOWN)
5483 break;
5484 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5485 asoc->base.dead)
5486 goto do_error;
5487 if (signal_pending(current))
5488 goto do_interrupted;
5490 if (sctp_state(asoc, ESTABLISHED))
5491 break;
5493 /* Let another process have a go. Since we are going
5494 * to sleep anyway.
5495 */
5496 sctp_release_sock(sk);
5497 current_timeo = schedule_timeout(current_timeo);
5498 sctp_lock_sock(sk);
5500 *timeo_p = current_timeo;
5503 out:
5504 finish_wait(&asoc->wait, &wait);
5506 /* Release the association's refcnt. */
5507 sctp_association_put(asoc);
5509 return err;
5511 do_error:
5512 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5513 err = -ETIMEDOUT;
5514 else
5515 err = -ECONNREFUSED;
5516 goto out;
5518 do_interrupted:
5519 err = sock_intr_errno(*timeo_p);
5520 goto out;
5522 do_nonblock:
5523 err = -EINPROGRESS;
5524 goto out;
5527 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5529 struct sctp_endpoint *ep;
5530 int err = 0;
5531 DEFINE_WAIT(wait);
5533 ep = sctp_sk(sk)->ep;
5536 for (;;) {
5537 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5538 TASK_INTERRUPTIBLE);
5540 if (list_empty(&ep->asocs)) {
5541 sctp_release_sock(sk);
5542 timeo = schedule_timeout(timeo);
5543 sctp_lock_sock(sk);
5546 err = -EINVAL;
5547 if (!sctp_sstate(sk, LISTENING))
5548 break;
5550 err = 0;
5551 if (!list_empty(&ep->asocs))
5552 break;
5554 err = sock_intr_errno(timeo);
5555 if (signal_pending(current))
5556 break;
5558 err = -EAGAIN;
5559 if (!timeo)
5560 break;
5563 finish_wait(sk->sk_sleep, &wait);
5565 return err;
5568 void sctp_wait_for_close(struct sock *sk, long timeout)
5570 DEFINE_WAIT(wait);
5572 do {
5573 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5574 if (list_empty(&sctp_sk(sk)->ep->asocs))
5575 break;
5576 sctp_release_sock(sk);
5577 timeout = schedule_timeout(timeout);
5578 sctp_lock_sock(sk);
5579 } while (!signal_pending(current) && timeout);
5581 finish_wait(sk->sk_sleep, &wait);
5584 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5585 * and its messages to the newsk.
5586 */
5587 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5588 struct sctp_association *assoc,
5589 sctp_socket_type_t type)
5591 struct sctp_sock *oldsp = sctp_sk(oldsk);
5592 struct sctp_sock *newsp = sctp_sk(newsk);
5593 struct sctp_bind_bucket *pp; /* hash list port iterator */
5594 struct sctp_endpoint *newep = newsp->ep;
5595 struct sk_buff *skb, *tmp;
5596 struct sctp_ulpevent *event;
5597 int flags = 0;
5599 /* Migrate socket buffer sizes and all the socket level options to the
5600 * new socket.
5601 */
5602 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5603 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5604 /* Brute force copy old sctp opt. */
5605 inet_sk_copy_descendant(newsk, oldsk);
5607 /* Restore the ep value that was overwritten with the above structure
5608 * copy.
5609 */
5610 newsp->ep = newep;
5611 newsp->hmac = NULL;
5613 /* Hook this new socket in to the bind_hash list. */
5614 pp = sctp_sk(oldsk)->bind_hash;
5615 sk_add_bind_node(newsk, &pp->owner);
5616 sctp_sk(newsk)->bind_hash = pp;
5617 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5619 /* Copy the bind_addr list from the original endpoint to the new
5620 * endpoint so that we can handle restarts properly
5621 */
5622 if (assoc->peer.ipv4_address)
5623 flags |= SCTP_ADDR4_PEERSUPP;
5624 if (assoc->peer.ipv6_address)
5625 flags |= SCTP_ADDR6_PEERSUPP;
5626 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5627 &oldsp->ep->base.bind_addr,
5628 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5630 /* Move any messages in the old socket's receive queue that are for the
5631 * peeled off association to the new socket's receive queue.
5632 */
5633 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5634 event = sctp_skb2event(skb);
5635 if (event->asoc == assoc) {
5636 sock_rfree(skb);
5637 __skb_unlink(skb, &oldsk->sk_receive_queue);
5638 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5639 skb_set_owner_r(skb, newsk);
5643 /* Clean up any messages pending delivery due to partial
5644 * delivery. Three cases:
5645 * 1) No partial deliver; no work.
5646 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5647 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5648 */
5649 skb_queue_head_init(&newsp->pd_lobby);
5650 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5652 if (sctp_sk(oldsk)->pd_mode) {
5653 struct sk_buff_head *queue;
5655 /* Decide which queue to move pd_lobby skbs to. */
5656 if (assoc->ulpq.pd_mode) {
5657 queue = &newsp->pd_lobby;
5658 } else
5659 queue = &newsk->sk_receive_queue;
5661 /* Walk through the pd_lobby, looking for skbs that
5662 * need moved to the new socket.
5663 */
5664 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5665 event = sctp_skb2event(skb);
5666 if (event->asoc == assoc) {
5667 sock_rfree(skb);
5668 __skb_unlink(skb, &oldsp->pd_lobby);
5669 __skb_queue_tail(queue, skb);
5670 skb_set_owner_r(skb, newsk);
5674 /* Clear up any skbs waiting for the partial
5675 * delivery to finish.
5676 */
5677 if (assoc->ulpq.pd_mode)
5678 sctp_clear_pd(oldsk);
5682 /* Set the type of socket to indicate that it is peeled off from the
5683 * original UDP-style socket or created with the accept() call on a
5684 * TCP-style socket..
5685 */
5686 newsp->type = type;
5688 /* Mark the new socket "in-use" by the user so that any packets
5689 * that may arrive on the association after we've moved it are
5690 * queued to the backlog. This prevents a potential race between
5691 * backlog processing on the old socket and new-packet processing
5692 * on the new socket.
5693 */
5694 sctp_lock_sock(newsk);
5695 sctp_assoc_migrate(assoc, newsk);
5697 /* If the association on the newsk is already closed before accept()
5698 * is called, set RCV_SHUTDOWN flag.
5699 */
5700 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5701 newsk->sk_shutdown |= RCV_SHUTDOWN;
5703 newsk->sk_state = SCTP_SS_ESTABLISHED;
5704 sctp_release_sock(newsk);
5707 /* This proto struct describes the ULP interface for SCTP. */
5708 struct proto sctp_prot = {
5709 .name = "SCTP",
5710 .owner = THIS_MODULE,
5711 .close = sctp_close,
5712 .connect = sctp_connect,
5713 .disconnect = sctp_disconnect,
5714 .accept = sctp_accept,
5715 .ioctl = sctp_ioctl,
5716 .init = sctp_init_sock,
5717 .destroy = sctp_destroy_sock,
5718 .shutdown = sctp_shutdown,
5719 .setsockopt = sctp_setsockopt,
5720 .getsockopt = sctp_getsockopt,
5721 .sendmsg = sctp_sendmsg,
5722 .recvmsg = sctp_recvmsg,
5723 .bind = sctp_bind,
5724 .backlog_rcv = sctp_backlog_rcv,
5725 .hash = sctp_hash,
5726 .unhash = sctp_unhash,
5727 .get_port = sctp_get_port,
5728 .obj_size = sizeof(struct sctp_sock),
5729 };
5731 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5732 struct proto sctpv6_prot = {
5733 .name = "SCTPv6",
5734 .owner = THIS_MODULE,
5735 .close = sctp_close,
5736 .connect = sctp_connect,
5737 .disconnect = sctp_disconnect,
5738 .accept = sctp_accept,
5739 .ioctl = sctp_ioctl,
5740 .init = sctp_init_sock,
5741 .destroy = sctp_destroy_sock,
5742 .shutdown = sctp_shutdown,
5743 .setsockopt = sctp_setsockopt,
5744 .getsockopt = sctp_getsockopt,
5745 .sendmsg = sctp_sendmsg,
5746 .recvmsg = sctp_recvmsg,
5747 .bind = sctp_bind,
5748 .backlog_rcv = sctp_backlog_rcv,
5749 .hash = sctp_hash,
5750 .unhash = sctp_unhash,
5751 .get_port = sctp_get_port,
5752 .obj_size = sizeof(struct sctp6_sock),
5753 };
5754 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */