[SCTP]: sctp_has_association() switched to net-endian.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blobe118e3c1285ef13615fb6d5ee9b29ac33604d6e1
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
9 * This file is part of the SCTP kernel reference Implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
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.
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.
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.
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.
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
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>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
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.
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)
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);
126 if (amt < 0)
127 amt = 0;
129 return amt;
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.
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.
141 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
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);
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)
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;
183 /* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
186 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
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.
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;
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;
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.
224 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
226 sctp_assoc_t id)
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 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
233 laddr,
234 &transport);
236 if (!addr_asoc)
237 return NULL;
239 id_asoc = sctp_id2assoc(sk, id);
240 if (id_asoc && (id_asoc != addr_asoc))
241 return NULL;
243 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
244 (union sctp_addr *)addr);
246 return transport;
249 /* API 3.1.2 bind() - UDP Style Syntax
250 * The syntax of bind() is,
252 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
254 * sd - the socket descriptor returned by socket().
255 * addr - the address structure (struct sockaddr_in or struct
256 * sockaddr_in6 [RFC 2553]),
257 * addr_len - the size of the address structure.
259 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
261 int retval = 0;
263 sctp_lock_sock(sk);
265 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
266 sk, addr, addr_len);
268 /* Disallow binding twice. */
269 if (!sctp_sk(sk)->ep->base.bind_addr.port)
270 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
271 addr_len);
272 else
273 retval = -EINVAL;
275 sctp_release_sock(sk);
277 return retval;
280 static long sctp_get_port_local(struct sock *, union sctp_addr *);
282 /* Verify this is a valid sockaddr. */
283 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
284 union sctp_addr *addr, int len)
286 struct sctp_af *af;
288 /* Check minimum size. */
289 if (len < sizeof (struct sockaddr))
290 return NULL;
292 /* Does this PF support this AF? */
293 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
294 return NULL;
296 /* If we get this far, af is valid. */
297 af = sctp_get_af_specific(addr->sa.sa_family);
299 if (len < af->sockaddr_len)
300 return NULL;
302 return af;
305 /* Bind a local address either to an endpoint or to an association. */
306 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
308 struct sctp_sock *sp = sctp_sk(sk);
309 struct sctp_endpoint *ep = sp->ep;
310 struct sctp_bind_addr *bp = &ep->base.bind_addr;
311 struct sctp_af *af;
312 unsigned short snum;
313 int ret = 0;
315 /* Common sockaddr verification. */
316 af = sctp_sockaddr_af(sp, addr, len);
317 if (!af) {
318 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
319 sk, addr, len);
320 return -EINVAL;
323 snum = ntohs(addr->v4.sin_port);
325 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
326 ", port: %d, new port: %d, len: %d)\n",
328 addr,
329 bp->port, snum,
330 len);
332 /* PF specific bind() address verification. */
333 if (!sp->pf->bind_verify(sp, addr))
334 return -EADDRNOTAVAIL;
336 /* We must either be unbound, or bind to the same port. */
337 if (bp->port && (snum != bp->port)) {
338 SCTP_DEBUG_PRINTK("sctp_do_bind:"
339 " New port %d does not match existing port "
340 "%d.\n", snum, bp->port);
341 return -EINVAL;
344 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
345 return -EACCES;
347 /* Make sure we are allowed to bind here.
348 * The function sctp_get_port_local() does duplicate address
349 * detection.
351 if ((ret = sctp_get_port_local(sk, addr))) {
352 if (ret == (long) sk) {
353 /* This endpoint has a conflicting address. */
354 return -EINVAL;
355 } else {
356 return -EADDRINUSE;
360 /* Refresh ephemeral port. */
361 if (!bp->port)
362 bp->port = inet_sk(sk)->num;
364 /* Add the address to the bind address list. */
365 sctp_local_bh_disable();
366 sctp_write_lock(&ep->base.addr_lock);
368 /* Use GFP_ATOMIC since BHs are disabled. */
369 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
370 sctp_write_unlock(&ep->base.addr_lock);
371 sctp_local_bh_enable();
373 /* Copy back into socket for getsockname() use. */
374 if (!ret) {
375 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
376 af->to_sk_saddr(addr, sk);
379 return ret;
382 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
384 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
385 * at any one time. If a sender, after sending an ASCONF chunk, decides
386 * it needs to transfer another ASCONF Chunk, it MUST wait until the
387 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
388 * subsequent ASCONF. Note this restriction binds each side, so at any
389 * time two ASCONF may be in-transit on any given association (one sent
390 * from each endpoint).
392 static int sctp_send_asconf(struct sctp_association *asoc,
393 struct sctp_chunk *chunk)
395 int retval = 0;
397 /* If there is an outstanding ASCONF chunk, queue it for later
398 * transmission.
400 if (asoc->addip_last_asconf) {
401 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
402 goto out;
405 /* Hold the chunk until an ASCONF_ACK is received. */
406 sctp_chunk_hold(chunk);
407 retval = sctp_primitive_ASCONF(asoc, chunk);
408 if (retval)
409 sctp_chunk_free(chunk);
410 else
411 asoc->addip_last_asconf = chunk;
413 out:
414 return retval;
417 /* Add a list of addresses as bind addresses to local endpoint or
418 * association.
420 * Basically run through each address specified in the addrs/addrcnt
421 * array/length pair, determine if it is IPv6 or IPv4 and call
422 * sctp_do_bind() on it.
424 * If any of them fails, then the operation will be reversed and the
425 * ones that were added will be removed.
427 * Only sctp_setsockopt_bindx() is supposed to call this function.
429 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
431 int cnt;
432 int retval = 0;
433 void *addr_buf;
434 struct sockaddr *sa_addr;
435 struct sctp_af *af;
437 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
438 sk, addrs, addrcnt);
440 addr_buf = addrs;
441 for (cnt = 0; cnt < addrcnt; cnt++) {
442 /* The list may contain either IPv4 or IPv6 address;
443 * determine the address length for walking thru the list.
445 sa_addr = (struct sockaddr *)addr_buf;
446 af = sctp_get_af_specific(sa_addr->sa_family);
447 if (!af) {
448 retval = -EINVAL;
449 goto err_bindx_add;
452 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
453 af->sockaddr_len);
455 addr_buf += af->sockaddr_len;
457 err_bindx_add:
458 if (retval < 0) {
459 /* Failed. Cleanup the ones that have been added */
460 if (cnt > 0)
461 sctp_bindx_rem(sk, addrs, cnt);
462 return retval;
466 return retval;
469 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
470 * associations that are part of the endpoint indicating that a list of local
471 * addresses are added to the endpoint.
473 * If any of the addresses is already in the bind address list of the
474 * association, we do not send the chunk for that association. But it will not
475 * affect other associations.
477 * Only sctp_setsockopt_bindx() is supposed to call this function.
479 static int sctp_send_asconf_add_ip(struct sock *sk,
480 struct sockaddr *addrs,
481 int addrcnt)
483 struct sctp_sock *sp;
484 struct sctp_endpoint *ep;
485 struct sctp_association *asoc;
486 struct sctp_bind_addr *bp;
487 struct sctp_chunk *chunk;
488 struct sctp_sockaddr_entry *laddr;
489 union sctp_addr *addr;
490 union sctp_addr saveaddr;
491 void *addr_buf;
492 struct sctp_af *af;
493 struct list_head *pos;
494 struct list_head *p;
495 int i;
496 int retval = 0;
498 if (!sctp_addip_enable)
499 return retval;
501 sp = sctp_sk(sk);
502 ep = sp->ep;
504 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
505 __FUNCTION__, sk, addrs, addrcnt);
507 list_for_each(pos, &ep->asocs) {
508 asoc = list_entry(pos, struct sctp_association, asocs);
510 if (!asoc->peer.asconf_capable)
511 continue;
513 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
514 continue;
516 if (!sctp_state(asoc, ESTABLISHED))
517 continue;
519 /* Check if any address in the packed array of addresses is
520 * in the bind address list of the association. If so,
521 * do not send the asconf chunk to its peer, but continue with
522 * other associations.
524 addr_buf = addrs;
525 for (i = 0; i < addrcnt; i++) {
526 addr = (union sctp_addr *)addr_buf;
527 af = sctp_get_af_specific(addr->v4.sin_family);
528 if (!af) {
529 retval = -EINVAL;
530 goto out;
533 if (sctp_assoc_lookup_laddr(asoc, addr))
534 break;
536 addr_buf += af->sockaddr_len;
538 if (i < addrcnt)
539 continue;
541 /* Use the first address in bind addr list of association as
542 * Address Parameter of ASCONF CHUNK.
544 sctp_read_lock(&asoc->base.addr_lock);
545 bp = &asoc->base.bind_addr;
546 p = bp->address_list.next;
547 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
548 sctp_read_unlock(&asoc->base.addr_lock);
550 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a_h, addrs,
551 addrcnt, SCTP_PARAM_ADD_IP);
552 if (!chunk) {
553 retval = -ENOMEM;
554 goto out;
557 retval = sctp_send_asconf(asoc, chunk);
558 if (retval)
559 goto out;
561 /* Add the new addresses to the bind address list with
562 * use_as_src set to 0.
564 sctp_local_bh_disable();
565 sctp_write_lock(&asoc->base.addr_lock);
566 addr_buf = addrs;
567 for (i = 0; i < addrcnt; i++) {
568 addr = (union sctp_addr *)addr_buf;
569 af = sctp_get_af_specific(addr->v4.sin_family);
570 memcpy(&saveaddr, addr, af->sockaddr_len);
571 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
572 GFP_ATOMIC);
573 addr_buf += af->sockaddr_len;
575 sctp_write_unlock(&asoc->base.addr_lock);
576 sctp_local_bh_enable();
579 out:
580 return retval;
583 /* Remove a list of addresses from bind addresses list. Do not remove the
584 * last address.
586 * Basically run through each address specified in the addrs/addrcnt
587 * array/length pair, determine if it is IPv6 or IPv4 and call
588 * sctp_del_bind() on it.
590 * If any of them fails, then the operation will be reversed and the
591 * ones that were removed will be added back.
593 * At least one address has to be left; if only one address is
594 * available, the operation will return -EBUSY.
596 * Only sctp_setsockopt_bindx() is supposed to call this function.
598 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
600 struct sctp_sock *sp = sctp_sk(sk);
601 struct sctp_endpoint *ep = sp->ep;
602 int cnt;
603 struct sctp_bind_addr *bp = &ep->base.bind_addr;
604 int retval = 0;
605 void *addr_buf;
606 union sctp_addr *sa_addr;
607 struct sctp_af *af;
609 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
610 sk, addrs, addrcnt);
612 addr_buf = addrs;
613 for (cnt = 0; cnt < addrcnt; cnt++) {
614 /* If the bind address list is empty or if there is only one
615 * bind address, there is nothing more to be removed (we need
616 * at least one address here).
618 if (list_empty(&bp->address_list) ||
619 (sctp_list_single_entry(&bp->address_list))) {
620 retval = -EBUSY;
621 goto err_bindx_rem;
624 sa_addr = (union sctp_addr *)addr_buf;
625 af = sctp_get_af_specific(sa_addr->sa.sa_family);
626 if (!af) {
627 retval = -EINVAL;
628 goto err_bindx_rem;
630 if (sa_addr->v4.sin_port != htons(bp->port)) {
631 retval = -EINVAL;
632 goto err_bindx_rem;
635 /* FIXME - There is probably a need to check if sk->sk_saddr and
636 * sk->sk_rcv_addr are currently set to one of the addresses to
637 * be removed. This is something which needs to be looked into
638 * when we are fixing the outstanding issues with multi-homing
639 * socket routing and failover schemes. Refer to comments in
640 * sctp_do_bind(). -daisy
642 sctp_local_bh_disable();
643 sctp_write_lock(&ep->base.addr_lock);
645 retval = sctp_del_bind_addr(bp, sa_addr);
647 sctp_write_unlock(&ep->base.addr_lock);
648 sctp_local_bh_enable();
650 addr_buf += af->sockaddr_len;
651 err_bindx_rem:
652 if (retval < 0) {
653 /* Failed. Add the ones that has been removed back */
654 if (cnt > 0)
655 sctp_bindx_add(sk, addrs, cnt);
656 return retval;
660 return retval;
663 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
664 * the associations that are part of the endpoint indicating that a list of
665 * local addresses are removed from the endpoint.
667 * If any of the addresses is already in the bind address list of the
668 * association, we do not send the chunk for that association. But it will not
669 * affect other associations.
671 * Only sctp_setsockopt_bindx() is supposed to call this function.
673 static int sctp_send_asconf_del_ip(struct sock *sk,
674 struct sockaddr *addrs,
675 int addrcnt)
677 struct sctp_sock *sp;
678 struct sctp_endpoint *ep;
679 struct sctp_association *asoc;
680 struct sctp_transport *transport;
681 struct sctp_bind_addr *bp;
682 struct sctp_chunk *chunk;
683 union sctp_addr *laddr;
684 void *addr_buf;
685 struct sctp_af *af;
686 struct list_head *pos, *pos1;
687 struct sctp_sockaddr_entry *saddr;
688 int i;
689 int retval = 0;
691 if (!sctp_addip_enable)
692 return retval;
694 sp = sctp_sk(sk);
695 ep = sp->ep;
697 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
698 __FUNCTION__, sk, addrs, addrcnt);
700 list_for_each(pos, &ep->asocs) {
701 asoc = list_entry(pos, struct sctp_association, asocs);
703 if (!asoc->peer.asconf_capable)
704 continue;
706 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
707 continue;
709 if (!sctp_state(asoc, ESTABLISHED))
710 continue;
712 /* Check if any address in the packed array of addresses is
713 * not present in the bind address list of the association.
714 * If so, do not send the asconf chunk to its peer, but
715 * continue with other associations.
717 addr_buf = addrs;
718 for (i = 0; i < addrcnt; i++) {
719 laddr = (union sctp_addr *)addr_buf;
720 af = sctp_get_af_specific(laddr->v4.sin_family);
721 if (!af) {
722 retval = -EINVAL;
723 goto out;
726 if (!sctp_assoc_lookup_laddr(asoc, laddr))
727 break;
729 addr_buf += af->sockaddr_len;
731 if (i < addrcnt)
732 continue;
734 /* Find one address in the association's bind address list
735 * that is not in the packed array of addresses. This is to
736 * make sure that we do not delete all the addresses in the
737 * association.
739 sctp_read_lock(&asoc->base.addr_lock);
740 bp = &asoc->base.bind_addr;
741 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
742 addrcnt, sp);
743 sctp_read_unlock(&asoc->base.addr_lock);
744 if (!laddr)
745 continue;
747 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
748 SCTP_PARAM_DEL_IP);
749 if (!chunk) {
750 retval = -ENOMEM;
751 goto out;
754 /* Reset use_as_src flag for the addresses in the bind address
755 * list that are to be deleted.
757 sctp_local_bh_disable();
758 sctp_write_lock(&asoc->base.addr_lock);
759 addr_buf = addrs;
760 for (i = 0; i < addrcnt; i++) {
761 laddr = (union sctp_addr *)addr_buf;
762 af = sctp_get_af_specific(laddr->v4.sin_family);
763 list_for_each(pos1, &bp->address_list) {
764 saddr = list_entry(pos1,
765 struct sctp_sockaddr_entry,
766 list);
767 if (sctp_cmp_addr_exact(&saddr->a, laddr))
768 saddr->use_as_src = 0;
770 addr_buf += af->sockaddr_len;
772 sctp_write_unlock(&asoc->base.addr_lock);
773 sctp_local_bh_enable();
775 /* Update the route and saddr entries for all the transports
776 * as some of the addresses in the bind address list are
777 * about to be deleted and cannot be used as source addresses.
779 list_for_each(pos1, &asoc->peer.transport_addr_list) {
780 transport = list_entry(pos1, struct sctp_transport,
781 transports);
782 dst_release(transport->dst);
783 sctp_transport_route(transport, NULL,
784 sctp_sk(asoc->base.sk));
787 retval = sctp_send_asconf(asoc, chunk);
789 out:
790 return retval;
793 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
795 * API 8.1
796 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
797 * int flags);
799 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
800 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
801 * or IPv6 addresses.
803 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
804 * Section 3.1.2 for this usage.
806 * addrs is a pointer to an array of one or more socket addresses. Each
807 * address is contained in its appropriate structure (i.e. struct
808 * sockaddr_in or struct sockaddr_in6) the family of the address type
809 * must be used to distinguish the address length (note that this
810 * representation is termed a "packed array" of addresses). The caller
811 * specifies the number of addresses in the array with addrcnt.
813 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
814 * -1, and sets errno to the appropriate error code.
816 * For SCTP, the port given in each socket address must be the same, or
817 * sctp_bindx() will fail, setting errno to EINVAL.
819 * The flags parameter is formed from the bitwise OR of zero or more of
820 * the following currently defined flags:
822 * SCTP_BINDX_ADD_ADDR
824 * SCTP_BINDX_REM_ADDR
826 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
827 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
828 * addresses from the association. The two flags are mutually exclusive;
829 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
830 * not remove all addresses from an association; sctp_bindx() will
831 * reject such an attempt with EINVAL.
833 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
834 * additional addresses with an endpoint after calling bind(). Or use
835 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
836 * socket is associated with so that no new association accepted will be
837 * associated with those addresses. If the endpoint supports dynamic
838 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
839 * endpoint to send the appropriate message to the peer to change the
840 * peers address lists.
842 * Adding and removing addresses from a connected association is
843 * optional functionality. Implementations that do not support this
844 * functionality should return EOPNOTSUPP.
846 * Basically do nothing but copying the addresses from user to kernel
847 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
848 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
849 * from userspace.
851 * We don't use copy_from_user() for optimization: we first do the
852 * sanity checks (buffer size -fast- and access check-healthy
853 * pointer); if all of those succeed, then we can alloc the memory
854 * (expensive operation) needed to copy the data to kernel. Then we do
855 * the copying without checking the user space area
856 * (__copy_from_user()).
858 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
859 * it.
861 * sk The sk of the socket
862 * addrs The pointer to the addresses in user land
863 * addrssize Size of the addrs buffer
864 * op Operation to perform (add or remove, see the flags of
865 * sctp_bindx)
867 * Returns 0 if ok, <0 errno code on error.
869 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
870 struct sockaddr __user *addrs,
871 int addrs_size, int op)
873 struct sockaddr *kaddrs;
874 int err;
875 int addrcnt = 0;
876 int walk_size = 0;
877 struct sockaddr *sa_addr;
878 void *addr_buf;
879 struct sctp_af *af;
881 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
882 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
884 if (unlikely(addrs_size <= 0))
885 return -EINVAL;
887 /* Check the user passed a healthy pointer. */
888 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
889 return -EFAULT;
891 /* Alloc space for the address array in kernel memory. */
892 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
893 if (unlikely(!kaddrs))
894 return -ENOMEM;
896 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
897 kfree(kaddrs);
898 return -EFAULT;
901 /* Walk through the addrs buffer and count the number of addresses. */
902 addr_buf = kaddrs;
903 while (walk_size < addrs_size) {
904 sa_addr = (struct sockaddr *)addr_buf;
905 af = sctp_get_af_specific(sa_addr->sa_family);
907 /* If the address family is not supported or if this address
908 * causes the address buffer to overflow return EINVAL.
910 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
911 kfree(kaddrs);
912 return -EINVAL;
914 addrcnt++;
915 addr_buf += af->sockaddr_len;
916 walk_size += af->sockaddr_len;
919 /* Do the work. */
920 switch (op) {
921 case SCTP_BINDX_ADD_ADDR:
922 err = sctp_bindx_add(sk, kaddrs, addrcnt);
923 if (err)
924 goto out;
925 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
926 break;
928 case SCTP_BINDX_REM_ADDR:
929 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
930 if (err)
931 goto out;
932 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
933 break;
935 default:
936 err = -EINVAL;
937 break;
940 out:
941 kfree(kaddrs);
943 return err;
946 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
948 * Common routine for handling connect() and sctp_connectx().
949 * Connect will come in with just a single address.
951 static int __sctp_connect(struct sock* sk,
952 struct sockaddr *kaddrs,
953 int addrs_size)
955 struct sctp_sock *sp;
956 struct sctp_endpoint *ep;
957 struct sctp_association *asoc = NULL;
958 struct sctp_association *asoc2;
959 struct sctp_transport *transport;
960 union sctp_addr to;
961 struct sctp_af *af;
962 sctp_scope_t scope;
963 long timeo;
964 int err = 0;
965 int addrcnt = 0;
966 int walk_size = 0;
967 union sctp_addr *sa_addr;
968 void *addr_buf;
970 sp = sctp_sk(sk);
971 ep = sp->ep;
973 /* connect() cannot be done on a socket that is already in ESTABLISHED
974 * state - UDP-style peeled off socket or a TCP-style socket that
975 * is already connected.
976 * It cannot be done even on a TCP-style listening socket.
978 if (sctp_sstate(sk, ESTABLISHED) ||
979 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
980 err = -EISCONN;
981 goto out_free;
984 /* Walk through the addrs buffer and count the number of addresses. */
985 addr_buf = kaddrs;
986 while (walk_size < addrs_size) {
987 sa_addr = (union sctp_addr *)addr_buf;
988 af = sctp_get_af_specific(sa_addr->sa.sa_family);
990 /* If the address family is not supported or if this address
991 * causes the address buffer to overflow return EINVAL.
993 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
994 err = -EINVAL;
995 goto out_free;
998 err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
999 if (err)
1000 goto out_free;
1002 memcpy(&to, sa_addr, af->sockaddr_len);
1003 to.v4.sin_port = ntohs(to.v4.sin_port);
1005 /* Check if there already is a matching association on the
1006 * endpoint (other than the one created here).
1008 asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1009 if (asoc2 && asoc2 != asoc) {
1010 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1011 err = -EISCONN;
1012 else
1013 err = -EALREADY;
1014 goto out_free;
1017 /* If we could not find a matching association on the endpoint,
1018 * make sure that there is no peeled-off association matching
1019 * the peer address even on another socket.
1021 if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1022 err = -EADDRNOTAVAIL;
1023 goto out_free;
1026 if (!asoc) {
1027 /* If a bind() or sctp_bindx() is not called prior to
1028 * an sctp_connectx() call, the system picks an
1029 * ephemeral port and will choose an address set
1030 * equivalent to binding with a wildcard address.
1032 if (!ep->base.bind_addr.port) {
1033 if (sctp_autobind(sk)) {
1034 err = -EAGAIN;
1035 goto out_free;
1037 } else {
1039 * If an unprivileged user inherits a 1-many
1040 * style socket with open associations on a
1041 * privileged port, it MAY be permitted to
1042 * accept new associations, but it SHOULD NOT
1043 * be permitted to open new associations.
1045 if (ep->base.bind_addr.port < PROT_SOCK &&
1046 !capable(CAP_NET_BIND_SERVICE)) {
1047 err = -EACCES;
1048 goto out_free;
1052 scope = sctp_scope(&to);
1053 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1054 if (!asoc) {
1055 err = -ENOMEM;
1056 goto out_free;
1060 /* Prime the peer's transport structures. */
1061 transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1062 SCTP_UNKNOWN);
1063 if (!transport) {
1064 err = -ENOMEM;
1065 goto out_free;
1068 addrcnt++;
1069 addr_buf += af->sockaddr_len;
1070 walk_size += af->sockaddr_len;
1073 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1074 if (err < 0) {
1075 goto out_free;
1078 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1079 if (err < 0) {
1080 goto out_free;
1083 /* Initialize sk's dport and daddr for getpeername() */
1084 inet_sk(sk)->dport = htons(asoc->peer.port);
1085 af = sctp_get_af_specific(to.sa.sa_family);
1086 af->to_sk_daddr(&to, sk);
1087 sk->sk_err = 0;
1089 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1090 err = sctp_wait_for_connect(asoc, &timeo);
1092 /* Don't free association on exit. */
1093 asoc = NULL;
1095 out_free:
1097 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1098 " kaddrs: %p err: %d\n",
1099 asoc, kaddrs, err);
1100 if (asoc)
1101 sctp_association_free(asoc);
1102 return err;
1105 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1107 * API 8.9
1108 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1110 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1111 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1112 * or IPv6 addresses.
1114 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1115 * Section 3.1.2 for this usage.
1117 * addrs is a pointer to an array of one or more socket addresses. Each
1118 * address is contained in its appropriate structure (i.e. struct
1119 * sockaddr_in or struct sockaddr_in6) the family of the address type
1120 * must be used to distengish the address length (note that this
1121 * representation is termed a "packed array" of addresses). The caller
1122 * specifies the number of addresses in the array with addrcnt.
1124 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1125 * -1, and sets errno to the appropriate error code.
1127 * For SCTP, the port given in each socket address must be the same, or
1128 * sctp_connectx() will fail, setting errno to EINVAL.
1130 * An application can use sctp_connectx to initiate an association with
1131 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1132 * allows a caller to specify multiple addresses at which a peer can be
1133 * reached. The way the SCTP stack uses the list of addresses to set up
1134 * the association is implementation dependant. This function only
1135 * specifies that the stack will try to make use of all the addresses in
1136 * the list when needed.
1138 * Note that the list of addresses passed in is only used for setting up
1139 * the association. It does not necessarily equal the set of addresses
1140 * the peer uses for the resulting association. If the caller wants to
1141 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1142 * retrieve them after the association has been set up.
1144 * Basically do nothing but copying the addresses from user to kernel
1145 * land and invoking either sctp_connectx(). This is used for tunneling
1146 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1148 * We don't use copy_from_user() for optimization: we first do the
1149 * sanity checks (buffer size -fast- and access check-healthy
1150 * pointer); if all of those succeed, then we can alloc the memory
1151 * (expensive operation) needed to copy the data to kernel. Then we do
1152 * the copying without checking the user space area
1153 * (__copy_from_user()).
1155 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1156 * it.
1158 * sk The sk of the socket
1159 * addrs The pointer to the addresses in user land
1160 * addrssize Size of the addrs buffer
1162 * Returns 0 if ok, <0 errno code on error.
1164 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1165 struct sockaddr __user *addrs,
1166 int addrs_size)
1168 int err = 0;
1169 struct sockaddr *kaddrs;
1171 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1172 __FUNCTION__, sk, addrs, addrs_size);
1174 if (unlikely(addrs_size <= 0))
1175 return -EINVAL;
1177 /* Check the user passed a healthy pointer. */
1178 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1179 return -EFAULT;
1181 /* Alloc space for the address array in kernel memory. */
1182 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1183 if (unlikely(!kaddrs))
1184 return -ENOMEM;
1186 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1187 err = -EFAULT;
1188 } else {
1189 err = __sctp_connect(sk, kaddrs, addrs_size);
1192 kfree(kaddrs);
1193 return err;
1196 /* API 3.1.4 close() - UDP Style Syntax
1197 * Applications use close() to perform graceful shutdown (as described in
1198 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1199 * by a UDP-style socket.
1201 * The syntax is
1203 * ret = close(int sd);
1205 * sd - the socket descriptor of the associations to be closed.
1207 * To gracefully shutdown a specific association represented by the
1208 * UDP-style socket, an application should use the sendmsg() call,
1209 * passing no user data, but including the appropriate flag in the
1210 * ancillary data (see Section xxxx).
1212 * If sd in the close() call is a branched-off socket representing only
1213 * one association, the shutdown is performed on that association only.
1215 * 4.1.6 close() - TCP Style Syntax
1217 * Applications use close() to gracefully close down an association.
1219 * The syntax is:
1221 * int close(int sd);
1223 * sd - the socket descriptor of the association to be closed.
1225 * After an application calls close() on a socket descriptor, no further
1226 * socket operations will succeed on that descriptor.
1228 * API 7.1.4 SO_LINGER
1230 * An application using the TCP-style socket can use this option to
1231 * perform the SCTP ABORT primitive. The linger option structure is:
1233 * struct linger {
1234 * int l_onoff; // option on/off
1235 * int l_linger; // linger time
1236 * };
1238 * To enable the option, set l_onoff to 1. If the l_linger value is set
1239 * to 0, calling close() is the same as the ABORT primitive. If the
1240 * value is set to a negative value, the setsockopt() call will return
1241 * an error. If the value is set to a positive value linger_time, the
1242 * close() can be blocked for at most linger_time ms. If the graceful
1243 * shutdown phase does not finish during this period, close() will
1244 * return but the graceful shutdown phase continues in the system.
1246 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1248 struct sctp_endpoint *ep;
1249 struct sctp_association *asoc;
1250 struct list_head *pos, *temp;
1252 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1254 sctp_lock_sock(sk);
1255 sk->sk_shutdown = SHUTDOWN_MASK;
1257 ep = sctp_sk(sk)->ep;
1259 /* Walk all associations on an endpoint. */
1260 list_for_each_safe(pos, temp, &ep->asocs) {
1261 asoc = list_entry(pos, struct sctp_association, asocs);
1263 if (sctp_style(sk, TCP)) {
1264 /* A closed association can still be in the list if
1265 * it belongs to a TCP-style listening socket that is
1266 * not yet accepted. If so, free it. If not, send an
1267 * ABORT or SHUTDOWN based on the linger options.
1269 if (sctp_state(asoc, CLOSED)) {
1270 sctp_unhash_established(asoc);
1271 sctp_association_free(asoc);
1272 continue;
1276 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1277 struct sctp_chunk *chunk;
1279 chunk = sctp_make_abort_user(asoc, NULL, 0);
1280 if (chunk)
1281 sctp_primitive_ABORT(asoc, chunk);
1282 } else
1283 sctp_primitive_SHUTDOWN(asoc, NULL);
1286 /* Clean up any skbs sitting on the receive queue. */
1287 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1288 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1290 /* On a TCP-style socket, block for at most linger_time if set. */
1291 if (sctp_style(sk, TCP) && timeout)
1292 sctp_wait_for_close(sk, timeout);
1294 /* This will run the backlog queue. */
1295 sctp_release_sock(sk);
1297 /* Supposedly, no process has access to the socket, but
1298 * the net layers still may.
1300 sctp_local_bh_disable();
1301 sctp_bh_lock_sock(sk);
1303 /* Hold the sock, since sk_common_release() will put sock_put()
1304 * and we have just a little more cleanup.
1306 sock_hold(sk);
1307 sk_common_release(sk);
1309 sctp_bh_unlock_sock(sk);
1310 sctp_local_bh_enable();
1312 sock_put(sk);
1314 SCTP_DBG_OBJCNT_DEC(sock);
1317 /* Handle EPIPE error. */
1318 static int sctp_error(struct sock *sk, int flags, int err)
1320 if (err == -EPIPE)
1321 err = sock_error(sk) ? : -EPIPE;
1322 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1323 send_sig(SIGPIPE, current, 0);
1324 return err;
1327 /* API 3.1.3 sendmsg() - UDP Style Syntax
1329 * An application uses sendmsg() and recvmsg() calls to transmit data to
1330 * and receive data from its peer.
1332 * ssize_t sendmsg(int socket, const struct msghdr *message,
1333 * int flags);
1335 * socket - the socket descriptor of the endpoint.
1336 * message - pointer to the msghdr structure which contains a single
1337 * user message and possibly some ancillary data.
1339 * See Section 5 for complete description of the data
1340 * structures.
1342 * flags - flags sent or received with the user message, see Section
1343 * 5 for complete description of the flags.
1345 * Note: This function could use a rewrite especially when explicit
1346 * connect support comes in.
1348 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1350 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1352 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1353 struct msghdr *msg, size_t msg_len)
1355 struct sctp_sock *sp;
1356 struct sctp_endpoint *ep;
1357 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1358 struct sctp_transport *transport, *chunk_tp;
1359 struct sctp_chunk *chunk;
1360 union sctp_addr to, tmp;
1361 struct sockaddr *msg_name = NULL;
1362 struct sctp_sndrcvinfo default_sinfo = { 0 };
1363 struct sctp_sndrcvinfo *sinfo;
1364 struct sctp_initmsg *sinit;
1365 sctp_assoc_t associd = 0;
1366 sctp_cmsgs_t cmsgs = { NULL };
1367 int err;
1368 sctp_scope_t scope;
1369 long timeo;
1370 __u16 sinfo_flags = 0;
1371 struct sctp_datamsg *datamsg;
1372 struct list_head *pos;
1373 int msg_flags = msg->msg_flags;
1375 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1376 sk, msg, msg_len);
1378 err = 0;
1379 sp = sctp_sk(sk);
1380 ep = sp->ep;
1382 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1384 /* We cannot send a message over a TCP-style listening socket. */
1385 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1386 err = -EPIPE;
1387 goto out_nounlock;
1390 /* Parse out the SCTP CMSGs. */
1391 err = sctp_msghdr_parse(msg, &cmsgs);
1393 if (err) {
1394 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1395 goto out_nounlock;
1398 /* Fetch the destination address for this packet. This
1399 * address only selects the association--it is not necessarily
1400 * the address we will send to.
1401 * For a peeled-off socket, msg_name is ignored.
1403 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1404 int msg_namelen = msg->msg_namelen;
1406 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1407 msg_namelen);
1408 if (err)
1409 return err;
1411 if (msg_namelen > sizeof(to))
1412 msg_namelen = sizeof(to);
1413 memcpy(&to, msg->msg_name, msg_namelen);
1414 memcpy(&tmp, msg->msg_name, msg_namelen);
1415 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1416 "0x%x:%u.\n",
1417 to.v4.sin_addr.s_addr, to.v4.sin_port);
1419 to.v4.sin_port = ntohs(to.v4.sin_port);
1420 msg_name = msg->msg_name;
1423 sinfo = cmsgs.info;
1424 sinit = cmsgs.init;
1426 /* Did the user specify SNDRCVINFO? */
1427 if (sinfo) {
1428 sinfo_flags = sinfo->sinfo_flags;
1429 associd = sinfo->sinfo_assoc_id;
1432 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1433 msg_len, sinfo_flags);
1435 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1436 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1437 err = -EINVAL;
1438 goto out_nounlock;
1441 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1442 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1443 * If SCTP_ABORT is set, the message length could be non zero with
1444 * the msg_iov set to the user abort reason.
1446 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1447 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1448 err = -EINVAL;
1449 goto out_nounlock;
1452 /* If SCTP_ADDR_OVER is set, there must be an address
1453 * specified in msg_name.
1455 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1456 err = -EINVAL;
1457 goto out_nounlock;
1460 transport = NULL;
1462 SCTP_DEBUG_PRINTK("About to look up association.\n");
1464 sctp_lock_sock(sk);
1466 /* If a msg_name has been specified, assume this is to be used. */
1467 if (msg_name) {
1468 /* Look for a matching association on the endpoint. */
1469 asoc = sctp_endpoint_lookup_assoc(ep, &tmp, &transport);
1470 if (!asoc) {
1471 /* If we could not find a matching association on the
1472 * endpoint, make sure that it is not a TCP-style
1473 * socket that already has an association or there is
1474 * no peeled-off association on another socket.
1476 if ((sctp_style(sk, TCP) &&
1477 sctp_sstate(sk, ESTABLISHED)) ||
1478 sctp_endpoint_is_peeled_off(ep, &tmp)) {
1479 err = -EADDRNOTAVAIL;
1480 goto out_unlock;
1483 } else {
1484 asoc = sctp_id2assoc(sk, associd);
1485 if (!asoc) {
1486 err = -EPIPE;
1487 goto out_unlock;
1491 if (asoc) {
1492 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1494 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1495 * socket that has an association in CLOSED state. This can
1496 * happen when an accepted socket has an association that is
1497 * already CLOSED.
1499 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1500 err = -EPIPE;
1501 goto out_unlock;
1504 if (sinfo_flags & SCTP_EOF) {
1505 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1506 asoc);
1507 sctp_primitive_SHUTDOWN(asoc, NULL);
1508 err = 0;
1509 goto out_unlock;
1511 if (sinfo_flags & SCTP_ABORT) {
1512 struct sctp_chunk *chunk;
1514 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1515 if (!chunk) {
1516 err = -ENOMEM;
1517 goto out_unlock;
1520 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1521 sctp_primitive_ABORT(asoc, chunk);
1522 err = 0;
1523 goto out_unlock;
1527 /* Do we need to create the association? */
1528 if (!asoc) {
1529 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1531 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1532 err = -EINVAL;
1533 goto out_unlock;
1536 /* Check for invalid stream against the stream counts,
1537 * either the default or the user specified stream counts.
1539 if (sinfo) {
1540 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1541 /* Check against the defaults. */
1542 if (sinfo->sinfo_stream >=
1543 sp->initmsg.sinit_num_ostreams) {
1544 err = -EINVAL;
1545 goto out_unlock;
1547 } else {
1548 /* Check against the requested. */
1549 if (sinfo->sinfo_stream >=
1550 sinit->sinit_num_ostreams) {
1551 err = -EINVAL;
1552 goto out_unlock;
1558 * API 3.1.2 bind() - UDP Style Syntax
1559 * If a bind() or sctp_bindx() is not called prior to a
1560 * sendmsg() call that initiates a new association, the
1561 * system picks an ephemeral port and will choose an address
1562 * set equivalent to binding with a wildcard address.
1564 if (!ep->base.bind_addr.port) {
1565 if (sctp_autobind(sk)) {
1566 err = -EAGAIN;
1567 goto out_unlock;
1569 } else {
1571 * If an unprivileged user inherits a one-to-many
1572 * style socket with open associations on a privileged
1573 * port, it MAY be permitted to accept new associations,
1574 * but it SHOULD NOT be permitted to open new
1575 * associations.
1577 if (ep->base.bind_addr.port < PROT_SOCK &&
1578 !capable(CAP_NET_BIND_SERVICE)) {
1579 err = -EACCES;
1580 goto out_unlock;
1584 scope = sctp_scope(&to);
1585 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1586 if (!new_asoc) {
1587 err = -ENOMEM;
1588 goto out_unlock;
1590 asoc = new_asoc;
1592 /* If the SCTP_INIT ancillary data is specified, set all
1593 * the association init values accordingly.
1595 if (sinit) {
1596 if (sinit->sinit_num_ostreams) {
1597 asoc->c.sinit_num_ostreams =
1598 sinit->sinit_num_ostreams;
1600 if (sinit->sinit_max_instreams) {
1601 asoc->c.sinit_max_instreams =
1602 sinit->sinit_max_instreams;
1604 if (sinit->sinit_max_attempts) {
1605 asoc->max_init_attempts
1606 = sinit->sinit_max_attempts;
1608 if (sinit->sinit_max_init_timeo) {
1609 asoc->max_init_timeo =
1610 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1614 /* Prime the peer's transport structures. */
1615 transport = sctp_assoc_add_peer(asoc, &tmp, GFP_KERNEL, SCTP_UNKNOWN);
1616 if (!transport) {
1617 err = -ENOMEM;
1618 goto out_free;
1620 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1621 if (err < 0) {
1622 err = -ENOMEM;
1623 goto out_free;
1627 /* ASSERT: we have a valid association at this point. */
1628 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1630 if (!sinfo) {
1631 /* If the user didn't specify SNDRCVINFO, make up one with
1632 * some defaults.
1634 default_sinfo.sinfo_stream = asoc->default_stream;
1635 default_sinfo.sinfo_flags = asoc->default_flags;
1636 default_sinfo.sinfo_ppid = asoc->default_ppid;
1637 default_sinfo.sinfo_context = asoc->default_context;
1638 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1639 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1640 sinfo = &default_sinfo;
1643 /* API 7.1.7, the sndbuf size per association bounds the
1644 * maximum size of data that can be sent in a single send call.
1646 if (msg_len > sk->sk_sndbuf) {
1647 err = -EMSGSIZE;
1648 goto out_free;
1651 /* If fragmentation is disabled and the message length exceeds the
1652 * association fragmentation point, return EMSGSIZE. The I-D
1653 * does not specify what this error is, but this looks like
1654 * a great fit.
1656 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1657 err = -EMSGSIZE;
1658 goto out_free;
1661 if (sinfo) {
1662 /* Check for invalid stream. */
1663 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1664 err = -EINVAL;
1665 goto out_free;
1669 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1670 if (!sctp_wspace(asoc)) {
1671 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1672 if (err)
1673 goto out_free;
1676 /* If an address is passed with the sendto/sendmsg call, it is used
1677 * to override the primary destination address in the TCP model, or
1678 * when SCTP_ADDR_OVER flag is set in the UDP model.
1680 if ((sctp_style(sk, TCP) && msg_name) ||
1681 (sinfo_flags & SCTP_ADDR_OVER)) {
1682 chunk_tp = sctp_assoc_lookup_paddr(asoc, &tmp);
1683 if (!chunk_tp) {
1684 err = -EINVAL;
1685 goto out_free;
1687 } else
1688 chunk_tp = NULL;
1690 /* Auto-connect, if we aren't connected already. */
1691 if (sctp_state(asoc, CLOSED)) {
1692 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1693 if (err < 0)
1694 goto out_free;
1695 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1698 /* Break the message into multiple chunks of maximum size. */
1699 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1700 if (!datamsg) {
1701 err = -ENOMEM;
1702 goto out_free;
1705 /* Now send the (possibly) fragmented message. */
1706 list_for_each(pos, &datamsg->chunks) {
1707 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1708 sctp_datamsg_track(chunk);
1710 /* Do accounting for the write space. */
1711 sctp_set_owner_w(chunk);
1713 chunk->transport = chunk_tp;
1715 /* Send it to the lower layers. Note: all chunks
1716 * must either fail or succeed. The lower layer
1717 * works that way today. Keep it that way or this
1718 * breaks.
1720 err = sctp_primitive_SEND(asoc, chunk);
1721 /* Did the lower layer accept the chunk? */
1722 if (err)
1723 sctp_chunk_free(chunk);
1724 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1727 sctp_datamsg_free(datamsg);
1728 if (err)
1729 goto out_free;
1730 else
1731 err = msg_len;
1733 /* If we are already past ASSOCIATE, the lower
1734 * layers are responsible for association cleanup.
1736 goto out_unlock;
1738 out_free:
1739 if (new_asoc)
1740 sctp_association_free(asoc);
1741 out_unlock:
1742 sctp_release_sock(sk);
1744 out_nounlock:
1745 return sctp_error(sk, msg_flags, err);
1747 #if 0
1748 do_sock_err:
1749 if (msg_len)
1750 err = msg_len;
1751 else
1752 err = sock_error(sk);
1753 goto out;
1755 do_interrupted:
1756 if (msg_len)
1757 err = msg_len;
1758 goto out;
1759 #endif /* 0 */
1762 /* This is an extended version of skb_pull() that removes the data from the
1763 * start of a skb even when data is spread across the list of skb's in the
1764 * frag_list. len specifies the total amount of data that needs to be removed.
1765 * when 'len' bytes could be removed from the skb, it returns 0.
1766 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1767 * could not be removed.
1769 static int sctp_skb_pull(struct sk_buff *skb, int len)
1771 struct sk_buff *list;
1772 int skb_len = skb_headlen(skb);
1773 int rlen;
1775 if (len <= skb_len) {
1776 __skb_pull(skb, len);
1777 return 0;
1779 len -= skb_len;
1780 __skb_pull(skb, skb_len);
1782 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1783 rlen = sctp_skb_pull(list, len);
1784 skb->len -= (len-rlen);
1785 skb->data_len -= (len-rlen);
1787 if (!rlen)
1788 return 0;
1790 len = rlen;
1793 return len;
1796 /* API 3.1.3 recvmsg() - UDP Style Syntax
1798 * ssize_t recvmsg(int socket, struct msghdr *message,
1799 * int flags);
1801 * socket - the socket descriptor of the endpoint.
1802 * message - pointer to the msghdr structure which contains a single
1803 * user message and possibly some ancillary data.
1805 * See Section 5 for complete description of the data
1806 * structures.
1808 * flags - flags sent or received with the user message, see Section
1809 * 5 for complete description of the flags.
1811 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1813 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1814 struct msghdr *msg, size_t len, int noblock,
1815 int flags, int *addr_len)
1817 struct sctp_ulpevent *event = NULL;
1818 struct sctp_sock *sp = sctp_sk(sk);
1819 struct sk_buff *skb;
1820 int copied;
1821 int err = 0;
1822 int skb_len;
1824 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1825 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1826 "len", len, "knoblauch", noblock,
1827 "flags", flags, "addr_len", addr_len);
1829 sctp_lock_sock(sk);
1831 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1832 err = -ENOTCONN;
1833 goto out;
1836 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1837 if (!skb)
1838 goto out;
1840 /* Get the total length of the skb including any skb's in the
1841 * frag_list.
1843 skb_len = skb->len;
1845 copied = skb_len;
1846 if (copied > len)
1847 copied = len;
1849 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1851 event = sctp_skb2event(skb);
1853 if (err)
1854 goto out_free;
1856 sock_recv_timestamp(msg, sk, skb);
1857 if (sctp_ulpevent_is_notification(event)) {
1858 msg->msg_flags |= MSG_NOTIFICATION;
1859 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1860 } else {
1861 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1864 /* Check if we allow SCTP_SNDRCVINFO. */
1865 if (sp->subscribe.sctp_data_io_event)
1866 sctp_ulpevent_read_sndrcvinfo(event, msg);
1867 #if 0
1868 /* FIXME: we should be calling IP/IPv6 layers. */
1869 if (sk->sk_protinfo.af_inet.cmsg_flags)
1870 ip_cmsg_recv(msg, skb);
1871 #endif
1873 err = copied;
1875 /* If skb's length exceeds the user's buffer, update the skb and
1876 * push it back to the receive_queue so that the next call to
1877 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1879 if (skb_len > copied) {
1880 msg->msg_flags &= ~MSG_EOR;
1881 if (flags & MSG_PEEK)
1882 goto out_free;
1883 sctp_skb_pull(skb, copied);
1884 skb_queue_head(&sk->sk_receive_queue, skb);
1886 /* When only partial message is copied to the user, increase
1887 * rwnd by that amount. If all the data in the skb is read,
1888 * rwnd is updated when the event is freed.
1890 sctp_assoc_rwnd_increase(event->asoc, copied);
1891 goto out;
1892 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1893 (event->msg_flags & MSG_EOR))
1894 msg->msg_flags |= MSG_EOR;
1895 else
1896 msg->msg_flags &= ~MSG_EOR;
1898 out_free:
1899 if (flags & MSG_PEEK) {
1900 /* Release the skb reference acquired after peeking the skb in
1901 * sctp_skb_recv_datagram().
1903 kfree_skb(skb);
1904 } else {
1905 /* Free the event which includes releasing the reference to
1906 * the owner of the skb, freeing the skb and updating the
1907 * rwnd.
1909 sctp_ulpevent_free(event);
1911 out:
1912 sctp_release_sock(sk);
1913 return err;
1916 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1918 * This option is a on/off flag. If enabled no SCTP message
1919 * fragmentation will be performed. Instead if a message being sent
1920 * exceeds the current PMTU size, the message will NOT be sent and
1921 * instead a error will be indicated to the user.
1923 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1924 char __user *optval, int optlen)
1926 int val;
1928 if (optlen < sizeof(int))
1929 return -EINVAL;
1931 if (get_user(val, (int __user *)optval))
1932 return -EFAULT;
1934 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1936 return 0;
1939 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1940 int optlen)
1942 if (optlen != sizeof(struct sctp_event_subscribe))
1943 return -EINVAL;
1944 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1945 return -EFAULT;
1946 return 0;
1949 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1951 * This socket option is applicable to the UDP-style socket only. When
1952 * set it will cause associations that are idle for more than the
1953 * specified number of seconds to automatically close. An association
1954 * being idle is defined an association that has NOT sent or received
1955 * user data. The special value of '0' indicates that no automatic
1956 * close of any associations should be performed. The option expects an
1957 * integer defining the number of seconds of idle time before an
1958 * association is closed.
1960 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1961 int optlen)
1963 struct sctp_sock *sp = sctp_sk(sk);
1965 /* Applicable to UDP-style socket only */
1966 if (sctp_style(sk, TCP))
1967 return -EOPNOTSUPP;
1968 if (optlen != sizeof(int))
1969 return -EINVAL;
1970 if (copy_from_user(&sp->autoclose, optval, optlen))
1971 return -EFAULT;
1973 return 0;
1976 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1978 * Applications can enable or disable heartbeats for any peer address of
1979 * an association, modify an address's heartbeat interval, force a
1980 * heartbeat to be sent immediately, and adjust the address's maximum
1981 * number of retransmissions sent before an address is considered
1982 * unreachable. The following structure is used to access and modify an
1983 * address's parameters:
1985 * struct sctp_paddrparams {
1986 * sctp_assoc_t spp_assoc_id;
1987 * struct sockaddr_storage spp_address;
1988 * uint32_t spp_hbinterval;
1989 * uint16_t spp_pathmaxrxt;
1990 * uint32_t spp_pathmtu;
1991 * uint32_t spp_sackdelay;
1992 * uint32_t spp_flags;
1993 * };
1995 * spp_assoc_id - (one-to-many style socket) This is filled in the
1996 * application, and identifies the association for
1997 * this query.
1998 * spp_address - This specifies which address is of interest.
1999 * spp_hbinterval - This contains the value of the heartbeat interval,
2000 * in milliseconds. If a value of zero
2001 * is present in this field then no changes are to
2002 * be made to this parameter.
2003 * spp_pathmaxrxt - This contains the maximum number of
2004 * retransmissions before this address shall be
2005 * considered unreachable. If a value of zero
2006 * is present in this field then no changes are to
2007 * be made to this parameter.
2008 * spp_pathmtu - When Path MTU discovery is disabled the value
2009 * specified here will be the "fixed" path mtu.
2010 * Note that if the spp_address field is empty
2011 * then all associations on this address will
2012 * have this fixed path mtu set upon them.
2014 * spp_sackdelay - When delayed sack is enabled, this value specifies
2015 * the number of milliseconds that sacks will be delayed
2016 * for. This value will apply to all addresses of an
2017 * association if the spp_address field is empty. Note
2018 * also, that if delayed sack is enabled and this
2019 * value is set to 0, no change is made to the last
2020 * recorded delayed sack timer value.
2022 * spp_flags - These flags are used to control various features
2023 * on an association. The flag field may contain
2024 * zero or more of the following options.
2026 * SPP_HB_ENABLE - Enable heartbeats on the
2027 * specified address. Note that if the address
2028 * field is empty all addresses for the association
2029 * have heartbeats enabled upon them.
2031 * SPP_HB_DISABLE - Disable heartbeats on the
2032 * speicifed address. Note that if the address
2033 * field is empty all addresses for the association
2034 * will have their heartbeats disabled. Note also
2035 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2036 * mutually exclusive, only one of these two should
2037 * be specified. Enabling both fields will have
2038 * undetermined results.
2040 * SPP_HB_DEMAND - Request a user initiated heartbeat
2041 * to be made immediately.
2043 * SPP_PMTUD_ENABLE - This field will enable PMTU
2044 * discovery upon the specified address. Note that
2045 * if the address feild is empty then all addresses
2046 * on the association are effected.
2048 * SPP_PMTUD_DISABLE - This field will disable PMTU
2049 * discovery upon the specified address. Note that
2050 * if the address feild is empty then all addresses
2051 * on the association are effected. Not also that
2052 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2053 * exclusive. Enabling both will have undetermined
2054 * results.
2056 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2057 * on delayed sack. The time specified in spp_sackdelay
2058 * is used to specify the sack delay for this address. Note
2059 * that if spp_address is empty then all addresses will
2060 * enable delayed sack and take on the sack delay
2061 * value specified in spp_sackdelay.
2062 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2063 * off delayed sack. If the spp_address field is blank then
2064 * delayed sack is disabled for the entire association. Note
2065 * also that this field is mutually exclusive to
2066 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2067 * results.
2069 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2070 struct sctp_transport *trans,
2071 struct sctp_association *asoc,
2072 struct sctp_sock *sp,
2073 int hb_change,
2074 int pmtud_change,
2075 int sackdelay_change)
2077 int error;
2079 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2080 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2081 if (error)
2082 return error;
2085 if (params->spp_hbinterval) {
2086 if (trans) {
2087 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2088 } else if (asoc) {
2089 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2090 } else {
2091 sp->hbinterval = params->spp_hbinterval;
2095 if (hb_change) {
2096 if (trans) {
2097 trans->param_flags =
2098 (trans->param_flags & ~SPP_HB) | hb_change;
2099 } else if (asoc) {
2100 asoc->param_flags =
2101 (asoc->param_flags & ~SPP_HB) | hb_change;
2102 } else {
2103 sp->param_flags =
2104 (sp->param_flags & ~SPP_HB) | hb_change;
2108 if (params->spp_pathmtu) {
2109 if (trans) {
2110 trans->pathmtu = params->spp_pathmtu;
2111 sctp_assoc_sync_pmtu(asoc);
2112 } else if (asoc) {
2113 asoc->pathmtu = params->spp_pathmtu;
2114 sctp_frag_point(sp, params->spp_pathmtu);
2115 } else {
2116 sp->pathmtu = params->spp_pathmtu;
2120 if (pmtud_change) {
2121 if (trans) {
2122 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2123 (params->spp_flags & SPP_PMTUD_ENABLE);
2124 trans->param_flags =
2125 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2126 if (update) {
2127 sctp_transport_pmtu(trans);
2128 sctp_assoc_sync_pmtu(asoc);
2130 } else if (asoc) {
2131 asoc->param_flags =
2132 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2133 } else {
2134 sp->param_flags =
2135 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2139 if (params->spp_sackdelay) {
2140 if (trans) {
2141 trans->sackdelay =
2142 msecs_to_jiffies(params->spp_sackdelay);
2143 } else if (asoc) {
2144 asoc->sackdelay =
2145 msecs_to_jiffies(params->spp_sackdelay);
2146 } else {
2147 sp->sackdelay = params->spp_sackdelay;
2151 if (sackdelay_change) {
2152 if (trans) {
2153 trans->param_flags =
2154 (trans->param_flags & ~SPP_SACKDELAY) |
2155 sackdelay_change;
2156 } else if (asoc) {
2157 asoc->param_flags =
2158 (asoc->param_flags & ~SPP_SACKDELAY) |
2159 sackdelay_change;
2160 } else {
2161 sp->param_flags =
2162 (sp->param_flags & ~SPP_SACKDELAY) |
2163 sackdelay_change;
2167 if (params->spp_pathmaxrxt) {
2168 if (trans) {
2169 trans->pathmaxrxt = params->spp_pathmaxrxt;
2170 } else if (asoc) {
2171 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2172 } else {
2173 sp->pathmaxrxt = params->spp_pathmaxrxt;
2177 return 0;
2180 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2181 char __user *optval, int optlen)
2183 struct sctp_paddrparams params;
2184 struct sctp_transport *trans = NULL;
2185 struct sctp_association *asoc = NULL;
2186 struct sctp_sock *sp = sctp_sk(sk);
2187 int error;
2188 int hb_change, pmtud_change, sackdelay_change;
2190 if (optlen != sizeof(struct sctp_paddrparams))
2191 return - EINVAL;
2193 if (copy_from_user(&params, optval, optlen))
2194 return -EFAULT;
2196 /* Validate flags and value parameters. */
2197 hb_change = params.spp_flags & SPP_HB;
2198 pmtud_change = params.spp_flags & SPP_PMTUD;
2199 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2201 if (hb_change == SPP_HB ||
2202 pmtud_change == SPP_PMTUD ||
2203 sackdelay_change == SPP_SACKDELAY ||
2204 params.spp_sackdelay > 500 ||
2205 (params.spp_pathmtu
2206 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2207 return -EINVAL;
2209 /* If an address other than INADDR_ANY is specified, and
2210 * no transport is found, then the request is invalid.
2212 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2213 trans = sctp_addr_id2transport(sk, &params.spp_address,
2214 params.spp_assoc_id);
2215 if (!trans)
2216 return -EINVAL;
2219 /* Get association, if assoc_id != 0 and the socket is a one
2220 * to many style socket, and an association was not found, then
2221 * the id was invalid.
2223 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2224 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2225 return -EINVAL;
2227 /* Heartbeat demand can only be sent on a transport or
2228 * association, but not a socket.
2230 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2231 return -EINVAL;
2233 /* Process parameters. */
2234 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2235 hb_change, pmtud_change,
2236 sackdelay_change);
2238 if (error)
2239 return error;
2241 /* If changes are for association, also apply parameters to each
2242 * transport.
2244 if (!trans && asoc) {
2245 struct list_head *pos;
2247 list_for_each(pos, &asoc->peer.transport_addr_list) {
2248 trans = list_entry(pos, struct sctp_transport,
2249 transports);
2250 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2251 hb_change, pmtud_change,
2252 sackdelay_change);
2256 return 0;
2259 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2261 * This options will get or set the delayed ack timer. The time is set
2262 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2263 * endpoints default delayed ack timer value. If the assoc_id field is
2264 * non-zero, then the set or get effects the specified association.
2266 * struct sctp_assoc_value {
2267 * sctp_assoc_t assoc_id;
2268 * uint32_t assoc_value;
2269 * };
2271 * assoc_id - This parameter, indicates which association the
2272 * user is preforming an action upon. Note that if
2273 * this field's value is zero then the endpoints
2274 * default value is changed (effecting future
2275 * associations only).
2277 * assoc_value - This parameter contains the number of milliseconds
2278 * that the user is requesting the delayed ACK timer
2279 * be set to. Note that this value is defined in
2280 * the standard to be between 200 and 500 milliseconds.
2282 * Note: a value of zero will leave the value alone,
2283 * but disable SACK delay. A non-zero value will also
2284 * enable SACK delay.
2287 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2288 char __user *optval, int optlen)
2290 struct sctp_assoc_value params;
2291 struct sctp_transport *trans = NULL;
2292 struct sctp_association *asoc = NULL;
2293 struct sctp_sock *sp = sctp_sk(sk);
2295 if (optlen != sizeof(struct sctp_assoc_value))
2296 return - EINVAL;
2298 if (copy_from_user(&params, optval, optlen))
2299 return -EFAULT;
2301 /* Validate value parameter. */
2302 if (params.assoc_value > 500)
2303 return -EINVAL;
2305 /* Get association, if assoc_id != 0 and the socket is a one
2306 * to many style socket, and an association was not found, then
2307 * the id was invalid.
2309 asoc = sctp_id2assoc(sk, params.assoc_id);
2310 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2311 return -EINVAL;
2313 if (params.assoc_value) {
2314 if (asoc) {
2315 asoc->sackdelay =
2316 msecs_to_jiffies(params.assoc_value);
2317 asoc->param_flags =
2318 (asoc->param_flags & ~SPP_SACKDELAY) |
2319 SPP_SACKDELAY_ENABLE;
2320 } else {
2321 sp->sackdelay = params.assoc_value;
2322 sp->param_flags =
2323 (sp->param_flags & ~SPP_SACKDELAY) |
2324 SPP_SACKDELAY_ENABLE;
2326 } else {
2327 if (asoc) {
2328 asoc->param_flags =
2329 (asoc->param_flags & ~SPP_SACKDELAY) |
2330 SPP_SACKDELAY_DISABLE;
2331 } else {
2332 sp->param_flags =
2333 (sp->param_flags & ~SPP_SACKDELAY) |
2334 SPP_SACKDELAY_DISABLE;
2338 /* If change is for association, also apply to each transport. */
2339 if (asoc) {
2340 struct list_head *pos;
2342 list_for_each(pos, &asoc->peer.transport_addr_list) {
2343 trans = list_entry(pos, struct sctp_transport,
2344 transports);
2345 if (params.assoc_value) {
2346 trans->sackdelay =
2347 msecs_to_jiffies(params.assoc_value);
2348 trans->param_flags =
2349 (trans->param_flags & ~SPP_SACKDELAY) |
2350 SPP_SACKDELAY_ENABLE;
2351 } else {
2352 trans->param_flags =
2353 (trans->param_flags & ~SPP_SACKDELAY) |
2354 SPP_SACKDELAY_DISABLE;
2359 return 0;
2362 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2364 * Applications can specify protocol parameters for the default association
2365 * initialization. The option name argument to setsockopt() and getsockopt()
2366 * is SCTP_INITMSG.
2368 * Setting initialization parameters is effective only on an unconnected
2369 * socket (for UDP-style sockets only future associations are effected
2370 * by the change). With TCP-style sockets, this option is inherited by
2371 * sockets derived from a listener socket.
2373 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2375 struct sctp_initmsg sinit;
2376 struct sctp_sock *sp = sctp_sk(sk);
2378 if (optlen != sizeof(struct sctp_initmsg))
2379 return -EINVAL;
2380 if (copy_from_user(&sinit, optval, optlen))
2381 return -EFAULT;
2383 if (sinit.sinit_num_ostreams)
2384 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2385 if (sinit.sinit_max_instreams)
2386 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2387 if (sinit.sinit_max_attempts)
2388 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2389 if (sinit.sinit_max_init_timeo)
2390 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2392 return 0;
2396 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2398 * Applications that wish to use the sendto() system call may wish to
2399 * specify a default set of parameters that would normally be supplied
2400 * through the inclusion of ancillary data. This socket option allows
2401 * such an application to set the default sctp_sndrcvinfo structure.
2402 * The application that wishes to use this socket option simply passes
2403 * in to this call the sctp_sndrcvinfo structure defined in Section
2404 * 5.2.2) The input parameters accepted by this call include
2405 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2406 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2407 * to this call if the caller is using the UDP model.
2409 static int sctp_setsockopt_default_send_param(struct sock *sk,
2410 char __user *optval, int optlen)
2412 struct sctp_sndrcvinfo info;
2413 struct sctp_association *asoc;
2414 struct sctp_sock *sp = sctp_sk(sk);
2416 if (optlen != sizeof(struct sctp_sndrcvinfo))
2417 return -EINVAL;
2418 if (copy_from_user(&info, optval, optlen))
2419 return -EFAULT;
2421 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2422 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2423 return -EINVAL;
2425 if (asoc) {
2426 asoc->default_stream = info.sinfo_stream;
2427 asoc->default_flags = info.sinfo_flags;
2428 asoc->default_ppid = info.sinfo_ppid;
2429 asoc->default_context = info.sinfo_context;
2430 asoc->default_timetolive = info.sinfo_timetolive;
2431 } else {
2432 sp->default_stream = info.sinfo_stream;
2433 sp->default_flags = info.sinfo_flags;
2434 sp->default_ppid = info.sinfo_ppid;
2435 sp->default_context = info.sinfo_context;
2436 sp->default_timetolive = info.sinfo_timetolive;
2439 return 0;
2442 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2444 * Requests that the local SCTP stack use the enclosed peer address as
2445 * the association primary. The enclosed address must be one of the
2446 * association peer's addresses.
2448 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2449 int optlen)
2451 struct sctp_prim prim;
2452 struct sctp_transport *trans;
2454 if (optlen != sizeof(struct sctp_prim))
2455 return -EINVAL;
2457 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2458 return -EFAULT;
2460 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2461 if (!trans)
2462 return -EINVAL;
2464 sctp_assoc_set_primary(trans->asoc, trans);
2466 return 0;
2470 * 7.1.5 SCTP_NODELAY
2472 * Turn on/off any Nagle-like algorithm. This means that packets are
2473 * generally sent as soon as possible and no unnecessary delays are
2474 * introduced, at the cost of more packets in the network. Expects an
2475 * integer boolean flag.
2477 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2478 int optlen)
2480 int val;
2482 if (optlen < sizeof(int))
2483 return -EINVAL;
2484 if (get_user(val, (int __user *)optval))
2485 return -EFAULT;
2487 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2488 return 0;
2493 * 7.1.1 SCTP_RTOINFO
2495 * The protocol parameters used to initialize and bound retransmission
2496 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2497 * and modify these parameters.
2498 * All parameters are time values, in milliseconds. A value of 0, when
2499 * modifying the parameters, indicates that the current value should not
2500 * be changed.
2503 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2504 struct sctp_rtoinfo rtoinfo;
2505 struct sctp_association *asoc;
2507 if (optlen != sizeof (struct sctp_rtoinfo))
2508 return -EINVAL;
2510 if (copy_from_user(&rtoinfo, optval, optlen))
2511 return -EFAULT;
2513 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2515 /* Set the values to the specific association */
2516 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2517 return -EINVAL;
2519 if (asoc) {
2520 if (rtoinfo.srto_initial != 0)
2521 asoc->rto_initial =
2522 msecs_to_jiffies(rtoinfo.srto_initial);
2523 if (rtoinfo.srto_max != 0)
2524 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2525 if (rtoinfo.srto_min != 0)
2526 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2527 } else {
2528 /* If there is no association or the association-id = 0
2529 * set the values to the endpoint.
2531 struct sctp_sock *sp = sctp_sk(sk);
2533 if (rtoinfo.srto_initial != 0)
2534 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2535 if (rtoinfo.srto_max != 0)
2536 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2537 if (rtoinfo.srto_min != 0)
2538 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2541 return 0;
2546 * 7.1.2 SCTP_ASSOCINFO
2548 * This option is used to tune the the maximum retransmission attempts
2549 * of the association.
2550 * Returns an error if the new association retransmission value is
2551 * greater than the sum of the retransmission value of the peer.
2552 * See [SCTP] for more information.
2555 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2558 struct sctp_assocparams assocparams;
2559 struct sctp_association *asoc;
2561 if (optlen != sizeof(struct sctp_assocparams))
2562 return -EINVAL;
2563 if (copy_from_user(&assocparams, optval, optlen))
2564 return -EFAULT;
2566 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2568 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2569 return -EINVAL;
2571 /* Set the values to the specific association */
2572 if (asoc) {
2573 if (assocparams.sasoc_asocmaxrxt != 0) {
2574 __u32 path_sum = 0;
2575 int paths = 0;
2576 struct list_head *pos;
2577 struct sctp_transport *peer_addr;
2579 list_for_each(pos, &asoc->peer.transport_addr_list) {
2580 peer_addr = list_entry(pos,
2581 struct sctp_transport,
2582 transports);
2583 path_sum += peer_addr->pathmaxrxt;
2584 paths++;
2587 /* Only validate asocmaxrxt if we have more then
2588 * one path/transport. We do this because path
2589 * retransmissions are only counted when we have more
2590 * then one path.
2592 if (paths > 1 &&
2593 assocparams.sasoc_asocmaxrxt > path_sum)
2594 return -EINVAL;
2596 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2599 if (assocparams.sasoc_cookie_life != 0) {
2600 asoc->cookie_life.tv_sec =
2601 assocparams.sasoc_cookie_life / 1000;
2602 asoc->cookie_life.tv_usec =
2603 (assocparams.sasoc_cookie_life % 1000)
2604 * 1000;
2606 } else {
2607 /* Set the values to the endpoint */
2608 struct sctp_sock *sp = sctp_sk(sk);
2610 if (assocparams.sasoc_asocmaxrxt != 0)
2611 sp->assocparams.sasoc_asocmaxrxt =
2612 assocparams.sasoc_asocmaxrxt;
2613 if (assocparams.sasoc_cookie_life != 0)
2614 sp->assocparams.sasoc_cookie_life =
2615 assocparams.sasoc_cookie_life;
2617 return 0;
2621 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2623 * This socket option is a boolean flag which turns on or off mapped V4
2624 * addresses. If this option is turned on and the socket is type
2625 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2626 * If this option is turned off, then no mapping will be done of V4
2627 * addresses and a user will receive both PF_INET6 and PF_INET type
2628 * addresses on the socket.
2630 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2632 int val;
2633 struct sctp_sock *sp = sctp_sk(sk);
2635 if (optlen < sizeof(int))
2636 return -EINVAL;
2637 if (get_user(val, (int __user *)optval))
2638 return -EFAULT;
2639 if (val)
2640 sp->v4mapped = 1;
2641 else
2642 sp->v4mapped = 0;
2644 return 0;
2648 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2650 * This socket option specifies the maximum size to put in any outgoing
2651 * SCTP chunk. If a message is larger than this size it will be
2652 * fragmented by SCTP into the specified size. Note that the underlying
2653 * SCTP implementation may fragment into smaller sized chunks when the
2654 * PMTU of the underlying association is smaller than the value set by
2655 * the user.
2657 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2659 struct sctp_association *asoc;
2660 struct list_head *pos;
2661 struct sctp_sock *sp = sctp_sk(sk);
2662 int val;
2664 if (optlen < sizeof(int))
2665 return -EINVAL;
2666 if (get_user(val, (int __user *)optval))
2667 return -EFAULT;
2668 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2669 return -EINVAL;
2670 sp->user_frag = val;
2672 /* Update the frag_point of the existing associations. */
2673 list_for_each(pos, &(sp->ep->asocs)) {
2674 asoc = list_entry(pos, struct sctp_association, asocs);
2675 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2678 return 0;
2683 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2685 * Requests that the peer mark the enclosed address as the association
2686 * primary. The enclosed address must be one of the association's
2687 * locally bound addresses. The following structure is used to make a
2688 * set primary request:
2690 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2691 int optlen)
2693 struct sctp_sock *sp;
2694 struct sctp_endpoint *ep;
2695 struct sctp_association *asoc = NULL;
2696 struct sctp_setpeerprim prim;
2697 struct sctp_chunk *chunk;
2698 int err;
2700 sp = sctp_sk(sk);
2701 ep = sp->ep;
2703 if (!sctp_addip_enable)
2704 return -EPERM;
2706 if (optlen != sizeof(struct sctp_setpeerprim))
2707 return -EINVAL;
2709 if (copy_from_user(&prim, optval, optlen))
2710 return -EFAULT;
2712 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2713 if (!asoc)
2714 return -EINVAL;
2716 if (!asoc->peer.asconf_capable)
2717 return -EPERM;
2719 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2720 return -EPERM;
2722 if (!sctp_state(asoc, ESTABLISHED))
2723 return -ENOTCONN;
2725 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2726 return -EADDRNOTAVAIL;
2728 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2729 chunk = sctp_make_asconf_set_prim(asoc,
2730 (union sctp_addr *)&prim.sspp_addr);
2731 if (!chunk)
2732 return -ENOMEM;
2734 err = sctp_send_asconf(asoc, chunk);
2736 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2738 return err;
2741 static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2742 int optlen)
2744 struct sctp_setadaption adaption;
2746 if (optlen != sizeof(struct sctp_setadaption))
2747 return -EINVAL;
2748 if (copy_from_user(&adaption, optval, optlen))
2749 return -EFAULT;
2751 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
2753 return 0;
2756 /* API 6.2 setsockopt(), getsockopt()
2758 * Applications use setsockopt() and getsockopt() to set or retrieve
2759 * socket options. Socket options are used to change the default
2760 * behavior of sockets calls. They are described in Section 7.
2762 * The syntax is:
2764 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2765 * int __user *optlen);
2766 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2767 * int optlen);
2769 * sd - the socket descript.
2770 * level - set to IPPROTO_SCTP for all SCTP options.
2771 * optname - the option name.
2772 * optval - the buffer to store the value of the option.
2773 * optlen - the size of the buffer.
2775 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2776 char __user *optval, int optlen)
2778 int retval = 0;
2780 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2781 sk, optname);
2783 /* I can hardly begin to describe how wrong this is. This is
2784 * so broken as to be worse than useless. The API draft
2785 * REALLY is NOT helpful here... I am not convinced that the
2786 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2787 * are at all well-founded.
2789 if (level != SOL_SCTP) {
2790 struct sctp_af *af = sctp_sk(sk)->pf->af;
2791 retval = af->setsockopt(sk, level, optname, optval, optlen);
2792 goto out_nounlock;
2795 sctp_lock_sock(sk);
2797 switch (optname) {
2798 case SCTP_SOCKOPT_BINDX_ADD:
2799 /* 'optlen' is the size of the addresses buffer. */
2800 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2801 optlen, SCTP_BINDX_ADD_ADDR);
2802 break;
2804 case SCTP_SOCKOPT_BINDX_REM:
2805 /* 'optlen' is the size of the addresses buffer. */
2806 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2807 optlen, SCTP_BINDX_REM_ADDR);
2808 break;
2810 case SCTP_SOCKOPT_CONNECTX:
2811 /* 'optlen' is the size of the addresses buffer. */
2812 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2813 optlen);
2814 break;
2816 case SCTP_DISABLE_FRAGMENTS:
2817 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2818 break;
2820 case SCTP_EVENTS:
2821 retval = sctp_setsockopt_events(sk, optval, optlen);
2822 break;
2824 case SCTP_AUTOCLOSE:
2825 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2826 break;
2828 case SCTP_PEER_ADDR_PARAMS:
2829 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2830 break;
2832 case SCTP_DELAYED_ACK_TIME:
2833 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2834 break;
2836 case SCTP_INITMSG:
2837 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2838 break;
2839 case SCTP_DEFAULT_SEND_PARAM:
2840 retval = sctp_setsockopt_default_send_param(sk, optval,
2841 optlen);
2842 break;
2843 case SCTP_PRIMARY_ADDR:
2844 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2845 break;
2846 case SCTP_SET_PEER_PRIMARY_ADDR:
2847 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2848 break;
2849 case SCTP_NODELAY:
2850 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2851 break;
2852 case SCTP_RTOINFO:
2853 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2854 break;
2855 case SCTP_ASSOCINFO:
2856 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2857 break;
2858 case SCTP_I_WANT_MAPPED_V4_ADDR:
2859 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2860 break;
2861 case SCTP_MAXSEG:
2862 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2863 break;
2864 case SCTP_ADAPTION_LAYER:
2865 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2866 break;
2868 default:
2869 retval = -ENOPROTOOPT;
2870 break;
2873 sctp_release_sock(sk);
2875 out_nounlock:
2876 return retval;
2879 /* API 3.1.6 connect() - UDP Style Syntax
2881 * An application may use the connect() call in the UDP model to initiate an
2882 * association without sending data.
2884 * The syntax is:
2886 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2888 * sd: the socket descriptor to have a new association added to.
2890 * nam: the address structure (either struct sockaddr_in or struct
2891 * sockaddr_in6 defined in RFC2553 [7]).
2893 * len: the size of the address.
2895 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2896 int addr_len)
2898 int err = 0;
2899 struct sctp_af *af;
2901 sctp_lock_sock(sk);
2903 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2904 __FUNCTION__, sk, addr, addr_len);
2906 /* Validate addr_len before calling common connect/connectx routine. */
2907 af = sctp_get_af_specific(addr->sa_family);
2908 if (!af || addr_len < af->sockaddr_len) {
2909 err = -EINVAL;
2910 } else {
2911 /* Pass correct addr len to common routine (so it knows there
2912 * is only one address being passed.
2914 err = __sctp_connect(sk, addr, af->sockaddr_len);
2917 sctp_release_sock(sk);
2918 return err;
2921 /* FIXME: Write comments. */
2922 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2924 return -EOPNOTSUPP; /* STUB */
2927 /* 4.1.4 accept() - TCP Style Syntax
2929 * Applications use accept() call to remove an established SCTP
2930 * association from the accept queue of the endpoint. A new socket
2931 * descriptor will be returned from accept() to represent the newly
2932 * formed association.
2934 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2936 struct sctp_sock *sp;
2937 struct sctp_endpoint *ep;
2938 struct sock *newsk = NULL;
2939 struct sctp_association *asoc;
2940 long timeo;
2941 int error = 0;
2943 sctp_lock_sock(sk);
2945 sp = sctp_sk(sk);
2946 ep = sp->ep;
2948 if (!sctp_style(sk, TCP)) {
2949 error = -EOPNOTSUPP;
2950 goto out;
2953 if (!sctp_sstate(sk, LISTENING)) {
2954 error = -EINVAL;
2955 goto out;
2958 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2960 error = sctp_wait_for_accept(sk, timeo);
2961 if (error)
2962 goto out;
2964 /* We treat the list of associations on the endpoint as the accept
2965 * queue and pick the first association on the list.
2967 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2969 newsk = sp->pf->create_accept_sk(sk, asoc);
2970 if (!newsk) {
2971 error = -ENOMEM;
2972 goto out;
2975 /* Populate the fields of the newsk from the oldsk and migrate the
2976 * asoc to the newsk.
2978 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2980 out:
2981 sctp_release_sock(sk);
2982 *err = error;
2983 return newsk;
2986 /* The SCTP ioctl handler. */
2987 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2989 return -ENOIOCTLCMD;
2992 /* This is the function which gets called during socket creation to
2993 * initialized the SCTP-specific portion of the sock.
2994 * The sock structure should already be zero-filled memory.
2996 SCTP_STATIC int sctp_init_sock(struct sock *sk)
2998 struct sctp_endpoint *ep;
2999 struct sctp_sock *sp;
3001 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3003 sp = sctp_sk(sk);
3005 /* Initialize the SCTP per socket area. */
3006 switch (sk->sk_type) {
3007 case SOCK_SEQPACKET:
3008 sp->type = SCTP_SOCKET_UDP;
3009 break;
3010 case SOCK_STREAM:
3011 sp->type = SCTP_SOCKET_TCP;
3012 break;
3013 default:
3014 return -ESOCKTNOSUPPORT;
3017 /* Initialize default send parameters. These parameters can be
3018 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3020 sp->default_stream = 0;
3021 sp->default_ppid = 0;
3022 sp->default_flags = 0;
3023 sp->default_context = 0;
3024 sp->default_timetolive = 0;
3026 /* Initialize default setup parameters. These parameters
3027 * can be modified with the SCTP_INITMSG socket option or
3028 * overridden by the SCTP_INIT CMSG.
3030 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3031 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3032 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3033 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3035 /* Initialize default RTO related parameters. These parameters can
3036 * be modified for with the SCTP_RTOINFO socket option.
3038 sp->rtoinfo.srto_initial = sctp_rto_initial;
3039 sp->rtoinfo.srto_max = sctp_rto_max;
3040 sp->rtoinfo.srto_min = sctp_rto_min;
3042 /* Initialize default association related parameters. These parameters
3043 * can be modified with the SCTP_ASSOCINFO socket option.
3045 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3046 sp->assocparams.sasoc_number_peer_destinations = 0;
3047 sp->assocparams.sasoc_peer_rwnd = 0;
3048 sp->assocparams.sasoc_local_rwnd = 0;
3049 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3051 /* Initialize default event subscriptions. By default, all the
3052 * options are off.
3054 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3056 /* Default Peer Address Parameters. These defaults can
3057 * be modified via SCTP_PEER_ADDR_PARAMS
3059 sp->hbinterval = sctp_hb_interval;
3060 sp->pathmaxrxt = sctp_max_retrans_path;
3061 sp->pathmtu = 0; // allow default discovery
3062 sp->sackdelay = sctp_sack_timeout;
3063 sp->param_flags = SPP_HB_ENABLE |
3064 SPP_PMTUD_ENABLE |
3065 SPP_SACKDELAY_ENABLE;
3067 /* If enabled no SCTP message fragmentation will be performed.
3068 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3070 sp->disable_fragments = 0;
3072 /* Enable Nagle algorithm by default. */
3073 sp->nodelay = 0;
3075 /* Enable by default. */
3076 sp->v4mapped = 1;
3078 /* Auto-close idle associations after the configured
3079 * number of seconds. A value of 0 disables this
3080 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3081 * for UDP-style sockets only.
3083 sp->autoclose = 0;
3085 /* User specified fragmentation limit. */
3086 sp->user_frag = 0;
3088 sp->adaption_ind = 0;
3090 sp->pf = sctp_get_pf_specific(sk->sk_family);
3092 /* Control variables for partial data delivery. */
3093 sp->pd_mode = 0;
3094 skb_queue_head_init(&sp->pd_lobby);
3096 /* Create a per socket endpoint structure. Even if we
3097 * change the data structure relationships, this may still
3098 * be useful for storing pre-connect address information.
3100 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3101 if (!ep)
3102 return -ENOMEM;
3104 sp->ep = ep;
3105 sp->hmac = NULL;
3107 SCTP_DBG_OBJCNT_INC(sock);
3108 return 0;
3111 /* Cleanup any SCTP per socket resources. */
3112 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3114 struct sctp_endpoint *ep;
3116 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3118 /* Release our hold on the endpoint. */
3119 ep = sctp_sk(sk)->ep;
3120 sctp_endpoint_free(ep);
3122 return 0;
3125 /* API 4.1.7 shutdown() - TCP Style Syntax
3126 * int shutdown(int socket, int how);
3128 * sd - the socket descriptor of the association to be closed.
3129 * how - Specifies the type of shutdown. The values are
3130 * as follows:
3131 * SHUT_RD
3132 * Disables further receive operations. No SCTP
3133 * protocol action is taken.
3134 * SHUT_WR
3135 * Disables further send operations, and initiates
3136 * the SCTP shutdown sequence.
3137 * SHUT_RDWR
3138 * Disables further send and receive operations
3139 * and initiates the SCTP shutdown sequence.
3141 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3143 struct sctp_endpoint *ep;
3144 struct sctp_association *asoc;
3146 if (!sctp_style(sk, TCP))
3147 return;
3149 if (how & SEND_SHUTDOWN) {
3150 ep = sctp_sk(sk)->ep;
3151 if (!list_empty(&ep->asocs)) {
3152 asoc = list_entry(ep->asocs.next,
3153 struct sctp_association, asocs);
3154 sctp_primitive_SHUTDOWN(asoc, NULL);
3159 /* 7.2.1 Association Status (SCTP_STATUS)
3161 * Applications can retrieve current status information about an
3162 * association, including association state, peer receiver window size,
3163 * number of unacked data chunks, and number of data chunks pending
3164 * receipt. This information is read-only.
3166 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3167 char __user *optval,
3168 int __user *optlen)
3170 struct sctp_status status;
3171 struct sctp_association *asoc = NULL;
3172 struct sctp_transport *transport;
3173 sctp_assoc_t associd;
3174 int retval = 0;
3176 if (len != sizeof(status)) {
3177 retval = -EINVAL;
3178 goto out;
3181 if (copy_from_user(&status, optval, sizeof(status))) {
3182 retval = -EFAULT;
3183 goto out;
3186 associd = status.sstat_assoc_id;
3187 asoc = sctp_id2assoc(sk, associd);
3188 if (!asoc) {
3189 retval = -EINVAL;
3190 goto out;
3193 transport = asoc->peer.primary_path;
3195 status.sstat_assoc_id = sctp_assoc2id(asoc);
3196 status.sstat_state = asoc->state;
3197 status.sstat_rwnd = asoc->peer.rwnd;
3198 status.sstat_unackdata = asoc->unack_data;
3200 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3201 status.sstat_instrms = asoc->c.sinit_max_instreams;
3202 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3203 status.sstat_fragmentation_point = asoc->frag_point;
3204 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3205 flip_to_n((union sctp_addr *)&status.sstat_primary.spinfo_address,
3206 &transport->ipaddr_h);
3207 /* Map ipv4 address into v4-mapped-on-v6 address. */
3208 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3209 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3210 status.sstat_primary.spinfo_state = transport->state;
3211 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3212 status.sstat_primary.spinfo_srtt = transport->srtt;
3213 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3214 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3216 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3217 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3219 if (put_user(len, optlen)) {
3220 retval = -EFAULT;
3221 goto out;
3224 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3225 len, status.sstat_state, status.sstat_rwnd,
3226 status.sstat_assoc_id);
3228 if (copy_to_user(optval, &status, len)) {
3229 retval = -EFAULT;
3230 goto out;
3233 out:
3234 return (retval);
3238 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3240 * Applications can retrieve information about a specific peer address
3241 * of an association, including its reachability state, congestion
3242 * window, and retransmission timer values. This information is
3243 * read-only.
3245 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3246 char __user *optval,
3247 int __user *optlen)
3249 struct sctp_paddrinfo pinfo;
3250 struct sctp_transport *transport;
3251 int retval = 0;
3253 if (len != sizeof(pinfo)) {
3254 retval = -EINVAL;
3255 goto out;
3258 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3259 retval = -EFAULT;
3260 goto out;
3263 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3264 pinfo.spinfo_assoc_id);
3265 if (!transport)
3266 return -EINVAL;
3268 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3269 pinfo.spinfo_state = transport->state;
3270 pinfo.spinfo_cwnd = transport->cwnd;
3271 pinfo.spinfo_srtt = transport->srtt;
3272 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3273 pinfo.spinfo_mtu = transport->pathmtu;
3275 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3276 pinfo.spinfo_state = SCTP_ACTIVE;
3278 if (put_user(len, optlen)) {
3279 retval = -EFAULT;
3280 goto out;
3283 if (copy_to_user(optval, &pinfo, len)) {
3284 retval = -EFAULT;
3285 goto out;
3288 out:
3289 return (retval);
3292 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3294 * This option is a on/off flag. If enabled no SCTP message
3295 * fragmentation will be performed. Instead if a message being sent
3296 * exceeds the current PMTU size, the message will NOT be sent and
3297 * instead a error will be indicated to the user.
3299 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3300 char __user *optval, int __user *optlen)
3302 int val;
3304 if (len < sizeof(int))
3305 return -EINVAL;
3307 len = sizeof(int);
3308 val = (sctp_sk(sk)->disable_fragments == 1);
3309 if (put_user(len, optlen))
3310 return -EFAULT;
3311 if (copy_to_user(optval, &val, len))
3312 return -EFAULT;
3313 return 0;
3316 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3318 * This socket option is used to specify various notifications and
3319 * ancillary data the user wishes to receive.
3321 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3322 int __user *optlen)
3324 if (len != sizeof(struct sctp_event_subscribe))
3325 return -EINVAL;
3326 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3327 return -EFAULT;
3328 return 0;
3331 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3333 * This socket option is applicable to the UDP-style socket only. When
3334 * set it will cause associations that are idle for more than the
3335 * specified number of seconds to automatically close. An association
3336 * being idle is defined an association that has NOT sent or received
3337 * user data. The special value of '0' indicates that no automatic
3338 * close of any associations should be performed. The option expects an
3339 * integer defining the number of seconds of idle time before an
3340 * association is closed.
3342 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3344 /* Applicable to UDP-style socket only */
3345 if (sctp_style(sk, TCP))
3346 return -EOPNOTSUPP;
3347 if (len != sizeof(int))
3348 return -EINVAL;
3349 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3350 return -EFAULT;
3351 return 0;
3354 /* Helper routine to branch off an association to a new socket. */
3355 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3356 struct socket **sockp)
3358 struct sock *sk = asoc->base.sk;
3359 struct socket *sock;
3360 struct inet_sock *inetsk;
3361 int err = 0;
3363 /* An association cannot be branched off from an already peeled-off
3364 * socket, nor is this supported for tcp style sockets.
3366 if (!sctp_style(sk, UDP))
3367 return -EINVAL;
3369 /* Create a new socket. */
3370 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3371 if (err < 0)
3372 return err;
3374 /* Populate the fields of the newsk from the oldsk and migrate the
3375 * asoc to the newsk.
3377 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3379 /* Make peeled-off sockets more like 1-1 accepted sockets.
3380 * Set the daddr and initialize id to something more random
3382 inetsk = inet_sk(sock->sk);
3383 inetsk->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
3384 inetsk->id = asoc->next_tsn ^ jiffies;
3386 *sockp = sock;
3388 return err;
3391 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3393 sctp_peeloff_arg_t peeloff;
3394 struct socket *newsock;
3395 int retval = 0;
3396 struct sctp_association *asoc;
3398 if (len != sizeof(sctp_peeloff_arg_t))
3399 return -EINVAL;
3400 if (copy_from_user(&peeloff, optval, len))
3401 return -EFAULT;
3403 asoc = sctp_id2assoc(sk, peeloff.associd);
3404 if (!asoc) {
3405 retval = -EINVAL;
3406 goto out;
3409 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3411 retval = sctp_do_peeloff(asoc, &newsock);
3412 if (retval < 0)
3413 goto out;
3415 /* Map the socket to an unused fd that can be returned to the user. */
3416 retval = sock_map_fd(newsock);
3417 if (retval < 0) {
3418 sock_release(newsock);
3419 goto out;
3422 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3423 __FUNCTION__, sk, asoc, newsock->sk, retval);
3425 /* Return the fd mapped to the new socket. */
3426 peeloff.sd = retval;
3427 if (copy_to_user(optval, &peeloff, len))
3428 retval = -EFAULT;
3430 out:
3431 return retval;
3434 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3436 * Applications can enable or disable heartbeats for any peer address of
3437 * an association, modify an address's heartbeat interval, force a
3438 * heartbeat to be sent immediately, and adjust the address's maximum
3439 * number of retransmissions sent before an address is considered
3440 * unreachable. The following structure is used to access and modify an
3441 * address's parameters:
3443 * struct sctp_paddrparams {
3444 * sctp_assoc_t spp_assoc_id;
3445 * struct sockaddr_storage spp_address;
3446 * uint32_t spp_hbinterval;
3447 * uint16_t spp_pathmaxrxt;
3448 * uint32_t spp_pathmtu;
3449 * uint32_t spp_sackdelay;
3450 * uint32_t spp_flags;
3451 * };
3453 * spp_assoc_id - (one-to-many style socket) This is filled in the
3454 * application, and identifies the association for
3455 * this query.
3456 * spp_address - This specifies which address is of interest.
3457 * spp_hbinterval - This contains the value of the heartbeat interval,
3458 * in milliseconds. If a value of zero
3459 * is present in this field then no changes are to
3460 * be made to this parameter.
3461 * spp_pathmaxrxt - This contains the maximum number of
3462 * retransmissions before this address shall be
3463 * considered unreachable. If a value of zero
3464 * is present in this field then no changes are to
3465 * be made to this parameter.
3466 * spp_pathmtu - When Path MTU discovery is disabled the value
3467 * specified here will be the "fixed" path mtu.
3468 * Note that if the spp_address field is empty
3469 * then all associations on this address will
3470 * have this fixed path mtu set upon them.
3472 * spp_sackdelay - When delayed sack is enabled, this value specifies
3473 * the number of milliseconds that sacks will be delayed
3474 * for. This value will apply to all addresses of an
3475 * association if the spp_address field is empty. Note
3476 * also, that if delayed sack is enabled and this
3477 * value is set to 0, no change is made to the last
3478 * recorded delayed sack timer value.
3480 * spp_flags - These flags are used to control various features
3481 * on an association. The flag field may contain
3482 * zero or more of the following options.
3484 * SPP_HB_ENABLE - Enable heartbeats on the
3485 * specified address. Note that if the address
3486 * field is empty all addresses for the association
3487 * have heartbeats enabled upon them.
3489 * SPP_HB_DISABLE - Disable heartbeats on the
3490 * speicifed address. Note that if the address
3491 * field is empty all addresses for the association
3492 * will have their heartbeats disabled. Note also
3493 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3494 * mutually exclusive, only one of these two should
3495 * be specified. Enabling both fields will have
3496 * undetermined results.
3498 * SPP_HB_DEMAND - Request a user initiated heartbeat
3499 * to be made immediately.
3501 * SPP_PMTUD_ENABLE - This field will enable PMTU
3502 * discovery upon the specified address. Note that
3503 * if the address feild is empty then all addresses
3504 * on the association are effected.
3506 * SPP_PMTUD_DISABLE - This field will disable PMTU
3507 * discovery upon the specified address. Note that
3508 * if the address feild is empty then all addresses
3509 * on the association are effected. Not also that
3510 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3511 * exclusive. Enabling both will have undetermined
3512 * results.
3514 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3515 * on delayed sack. The time specified in spp_sackdelay
3516 * is used to specify the sack delay for this address. Note
3517 * that if spp_address is empty then all addresses will
3518 * enable delayed sack and take on the sack delay
3519 * value specified in spp_sackdelay.
3520 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3521 * off delayed sack. If the spp_address field is blank then
3522 * delayed sack is disabled for the entire association. Note
3523 * also that this field is mutually exclusive to
3524 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3525 * results.
3527 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3528 char __user *optval, int __user *optlen)
3530 struct sctp_paddrparams params;
3531 struct sctp_transport *trans = NULL;
3532 struct sctp_association *asoc = NULL;
3533 struct sctp_sock *sp = sctp_sk(sk);
3535 if (len != sizeof(struct sctp_paddrparams))
3536 return -EINVAL;
3538 if (copy_from_user(&params, optval, len))
3539 return -EFAULT;
3541 /* If an address other than INADDR_ANY is specified, and
3542 * no transport is found, then the request is invalid.
3544 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3545 trans = sctp_addr_id2transport(sk, &params.spp_address,
3546 params.spp_assoc_id);
3547 if (!trans) {
3548 SCTP_DEBUG_PRINTK("Failed no transport\n");
3549 return -EINVAL;
3553 /* Get association, if assoc_id != 0 and the socket is a one
3554 * to many style socket, and an association was not found, then
3555 * the id was invalid.
3557 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3558 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3559 SCTP_DEBUG_PRINTK("Failed no association\n");
3560 return -EINVAL;
3563 if (trans) {
3564 /* Fetch transport values. */
3565 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3566 params.spp_pathmtu = trans->pathmtu;
3567 params.spp_pathmaxrxt = trans->pathmaxrxt;
3568 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3570 /*draft-11 doesn't say what to return in spp_flags*/
3571 params.spp_flags = trans->param_flags;
3572 } else if (asoc) {
3573 /* Fetch association values. */
3574 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3575 params.spp_pathmtu = asoc->pathmtu;
3576 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3577 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3579 /*draft-11 doesn't say what to return in spp_flags*/
3580 params.spp_flags = asoc->param_flags;
3581 } else {
3582 /* Fetch socket values. */
3583 params.spp_hbinterval = sp->hbinterval;
3584 params.spp_pathmtu = sp->pathmtu;
3585 params.spp_sackdelay = sp->sackdelay;
3586 params.spp_pathmaxrxt = sp->pathmaxrxt;
3588 /*draft-11 doesn't say what to return in spp_flags*/
3589 params.spp_flags = sp->param_flags;
3592 if (copy_to_user(optval, &params, len))
3593 return -EFAULT;
3595 if (put_user(len, optlen))
3596 return -EFAULT;
3598 return 0;
3601 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3603 * This options will get or set the delayed ack timer. The time is set
3604 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3605 * endpoints default delayed ack timer value. If the assoc_id field is
3606 * non-zero, then the set or get effects the specified association.
3608 * struct sctp_assoc_value {
3609 * sctp_assoc_t assoc_id;
3610 * uint32_t assoc_value;
3611 * };
3613 * assoc_id - This parameter, indicates which association the
3614 * user is preforming an action upon. Note that if
3615 * this field's value is zero then the endpoints
3616 * default value is changed (effecting future
3617 * associations only).
3619 * assoc_value - This parameter contains the number of milliseconds
3620 * that the user is requesting the delayed ACK timer
3621 * be set to. Note that this value is defined in
3622 * the standard to be between 200 and 500 milliseconds.
3624 * Note: a value of zero will leave the value alone,
3625 * but disable SACK delay. A non-zero value will also
3626 * enable SACK delay.
3628 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3629 char __user *optval,
3630 int __user *optlen)
3632 struct sctp_assoc_value params;
3633 struct sctp_association *asoc = NULL;
3634 struct sctp_sock *sp = sctp_sk(sk);
3636 if (len != sizeof(struct sctp_assoc_value))
3637 return - EINVAL;
3639 if (copy_from_user(&params, optval, len))
3640 return -EFAULT;
3642 /* Get association, if assoc_id != 0 and the socket is a one
3643 * to many style socket, and an association was not found, then
3644 * the id was invalid.
3646 asoc = sctp_id2assoc(sk, params.assoc_id);
3647 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3648 return -EINVAL;
3650 if (asoc) {
3651 /* Fetch association values. */
3652 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3653 params.assoc_value = jiffies_to_msecs(
3654 asoc->sackdelay);
3655 else
3656 params.assoc_value = 0;
3657 } else {
3658 /* Fetch socket values. */
3659 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3660 params.assoc_value = sp->sackdelay;
3661 else
3662 params.assoc_value = 0;
3665 if (copy_to_user(optval, &params, len))
3666 return -EFAULT;
3668 if (put_user(len, optlen))
3669 return -EFAULT;
3671 return 0;
3674 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3676 * Applications can specify protocol parameters for the default association
3677 * initialization. The option name argument to setsockopt() and getsockopt()
3678 * is SCTP_INITMSG.
3680 * Setting initialization parameters is effective only on an unconnected
3681 * socket (for UDP-style sockets only future associations are effected
3682 * by the change). With TCP-style sockets, this option is inherited by
3683 * sockets derived from a listener socket.
3685 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3687 if (len != sizeof(struct sctp_initmsg))
3688 return -EINVAL;
3689 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3690 return -EFAULT;
3691 return 0;
3694 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3695 char __user *optval,
3696 int __user *optlen)
3698 sctp_assoc_t id;
3699 struct sctp_association *asoc;
3700 struct list_head *pos;
3701 int cnt = 0;
3703 if (len != sizeof(sctp_assoc_t))
3704 return -EINVAL;
3706 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3707 return -EFAULT;
3709 /* For UDP-style sockets, id specifies the association to query. */
3710 asoc = sctp_id2assoc(sk, id);
3711 if (!asoc)
3712 return -EINVAL;
3714 list_for_each(pos, &asoc->peer.transport_addr_list) {
3715 cnt ++;
3718 return cnt;
3722 * Old API for getting list of peer addresses. Does not work for 32-bit
3723 * programs running on a 64-bit kernel
3725 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3726 char __user *optval,
3727 int __user *optlen)
3729 struct sctp_association *asoc;
3730 struct list_head *pos;
3731 int cnt = 0;
3732 struct sctp_getaddrs_old getaddrs;
3733 struct sctp_transport *from;
3734 void __user *to;
3735 union sctp_addr temp;
3736 struct sctp_sock *sp = sctp_sk(sk);
3737 int addrlen;
3739 if (len != sizeof(struct sctp_getaddrs_old))
3740 return -EINVAL;
3742 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3743 return -EFAULT;
3745 if (getaddrs.addr_num <= 0) return -EINVAL;
3747 /* For UDP-style sockets, id specifies the association to query. */
3748 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3749 if (!asoc)
3750 return -EINVAL;
3752 to = (void __user *)getaddrs.addrs;
3753 list_for_each(pos, &asoc->peer.transport_addr_list) {
3754 from = list_entry(pos, struct sctp_transport, transports);
3755 memcpy(&temp, &from->ipaddr_h, sizeof(temp));
3756 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3757 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3758 temp.v4.sin_port = htons(temp.v4.sin_port);
3759 if (copy_to_user(to, &temp, addrlen))
3760 return -EFAULT;
3761 to += addrlen ;
3762 cnt ++;
3763 if (cnt >= getaddrs.addr_num) break;
3765 getaddrs.addr_num = cnt;
3766 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3767 return -EFAULT;
3769 return 0;
3772 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3773 char __user *optval, int __user *optlen)
3775 struct sctp_association *asoc;
3776 struct list_head *pos;
3777 int cnt = 0;
3778 struct sctp_getaddrs getaddrs;
3779 struct sctp_transport *from;
3780 void __user *to;
3781 union sctp_addr temp;
3782 struct sctp_sock *sp = sctp_sk(sk);
3783 int addrlen;
3784 size_t space_left;
3785 int bytes_copied;
3787 if (len < sizeof(struct sctp_getaddrs))
3788 return -EINVAL;
3790 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3791 return -EFAULT;
3793 /* For UDP-style sockets, id specifies the association to query. */
3794 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3795 if (!asoc)
3796 return -EINVAL;
3798 to = optval + offsetof(struct sctp_getaddrs,addrs);
3799 space_left = len - sizeof(struct sctp_getaddrs) -
3800 offsetof(struct sctp_getaddrs,addrs);
3802 list_for_each(pos, &asoc->peer.transport_addr_list) {
3803 from = list_entry(pos, struct sctp_transport, transports);
3804 memcpy(&temp, &from->ipaddr_h, sizeof(temp));
3805 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3806 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3807 if(space_left < addrlen)
3808 return -ENOMEM;
3809 temp.v4.sin_port = htons(temp.v4.sin_port);
3810 if (copy_to_user(to, &temp, addrlen))
3811 return -EFAULT;
3812 to += addrlen;
3813 cnt++;
3814 space_left -= addrlen;
3817 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3818 return -EFAULT;
3819 bytes_copied = ((char __user *)to) - optval;
3820 if (put_user(bytes_copied, optlen))
3821 return -EFAULT;
3823 return 0;
3826 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3827 char __user *optval,
3828 int __user *optlen)
3830 sctp_assoc_t id;
3831 struct sctp_bind_addr *bp;
3832 struct sctp_association *asoc;
3833 struct list_head *pos;
3834 struct sctp_sockaddr_entry *addr;
3835 rwlock_t *addr_lock;
3836 unsigned long flags;
3837 int cnt = 0;
3839 if (len != sizeof(sctp_assoc_t))
3840 return -EINVAL;
3842 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3843 return -EFAULT;
3846 * For UDP-style sockets, id specifies the association to query.
3847 * If the id field is set to the value '0' then the locally bound
3848 * addresses are returned without regard to any particular
3849 * association.
3851 if (0 == id) {
3852 bp = &sctp_sk(sk)->ep->base.bind_addr;
3853 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3854 } else {
3855 asoc = sctp_id2assoc(sk, id);
3856 if (!asoc)
3857 return -EINVAL;
3858 bp = &asoc->base.bind_addr;
3859 addr_lock = &asoc->base.addr_lock;
3862 sctp_read_lock(addr_lock);
3864 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3865 * addresses from the global local address list.
3867 if (sctp_list_single_entry(&bp->address_list)) {
3868 addr = list_entry(bp->address_list.next,
3869 struct sctp_sockaddr_entry, list);
3870 if (sctp_is_any(&addr->a_h)) {
3871 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3872 list_for_each(pos, &sctp_local_addr_list) {
3873 addr = list_entry(pos,
3874 struct sctp_sockaddr_entry,
3875 list);
3876 if ((PF_INET == sk->sk_family) &&
3877 (AF_INET6 == addr->a_h.sa.sa_family))
3878 continue;
3879 cnt++;
3881 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3882 flags);
3883 } else {
3884 cnt = 1;
3886 goto done;
3889 list_for_each(pos, &bp->address_list) {
3890 cnt ++;
3893 done:
3894 sctp_read_unlock(addr_lock);
3895 return cnt;
3898 /* Helper function that copies local addresses to user and returns the number
3899 * of addresses copied.
3901 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3902 void __user *to)
3904 struct list_head *pos;
3905 struct sctp_sockaddr_entry *addr;
3906 unsigned long flags;
3907 union sctp_addr temp;
3908 int cnt = 0;
3909 int addrlen;
3911 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3912 list_for_each(pos, &sctp_local_addr_list) {
3913 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3914 if ((PF_INET == sk->sk_family) &&
3915 (AF_INET6 == addr->a_h.sa.sa_family))
3916 continue;
3917 memcpy(&temp, &addr->a_h, sizeof(temp));
3918 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3919 &temp);
3920 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3921 temp.v4.sin_port = htons(port);
3922 if (copy_to_user(to, &temp, addrlen)) {
3923 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3924 flags);
3925 return -EFAULT;
3927 to += addrlen;
3928 cnt ++;
3929 if (cnt >= max_addrs) break;
3931 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3933 return cnt;
3936 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
3937 void __user **to, size_t space_left)
3939 struct list_head *pos;
3940 struct sctp_sockaddr_entry *addr;
3941 unsigned long flags;
3942 union sctp_addr temp;
3943 int cnt = 0;
3944 int addrlen;
3946 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3947 list_for_each(pos, &sctp_local_addr_list) {
3948 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3949 if ((PF_INET == sk->sk_family) &&
3950 (AF_INET6 == addr->a_h.sa.sa_family))
3951 continue;
3952 memcpy(&temp, &addr->a_h, sizeof(temp));
3953 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3954 &temp);
3955 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3956 if(space_left<addrlen)
3957 return -ENOMEM;
3958 temp.v4.sin_port = htons(port);
3959 if (copy_to_user(*to, &temp, addrlen)) {
3960 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3961 flags);
3962 return -EFAULT;
3964 *to += addrlen;
3965 cnt ++;
3966 space_left -= addrlen;
3968 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3970 return cnt;
3973 /* Old API for getting list of local addresses. Does not work for 32-bit
3974 * programs running on a 64-bit kernel
3976 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3977 char __user *optval, int __user *optlen)
3979 struct sctp_bind_addr *bp;
3980 struct sctp_association *asoc;
3981 struct list_head *pos;
3982 int cnt = 0;
3983 struct sctp_getaddrs_old getaddrs;
3984 struct sctp_sockaddr_entry *addr;
3985 void __user *to;
3986 union sctp_addr temp;
3987 struct sctp_sock *sp = sctp_sk(sk);
3988 int addrlen;
3989 rwlock_t *addr_lock;
3990 int err = 0;
3992 if (len != sizeof(struct sctp_getaddrs_old))
3993 return -EINVAL;
3995 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3996 return -EFAULT;
3998 if (getaddrs.addr_num <= 0) return -EINVAL;
4000 * For UDP-style sockets, id specifies the association to query.
4001 * If the id field is set to the value '0' then the locally bound
4002 * addresses are returned without regard to any particular
4003 * association.
4005 if (0 == getaddrs.assoc_id) {
4006 bp = &sctp_sk(sk)->ep->base.bind_addr;
4007 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4008 } else {
4009 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4010 if (!asoc)
4011 return -EINVAL;
4012 bp = &asoc->base.bind_addr;
4013 addr_lock = &asoc->base.addr_lock;
4016 to = getaddrs.addrs;
4018 sctp_read_lock(addr_lock);
4020 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4021 * addresses from the global local address list.
4023 if (sctp_list_single_entry(&bp->address_list)) {
4024 addr = list_entry(bp->address_list.next,
4025 struct sctp_sockaddr_entry, list);
4026 if (sctp_is_any(&addr->a_h)) {
4027 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4028 getaddrs.addr_num,
4029 to);
4030 if (cnt < 0) {
4031 err = cnt;
4032 goto unlock;
4034 goto copy_getaddrs;
4038 list_for_each(pos, &bp->address_list) {
4039 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4040 memcpy(&temp, &addr->a_h, sizeof(temp));
4041 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4042 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4043 temp.v4.sin_port = htons(temp.v4.sin_port);
4044 if (copy_to_user(to, &temp, addrlen)) {
4045 err = -EFAULT;
4046 goto unlock;
4048 to += addrlen;
4049 cnt ++;
4050 if (cnt >= getaddrs.addr_num) break;
4053 copy_getaddrs:
4054 getaddrs.addr_num = cnt;
4055 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4056 err = -EFAULT;
4058 unlock:
4059 sctp_read_unlock(addr_lock);
4060 return err;
4063 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4064 char __user *optval, int __user *optlen)
4066 struct sctp_bind_addr *bp;
4067 struct sctp_association *asoc;
4068 struct list_head *pos;
4069 int cnt = 0;
4070 struct sctp_getaddrs getaddrs;
4071 struct sctp_sockaddr_entry *addr;
4072 void __user *to;
4073 union sctp_addr temp;
4074 struct sctp_sock *sp = sctp_sk(sk);
4075 int addrlen;
4076 rwlock_t *addr_lock;
4077 int err = 0;
4078 size_t space_left;
4079 int bytes_copied;
4081 if (len <= sizeof(struct sctp_getaddrs))
4082 return -EINVAL;
4084 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4085 return -EFAULT;
4088 * For UDP-style sockets, id specifies the association to query.
4089 * If the id field is set to the value '0' then the locally bound
4090 * addresses are returned without regard to any particular
4091 * association.
4093 if (0 == getaddrs.assoc_id) {
4094 bp = &sctp_sk(sk)->ep->base.bind_addr;
4095 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4096 } else {
4097 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4098 if (!asoc)
4099 return -EINVAL;
4100 bp = &asoc->base.bind_addr;
4101 addr_lock = &asoc->base.addr_lock;
4104 to = optval + offsetof(struct sctp_getaddrs,addrs);
4105 space_left = len - sizeof(struct sctp_getaddrs) -
4106 offsetof(struct sctp_getaddrs,addrs);
4108 sctp_read_lock(addr_lock);
4110 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4111 * addresses from the global local address list.
4113 if (sctp_list_single_entry(&bp->address_list)) {
4114 addr = list_entry(bp->address_list.next,
4115 struct sctp_sockaddr_entry, list);
4116 if (sctp_is_any(&addr->a_h)) {
4117 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4118 &to, space_left);
4119 if (cnt < 0) {
4120 err = cnt;
4121 goto unlock;
4123 goto copy_getaddrs;
4127 list_for_each(pos, &bp->address_list) {
4128 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4129 memcpy(&temp, &addr->a_h, sizeof(temp));
4130 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4131 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4132 if(space_left < addrlen)
4133 return -ENOMEM; /*fixme: right error?*/
4134 temp.v4.sin_port = htons(temp.v4.sin_port);
4135 if (copy_to_user(to, &temp, addrlen)) {
4136 err = -EFAULT;
4137 goto unlock;
4139 to += addrlen;
4140 cnt ++;
4141 space_left -= addrlen;
4144 copy_getaddrs:
4145 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4146 return -EFAULT;
4147 bytes_copied = ((char __user *)to) - optval;
4148 if (put_user(bytes_copied, optlen))
4149 return -EFAULT;
4151 unlock:
4152 sctp_read_unlock(addr_lock);
4153 return err;
4156 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4158 * Requests that the local SCTP stack use the enclosed peer address as
4159 * the association primary. The enclosed address must be one of the
4160 * association peer's addresses.
4162 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4163 char __user *optval, int __user *optlen)
4165 struct sctp_prim prim;
4166 struct sctp_association *asoc;
4167 struct sctp_sock *sp = sctp_sk(sk);
4169 if (len != sizeof(struct sctp_prim))
4170 return -EINVAL;
4172 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4173 return -EFAULT;
4175 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4176 if (!asoc)
4177 return -EINVAL;
4179 if (!asoc->peer.primary_path)
4180 return -ENOTCONN;
4182 flip_to_n((union sctp_addr *)&prim.ssp_addr,
4183 &asoc->peer.primary_path->ipaddr_h);
4185 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4186 (union sctp_addr *)&prim.ssp_addr);
4188 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4189 return -EFAULT;
4191 return 0;
4195 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
4197 * Requests that the local endpoint set the specified Adaption Layer
4198 * Indication parameter for all future INIT and INIT-ACK exchanges.
4200 static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
4201 char __user *optval, int __user *optlen)
4203 struct sctp_setadaption adaption;
4205 if (len != sizeof(struct sctp_setadaption))
4206 return -EINVAL;
4208 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
4209 if (copy_to_user(optval, &adaption, len))
4210 return -EFAULT;
4212 return 0;
4217 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4219 * Applications that wish to use the sendto() system call may wish to
4220 * specify a default set of parameters that would normally be supplied
4221 * through the inclusion of ancillary data. This socket option allows
4222 * such an application to set the default sctp_sndrcvinfo structure.
4225 * The application that wishes to use this socket option simply passes
4226 * in to this call the sctp_sndrcvinfo structure defined in Section
4227 * 5.2.2) The input parameters accepted by this call include
4228 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4229 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4230 * to this call if the caller is using the UDP model.
4232 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4234 static int sctp_getsockopt_default_send_param(struct sock *sk,
4235 int len, char __user *optval,
4236 int __user *optlen)
4238 struct sctp_sndrcvinfo info;
4239 struct sctp_association *asoc;
4240 struct sctp_sock *sp = sctp_sk(sk);
4242 if (len != sizeof(struct sctp_sndrcvinfo))
4243 return -EINVAL;
4244 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4245 return -EFAULT;
4247 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4248 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4249 return -EINVAL;
4251 if (asoc) {
4252 info.sinfo_stream = asoc->default_stream;
4253 info.sinfo_flags = asoc->default_flags;
4254 info.sinfo_ppid = asoc->default_ppid;
4255 info.sinfo_context = asoc->default_context;
4256 info.sinfo_timetolive = asoc->default_timetolive;
4257 } else {
4258 info.sinfo_stream = sp->default_stream;
4259 info.sinfo_flags = sp->default_flags;
4260 info.sinfo_ppid = sp->default_ppid;
4261 info.sinfo_context = sp->default_context;
4262 info.sinfo_timetolive = sp->default_timetolive;
4265 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4266 return -EFAULT;
4268 return 0;
4273 * 7.1.5 SCTP_NODELAY
4275 * Turn on/off any Nagle-like algorithm. This means that packets are
4276 * generally sent as soon as possible and no unnecessary delays are
4277 * introduced, at the cost of more packets in the network. Expects an
4278 * integer boolean flag.
4281 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4282 char __user *optval, int __user *optlen)
4284 int val;
4286 if (len < sizeof(int))
4287 return -EINVAL;
4289 len = sizeof(int);
4290 val = (sctp_sk(sk)->nodelay == 1);
4291 if (put_user(len, optlen))
4292 return -EFAULT;
4293 if (copy_to_user(optval, &val, len))
4294 return -EFAULT;
4295 return 0;
4300 * 7.1.1 SCTP_RTOINFO
4302 * The protocol parameters used to initialize and bound retransmission
4303 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4304 * and modify these parameters.
4305 * All parameters are time values, in milliseconds. A value of 0, when
4306 * modifying the parameters, indicates that the current value should not
4307 * be changed.
4310 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4311 char __user *optval,
4312 int __user *optlen) {
4313 struct sctp_rtoinfo rtoinfo;
4314 struct sctp_association *asoc;
4316 if (len != sizeof (struct sctp_rtoinfo))
4317 return -EINVAL;
4319 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4320 return -EFAULT;
4322 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4324 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4325 return -EINVAL;
4327 /* Values corresponding to the specific association. */
4328 if (asoc) {
4329 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4330 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4331 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4332 } else {
4333 /* Values corresponding to the endpoint. */
4334 struct sctp_sock *sp = sctp_sk(sk);
4336 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4337 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4338 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4341 if (put_user(len, optlen))
4342 return -EFAULT;
4344 if (copy_to_user(optval, &rtoinfo, len))
4345 return -EFAULT;
4347 return 0;
4352 * 7.1.2 SCTP_ASSOCINFO
4354 * This option is used to tune the the maximum retransmission attempts
4355 * of the association.
4356 * Returns an error if the new association retransmission value is
4357 * greater than the sum of the retransmission value of the peer.
4358 * See [SCTP] for more information.
4361 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4362 char __user *optval,
4363 int __user *optlen)
4366 struct sctp_assocparams assocparams;
4367 struct sctp_association *asoc;
4368 struct list_head *pos;
4369 int cnt = 0;
4371 if (len != sizeof (struct sctp_assocparams))
4372 return -EINVAL;
4374 if (copy_from_user(&assocparams, optval,
4375 sizeof (struct sctp_assocparams)))
4376 return -EFAULT;
4378 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4380 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4381 return -EINVAL;
4383 /* Values correspoinding to the specific association */
4384 if (asoc) {
4385 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4386 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4387 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4388 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4389 * 1000) +
4390 (asoc->cookie_life.tv_usec
4391 / 1000);
4393 list_for_each(pos, &asoc->peer.transport_addr_list) {
4394 cnt ++;
4397 assocparams.sasoc_number_peer_destinations = cnt;
4398 } else {
4399 /* Values corresponding to the endpoint */
4400 struct sctp_sock *sp = sctp_sk(sk);
4402 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4403 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4404 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4405 assocparams.sasoc_cookie_life =
4406 sp->assocparams.sasoc_cookie_life;
4407 assocparams.sasoc_number_peer_destinations =
4408 sp->assocparams.
4409 sasoc_number_peer_destinations;
4412 if (put_user(len, optlen))
4413 return -EFAULT;
4415 if (copy_to_user(optval, &assocparams, len))
4416 return -EFAULT;
4418 return 0;
4422 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4424 * This socket option is a boolean flag which turns on or off mapped V4
4425 * addresses. If this option is turned on and the socket is type
4426 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4427 * If this option is turned off, then no mapping will be done of V4
4428 * addresses and a user will receive both PF_INET6 and PF_INET type
4429 * addresses on the socket.
4431 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4432 char __user *optval, int __user *optlen)
4434 int val;
4435 struct sctp_sock *sp = sctp_sk(sk);
4437 if (len < sizeof(int))
4438 return -EINVAL;
4440 len = sizeof(int);
4441 val = sp->v4mapped;
4442 if (put_user(len, optlen))
4443 return -EFAULT;
4444 if (copy_to_user(optval, &val, len))
4445 return -EFAULT;
4447 return 0;
4451 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4453 * This socket option specifies the maximum size to put in any outgoing
4454 * SCTP chunk. If a message is larger than this size it will be
4455 * fragmented by SCTP into the specified size. Note that the underlying
4456 * SCTP implementation may fragment into smaller sized chunks when the
4457 * PMTU of the underlying association is smaller than the value set by
4458 * the user.
4460 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4461 char __user *optval, int __user *optlen)
4463 int val;
4465 if (len < sizeof(int))
4466 return -EINVAL;
4468 len = sizeof(int);
4470 val = sctp_sk(sk)->user_frag;
4471 if (put_user(len, optlen))
4472 return -EFAULT;
4473 if (copy_to_user(optval, &val, len))
4474 return -EFAULT;
4476 return 0;
4479 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4480 char __user *optval, int __user *optlen)
4482 int retval = 0;
4483 int len;
4485 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4486 sk, optname);
4488 /* I can hardly begin to describe how wrong this is. This is
4489 * so broken as to be worse than useless. The API draft
4490 * REALLY is NOT helpful here... I am not convinced that the
4491 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4492 * are at all well-founded.
4494 if (level != SOL_SCTP) {
4495 struct sctp_af *af = sctp_sk(sk)->pf->af;
4497 retval = af->getsockopt(sk, level, optname, optval, optlen);
4498 return retval;
4501 if (get_user(len, optlen))
4502 return -EFAULT;
4504 sctp_lock_sock(sk);
4506 switch (optname) {
4507 case SCTP_STATUS:
4508 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4509 break;
4510 case SCTP_DISABLE_FRAGMENTS:
4511 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4512 optlen);
4513 break;
4514 case SCTP_EVENTS:
4515 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4516 break;
4517 case SCTP_AUTOCLOSE:
4518 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4519 break;
4520 case SCTP_SOCKOPT_PEELOFF:
4521 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4522 break;
4523 case SCTP_PEER_ADDR_PARAMS:
4524 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4525 optlen);
4526 break;
4527 case SCTP_DELAYED_ACK_TIME:
4528 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4529 optlen);
4530 break;
4531 case SCTP_INITMSG:
4532 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4533 break;
4534 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4535 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4536 optlen);
4537 break;
4538 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4539 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4540 optlen);
4541 break;
4542 case SCTP_GET_PEER_ADDRS_OLD:
4543 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4544 optlen);
4545 break;
4546 case SCTP_GET_LOCAL_ADDRS_OLD:
4547 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4548 optlen);
4549 break;
4550 case SCTP_GET_PEER_ADDRS:
4551 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4552 optlen);
4553 break;
4554 case SCTP_GET_LOCAL_ADDRS:
4555 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4556 optlen);
4557 break;
4558 case SCTP_DEFAULT_SEND_PARAM:
4559 retval = sctp_getsockopt_default_send_param(sk, len,
4560 optval, optlen);
4561 break;
4562 case SCTP_PRIMARY_ADDR:
4563 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4564 break;
4565 case SCTP_NODELAY:
4566 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4567 break;
4568 case SCTP_RTOINFO:
4569 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4570 break;
4571 case SCTP_ASSOCINFO:
4572 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4573 break;
4574 case SCTP_I_WANT_MAPPED_V4_ADDR:
4575 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4576 break;
4577 case SCTP_MAXSEG:
4578 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4579 break;
4580 case SCTP_GET_PEER_ADDR_INFO:
4581 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4582 optlen);
4583 break;
4584 case SCTP_ADAPTION_LAYER:
4585 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4586 optlen);
4587 break;
4588 default:
4589 retval = -ENOPROTOOPT;
4590 break;
4593 sctp_release_sock(sk);
4594 return retval;
4597 static void sctp_hash(struct sock *sk)
4599 /* STUB */
4602 static void sctp_unhash(struct sock *sk)
4604 /* STUB */
4607 /* Check if port is acceptable. Possibly find first available port.
4609 * The port hash table (contained in the 'global' SCTP protocol storage
4610 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4611 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4612 * list (the list number is the port number hashed out, so as you
4613 * would expect from a hash function, all the ports in a given list have
4614 * such a number that hashes out to the same list number; you were
4615 * expecting that, right?); so each list has a set of ports, with a
4616 * link to the socket (struct sock) that uses it, the port number and
4617 * a fastreuse flag (FIXME: NPI ipg).
4619 static struct sctp_bind_bucket *sctp_bucket_create(
4620 struct sctp_bind_hashbucket *head, unsigned short snum);
4622 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4624 struct sctp_bind_hashbucket *head; /* hash list */
4625 struct sctp_bind_bucket *pp; /* hash list port iterator */
4626 unsigned short snum;
4627 int ret;
4629 snum = ntohs(addr->v4.sin_port);
4631 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4632 sctp_local_bh_disable();
4634 if (snum == 0) {
4635 /* Search for an available port.
4637 * 'sctp_port_rover' was the last port assigned, so
4638 * we start to search from 'sctp_port_rover +
4639 * 1'. What we do is first check if port 'rover' is
4640 * already in the hash table; if not, we use that; if
4641 * it is, we try next.
4643 int low = sysctl_local_port_range[0];
4644 int high = sysctl_local_port_range[1];
4645 int remaining = (high - low) + 1;
4646 int rover;
4647 int index;
4649 sctp_spin_lock(&sctp_port_alloc_lock);
4650 rover = sctp_port_rover;
4651 do {
4652 rover++;
4653 if ((rover < low) || (rover > high))
4654 rover = low;
4655 index = sctp_phashfn(rover);
4656 head = &sctp_port_hashtable[index];
4657 sctp_spin_lock(&head->lock);
4658 for (pp = head->chain; pp; pp = pp->next)
4659 if (pp->port == rover)
4660 goto next;
4661 break;
4662 next:
4663 sctp_spin_unlock(&head->lock);
4664 } while (--remaining > 0);
4665 sctp_port_rover = rover;
4666 sctp_spin_unlock(&sctp_port_alloc_lock);
4668 /* Exhausted local port range during search? */
4669 ret = 1;
4670 if (remaining <= 0)
4671 goto fail;
4673 /* OK, here is the one we will use. HEAD (the port
4674 * hash table list entry) is non-NULL and we hold it's
4675 * mutex.
4677 snum = rover;
4678 } else {
4679 /* We are given an specific port number; we verify
4680 * that it is not being used. If it is used, we will
4681 * exahust the search in the hash list corresponding
4682 * to the port number (snum) - we detect that with the
4683 * port iterator, pp being NULL.
4685 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4686 sctp_spin_lock(&head->lock);
4687 for (pp = head->chain; pp; pp = pp->next) {
4688 if (pp->port == snum)
4689 goto pp_found;
4692 pp = NULL;
4693 goto pp_not_found;
4694 pp_found:
4695 if (!hlist_empty(&pp->owner)) {
4696 /* We had a port hash table hit - there is an
4697 * available port (pp != NULL) and it is being
4698 * used by other socket (pp->owner not empty); that other
4699 * socket is going to be sk2.
4701 int reuse = sk->sk_reuse;
4702 struct sock *sk2;
4703 struct hlist_node *node;
4705 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4706 if (pp->fastreuse && sk->sk_reuse)
4707 goto success;
4709 /* Run through the list of sockets bound to the port
4710 * (pp->port) [via the pointers bind_next and
4711 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4712 * we get the endpoint they describe and run through
4713 * the endpoint's list of IP (v4 or v6) addresses,
4714 * comparing each of the addresses with the address of
4715 * the socket sk. If we find a match, then that means
4716 * that this port/socket (sk) combination are already
4717 * in an endpoint.
4719 sk_for_each_bound(sk2, node, &pp->owner) {
4720 struct sctp_endpoint *ep2;
4721 ep2 = sctp_sk(sk2)->ep;
4723 if (reuse && sk2->sk_reuse)
4724 continue;
4726 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4727 sctp_sk(sk))) {
4728 ret = (long)sk2;
4729 goto fail_unlock;
4732 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4734 pp_not_found:
4735 /* If there was a hash table miss, create a new port. */
4736 ret = 1;
4737 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4738 goto fail_unlock;
4740 /* In either case (hit or miss), make sure fastreuse is 1 only
4741 * if sk->sk_reuse is too (that is, if the caller requested
4742 * SO_REUSEADDR on this socket -sk-).
4744 if (hlist_empty(&pp->owner))
4745 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4746 else if (pp->fastreuse && !sk->sk_reuse)
4747 pp->fastreuse = 0;
4749 /* We are set, so fill up all the data in the hash table
4750 * entry, tie the socket list information with the rest of the
4751 * sockets FIXME: Blurry, NPI (ipg).
4753 success:
4754 inet_sk(sk)->num = snum;
4755 if (!sctp_sk(sk)->bind_hash) {
4756 sk_add_bind_node(sk, &pp->owner);
4757 sctp_sk(sk)->bind_hash = pp;
4759 ret = 0;
4761 fail_unlock:
4762 sctp_spin_unlock(&head->lock);
4764 fail:
4765 sctp_local_bh_enable();
4766 return ret;
4769 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4770 * port is requested.
4772 static int sctp_get_port(struct sock *sk, unsigned short snum)
4774 long ret;
4775 union sctp_addr addr;
4776 struct sctp_af *af = sctp_sk(sk)->pf->af;
4778 /* Set up a dummy address struct from the sk. */
4779 af->from_sk(&addr, sk);
4780 addr.v4.sin_port = htons(snum);
4782 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4783 ret = sctp_get_port_local(sk, &addr);
4785 return (ret ? 1 : 0);
4789 * 3.1.3 listen() - UDP Style Syntax
4791 * By default, new associations are not accepted for UDP style sockets.
4792 * An application uses listen() to mark a socket as being able to
4793 * accept new associations.
4795 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4797 struct sctp_sock *sp = sctp_sk(sk);
4798 struct sctp_endpoint *ep = sp->ep;
4800 /* Only UDP style sockets that are not peeled off are allowed to
4801 * listen().
4803 if (!sctp_style(sk, UDP))
4804 return -EINVAL;
4806 /* If backlog is zero, disable listening. */
4807 if (!backlog) {
4808 if (sctp_sstate(sk, CLOSED))
4809 return 0;
4811 sctp_unhash_endpoint(ep);
4812 sk->sk_state = SCTP_SS_CLOSED;
4815 /* Return if we are already listening. */
4816 if (sctp_sstate(sk, LISTENING))
4817 return 0;
4820 * If a bind() or sctp_bindx() is not called prior to a listen()
4821 * call that allows new associations to be accepted, the system
4822 * picks an ephemeral port and will choose an address set equivalent
4823 * to binding with a wildcard address.
4825 * This is not currently spelled out in the SCTP sockets
4826 * extensions draft, but follows the practice as seen in TCP
4827 * sockets.
4829 if (!ep->base.bind_addr.port) {
4830 if (sctp_autobind(sk))
4831 return -EAGAIN;
4833 sk->sk_state = SCTP_SS_LISTENING;
4834 sctp_hash_endpoint(ep);
4835 return 0;
4839 * 4.1.3 listen() - TCP Style Syntax
4841 * Applications uses listen() to ready the SCTP endpoint for accepting
4842 * inbound associations.
4844 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4846 struct sctp_sock *sp = sctp_sk(sk);
4847 struct sctp_endpoint *ep = sp->ep;
4849 /* If backlog is zero, disable listening. */
4850 if (!backlog) {
4851 if (sctp_sstate(sk, CLOSED))
4852 return 0;
4854 sctp_unhash_endpoint(ep);
4855 sk->sk_state = SCTP_SS_CLOSED;
4858 if (sctp_sstate(sk, LISTENING))
4859 return 0;
4862 * If a bind() or sctp_bindx() is not called prior to a listen()
4863 * call that allows new associations to be accepted, the system
4864 * picks an ephemeral port and will choose an address set equivalent
4865 * to binding with a wildcard address.
4867 * This is not currently spelled out in the SCTP sockets
4868 * extensions draft, but follows the practice as seen in TCP
4869 * sockets.
4871 if (!ep->base.bind_addr.port) {
4872 if (sctp_autobind(sk))
4873 return -EAGAIN;
4875 sk->sk_state = SCTP_SS_LISTENING;
4876 sk->sk_max_ack_backlog = backlog;
4877 sctp_hash_endpoint(ep);
4878 return 0;
4882 * Move a socket to LISTENING state.
4884 int sctp_inet_listen(struct socket *sock, int backlog)
4886 struct sock *sk = sock->sk;
4887 struct crypto_hash *tfm = NULL;
4888 int err = -EINVAL;
4890 if (unlikely(backlog < 0))
4891 goto out;
4893 sctp_lock_sock(sk);
4895 if (sock->state != SS_UNCONNECTED)
4896 goto out;
4898 /* Allocate HMAC for generating cookie. */
4899 if (sctp_hmac_alg) {
4900 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
4901 if (!tfm) {
4902 err = -ENOSYS;
4903 goto out;
4907 switch (sock->type) {
4908 case SOCK_SEQPACKET:
4909 err = sctp_seqpacket_listen(sk, backlog);
4910 break;
4911 case SOCK_STREAM:
4912 err = sctp_stream_listen(sk, backlog);
4913 break;
4914 default:
4915 break;
4917 if (err)
4918 goto cleanup;
4920 /* Store away the transform reference. */
4921 sctp_sk(sk)->hmac = tfm;
4922 out:
4923 sctp_release_sock(sk);
4924 return err;
4925 cleanup:
4926 crypto_free_hash(tfm);
4927 goto out;
4931 * This function is done by modeling the current datagram_poll() and the
4932 * tcp_poll(). Note that, based on these implementations, we don't
4933 * lock the socket in this function, even though it seems that,
4934 * ideally, locking or some other mechanisms can be used to ensure
4935 * the integrity of the counters (sndbuf and wmem_alloc) used
4936 * in this place. We assume that we don't need locks either until proven
4937 * otherwise.
4939 * Another thing to note is that we include the Async I/O support
4940 * here, again, by modeling the current TCP/UDP code. We don't have
4941 * a good way to test with it yet.
4943 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4945 struct sock *sk = sock->sk;
4946 struct sctp_sock *sp = sctp_sk(sk);
4947 unsigned int mask;
4949 poll_wait(file, sk->sk_sleep, wait);
4951 /* A TCP-style listening socket becomes readable when the accept queue
4952 * is not empty.
4954 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4955 return (!list_empty(&sp->ep->asocs)) ?
4956 (POLLIN | POLLRDNORM) : 0;
4958 mask = 0;
4960 /* Is there any exceptional events? */
4961 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4962 mask |= POLLERR;
4963 if (sk->sk_shutdown & RCV_SHUTDOWN)
4964 mask |= POLLRDHUP;
4965 if (sk->sk_shutdown == SHUTDOWN_MASK)
4966 mask |= POLLHUP;
4968 /* Is it readable? Reconsider this code with TCP-style support. */
4969 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4970 (sk->sk_shutdown & RCV_SHUTDOWN))
4971 mask |= POLLIN | POLLRDNORM;
4973 /* The association is either gone or not ready. */
4974 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4975 return mask;
4977 /* Is it writable? */
4978 if (sctp_writeable(sk)) {
4979 mask |= POLLOUT | POLLWRNORM;
4980 } else {
4981 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4983 * Since the socket is not locked, the buffer
4984 * might be made available after the writeable check and
4985 * before the bit is set. This could cause a lost I/O
4986 * signal. tcp_poll() has a race breaker for this race
4987 * condition. Based on their implementation, we put
4988 * in the following code to cover it as well.
4990 if (sctp_writeable(sk))
4991 mask |= POLLOUT | POLLWRNORM;
4993 return mask;
4996 /********************************************************************
4997 * 2nd Level Abstractions
4998 ********************************************************************/
5000 static struct sctp_bind_bucket *sctp_bucket_create(
5001 struct sctp_bind_hashbucket *head, unsigned short snum)
5003 struct sctp_bind_bucket *pp;
5005 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
5006 SCTP_DBG_OBJCNT_INC(bind_bucket);
5007 if (pp) {
5008 pp->port = snum;
5009 pp->fastreuse = 0;
5010 INIT_HLIST_HEAD(&pp->owner);
5011 if ((pp->next = head->chain) != NULL)
5012 pp->next->pprev = &pp->next;
5013 head->chain = pp;
5014 pp->pprev = &head->chain;
5016 return pp;
5019 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5020 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5022 if (pp && hlist_empty(&pp->owner)) {
5023 if (pp->next)
5024 pp->next->pprev = pp->pprev;
5025 *(pp->pprev) = pp->next;
5026 kmem_cache_free(sctp_bucket_cachep, pp);
5027 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5031 /* Release this socket's reference to a local port. */
5032 static inline void __sctp_put_port(struct sock *sk)
5034 struct sctp_bind_hashbucket *head =
5035 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5036 struct sctp_bind_bucket *pp;
5038 sctp_spin_lock(&head->lock);
5039 pp = sctp_sk(sk)->bind_hash;
5040 __sk_del_bind_node(sk);
5041 sctp_sk(sk)->bind_hash = NULL;
5042 inet_sk(sk)->num = 0;
5043 sctp_bucket_destroy(pp);
5044 sctp_spin_unlock(&head->lock);
5047 void sctp_put_port(struct sock *sk)
5049 sctp_local_bh_disable();
5050 __sctp_put_port(sk);
5051 sctp_local_bh_enable();
5055 * The system picks an ephemeral port and choose an address set equivalent
5056 * to binding with a wildcard address.
5057 * One of those addresses will be the primary address for the association.
5058 * This automatically enables the multihoming capability of SCTP.
5060 static int sctp_autobind(struct sock *sk)
5062 union sctp_addr autoaddr;
5063 struct sctp_af *af;
5064 unsigned short port;
5066 /* Initialize a local sockaddr structure to INADDR_ANY. */
5067 af = sctp_sk(sk)->pf->af;
5069 port = htons(inet_sk(sk)->num);
5070 af->inaddr_any(&autoaddr, port);
5072 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5075 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5077 * From RFC 2292
5078 * 4.2 The cmsghdr Structure *
5080 * When ancillary data is sent or received, any number of ancillary data
5081 * objects can be specified by the msg_control and msg_controllen members of
5082 * the msghdr structure, because each object is preceded by
5083 * a cmsghdr structure defining the object's length (the cmsg_len member).
5084 * Historically Berkeley-derived implementations have passed only one object
5085 * at a time, but this API allows multiple objects to be
5086 * passed in a single call to sendmsg() or recvmsg(). The following example
5087 * shows two ancillary data objects in a control buffer.
5089 * |<--------------------------- msg_controllen -------------------------->|
5090 * | |
5092 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5094 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5095 * | | |
5097 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5099 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5100 * | | | | |
5102 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5103 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5105 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5107 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5111 * msg_control
5112 * points here
5114 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5115 sctp_cmsgs_t *cmsgs)
5117 struct cmsghdr *cmsg;
5119 for (cmsg = CMSG_FIRSTHDR(msg);
5120 cmsg != NULL;
5121 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5122 if (!CMSG_OK(msg, cmsg))
5123 return -EINVAL;
5125 /* Should we parse this header or ignore? */
5126 if (cmsg->cmsg_level != IPPROTO_SCTP)
5127 continue;
5129 /* Strictly check lengths following example in SCM code. */
5130 switch (cmsg->cmsg_type) {
5131 case SCTP_INIT:
5132 /* SCTP Socket API Extension
5133 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5135 * This cmsghdr structure provides information for
5136 * initializing new SCTP associations with sendmsg().
5137 * The SCTP_INITMSG socket option uses this same data
5138 * structure. This structure is not used for
5139 * recvmsg().
5141 * cmsg_level cmsg_type cmsg_data[]
5142 * ------------ ------------ ----------------------
5143 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5145 if (cmsg->cmsg_len !=
5146 CMSG_LEN(sizeof(struct sctp_initmsg)))
5147 return -EINVAL;
5148 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5149 break;
5151 case SCTP_SNDRCV:
5152 /* SCTP Socket API Extension
5153 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5155 * This cmsghdr structure specifies SCTP options for
5156 * sendmsg() and describes SCTP header information
5157 * about a received message through recvmsg().
5159 * cmsg_level cmsg_type cmsg_data[]
5160 * ------------ ------------ ----------------------
5161 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5163 if (cmsg->cmsg_len !=
5164 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5165 return -EINVAL;
5167 cmsgs->info =
5168 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5170 /* Minimally, validate the sinfo_flags. */
5171 if (cmsgs->info->sinfo_flags &
5172 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5173 SCTP_ABORT | SCTP_EOF))
5174 return -EINVAL;
5175 break;
5177 default:
5178 return -EINVAL;
5181 return 0;
5185 * Wait for a packet..
5186 * Note: This function is the same function as in core/datagram.c
5187 * with a few modifications to make lksctp work.
5189 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5191 int error;
5192 DEFINE_WAIT(wait);
5194 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5196 /* Socket errors? */
5197 error = sock_error(sk);
5198 if (error)
5199 goto out;
5201 if (!skb_queue_empty(&sk->sk_receive_queue))
5202 goto ready;
5204 /* Socket shut down? */
5205 if (sk->sk_shutdown & RCV_SHUTDOWN)
5206 goto out;
5208 /* Sequenced packets can come disconnected. If so we report the
5209 * problem.
5211 error = -ENOTCONN;
5213 /* Is there a good reason to think that we may receive some data? */
5214 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5215 goto out;
5217 /* Handle signals. */
5218 if (signal_pending(current))
5219 goto interrupted;
5221 /* Let another process have a go. Since we are going to sleep
5222 * anyway. Note: This may cause odd behaviors if the message
5223 * does not fit in the user's buffer, but this seems to be the
5224 * only way to honor MSG_DONTWAIT realistically.
5226 sctp_release_sock(sk);
5227 *timeo_p = schedule_timeout(*timeo_p);
5228 sctp_lock_sock(sk);
5230 ready:
5231 finish_wait(sk->sk_sleep, &wait);
5232 return 0;
5234 interrupted:
5235 error = sock_intr_errno(*timeo_p);
5237 out:
5238 finish_wait(sk->sk_sleep, &wait);
5239 *err = error;
5240 return error;
5243 /* Receive a datagram.
5244 * Note: This is pretty much the same routine as in core/datagram.c
5245 * with a few changes to make lksctp work.
5247 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5248 int noblock, int *err)
5250 int error;
5251 struct sk_buff *skb;
5252 long timeo;
5254 timeo = sock_rcvtimeo(sk, noblock);
5256 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5257 timeo, MAX_SCHEDULE_TIMEOUT);
5259 do {
5260 /* Again only user level code calls this function,
5261 * so nothing interrupt level
5262 * will suddenly eat the receive_queue.
5264 * Look at current nfs client by the way...
5265 * However, this function was corrent in any case. 8)
5267 if (flags & MSG_PEEK) {
5268 spin_lock_bh(&sk->sk_receive_queue.lock);
5269 skb = skb_peek(&sk->sk_receive_queue);
5270 if (skb)
5271 atomic_inc(&skb->users);
5272 spin_unlock_bh(&sk->sk_receive_queue.lock);
5273 } else {
5274 skb = skb_dequeue(&sk->sk_receive_queue);
5277 if (skb)
5278 return skb;
5280 /* Caller is allowed not to check sk->sk_err before calling. */
5281 error = sock_error(sk);
5282 if (error)
5283 goto no_packet;
5285 if (sk->sk_shutdown & RCV_SHUTDOWN)
5286 break;
5288 /* User doesn't want to wait. */
5289 error = -EAGAIN;
5290 if (!timeo)
5291 goto no_packet;
5292 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5294 return NULL;
5296 no_packet:
5297 *err = error;
5298 return NULL;
5301 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5302 static void __sctp_write_space(struct sctp_association *asoc)
5304 struct sock *sk = asoc->base.sk;
5305 struct socket *sock = sk->sk_socket;
5307 if ((sctp_wspace(asoc) > 0) && sock) {
5308 if (waitqueue_active(&asoc->wait))
5309 wake_up_interruptible(&asoc->wait);
5311 if (sctp_writeable(sk)) {
5312 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5313 wake_up_interruptible(sk->sk_sleep);
5315 /* Note that we try to include the Async I/O support
5316 * here by modeling from the current TCP/UDP code.
5317 * We have not tested with it yet.
5319 if (sock->fasync_list &&
5320 !(sk->sk_shutdown & SEND_SHUTDOWN))
5321 sock_wake_async(sock, 2, POLL_OUT);
5326 /* Do accounting for the sndbuf space.
5327 * Decrement the used sndbuf space of the corresponding association by the
5328 * data size which was just transmitted(freed).
5330 static void sctp_wfree(struct sk_buff *skb)
5332 struct sctp_association *asoc;
5333 struct sctp_chunk *chunk;
5334 struct sock *sk;
5336 /* Get the saved chunk pointer. */
5337 chunk = *((struct sctp_chunk **)(skb->cb));
5338 asoc = chunk->asoc;
5339 sk = asoc->base.sk;
5340 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5341 sizeof(struct sk_buff) +
5342 sizeof(struct sctp_chunk);
5344 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5346 sock_wfree(skb);
5347 __sctp_write_space(asoc);
5349 sctp_association_put(asoc);
5352 /* Do accounting for the receive space on the socket.
5353 * Accounting for the association is done in ulpevent.c
5354 * We set this as a destructor for the cloned data skbs so that
5355 * accounting is done at the correct time.
5357 void sctp_sock_rfree(struct sk_buff *skb)
5359 struct sock *sk = skb->sk;
5360 struct sctp_ulpevent *event = sctp_skb2event(skb);
5362 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
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.
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
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.
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.
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.
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.
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.
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
5622 if (PF_INET6 == assoc->base.sk->sk_family)
5623 flags = SCTP_ADDR6_ALLOWED;
5624 if (assoc->peer.ipv4_address)
5625 flags |= SCTP_ADDR4_PEERSUPP;
5626 if (assoc->peer.ipv6_address)
5627 flags |= SCTP_ADDR6_PEERSUPP;
5628 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5629 &oldsp->ep->base.bind_addr,
5630 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5632 /* Move any messages in the old socket's receive queue that are for the
5633 * peeled off association to the new socket's receive queue.
5635 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5636 event = sctp_skb2event(skb);
5637 if (event->asoc == assoc) {
5638 sctp_sock_rfree(skb);
5639 __skb_unlink(skb, &oldsk->sk_receive_queue);
5640 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5641 sctp_skb_set_owner_r(skb, newsk);
5645 /* Clean up any messages pending delivery due to partial
5646 * delivery. Three cases:
5647 * 1) No partial deliver; no work.
5648 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5649 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5651 skb_queue_head_init(&newsp->pd_lobby);
5652 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5654 if (sctp_sk(oldsk)->pd_mode) {
5655 struct sk_buff_head *queue;
5657 /* Decide which queue to move pd_lobby skbs to. */
5658 if (assoc->ulpq.pd_mode) {
5659 queue = &newsp->pd_lobby;
5660 } else
5661 queue = &newsk->sk_receive_queue;
5663 /* Walk through the pd_lobby, looking for skbs that
5664 * need moved to the new socket.
5666 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5667 event = sctp_skb2event(skb);
5668 if (event->asoc == assoc) {
5669 sctp_sock_rfree(skb);
5670 __skb_unlink(skb, &oldsp->pd_lobby);
5671 __skb_queue_tail(queue, skb);
5672 sctp_skb_set_owner_r(skb, newsk);
5676 /* Clear up any skbs waiting for the partial
5677 * delivery to finish.
5679 if (assoc->ulpq.pd_mode)
5680 sctp_clear_pd(oldsk);
5684 /* Set the type of socket to indicate that it is peeled off from the
5685 * original UDP-style socket or created with the accept() call on a
5686 * TCP-style socket..
5688 newsp->type = type;
5690 /* Mark the new socket "in-use" by the user so that any packets
5691 * that may arrive on the association after we've moved it are
5692 * queued to the backlog. This prevents a potential race between
5693 * backlog processing on the old socket and new-packet processing
5694 * on the new socket.
5696 sctp_lock_sock(newsk);
5697 sctp_assoc_migrate(assoc, newsk);
5699 /* If the association on the newsk is already closed before accept()
5700 * is called, set RCV_SHUTDOWN flag.
5702 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5703 newsk->sk_shutdown |= RCV_SHUTDOWN;
5705 newsk->sk_state = SCTP_SS_ESTABLISHED;
5706 sctp_release_sock(newsk);
5709 /* This proto struct describes the ULP interface for SCTP. */
5710 struct proto sctp_prot = {
5711 .name = "SCTP",
5712 .owner = THIS_MODULE,
5713 .close = sctp_close,
5714 .connect = sctp_connect,
5715 .disconnect = sctp_disconnect,
5716 .accept = sctp_accept,
5717 .ioctl = sctp_ioctl,
5718 .init = sctp_init_sock,
5719 .destroy = sctp_destroy_sock,
5720 .shutdown = sctp_shutdown,
5721 .setsockopt = sctp_setsockopt,
5722 .getsockopt = sctp_getsockopt,
5723 .sendmsg = sctp_sendmsg,
5724 .recvmsg = sctp_recvmsg,
5725 .bind = sctp_bind,
5726 .backlog_rcv = sctp_backlog_rcv,
5727 .hash = sctp_hash,
5728 .unhash = sctp_unhash,
5729 .get_port = sctp_get_port,
5730 .obj_size = sizeof(struct sctp_sock),
5733 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5734 struct proto sctpv6_prot = {
5735 .name = "SCTPv6",
5736 .owner = THIS_MODULE,
5737 .close = sctp_close,
5738 .connect = sctp_connect,
5739 .disconnect = sctp_disconnect,
5740 .accept = sctp_accept,
5741 .ioctl = sctp_ioctl,
5742 .init = sctp_init_sock,
5743 .destroy = sctp_destroy_sock,
5744 .shutdown = sctp_shutdown,
5745 .setsockopt = sctp_setsockopt,
5746 .getsockopt = sctp_getsockopt,
5747 .sendmsg = sctp_sendmsg,
5748 .recvmsg = sctp_recvmsg,
5749 .bind = sctp_bind,
5750 .backlog_rcv = sctp_backlog_rcv,
5751 .hash = sctp_hash,
5752 .unhash = sctp_unhash,
5753 .get_port = sctp_get_port,
5754 .obj_size = sizeof(struct sctp6_sock),
5756 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */