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)
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
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>
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>
73 #include <net/route.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
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,
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 struct kmem_cache *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;
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;
122 /* do socket level accounting */
123 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
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
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,
168 /* Verify basic sockaddr. */
169 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
173 /* Is this a valid SCTP address? */
174 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
177 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
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))
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);
206 /* Otherwise this is a UDP-style socket. */
207 if (!id || (id == (sctp_assoc_t)-1))
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)
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
224 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
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,
239 id_asoc = sctp_id2assoc(sk, id);
240 if (id_asoc && (id_asoc != addr_asoc))
243 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
244 (union sctp_addr *)addr);
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)
265 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
268 /* Disallow binding twice. */
269 if (!sctp_sk(sk)->ep->base.bind_addr.port)
270 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
275 sctp_release_sock(sk);
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)
288 /* Check minimum size. */
289 if (len < sizeof (struct sockaddr))
292 /* Does this PF support this AF? */
293 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
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)
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;
315 /* Common sockaddr verification. */
316 af = sctp_sockaddr_af(sp, addr, len);
318 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
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",
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 * It's OK to allow 0 ports if we are already bound.
338 * We'll just inhert an already bound port in this case
343 else if (snum != bp->port) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind:"
345 " New port %d does not match existing port "
346 "%d.\n", snum, bp->port);
351 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
354 /* Make sure we are allowed to bind here.
355 * The function sctp_get_port_local() does duplicate address
358 if ((ret = sctp_get_port_local(sk, addr))) {
359 if (ret == (long) sk) {
360 /* This endpoint has a conflicting address. */
367 /* Refresh ephemeral port. */
369 bp->port = inet_sk(sk)->num;
371 /* Add the address to the bind address list. */
372 sctp_local_bh_disable();
373 sctp_write_lock(&ep->base.addr_lock);
375 /* Use GFP_ATOMIC since BHs are disabled. */
376 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
377 sctp_write_unlock(&ep->base.addr_lock);
378 sctp_local_bh_enable();
380 /* Copy back into socket for getsockname() use. */
382 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
383 af->to_sk_saddr(addr, sk);
389 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
391 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
392 * at any one time. If a sender, after sending an ASCONF chunk, decides
393 * it needs to transfer another ASCONF Chunk, it MUST wait until the
394 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
395 * subsequent ASCONF. Note this restriction binds each side, so at any
396 * time two ASCONF may be in-transit on any given association (one sent
397 * from each endpoint).
399 static int sctp_send_asconf(struct sctp_association *asoc,
400 struct sctp_chunk *chunk)
404 /* If there is an outstanding ASCONF chunk, queue it for later
407 if (asoc->addip_last_asconf) {
408 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
412 /* Hold the chunk until an ASCONF_ACK is received. */
413 sctp_chunk_hold(chunk);
414 retval = sctp_primitive_ASCONF(asoc, chunk);
416 sctp_chunk_free(chunk);
418 asoc->addip_last_asconf = chunk;
424 /* Add a list of addresses as bind addresses to local endpoint or
427 * Basically run through each address specified in the addrs/addrcnt
428 * array/length pair, determine if it is IPv6 or IPv4 and call
429 * sctp_do_bind() on it.
431 * If any of them fails, then the operation will be reversed and the
432 * ones that were added will be removed.
434 * Only sctp_setsockopt_bindx() is supposed to call this function.
436 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
441 struct sockaddr *sa_addr;
444 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
448 for (cnt = 0; cnt < addrcnt; cnt++) {
449 /* The list may contain either IPv4 or IPv6 address;
450 * determine the address length for walking thru the list.
452 sa_addr = (struct sockaddr *)addr_buf;
453 af = sctp_get_af_specific(sa_addr->sa_family);
459 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
462 addr_buf += af->sockaddr_len;
466 /* Failed. Cleanup the ones that have been added */
468 sctp_bindx_rem(sk, addrs, cnt);
476 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
477 * associations that are part of the endpoint indicating that a list of local
478 * addresses are added to the endpoint.
480 * If any of the addresses is already in the bind address list of the
481 * association, we do not send the chunk for that association. But it will not
482 * affect other associations.
484 * Only sctp_setsockopt_bindx() is supposed to call this function.
486 static int sctp_send_asconf_add_ip(struct sock *sk,
487 struct sockaddr *addrs,
490 struct sctp_sock *sp;
491 struct sctp_endpoint *ep;
492 struct sctp_association *asoc;
493 struct sctp_bind_addr *bp;
494 struct sctp_chunk *chunk;
495 struct sctp_sockaddr_entry *laddr;
496 union sctp_addr *addr;
497 union sctp_addr saveaddr;
500 struct list_head *pos;
505 if (!sctp_addip_enable)
511 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
512 __FUNCTION__, sk, addrs, addrcnt);
514 list_for_each(pos, &ep->asocs) {
515 asoc = list_entry(pos, struct sctp_association, asocs);
517 if (!asoc->peer.asconf_capable)
520 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
523 if (!sctp_state(asoc, ESTABLISHED))
526 /* Check if any address in the packed array of addresses is
527 * in the bind address list of the association. If so,
528 * do not send the asconf chunk to its peer, but continue with
529 * other associations.
532 for (i = 0; i < addrcnt; i++) {
533 addr = (union sctp_addr *)addr_buf;
534 af = sctp_get_af_specific(addr->v4.sin_family);
540 if (sctp_assoc_lookup_laddr(asoc, addr))
543 addr_buf += af->sockaddr_len;
548 /* Use the first address in bind addr list of association as
549 * Address Parameter of ASCONF CHUNK.
551 sctp_read_lock(&asoc->base.addr_lock);
552 bp = &asoc->base.bind_addr;
553 p = bp->address_list.next;
554 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
555 sctp_read_unlock(&asoc->base.addr_lock);
557 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
558 addrcnt, SCTP_PARAM_ADD_IP);
564 retval = sctp_send_asconf(asoc, chunk);
568 /* Add the new addresses to the bind address list with
569 * use_as_src set to 0.
571 sctp_local_bh_disable();
572 sctp_write_lock(&asoc->base.addr_lock);
574 for (i = 0; i < addrcnt; i++) {
575 addr = (union sctp_addr *)addr_buf;
576 af = sctp_get_af_specific(addr->v4.sin_family);
577 memcpy(&saveaddr, addr, af->sockaddr_len);
578 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
580 addr_buf += af->sockaddr_len;
582 sctp_write_unlock(&asoc->base.addr_lock);
583 sctp_local_bh_enable();
590 /* Remove a list of addresses from bind addresses list. Do not remove the
593 * Basically run through each address specified in the addrs/addrcnt
594 * array/length pair, determine if it is IPv6 or IPv4 and call
595 * sctp_del_bind() on it.
597 * If any of them fails, then the operation will be reversed and the
598 * ones that were removed will be added back.
600 * At least one address has to be left; if only one address is
601 * available, the operation will return -EBUSY.
603 * Only sctp_setsockopt_bindx() is supposed to call this function.
605 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
607 struct sctp_sock *sp = sctp_sk(sk);
608 struct sctp_endpoint *ep = sp->ep;
610 struct sctp_bind_addr *bp = &ep->base.bind_addr;
613 union sctp_addr *sa_addr;
616 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
620 for (cnt = 0; cnt < addrcnt; cnt++) {
621 /* If the bind address list is empty or if there is only one
622 * bind address, there is nothing more to be removed (we need
623 * at least one address here).
625 if (list_empty(&bp->address_list) ||
626 (sctp_list_single_entry(&bp->address_list))) {
631 sa_addr = (union sctp_addr *)addr_buf;
632 af = sctp_get_af_specific(sa_addr->sa.sa_family);
638 if (!af->addr_valid(sa_addr, sp, NULL)) {
639 retval = -EADDRNOTAVAIL;
643 if (sa_addr->v4.sin_port != htons(bp->port)) {
648 /* FIXME - There is probably a need to check if sk->sk_saddr and
649 * sk->sk_rcv_addr are currently set to one of the addresses to
650 * be removed. This is something which needs to be looked into
651 * when we are fixing the outstanding issues with multi-homing
652 * socket routing and failover schemes. Refer to comments in
653 * sctp_do_bind(). -daisy
655 sctp_local_bh_disable();
656 sctp_write_lock(&ep->base.addr_lock);
658 retval = sctp_del_bind_addr(bp, sa_addr);
660 sctp_write_unlock(&ep->base.addr_lock);
661 sctp_local_bh_enable();
663 addr_buf += af->sockaddr_len;
666 /* Failed. Add the ones that has been removed back */
668 sctp_bindx_add(sk, addrs, cnt);
676 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
677 * the associations that are part of the endpoint indicating that a list of
678 * local addresses are removed from the endpoint.
680 * If any of the addresses is already in the bind address list of the
681 * association, we do not send the chunk for that association. But it will not
682 * affect other associations.
684 * Only sctp_setsockopt_bindx() is supposed to call this function.
686 static int sctp_send_asconf_del_ip(struct sock *sk,
687 struct sockaddr *addrs,
690 struct sctp_sock *sp;
691 struct sctp_endpoint *ep;
692 struct sctp_association *asoc;
693 struct sctp_transport *transport;
694 struct sctp_bind_addr *bp;
695 struct sctp_chunk *chunk;
696 union sctp_addr *laddr;
699 struct list_head *pos, *pos1;
700 struct sctp_sockaddr_entry *saddr;
704 if (!sctp_addip_enable)
710 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 __FUNCTION__, sk, addrs, addrcnt);
713 list_for_each(pos, &ep->asocs) {
714 asoc = list_entry(pos, struct sctp_association, asocs);
716 if (!asoc->peer.asconf_capable)
719 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
722 if (!sctp_state(asoc, ESTABLISHED))
725 /* Check if any address in the packed array of addresses is
726 * not present in the bind address list of the association.
727 * If so, do not send the asconf chunk to its peer, but
728 * continue with other associations.
731 for (i = 0; i < addrcnt; i++) {
732 laddr = (union sctp_addr *)addr_buf;
733 af = sctp_get_af_specific(laddr->v4.sin_family);
739 if (!sctp_assoc_lookup_laddr(asoc, laddr))
742 addr_buf += af->sockaddr_len;
747 /* Find one address in the association's bind address list
748 * that is not in the packed array of addresses. This is to
749 * make sure that we do not delete all the addresses in the
752 sctp_read_lock(&asoc->base.addr_lock);
753 bp = &asoc->base.bind_addr;
754 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
756 sctp_read_unlock(&asoc->base.addr_lock);
760 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
767 /* Reset use_as_src flag for the addresses in the bind address
768 * list that are to be deleted.
770 sctp_local_bh_disable();
771 sctp_write_lock(&asoc->base.addr_lock);
773 for (i = 0; i < addrcnt; i++) {
774 laddr = (union sctp_addr *)addr_buf;
775 af = sctp_get_af_specific(laddr->v4.sin_family);
776 list_for_each(pos1, &bp->address_list) {
777 saddr = list_entry(pos1,
778 struct sctp_sockaddr_entry,
780 if (sctp_cmp_addr_exact(&saddr->a, laddr))
781 saddr->use_as_src = 0;
783 addr_buf += af->sockaddr_len;
785 sctp_write_unlock(&asoc->base.addr_lock);
786 sctp_local_bh_enable();
788 /* Update the route and saddr entries for all the transports
789 * as some of the addresses in the bind address list are
790 * about to be deleted and cannot be used as source addresses.
792 list_for_each(pos1, &asoc->peer.transport_addr_list) {
793 transport = list_entry(pos1, struct sctp_transport,
795 dst_release(transport->dst);
796 sctp_transport_route(transport, NULL,
797 sctp_sk(asoc->base.sk));
800 retval = sctp_send_asconf(asoc, chunk);
806 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
812 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
813 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
816 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
817 * Section 3.1.2 for this usage.
819 * addrs is a pointer to an array of one or more socket addresses. Each
820 * address is contained in its appropriate structure (i.e. struct
821 * sockaddr_in or struct sockaddr_in6) the family of the address type
822 * must be used to distinguish the address length (note that this
823 * representation is termed a "packed array" of addresses). The caller
824 * specifies the number of addresses in the array with addrcnt.
826 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
827 * -1, and sets errno to the appropriate error code.
829 * For SCTP, the port given in each socket address must be the same, or
830 * sctp_bindx() will fail, setting errno to EINVAL.
832 * The flags parameter is formed from the bitwise OR of zero or more of
833 * the following currently defined flags:
835 * SCTP_BINDX_ADD_ADDR
837 * SCTP_BINDX_REM_ADDR
839 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
840 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
841 * addresses from the association. The two flags are mutually exclusive;
842 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
843 * not remove all addresses from an association; sctp_bindx() will
844 * reject such an attempt with EINVAL.
846 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
847 * additional addresses with an endpoint after calling bind(). Or use
848 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
849 * socket is associated with so that no new association accepted will be
850 * associated with those addresses. If the endpoint supports dynamic
851 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
852 * endpoint to send the appropriate message to the peer to change the
853 * peers address lists.
855 * Adding and removing addresses from a connected association is
856 * optional functionality. Implementations that do not support this
857 * functionality should return EOPNOTSUPP.
859 * Basically do nothing but copying the addresses from user to kernel
860 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
861 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
864 * We don't use copy_from_user() for optimization: we first do the
865 * sanity checks (buffer size -fast- and access check-healthy
866 * pointer); if all of those succeed, then we can alloc the memory
867 * (expensive operation) needed to copy the data to kernel. Then we do
868 * the copying without checking the user space area
869 * (__copy_from_user()).
871 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
874 * sk The sk of the socket
875 * addrs The pointer to the addresses in user land
876 * addrssize Size of the addrs buffer
877 * op Operation to perform (add or remove, see the flags of
880 * Returns 0 if ok, <0 errno code on error.
882 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
883 struct sockaddr __user *addrs,
884 int addrs_size, int op)
886 struct sockaddr *kaddrs;
890 struct sockaddr *sa_addr;
894 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
895 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
897 if (unlikely(addrs_size <= 0))
900 /* Check the user passed a healthy pointer. */
901 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
904 /* Alloc space for the address array in kernel memory. */
905 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
906 if (unlikely(!kaddrs))
909 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
914 /* Walk through the addrs buffer and count the number of addresses. */
916 while (walk_size < addrs_size) {
917 sa_addr = (struct sockaddr *)addr_buf;
918 af = sctp_get_af_specific(sa_addr->sa_family);
920 /* If the address family is not supported or if this address
921 * causes the address buffer to overflow return EINVAL.
923 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
928 addr_buf += af->sockaddr_len;
929 walk_size += af->sockaddr_len;
934 case SCTP_BINDX_ADD_ADDR:
935 err = sctp_bindx_add(sk, kaddrs, addrcnt);
938 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
941 case SCTP_BINDX_REM_ADDR:
942 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
945 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
959 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
961 * Common routine for handling connect() and sctp_connectx().
962 * Connect will come in with just a single address.
964 static int __sctp_connect(struct sock* sk,
965 struct sockaddr *kaddrs,
968 struct sctp_sock *sp;
969 struct sctp_endpoint *ep;
970 struct sctp_association *asoc = NULL;
971 struct sctp_association *asoc2;
972 struct sctp_transport *transport;
980 union sctp_addr *sa_addr;
987 /* connect() cannot be done on a socket that is already in ESTABLISHED
988 * state - UDP-style peeled off socket or a TCP-style socket that
989 * is already connected.
990 * It cannot be done even on a TCP-style listening socket.
992 if (sctp_sstate(sk, ESTABLISHED) ||
993 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
998 /* Walk through the addrs buffer and count the number of addresses. */
1000 while (walk_size < addrs_size) {
1001 sa_addr = (union sctp_addr *)addr_buf;
1002 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1003 port = ntohs(sa_addr->v4.sin_port);
1005 /* If the address family is not supported or if this address
1006 * causes the address buffer to overflow return EINVAL.
1008 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1013 err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
1017 /* Make sure the destination port is correctly set
1020 if (asoc && asoc->peer.port && asoc->peer.port != port)
1023 memcpy(&to, sa_addr, af->sockaddr_len);
1025 /* Check if there already is a matching association on the
1026 * endpoint (other than the one created here).
1028 asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1029 if (asoc2 && asoc2 != asoc) {
1030 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1037 /* If we could not find a matching association on the endpoint,
1038 * make sure that there is no peeled-off association matching
1039 * the peer address even on another socket.
1041 if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1042 err = -EADDRNOTAVAIL;
1047 /* If a bind() or sctp_bindx() is not called prior to
1048 * an sctp_connectx() call, the system picks an
1049 * ephemeral port and will choose an address set
1050 * equivalent to binding with a wildcard address.
1052 if (!ep->base.bind_addr.port) {
1053 if (sctp_autobind(sk)) {
1059 * If an unprivileged user inherits a 1-many
1060 * style socket with open associations on a
1061 * privileged port, it MAY be permitted to
1062 * accept new associations, but it SHOULD NOT
1063 * be permitted to open new associations.
1065 if (ep->base.bind_addr.port < PROT_SOCK &&
1066 !capable(CAP_NET_BIND_SERVICE)) {
1072 scope = sctp_scope(sa_addr);
1073 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1080 /* Prime the peer's transport structures. */
1081 transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1089 addr_buf += af->sockaddr_len;
1090 walk_size += af->sockaddr_len;
1093 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1098 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1103 /* Initialize sk's dport and daddr for getpeername() */
1104 inet_sk(sk)->dport = htons(asoc->peer.port);
1105 af = sctp_get_af_specific(to.sa.sa_family);
1106 af->to_sk_daddr(&to, sk);
1109 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1110 err = sctp_wait_for_connect(asoc, &timeo);
1112 /* Don't free association on exit. */
1117 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1118 " kaddrs: %p err: %d\n",
1121 sctp_association_free(asoc);
1125 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1128 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1130 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1131 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1132 * or IPv6 addresses.
1134 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1135 * Section 3.1.2 for this usage.
1137 * addrs is a pointer to an array of one or more socket addresses. Each
1138 * address is contained in its appropriate structure (i.e. struct
1139 * sockaddr_in or struct sockaddr_in6) the family of the address type
1140 * must be used to distengish the address length (note that this
1141 * representation is termed a "packed array" of addresses). The caller
1142 * specifies the number of addresses in the array with addrcnt.
1144 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1145 * -1, and sets errno to the appropriate error code.
1147 * For SCTP, the port given in each socket address must be the same, or
1148 * sctp_connectx() will fail, setting errno to EINVAL.
1150 * An application can use sctp_connectx to initiate an association with
1151 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1152 * allows a caller to specify multiple addresses at which a peer can be
1153 * reached. The way the SCTP stack uses the list of addresses to set up
1154 * the association is implementation dependant. This function only
1155 * specifies that the stack will try to make use of all the addresses in
1156 * the list when needed.
1158 * Note that the list of addresses passed in is only used for setting up
1159 * the association. It does not necessarily equal the set of addresses
1160 * the peer uses for the resulting association. If the caller wants to
1161 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1162 * retrieve them after the association has been set up.
1164 * Basically do nothing but copying the addresses from user to kernel
1165 * land and invoking either sctp_connectx(). This is used for tunneling
1166 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1168 * We don't use copy_from_user() for optimization: we first do the
1169 * sanity checks (buffer size -fast- and access check-healthy
1170 * pointer); if all of those succeed, then we can alloc the memory
1171 * (expensive operation) needed to copy the data to kernel. Then we do
1172 * the copying without checking the user space area
1173 * (__copy_from_user()).
1175 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1178 * sk The sk of the socket
1179 * addrs The pointer to the addresses in user land
1180 * addrssize Size of the addrs buffer
1182 * Returns 0 if ok, <0 errno code on error.
1184 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1185 struct sockaddr __user *addrs,
1189 struct sockaddr *kaddrs;
1191 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1192 __FUNCTION__, sk, addrs, addrs_size);
1194 if (unlikely(addrs_size <= 0))
1197 /* Check the user passed a healthy pointer. */
1198 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1201 /* Alloc space for the address array in kernel memory. */
1202 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1203 if (unlikely(!kaddrs))
1206 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1209 err = __sctp_connect(sk, kaddrs, addrs_size);
1216 /* API 3.1.4 close() - UDP Style Syntax
1217 * Applications use close() to perform graceful shutdown (as described in
1218 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1219 * by a UDP-style socket.
1223 * ret = close(int sd);
1225 * sd - the socket descriptor of the associations to be closed.
1227 * To gracefully shutdown a specific association represented by the
1228 * UDP-style socket, an application should use the sendmsg() call,
1229 * passing no user data, but including the appropriate flag in the
1230 * ancillary data (see Section xxxx).
1232 * If sd in the close() call is a branched-off socket representing only
1233 * one association, the shutdown is performed on that association only.
1235 * 4.1.6 close() - TCP Style Syntax
1237 * Applications use close() to gracefully close down an association.
1241 * int close(int sd);
1243 * sd - the socket descriptor of the association to be closed.
1245 * After an application calls close() on a socket descriptor, no further
1246 * socket operations will succeed on that descriptor.
1248 * API 7.1.4 SO_LINGER
1250 * An application using the TCP-style socket can use this option to
1251 * perform the SCTP ABORT primitive. The linger option structure is:
1254 * int l_onoff; // option on/off
1255 * int l_linger; // linger time
1258 * To enable the option, set l_onoff to 1. If the l_linger value is set
1259 * to 0, calling close() is the same as the ABORT primitive. If the
1260 * value is set to a negative value, the setsockopt() call will return
1261 * an error. If the value is set to a positive value linger_time, the
1262 * close() can be blocked for at most linger_time ms. If the graceful
1263 * shutdown phase does not finish during this period, close() will
1264 * return but the graceful shutdown phase continues in the system.
1266 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1268 struct sctp_endpoint *ep;
1269 struct sctp_association *asoc;
1270 struct list_head *pos, *temp;
1272 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1275 sk->sk_shutdown = SHUTDOWN_MASK;
1277 ep = sctp_sk(sk)->ep;
1279 /* Walk all associations on an endpoint. */
1280 list_for_each_safe(pos, temp, &ep->asocs) {
1281 asoc = list_entry(pos, struct sctp_association, asocs);
1283 if (sctp_style(sk, TCP)) {
1284 /* A closed association can still be in the list if
1285 * it belongs to a TCP-style listening socket that is
1286 * not yet accepted. If so, free it. If not, send an
1287 * ABORT or SHUTDOWN based on the linger options.
1289 if (sctp_state(asoc, CLOSED)) {
1290 sctp_unhash_established(asoc);
1291 sctp_association_free(asoc);
1296 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1297 struct sctp_chunk *chunk;
1299 chunk = sctp_make_abort_user(asoc, NULL, 0);
1301 sctp_primitive_ABORT(asoc, chunk);
1303 sctp_primitive_SHUTDOWN(asoc, NULL);
1306 /* Clean up any skbs sitting on the receive queue. */
1307 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1308 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1310 /* On a TCP-style socket, block for at most linger_time if set. */
1311 if (sctp_style(sk, TCP) && timeout)
1312 sctp_wait_for_close(sk, timeout);
1314 /* This will run the backlog queue. */
1315 sctp_release_sock(sk);
1317 /* Supposedly, no process has access to the socket, but
1318 * the net layers still may.
1320 sctp_local_bh_disable();
1321 sctp_bh_lock_sock(sk);
1323 /* Hold the sock, since sk_common_release() will put sock_put()
1324 * and we have just a little more cleanup.
1327 sk_common_release(sk);
1329 sctp_bh_unlock_sock(sk);
1330 sctp_local_bh_enable();
1334 SCTP_DBG_OBJCNT_DEC(sock);
1337 /* Handle EPIPE error. */
1338 static int sctp_error(struct sock *sk, int flags, int err)
1341 err = sock_error(sk) ? : -EPIPE;
1342 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1343 send_sig(SIGPIPE, current, 0);
1347 /* API 3.1.3 sendmsg() - UDP Style Syntax
1349 * An application uses sendmsg() and recvmsg() calls to transmit data to
1350 * and receive data from its peer.
1352 * ssize_t sendmsg(int socket, const struct msghdr *message,
1355 * socket - the socket descriptor of the endpoint.
1356 * message - pointer to the msghdr structure which contains a single
1357 * user message and possibly some ancillary data.
1359 * See Section 5 for complete description of the data
1362 * flags - flags sent or received with the user message, see Section
1363 * 5 for complete description of the flags.
1365 * Note: This function could use a rewrite especially when explicit
1366 * connect support comes in.
1368 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1370 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1372 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1373 struct msghdr *msg, size_t msg_len)
1375 struct sctp_sock *sp;
1376 struct sctp_endpoint *ep;
1377 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1378 struct sctp_transport *transport, *chunk_tp;
1379 struct sctp_chunk *chunk;
1381 struct sockaddr *msg_name = NULL;
1382 struct sctp_sndrcvinfo default_sinfo = { 0 };
1383 struct sctp_sndrcvinfo *sinfo;
1384 struct sctp_initmsg *sinit;
1385 sctp_assoc_t associd = 0;
1386 sctp_cmsgs_t cmsgs = { NULL };
1390 __u16 sinfo_flags = 0;
1391 struct sctp_datamsg *datamsg;
1392 struct list_head *pos;
1393 int msg_flags = msg->msg_flags;
1395 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1402 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1404 /* We cannot send a message over a TCP-style listening socket. */
1405 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1410 /* Parse out the SCTP CMSGs. */
1411 err = sctp_msghdr_parse(msg, &cmsgs);
1414 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1418 /* Fetch the destination address for this packet. This
1419 * address only selects the association--it is not necessarily
1420 * the address we will send to.
1421 * For a peeled-off socket, msg_name is ignored.
1423 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1424 int msg_namelen = msg->msg_namelen;
1426 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1431 if (msg_namelen > sizeof(to))
1432 msg_namelen = sizeof(to);
1433 memcpy(&to, msg->msg_name, msg_namelen);
1434 msg_name = msg->msg_name;
1440 /* Did the user specify SNDRCVINFO? */
1442 sinfo_flags = sinfo->sinfo_flags;
1443 associd = sinfo->sinfo_assoc_id;
1446 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1447 msg_len, sinfo_flags);
1449 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1450 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1455 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1456 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1457 * If SCTP_ABORT is set, the message length could be non zero with
1458 * the msg_iov set to the user abort reason.
1460 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1461 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1466 /* If SCTP_ADDR_OVER is set, there must be an address
1467 * specified in msg_name.
1469 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1476 SCTP_DEBUG_PRINTK("About to look up association.\n");
1480 /* If a msg_name has been specified, assume this is to be used. */
1482 /* Look for a matching association on the endpoint. */
1483 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1485 /* If we could not find a matching association on the
1486 * endpoint, make sure that it is not a TCP-style
1487 * socket that already has an association or there is
1488 * no peeled-off association on another socket.
1490 if ((sctp_style(sk, TCP) &&
1491 sctp_sstate(sk, ESTABLISHED)) ||
1492 sctp_endpoint_is_peeled_off(ep, &to)) {
1493 err = -EADDRNOTAVAIL;
1498 asoc = sctp_id2assoc(sk, associd);
1506 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1508 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1509 * socket that has an association in CLOSED state. This can
1510 * happen when an accepted socket has an association that is
1513 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1518 if (sinfo_flags & SCTP_EOF) {
1519 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1521 sctp_primitive_SHUTDOWN(asoc, NULL);
1525 if (sinfo_flags & SCTP_ABORT) {
1526 struct sctp_chunk *chunk;
1528 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1534 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1535 sctp_primitive_ABORT(asoc, chunk);
1541 /* Do we need to create the association? */
1543 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1545 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1550 /* Check for invalid stream against the stream counts,
1551 * either the default or the user specified stream counts.
1554 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1555 /* Check against the defaults. */
1556 if (sinfo->sinfo_stream >=
1557 sp->initmsg.sinit_num_ostreams) {
1562 /* Check against the requested. */
1563 if (sinfo->sinfo_stream >=
1564 sinit->sinit_num_ostreams) {
1572 * API 3.1.2 bind() - UDP Style Syntax
1573 * If a bind() or sctp_bindx() is not called prior to a
1574 * sendmsg() call that initiates a new association, the
1575 * system picks an ephemeral port and will choose an address
1576 * set equivalent to binding with a wildcard address.
1578 if (!ep->base.bind_addr.port) {
1579 if (sctp_autobind(sk)) {
1585 * If an unprivileged user inherits a one-to-many
1586 * style socket with open associations on a privileged
1587 * port, it MAY be permitted to accept new associations,
1588 * but it SHOULD NOT be permitted to open new
1591 if (ep->base.bind_addr.port < PROT_SOCK &&
1592 !capable(CAP_NET_BIND_SERVICE)) {
1598 scope = sctp_scope(&to);
1599 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1606 /* If the SCTP_INIT ancillary data is specified, set all
1607 * the association init values accordingly.
1610 if (sinit->sinit_num_ostreams) {
1611 asoc->c.sinit_num_ostreams =
1612 sinit->sinit_num_ostreams;
1614 if (sinit->sinit_max_instreams) {
1615 asoc->c.sinit_max_instreams =
1616 sinit->sinit_max_instreams;
1618 if (sinit->sinit_max_attempts) {
1619 asoc->max_init_attempts
1620 = sinit->sinit_max_attempts;
1622 if (sinit->sinit_max_init_timeo) {
1623 asoc->max_init_timeo =
1624 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1628 /* Prime the peer's transport structures. */
1629 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1634 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1641 /* ASSERT: we have a valid association at this point. */
1642 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1645 /* If the user didn't specify SNDRCVINFO, make up one with
1648 default_sinfo.sinfo_stream = asoc->default_stream;
1649 default_sinfo.sinfo_flags = asoc->default_flags;
1650 default_sinfo.sinfo_ppid = asoc->default_ppid;
1651 default_sinfo.sinfo_context = asoc->default_context;
1652 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1653 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1654 sinfo = &default_sinfo;
1657 /* API 7.1.7, the sndbuf size per association bounds the
1658 * maximum size of data that can be sent in a single send call.
1660 if (msg_len > sk->sk_sndbuf) {
1665 /* If fragmentation is disabled and the message length exceeds the
1666 * association fragmentation point, return EMSGSIZE. The I-D
1667 * does not specify what this error is, but this looks like
1670 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1676 /* Check for invalid stream. */
1677 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1683 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1684 if (!sctp_wspace(asoc)) {
1685 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1690 /* If an address is passed with the sendto/sendmsg call, it is used
1691 * to override the primary destination address in the TCP model, or
1692 * when SCTP_ADDR_OVER flag is set in the UDP model.
1694 if ((sctp_style(sk, TCP) && msg_name) ||
1695 (sinfo_flags & SCTP_ADDR_OVER)) {
1696 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1704 /* Auto-connect, if we aren't connected already. */
1705 if (sctp_state(asoc, CLOSED)) {
1706 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1709 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1712 /* Break the message into multiple chunks of maximum size. */
1713 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1719 /* Now send the (possibly) fragmented message. */
1720 list_for_each(pos, &datamsg->chunks) {
1721 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1722 sctp_datamsg_track(chunk);
1724 /* Do accounting for the write space. */
1725 sctp_set_owner_w(chunk);
1727 chunk->transport = chunk_tp;
1729 /* Send it to the lower layers. Note: all chunks
1730 * must either fail or succeed. The lower layer
1731 * works that way today. Keep it that way or this
1734 err = sctp_primitive_SEND(asoc, chunk);
1735 /* Did the lower layer accept the chunk? */
1737 sctp_chunk_free(chunk);
1738 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1741 sctp_datamsg_free(datamsg);
1747 /* If we are already past ASSOCIATE, the lower
1748 * layers are responsible for association cleanup.
1754 sctp_association_free(asoc);
1756 sctp_release_sock(sk);
1759 return sctp_error(sk, msg_flags, err);
1766 err = sock_error(sk);
1776 /* This is an extended version of skb_pull() that removes the data from the
1777 * start of a skb even when data is spread across the list of skb's in the
1778 * frag_list. len specifies the total amount of data that needs to be removed.
1779 * when 'len' bytes could be removed from the skb, it returns 0.
1780 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1781 * could not be removed.
1783 static int sctp_skb_pull(struct sk_buff *skb, int len)
1785 struct sk_buff *list;
1786 int skb_len = skb_headlen(skb);
1789 if (len <= skb_len) {
1790 __skb_pull(skb, len);
1794 __skb_pull(skb, skb_len);
1796 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1797 rlen = sctp_skb_pull(list, len);
1798 skb->len -= (len-rlen);
1799 skb->data_len -= (len-rlen);
1810 /* API 3.1.3 recvmsg() - UDP Style Syntax
1812 * ssize_t recvmsg(int socket, struct msghdr *message,
1815 * socket - the socket descriptor of the endpoint.
1816 * message - pointer to the msghdr structure which contains a single
1817 * user message and possibly some ancillary data.
1819 * See Section 5 for complete description of the data
1822 * flags - flags sent or received with the user message, see Section
1823 * 5 for complete description of the flags.
1825 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1827 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1828 struct msghdr *msg, size_t len, int noblock,
1829 int flags, int *addr_len)
1831 struct sctp_ulpevent *event = NULL;
1832 struct sctp_sock *sp = sctp_sk(sk);
1833 struct sk_buff *skb;
1838 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1839 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1840 "len", len, "knoblauch", noblock,
1841 "flags", flags, "addr_len", addr_len);
1845 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1850 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1854 /* Get the total length of the skb including any skb's in the
1863 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1865 event = sctp_skb2event(skb);
1870 sock_recv_timestamp(msg, sk, skb);
1871 if (sctp_ulpevent_is_notification(event)) {
1872 msg->msg_flags |= MSG_NOTIFICATION;
1873 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1875 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1878 /* Check if we allow SCTP_SNDRCVINFO. */
1879 if (sp->subscribe.sctp_data_io_event)
1880 sctp_ulpevent_read_sndrcvinfo(event, msg);
1882 /* FIXME: we should be calling IP/IPv6 layers. */
1883 if (sk->sk_protinfo.af_inet.cmsg_flags)
1884 ip_cmsg_recv(msg, skb);
1889 /* If skb's length exceeds the user's buffer, update the skb and
1890 * push it back to the receive_queue so that the next call to
1891 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1893 if (skb_len > copied) {
1894 msg->msg_flags &= ~MSG_EOR;
1895 if (flags & MSG_PEEK)
1897 sctp_skb_pull(skb, copied);
1898 skb_queue_head(&sk->sk_receive_queue, skb);
1900 /* When only partial message is copied to the user, increase
1901 * rwnd by that amount. If all the data in the skb is read,
1902 * rwnd is updated when the event is freed.
1904 sctp_assoc_rwnd_increase(event->asoc, copied);
1906 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1907 (event->msg_flags & MSG_EOR))
1908 msg->msg_flags |= MSG_EOR;
1910 msg->msg_flags &= ~MSG_EOR;
1913 if (flags & MSG_PEEK) {
1914 /* Release the skb reference acquired after peeking the skb in
1915 * sctp_skb_recv_datagram().
1919 /* Free the event which includes releasing the reference to
1920 * the owner of the skb, freeing the skb and updating the
1923 sctp_ulpevent_free(event);
1926 sctp_release_sock(sk);
1930 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1932 * This option is a on/off flag. If enabled no SCTP message
1933 * fragmentation will be performed. Instead if a message being sent
1934 * exceeds the current PMTU size, the message will NOT be sent and
1935 * instead a error will be indicated to the user.
1937 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1938 char __user *optval, int optlen)
1942 if (optlen < sizeof(int))
1945 if (get_user(val, (int __user *)optval))
1948 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1953 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1956 if (optlen != sizeof(struct sctp_event_subscribe))
1958 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1963 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1965 * This socket option is applicable to the UDP-style socket only. When
1966 * set it will cause associations that are idle for more than the
1967 * specified number of seconds to automatically close. An association
1968 * being idle is defined an association that has NOT sent or received
1969 * user data. The special value of '0' indicates that no automatic
1970 * close of any associations should be performed. The option expects an
1971 * integer defining the number of seconds of idle time before an
1972 * association is closed.
1974 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1977 struct sctp_sock *sp = sctp_sk(sk);
1979 /* Applicable to UDP-style socket only */
1980 if (sctp_style(sk, TCP))
1982 if (optlen != sizeof(int))
1984 if (copy_from_user(&sp->autoclose, optval, optlen))
1990 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1992 * Applications can enable or disable heartbeats for any peer address of
1993 * an association, modify an address's heartbeat interval, force a
1994 * heartbeat to be sent immediately, and adjust the address's maximum
1995 * number of retransmissions sent before an address is considered
1996 * unreachable. The following structure is used to access and modify an
1997 * address's parameters:
1999 * struct sctp_paddrparams {
2000 * sctp_assoc_t spp_assoc_id;
2001 * struct sockaddr_storage spp_address;
2002 * uint32_t spp_hbinterval;
2003 * uint16_t spp_pathmaxrxt;
2004 * uint32_t spp_pathmtu;
2005 * uint32_t spp_sackdelay;
2006 * uint32_t spp_flags;
2009 * spp_assoc_id - (one-to-many style socket) This is filled in the
2010 * application, and identifies the association for
2012 * spp_address - This specifies which address is of interest.
2013 * spp_hbinterval - This contains the value of the heartbeat interval,
2014 * in milliseconds. If a value of zero
2015 * is present in this field then no changes are to
2016 * be made to this parameter.
2017 * spp_pathmaxrxt - This contains the maximum number of
2018 * retransmissions before this address shall be
2019 * considered unreachable. If a value of zero
2020 * is present in this field then no changes are to
2021 * be made to this parameter.
2022 * spp_pathmtu - When Path MTU discovery is disabled the value
2023 * specified here will be the "fixed" path mtu.
2024 * Note that if the spp_address field is empty
2025 * then all associations on this address will
2026 * have this fixed path mtu set upon them.
2028 * spp_sackdelay - When delayed sack is enabled, this value specifies
2029 * the number of milliseconds that sacks will be delayed
2030 * for. This value will apply to all addresses of an
2031 * association if the spp_address field is empty. Note
2032 * also, that if delayed sack is enabled and this
2033 * value is set to 0, no change is made to the last
2034 * recorded delayed sack timer value.
2036 * spp_flags - These flags are used to control various features
2037 * on an association. The flag field may contain
2038 * zero or more of the following options.
2040 * SPP_HB_ENABLE - Enable heartbeats on the
2041 * specified address. Note that if the address
2042 * field is empty all addresses for the association
2043 * have heartbeats enabled upon them.
2045 * SPP_HB_DISABLE - Disable heartbeats on the
2046 * speicifed address. Note that if the address
2047 * field is empty all addresses for the association
2048 * will have their heartbeats disabled. Note also
2049 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2050 * mutually exclusive, only one of these two should
2051 * be specified. Enabling both fields will have
2052 * undetermined results.
2054 * SPP_HB_DEMAND - Request a user initiated heartbeat
2055 * to be made immediately.
2057 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2058 * heartbeat delayis to be set to the value of 0
2061 * SPP_PMTUD_ENABLE - This field will enable PMTU
2062 * discovery upon the specified address. Note that
2063 * if the address feild is empty then all addresses
2064 * on the association are effected.
2066 * SPP_PMTUD_DISABLE - This field will disable PMTU
2067 * discovery upon the specified address. Note that
2068 * if the address feild is empty then all addresses
2069 * on the association are effected. Not also that
2070 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2071 * exclusive. Enabling both will have undetermined
2074 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2075 * on delayed sack. The time specified in spp_sackdelay
2076 * is used to specify the sack delay for this address. Note
2077 * that if spp_address is empty then all addresses will
2078 * enable delayed sack and take on the sack delay
2079 * value specified in spp_sackdelay.
2080 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2081 * off delayed sack. If the spp_address field is blank then
2082 * delayed sack is disabled for the entire association. Note
2083 * also that this field is mutually exclusive to
2084 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2087 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2088 struct sctp_transport *trans,
2089 struct sctp_association *asoc,
2090 struct sctp_sock *sp,
2093 int sackdelay_change)
2097 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2098 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2103 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2104 * this field is ignored. Note also that a value of zero indicates
2105 * the current setting should be left unchanged.
2107 if (params->spp_flags & SPP_HB_ENABLE) {
2109 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2110 * set. This lets us use 0 value when this flag
2113 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2114 params->spp_hbinterval = 0;
2116 if (params->spp_hbinterval ||
2117 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2120 msecs_to_jiffies(params->spp_hbinterval);
2123 msecs_to_jiffies(params->spp_hbinterval);
2125 sp->hbinterval = params->spp_hbinterval;
2132 trans->param_flags =
2133 (trans->param_flags & ~SPP_HB) | hb_change;
2136 (asoc->param_flags & ~SPP_HB) | hb_change;
2139 (sp->param_flags & ~SPP_HB) | hb_change;
2143 /* When Path MTU discovery is disabled the value specified here will
2144 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2145 * include the flag SPP_PMTUD_DISABLE for this field to have any
2148 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2150 trans->pathmtu = params->spp_pathmtu;
2151 sctp_assoc_sync_pmtu(asoc);
2153 asoc->pathmtu = params->spp_pathmtu;
2154 sctp_frag_point(sp, params->spp_pathmtu);
2156 sp->pathmtu = params->spp_pathmtu;
2162 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2163 (params->spp_flags & SPP_PMTUD_ENABLE);
2164 trans->param_flags =
2165 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2167 sctp_transport_pmtu(trans);
2168 sctp_assoc_sync_pmtu(asoc);
2172 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2175 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2179 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2180 * value of this field is ignored. Note also that a value of zero
2181 * indicates the current setting should be left unchanged.
2183 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2186 msecs_to_jiffies(params->spp_sackdelay);
2189 msecs_to_jiffies(params->spp_sackdelay);
2191 sp->sackdelay = params->spp_sackdelay;
2195 if (sackdelay_change) {
2197 trans->param_flags =
2198 (trans->param_flags & ~SPP_SACKDELAY) |
2202 (asoc->param_flags & ~SPP_SACKDELAY) |
2206 (sp->param_flags & ~SPP_SACKDELAY) |
2211 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2212 * of this field is ignored. Note also that a value of zero
2213 * indicates the current setting should be left unchanged.
2215 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2217 trans->pathmaxrxt = params->spp_pathmaxrxt;
2219 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2221 sp->pathmaxrxt = params->spp_pathmaxrxt;
2228 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2229 char __user *optval, int optlen)
2231 struct sctp_paddrparams params;
2232 struct sctp_transport *trans = NULL;
2233 struct sctp_association *asoc = NULL;
2234 struct sctp_sock *sp = sctp_sk(sk);
2236 int hb_change, pmtud_change, sackdelay_change;
2238 if (optlen != sizeof(struct sctp_paddrparams))
2241 if (copy_from_user(¶ms, optval, optlen))
2244 /* Validate flags and value parameters. */
2245 hb_change = params.spp_flags & SPP_HB;
2246 pmtud_change = params.spp_flags & SPP_PMTUD;
2247 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2249 if (hb_change == SPP_HB ||
2250 pmtud_change == SPP_PMTUD ||
2251 sackdelay_change == SPP_SACKDELAY ||
2252 params.spp_sackdelay > 500 ||
2254 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2257 /* If an address other than INADDR_ANY is specified, and
2258 * no transport is found, then the request is invalid.
2260 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
2261 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2262 params.spp_assoc_id);
2267 /* Get association, if assoc_id != 0 and the socket is a one
2268 * to many style socket, and an association was not found, then
2269 * the id was invalid.
2271 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2272 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2275 /* Heartbeat demand can only be sent on a transport or
2276 * association, but not a socket.
2278 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2281 /* Process parameters. */
2282 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2283 hb_change, pmtud_change,
2289 /* If changes are for association, also apply parameters to each
2292 if (!trans && asoc) {
2293 struct list_head *pos;
2295 list_for_each(pos, &asoc->peer.transport_addr_list) {
2296 trans = list_entry(pos, struct sctp_transport,
2298 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2299 hb_change, pmtud_change,
2307 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2309 * This options will get or set the delayed ack timer. The time is set
2310 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2311 * endpoints default delayed ack timer value. If the assoc_id field is
2312 * non-zero, then the set or get effects the specified association.
2314 * struct sctp_assoc_value {
2315 * sctp_assoc_t assoc_id;
2316 * uint32_t assoc_value;
2319 * assoc_id - This parameter, indicates which association the
2320 * user is preforming an action upon. Note that if
2321 * this field's value is zero then the endpoints
2322 * default value is changed (effecting future
2323 * associations only).
2325 * assoc_value - This parameter contains the number of milliseconds
2326 * that the user is requesting the delayed ACK timer
2327 * be set to. Note that this value is defined in
2328 * the standard to be between 200 and 500 milliseconds.
2330 * Note: a value of zero will leave the value alone,
2331 * but disable SACK delay. A non-zero value will also
2332 * enable SACK delay.
2335 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2336 char __user *optval, int optlen)
2338 struct sctp_assoc_value params;
2339 struct sctp_transport *trans = NULL;
2340 struct sctp_association *asoc = NULL;
2341 struct sctp_sock *sp = sctp_sk(sk);
2343 if (optlen != sizeof(struct sctp_assoc_value))
2346 if (copy_from_user(¶ms, optval, optlen))
2349 /* Validate value parameter. */
2350 if (params.assoc_value > 500)
2353 /* Get association, if assoc_id != 0 and the socket is a one
2354 * to many style socket, and an association was not found, then
2355 * the id was invalid.
2357 asoc = sctp_id2assoc(sk, params.assoc_id);
2358 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2361 if (params.assoc_value) {
2364 msecs_to_jiffies(params.assoc_value);
2366 (asoc->param_flags & ~SPP_SACKDELAY) |
2367 SPP_SACKDELAY_ENABLE;
2369 sp->sackdelay = params.assoc_value;
2371 (sp->param_flags & ~SPP_SACKDELAY) |
2372 SPP_SACKDELAY_ENABLE;
2377 (asoc->param_flags & ~SPP_SACKDELAY) |
2378 SPP_SACKDELAY_DISABLE;
2381 (sp->param_flags & ~SPP_SACKDELAY) |
2382 SPP_SACKDELAY_DISABLE;
2386 /* If change is for association, also apply to each transport. */
2388 struct list_head *pos;
2390 list_for_each(pos, &asoc->peer.transport_addr_list) {
2391 trans = list_entry(pos, struct sctp_transport,
2393 if (params.assoc_value) {
2395 msecs_to_jiffies(params.assoc_value);
2396 trans->param_flags =
2397 (trans->param_flags & ~SPP_SACKDELAY) |
2398 SPP_SACKDELAY_ENABLE;
2400 trans->param_flags =
2401 (trans->param_flags & ~SPP_SACKDELAY) |
2402 SPP_SACKDELAY_DISABLE;
2410 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2412 * Applications can specify protocol parameters for the default association
2413 * initialization. The option name argument to setsockopt() and getsockopt()
2416 * Setting initialization parameters is effective only on an unconnected
2417 * socket (for UDP-style sockets only future associations are effected
2418 * by the change). With TCP-style sockets, this option is inherited by
2419 * sockets derived from a listener socket.
2421 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2423 struct sctp_initmsg sinit;
2424 struct sctp_sock *sp = sctp_sk(sk);
2426 if (optlen != sizeof(struct sctp_initmsg))
2428 if (copy_from_user(&sinit, optval, optlen))
2431 if (sinit.sinit_num_ostreams)
2432 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2433 if (sinit.sinit_max_instreams)
2434 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2435 if (sinit.sinit_max_attempts)
2436 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2437 if (sinit.sinit_max_init_timeo)
2438 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2444 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2446 * Applications that wish to use the sendto() system call may wish to
2447 * specify a default set of parameters that would normally be supplied
2448 * through the inclusion of ancillary data. This socket option allows
2449 * such an application to set the default sctp_sndrcvinfo structure.
2450 * The application that wishes to use this socket option simply passes
2451 * in to this call the sctp_sndrcvinfo structure defined in Section
2452 * 5.2.2) The input parameters accepted by this call include
2453 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2454 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2455 * to this call if the caller is using the UDP model.
2457 static int sctp_setsockopt_default_send_param(struct sock *sk,
2458 char __user *optval, int optlen)
2460 struct sctp_sndrcvinfo info;
2461 struct sctp_association *asoc;
2462 struct sctp_sock *sp = sctp_sk(sk);
2464 if (optlen != sizeof(struct sctp_sndrcvinfo))
2466 if (copy_from_user(&info, optval, optlen))
2469 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2470 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2474 asoc->default_stream = info.sinfo_stream;
2475 asoc->default_flags = info.sinfo_flags;
2476 asoc->default_ppid = info.sinfo_ppid;
2477 asoc->default_context = info.sinfo_context;
2478 asoc->default_timetolive = info.sinfo_timetolive;
2480 sp->default_stream = info.sinfo_stream;
2481 sp->default_flags = info.sinfo_flags;
2482 sp->default_ppid = info.sinfo_ppid;
2483 sp->default_context = info.sinfo_context;
2484 sp->default_timetolive = info.sinfo_timetolive;
2490 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2492 * Requests that the local SCTP stack use the enclosed peer address as
2493 * the association primary. The enclosed address must be one of the
2494 * association peer's addresses.
2496 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2499 struct sctp_prim prim;
2500 struct sctp_transport *trans;
2502 if (optlen != sizeof(struct sctp_prim))
2505 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2508 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2512 sctp_assoc_set_primary(trans->asoc, trans);
2518 * 7.1.5 SCTP_NODELAY
2520 * Turn on/off any Nagle-like algorithm. This means that packets are
2521 * generally sent as soon as possible and no unnecessary delays are
2522 * introduced, at the cost of more packets in the network. Expects an
2523 * integer boolean flag.
2525 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2530 if (optlen < sizeof(int))
2532 if (get_user(val, (int __user *)optval))
2535 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2541 * 7.1.1 SCTP_RTOINFO
2543 * The protocol parameters used to initialize and bound retransmission
2544 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2545 * and modify these parameters.
2546 * All parameters are time values, in milliseconds. A value of 0, when
2547 * modifying the parameters, indicates that the current value should not
2551 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2552 struct sctp_rtoinfo rtoinfo;
2553 struct sctp_association *asoc;
2555 if (optlen != sizeof (struct sctp_rtoinfo))
2558 if (copy_from_user(&rtoinfo, optval, optlen))
2561 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2563 /* Set the values to the specific association */
2564 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2568 if (rtoinfo.srto_initial != 0)
2570 msecs_to_jiffies(rtoinfo.srto_initial);
2571 if (rtoinfo.srto_max != 0)
2572 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2573 if (rtoinfo.srto_min != 0)
2574 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2576 /* If there is no association or the association-id = 0
2577 * set the values to the endpoint.
2579 struct sctp_sock *sp = sctp_sk(sk);
2581 if (rtoinfo.srto_initial != 0)
2582 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2583 if (rtoinfo.srto_max != 0)
2584 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2585 if (rtoinfo.srto_min != 0)
2586 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2594 * 7.1.2 SCTP_ASSOCINFO
2596 * This option is used to tune the maximum retransmission attempts
2597 * of the association.
2598 * Returns an error if the new association retransmission value is
2599 * greater than the sum of the retransmission value of the peer.
2600 * See [SCTP] for more information.
2603 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2606 struct sctp_assocparams assocparams;
2607 struct sctp_association *asoc;
2609 if (optlen != sizeof(struct sctp_assocparams))
2611 if (copy_from_user(&assocparams, optval, optlen))
2614 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2616 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2619 /* Set the values to the specific association */
2621 if (assocparams.sasoc_asocmaxrxt != 0) {
2624 struct list_head *pos;
2625 struct sctp_transport *peer_addr;
2627 list_for_each(pos, &asoc->peer.transport_addr_list) {
2628 peer_addr = list_entry(pos,
2629 struct sctp_transport,
2631 path_sum += peer_addr->pathmaxrxt;
2635 /* Only validate asocmaxrxt if we have more then
2636 * one path/transport. We do this because path
2637 * retransmissions are only counted when we have more
2641 assocparams.sasoc_asocmaxrxt > path_sum)
2644 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2647 if (assocparams.sasoc_cookie_life != 0) {
2648 asoc->cookie_life.tv_sec =
2649 assocparams.sasoc_cookie_life / 1000;
2650 asoc->cookie_life.tv_usec =
2651 (assocparams.sasoc_cookie_life % 1000)
2655 /* Set the values to the endpoint */
2656 struct sctp_sock *sp = sctp_sk(sk);
2658 if (assocparams.sasoc_asocmaxrxt != 0)
2659 sp->assocparams.sasoc_asocmaxrxt =
2660 assocparams.sasoc_asocmaxrxt;
2661 if (assocparams.sasoc_cookie_life != 0)
2662 sp->assocparams.sasoc_cookie_life =
2663 assocparams.sasoc_cookie_life;
2669 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2671 * This socket option is a boolean flag which turns on or off mapped V4
2672 * addresses. If this option is turned on and the socket is type
2673 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2674 * If this option is turned off, then no mapping will be done of V4
2675 * addresses and a user will receive both PF_INET6 and PF_INET type
2676 * addresses on the socket.
2678 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2681 struct sctp_sock *sp = sctp_sk(sk);
2683 if (optlen < sizeof(int))
2685 if (get_user(val, (int __user *)optval))
2696 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2698 * This socket option specifies the maximum size to put in any outgoing
2699 * SCTP chunk. If a message is larger than this size it will be
2700 * fragmented by SCTP into the specified size. Note that the underlying
2701 * SCTP implementation may fragment into smaller sized chunks when the
2702 * PMTU of the underlying association is smaller than the value set by
2705 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2707 struct sctp_association *asoc;
2708 struct list_head *pos;
2709 struct sctp_sock *sp = sctp_sk(sk);
2712 if (optlen < sizeof(int))
2714 if (get_user(val, (int __user *)optval))
2716 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2718 sp->user_frag = val;
2720 /* Update the frag_point of the existing associations. */
2721 list_for_each(pos, &(sp->ep->asocs)) {
2722 asoc = list_entry(pos, struct sctp_association, asocs);
2723 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2731 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2733 * Requests that the peer mark the enclosed address as the association
2734 * primary. The enclosed address must be one of the association's
2735 * locally bound addresses. The following structure is used to make a
2736 * set primary request:
2738 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2741 struct sctp_sock *sp;
2742 struct sctp_endpoint *ep;
2743 struct sctp_association *asoc = NULL;
2744 struct sctp_setpeerprim prim;
2745 struct sctp_chunk *chunk;
2751 if (!sctp_addip_enable)
2754 if (optlen != sizeof(struct sctp_setpeerprim))
2757 if (copy_from_user(&prim, optval, optlen))
2760 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2764 if (!asoc->peer.asconf_capable)
2767 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2770 if (!sctp_state(asoc, ESTABLISHED))
2773 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2774 return -EADDRNOTAVAIL;
2776 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2777 chunk = sctp_make_asconf_set_prim(asoc,
2778 (union sctp_addr *)&prim.sspp_addr);
2782 err = sctp_send_asconf(asoc, chunk);
2784 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2789 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2792 struct sctp_setadaptation adaptation;
2794 if (optlen != sizeof(struct sctp_setadaptation))
2796 if (copy_from_user(&adaptation, optval, optlen))
2799 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2805 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2807 * The context field in the sctp_sndrcvinfo structure is normally only
2808 * used when a failed message is retrieved holding the value that was
2809 * sent down on the actual send call. This option allows the setting of
2810 * a default context on an association basis that will be received on
2811 * reading messages from the peer. This is especially helpful in the
2812 * one-2-many model for an application to keep some reference to an
2813 * internal state machine that is processing messages on the
2814 * association. Note that the setting of this value only effects
2815 * received messages from the peer and does not effect the value that is
2816 * saved with outbound messages.
2818 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2821 struct sctp_assoc_value params;
2822 struct sctp_sock *sp;
2823 struct sctp_association *asoc;
2825 if (optlen != sizeof(struct sctp_assoc_value))
2827 if (copy_from_user(¶ms, optval, optlen))
2832 if (params.assoc_id != 0) {
2833 asoc = sctp_id2assoc(sk, params.assoc_id);
2836 asoc->default_rcv_context = params.assoc_value;
2838 sp->default_rcv_context = params.assoc_value;
2845 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2847 * This options will at a minimum specify if the implementation is doing
2848 * fragmented interleave. Fragmented interleave, for a one to many
2849 * socket, is when subsequent calls to receive a message may return
2850 * parts of messages from different associations. Some implementations
2851 * may allow you to turn this value on or off. If so, when turned off,
2852 * no fragment interleave will occur (which will cause a head of line
2853 * blocking amongst multiple associations sharing the same one to many
2854 * socket). When this option is turned on, then each receive call may
2855 * come from a different association (thus the user must receive data
2856 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2857 * association each receive belongs to.
2859 * This option takes a boolean value. A non-zero value indicates that
2860 * fragmented interleave is on. A value of zero indicates that
2861 * fragmented interleave is off.
2863 * Note that it is important that an implementation that allows this
2864 * option to be turned on, have it off by default. Otherwise an unaware
2865 * application using the one to many model may become confused and act
2868 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2869 char __user *optval,
2874 if (optlen != sizeof(int))
2876 if (get_user(val, (int __user *)optval))
2879 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2885 * 7.1.25. Set or Get the sctp partial delivery point
2886 * (SCTP_PARTIAL_DELIVERY_POINT)
2887 * This option will set or get the SCTP partial delivery point. This
2888 * point is the size of a message where the partial delivery API will be
2889 * invoked to help free up rwnd space for the peer. Setting this to a
2890 * lower value will cause partial delivery's to happen more often. The
2891 * calls argument is an integer that sets or gets the partial delivery
2894 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2895 char __user *optval,
2900 if (optlen != sizeof(u32))
2902 if (get_user(val, (int __user *)optval))
2905 sctp_sk(sk)->pd_point = val;
2907 return 0; /* is this the right error code? */
2911 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2913 * This option will allow a user to change the maximum burst of packets
2914 * that can be emitted by this association. Note that the default value
2915 * is 4, and some implementations may restrict this setting so that it
2916 * can only be lowered.
2918 * NOTE: This text doesn't seem right. Do this on a socket basis with
2919 * future associations inheriting the socket value.
2921 static int sctp_setsockopt_maxburst(struct sock *sk,
2922 char __user *optval,
2927 if (optlen != sizeof(int))
2929 if (get_user(val, (int __user *)optval))
2935 sctp_sk(sk)->max_burst = val;
2940 /* API 6.2 setsockopt(), getsockopt()
2942 * Applications use setsockopt() and getsockopt() to set or retrieve
2943 * socket options. Socket options are used to change the default
2944 * behavior of sockets calls. They are described in Section 7.
2948 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2949 * int __user *optlen);
2950 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2953 * sd - the socket descript.
2954 * level - set to IPPROTO_SCTP for all SCTP options.
2955 * optname - the option name.
2956 * optval - the buffer to store the value of the option.
2957 * optlen - the size of the buffer.
2959 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2960 char __user *optval, int optlen)
2964 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2967 /* I can hardly begin to describe how wrong this is. This is
2968 * so broken as to be worse than useless. The API draft
2969 * REALLY is NOT helpful here... I am not convinced that the
2970 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2971 * are at all well-founded.
2973 if (level != SOL_SCTP) {
2974 struct sctp_af *af = sctp_sk(sk)->pf->af;
2975 retval = af->setsockopt(sk, level, optname, optval, optlen);
2982 case SCTP_SOCKOPT_BINDX_ADD:
2983 /* 'optlen' is the size of the addresses buffer. */
2984 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2985 optlen, SCTP_BINDX_ADD_ADDR);
2988 case SCTP_SOCKOPT_BINDX_REM:
2989 /* 'optlen' is the size of the addresses buffer. */
2990 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2991 optlen, SCTP_BINDX_REM_ADDR);
2994 case SCTP_SOCKOPT_CONNECTX:
2995 /* 'optlen' is the size of the addresses buffer. */
2996 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3000 case SCTP_DISABLE_FRAGMENTS:
3001 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3005 retval = sctp_setsockopt_events(sk, optval, optlen);
3008 case SCTP_AUTOCLOSE:
3009 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3012 case SCTP_PEER_ADDR_PARAMS:
3013 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3016 case SCTP_DELAYED_ACK_TIME:
3017 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3019 case SCTP_PARTIAL_DELIVERY_POINT:
3020 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3024 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3026 case SCTP_DEFAULT_SEND_PARAM:
3027 retval = sctp_setsockopt_default_send_param(sk, optval,
3030 case SCTP_PRIMARY_ADDR:
3031 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3033 case SCTP_SET_PEER_PRIMARY_ADDR:
3034 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3037 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3040 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3042 case SCTP_ASSOCINFO:
3043 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3045 case SCTP_I_WANT_MAPPED_V4_ADDR:
3046 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3049 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3051 case SCTP_ADAPTATION_LAYER:
3052 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3055 retval = sctp_setsockopt_context(sk, optval, optlen);
3057 case SCTP_FRAGMENT_INTERLEAVE:
3058 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3060 case SCTP_MAX_BURST:
3061 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3064 retval = -ENOPROTOOPT;
3068 sctp_release_sock(sk);
3074 /* API 3.1.6 connect() - UDP Style Syntax
3076 * An application may use the connect() call in the UDP model to initiate an
3077 * association without sending data.
3081 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3083 * sd: the socket descriptor to have a new association added to.
3085 * nam: the address structure (either struct sockaddr_in or struct
3086 * sockaddr_in6 defined in RFC2553 [7]).
3088 * len: the size of the address.
3090 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3098 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3099 __FUNCTION__, sk, addr, addr_len);
3101 /* Validate addr_len before calling common connect/connectx routine. */
3102 af = sctp_get_af_specific(addr->sa_family);
3103 if (!af || addr_len < af->sockaddr_len) {
3106 /* Pass correct addr len to common routine (so it knows there
3107 * is only one address being passed.
3109 err = __sctp_connect(sk, addr, af->sockaddr_len);
3112 sctp_release_sock(sk);
3116 /* FIXME: Write comments. */
3117 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3119 return -EOPNOTSUPP; /* STUB */
3122 /* 4.1.4 accept() - TCP Style Syntax
3124 * Applications use accept() call to remove an established SCTP
3125 * association from the accept queue of the endpoint. A new socket
3126 * descriptor will be returned from accept() to represent the newly
3127 * formed association.
3129 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3131 struct sctp_sock *sp;
3132 struct sctp_endpoint *ep;
3133 struct sock *newsk = NULL;
3134 struct sctp_association *asoc;
3143 if (!sctp_style(sk, TCP)) {
3144 error = -EOPNOTSUPP;
3148 if (!sctp_sstate(sk, LISTENING)) {
3153 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3155 error = sctp_wait_for_accept(sk, timeo);
3159 /* We treat the list of associations on the endpoint as the accept
3160 * queue and pick the first association on the list.
3162 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3164 newsk = sp->pf->create_accept_sk(sk, asoc);
3170 /* Populate the fields of the newsk from the oldsk and migrate the
3171 * asoc to the newsk.
3173 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3176 sctp_release_sock(sk);
3181 /* The SCTP ioctl handler. */
3182 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3184 return -ENOIOCTLCMD;
3187 /* This is the function which gets called during socket creation to
3188 * initialized the SCTP-specific portion of the sock.
3189 * The sock structure should already be zero-filled memory.
3191 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3193 struct sctp_endpoint *ep;
3194 struct sctp_sock *sp;
3196 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3200 /* Initialize the SCTP per socket area. */
3201 switch (sk->sk_type) {
3202 case SOCK_SEQPACKET:
3203 sp->type = SCTP_SOCKET_UDP;
3206 sp->type = SCTP_SOCKET_TCP;
3209 return -ESOCKTNOSUPPORT;
3212 /* Initialize default send parameters. These parameters can be
3213 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3215 sp->default_stream = 0;
3216 sp->default_ppid = 0;
3217 sp->default_flags = 0;
3218 sp->default_context = 0;
3219 sp->default_timetolive = 0;
3221 sp->default_rcv_context = 0;
3222 sp->max_burst = sctp_max_burst;
3224 /* Initialize default setup parameters. These parameters
3225 * can be modified with the SCTP_INITMSG socket option or
3226 * overridden by the SCTP_INIT CMSG.
3228 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3229 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3230 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3231 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3233 /* Initialize default RTO related parameters. These parameters can
3234 * be modified for with the SCTP_RTOINFO socket option.
3236 sp->rtoinfo.srto_initial = sctp_rto_initial;
3237 sp->rtoinfo.srto_max = sctp_rto_max;
3238 sp->rtoinfo.srto_min = sctp_rto_min;
3240 /* Initialize default association related parameters. These parameters
3241 * can be modified with the SCTP_ASSOCINFO socket option.
3243 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3244 sp->assocparams.sasoc_number_peer_destinations = 0;
3245 sp->assocparams.sasoc_peer_rwnd = 0;
3246 sp->assocparams.sasoc_local_rwnd = 0;
3247 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3249 /* Initialize default event subscriptions. By default, all the
3252 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3254 /* Default Peer Address Parameters. These defaults can
3255 * be modified via SCTP_PEER_ADDR_PARAMS
3257 sp->hbinterval = sctp_hb_interval;
3258 sp->pathmaxrxt = sctp_max_retrans_path;
3259 sp->pathmtu = 0; // allow default discovery
3260 sp->sackdelay = sctp_sack_timeout;
3261 sp->param_flags = SPP_HB_ENABLE |
3263 SPP_SACKDELAY_ENABLE;
3265 /* If enabled no SCTP message fragmentation will be performed.
3266 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3268 sp->disable_fragments = 0;
3270 /* Enable Nagle algorithm by default. */
3273 /* Enable by default. */
3276 /* Auto-close idle associations after the configured
3277 * number of seconds. A value of 0 disables this
3278 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3279 * for UDP-style sockets only.
3283 /* User specified fragmentation limit. */
3286 sp->adaptation_ind = 0;
3288 sp->pf = sctp_get_pf_specific(sk->sk_family);
3290 /* Control variables for partial data delivery. */
3291 atomic_set(&sp->pd_mode, 0);
3292 skb_queue_head_init(&sp->pd_lobby);
3293 sp->frag_interleave = 0;
3295 /* Create a per socket endpoint structure. Even if we
3296 * change the data structure relationships, this may still
3297 * be useful for storing pre-connect address information.
3299 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3306 SCTP_DBG_OBJCNT_INC(sock);
3310 /* Cleanup any SCTP per socket resources. */
3311 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3313 struct sctp_endpoint *ep;
3315 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3317 /* Release our hold on the endpoint. */
3318 ep = sctp_sk(sk)->ep;
3319 sctp_endpoint_free(ep);
3324 /* API 4.1.7 shutdown() - TCP Style Syntax
3325 * int shutdown(int socket, int how);
3327 * sd - the socket descriptor of the association to be closed.
3328 * how - Specifies the type of shutdown. The values are
3331 * Disables further receive operations. No SCTP
3332 * protocol action is taken.
3334 * Disables further send operations, and initiates
3335 * the SCTP shutdown sequence.
3337 * Disables further send and receive operations
3338 * and initiates the SCTP shutdown sequence.
3340 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3342 struct sctp_endpoint *ep;
3343 struct sctp_association *asoc;
3345 if (!sctp_style(sk, TCP))
3348 if (how & SEND_SHUTDOWN) {
3349 ep = sctp_sk(sk)->ep;
3350 if (!list_empty(&ep->asocs)) {
3351 asoc = list_entry(ep->asocs.next,
3352 struct sctp_association, asocs);
3353 sctp_primitive_SHUTDOWN(asoc, NULL);
3358 /* 7.2.1 Association Status (SCTP_STATUS)
3360 * Applications can retrieve current status information about an
3361 * association, including association state, peer receiver window size,
3362 * number of unacked data chunks, and number of data chunks pending
3363 * receipt. This information is read-only.
3365 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3366 char __user *optval,
3369 struct sctp_status status;
3370 struct sctp_association *asoc = NULL;
3371 struct sctp_transport *transport;
3372 sctp_assoc_t associd;
3375 if (len != sizeof(status)) {
3380 if (copy_from_user(&status, optval, sizeof(status))) {
3385 associd = status.sstat_assoc_id;
3386 asoc = sctp_id2assoc(sk, associd);
3392 transport = asoc->peer.primary_path;
3394 status.sstat_assoc_id = sctp_assoc2id(asoc);
3395 status.sstat_state = asoc->state;
3396 status.sstat_rwnd = asoc->peer.rwnd;
3397 status.sstat_unackdata = asoc->unack_data;
3399 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3400 status.sstat_instrms = asoc->c.sinit_max_instreams;
3401 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3402 status.sstat_fragmentation_point = asoc->frag_point;
3403 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3404 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3405 transport->af_specific->sockaddr_len);
3406 /* Map ipv4 address into v4-mapped-on-v6 address. */
3407 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3408 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3409 status.sstat_primary.spinfo_state = transport->state;
3410 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3411 status.sstat_primary.spinfo_srtt = transport->srtt;
3412 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3413 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3415 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3416 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3418 if (put_user(len, optlen)) {
3423 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3424 len, status.sstat_state, status.sstat_rwnd,
3425 status.sstat_assoc_id);
3427 if (copy_to_user(optval, &status, len)) {
3437 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3439 * Applications can retrieve information about a specific peer address
3440 * of an association, including its reachability state, congestion
3441 * window, and retransmission timer values. This information is
3444 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3445 char __user *optval,
3448 struct sctp_paddrinfo pinfo;
3449 struct sctp_transport *transport;
3452 if (len != sizeof(pinfo)) {
3457 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3462 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3463 pinfo.spinfo_assoc_id);
3467 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3468 pinfo.spinfo_state = transport->state;
3469 pinfo.spinfo_cwnd = transport->cwnd;
3470 pinfo.spinfo_srtt = transport->srtt;
3471 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3472 pinfo.spinfo_mtu = transport->pathmtu;
3474 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3475 pinfo.spinfo_state = SCTP_ACTIVE;
3477 if (put_user(len, optlen)) {
3482 if (copy_to_user(optval, &pinfo, len)) {
3491 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3493 * This option is a on/off flag. If enabled no SCTP message
3494 * fragmentation will be performed. Instead if a message being sent
3495 * exceeds the current PMTU size, the message will NOT be sent and
3496 * instead a error will be indicated to the user.
3498 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3499 char __user *optval, int __user *optlen)
3503 if (len < sizeof(int))
3507 val = (sctp_sk(sk)->disable_fragments == 1);
3508 if (put_user(len, optlen))
3510 if (copy_to_user(optval, &val, len))
3515 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3517 * This socket option is used to specify various notifications and
3518 * ancillary data the user wishes to receive.
3520 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3523 if (len != sizeof(struct sctp_event_subscribe))
3525 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3530 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3532 * This socket option is applicable to the UDP-style socket only. When
3533 * set it will cause associations that are idle for more than the
3534 * specified number of seconds to automatically close. An association
3535 * being idle is defined an association that has NOT sent or received
3536 * user data. The special value of '0' indicates that no automatic
3537 * close of any associations should be performed. The option expects an
3538 * integer defining the number of seconds of idle time before an
3539 * association is closed.
3541 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3543 /* Applicable to UDP-style socket only */
3544 if (sctp_style(sk, TCP))
3546 if (len != sizeof(int))
3548 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3553 /* Helper routine to branch off an association to a new socket. */
3554 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3555 struct socket **sockp)
3557 struct sock *sk = asoc->base.sk;
3558 struct socket *sock;
3559 struct inet_sock *inetsk;
3563 /* An association cannot be branched off from an already peeled-off
3564 * socket, nor is this supported for tcp style sockets.
3566 if (!sctp_style(sk, UDP))
3569 /* Create a new socket. */
3570 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3574 /* Populate the fields of the newsk from the oldsk and migrate the
3575 * asoc to the newsk.
3577 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3579 /* Make peeled-off sockets more like 1-1 accepted sockets.
3580 * Set the daddr and initialize id to something more random
3582 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3583 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3584 inetsk = inet_sk(sock->sk);
3585 inetsk->id = asoc->next_tsn ^ jiffies;
3592 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3594 sctp_peeloff_arg_t peeloff;
3595 struct socket *newsock;
3597 struct sctp_association *asoc;
3599 if (len != sizeof(sctp_peeloff_arg_t))
3601 if (copy_from_user(&peeloff, optval, len))
3604 asoc = sctp_id2assoc(sk, peeloff.associd);
3610 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3612 retval = sctp_do_peeloff(asoc, &newsock);
3616 /* Map the socket to an unused fd that can be returned to the user. */
3617 retval = sock_map_fd(newsock);
3619 sock_release(newsock);
3623 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3624 __FUNCTION__, sk, asoc, newsock->sk, retval);
3626 /* Return the fd mapped to the new socket. */
3627 peeloff.sd = retval;
3628 if (copy_to_user(optval, &peeloff, len))
3635 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3637 * Applications can enable or disable heartbeats for any peer address of
3638 * an association, modify an address's heartbeat interval, force a
3639 * heartbeat to be sent immediately, and adjust the address's maximum
3640 * number of retransmissions sent before an address is considered
3641 * unreachable. The following structure is used to access and modify an
3642 * address's parameters:
3644 * struct sctp_paddrparams {
3645 * sctp_assoc_t spp_assoc_id;
3646 * struct sockaddr_storage spp_address;
3647 * uint32_t spp_hbinterval;
3648 * uint16_t spp_pathmaxrxt;
3649 * uint32_t spp_pathmtu;
3650 * uint32_t spp_sackdelay;
3651 * uint32_t spp_flags;
3654 * spp_assoc_id - (one-to-many style socket) This is filled in the
3655 * application, and identifies the association for
3657 * spp_address - This specifies which address is of interest.
3658 * spp_hbinterval - This contains the value of the heartbeat interval,
3659 * in milliseconds. If a value of zero
3660 * is present in this field then no changes are to
3661 * be made to this parameter.
3662 * spp_pathmaxrxt - This contains the maximum number of
3663 * retransmissions before this address shall be
3664 * considered unreachable. If a value of zero
3665 * is present in this field then no changes are to
3666 * be made to this parameter.
3667 * spp_pathmtu - When Path MTU discovery is disabled the value
3668 * specified here will be the "fixed" path mtu.
3669 * Note that if the spp_address field is empty
3670 * then all associations on this address will
3671 * have this fixed path mtu set upon them.
3673 * spp_sackdelay - When delayed sack is enabled, this value specifies
3674 * the number of milliseconds that sacks will be delayed
3675 * for. This value will apply to all addresses of an
3676 * association if the spp_address field is empty. Note
3677 * also, that if delayed sack is enabled and this
3678 * value is set to 0, no change is made to the last
3679 * recorded delayed sack timer value.
3681 * spp_flags - These flags are used to control various features
3682 * on an association. The flag field may contain
3683 * zero or more of the following options.
3685 * SPP_HB_ENABLE - Enable heartbeats on the
3686 * specified address. Note that if the address
3687 * field is empty all addresses for the association
3688 * have heartbeats enabled upon them.
3690 * SPP_HB_DISABLE - Disable heartbeats on the
3691 * speicifed address. Note that if the address
3692 * field is empty all addresses for the association
3693 * will have their heartbeats disabled. Note also
3694 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3695 * mutually exclusive, only one of these two should
3696 * be specified. Enabling both fields will have
3697 * undetermined results.
3699 * SPP_HB_DEMAND - Request a user initiated heartbeat
3700 * to be made immediately.
3702 * SPP_PMTUD_ENABLE - This field will enable PMTU
3703 * discovery upon the specified address. Note that
3704 * if the address feild is empty then all addresses
3705 * on the association are effected.
3707 * SPP_PMTUD_DISABLE - This field will disable PMTU
3708 * discovery upon the specified address. Note that
3709 * if the address feild is empty then all addresses
3710 * on the association are effected. Not also that
3711 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3712 * exclusive. Enabling both will have undetermined
3715 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3716 * on delayed sack. The time specified in spp_sackdelay
3717 * is used to specify the sack delay for this address. Note
3718 * that if spp_address is empty then all addresses will
3719 * enable delayed sack and take on the sack delay
3720 * value specified in spp_sackdelay.
3721 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3722 * off delayed sack. If the spp_address field is blank then
3723 * delayed sack is disabled for the entire association. Note
3724 * also that this field is mutually exclusive to
3725 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3728 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3729 char __user *optval, int __user *optlen)
3731 struct sctp_paddrparams params;
3732 struct sctp_transport *trans = NULL;
3733 struct sctp_association *asoc = NULL;
3734 struct sctp_sock *sp = sctp_sk(sk);
3736 if (len != sizeof(struct sctp_paddrparams))
3739 if (copy_from_user(¶ms, optval, len))
3742 /* If an address other than INADDR_ANY is specified, and
3743 * no transport is found, then the request is invalid.
3745 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3746 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3747 params.spp_assoc_id);
3749 SCTP_DEBUG_PRINTK("Failed no transport\n");
3754 /* Get association, if assoc_id != 0 and the socket is a one
3755 * to many style socket, and an association was not found, then
3756 * the id was invalid.
3758 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3759 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3760 SCTP_DEBUG_PRINTK("Failed no association\n");
3765 /* Fetch transport values. */
3766 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3767 params.spp_pathmtu = trans->pathmtu;
3768 params.spp_pathmaxrxt = trans->pathmaxrxt;
3769 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3771 /*draft-11 doesn't say what to return in spp_flags*/
3772 params.spp_flags = trans->param_flags;
3774 /* Fetch association values. */
3775 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3776 params.spp_pathmtu = asoc->pathmtu;
3777 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3778 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3780 /*draft-11 doesn't say what to return in spp_flags*/
3781 params.spp_flags = asoc->param_flags;
3783 /* Fetch socket values. */
3784 params.spp_hbinterval = sp->hbinterval;
3785 params.spp_pathmtu = sp->pathmtu;
3786 params.spp_sackdelay = sp->sackdelay;
3787 params.spp_pathmaxrxt = sp->pathmaxrxt;
3789 /*draft-11 doesn't say what to return in spp_flags*/
3790 params.spp_flags = sp->param_flags;
3793 if (copy_to_user(optval, ¶ms, len))
3796 if (put_user(len, optlen))
3802 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3804 * This options will get or set the delayed ack timer. The time is set
3805 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3806 * endpoints default delayed ack timer value. If the assoc_id field is
3807 * non-zero, then the set or get effects the specified association.
3809 * struct sctp_assoc_value {
3810 * sctp_assoc_t assoc_id;
3811 * uint32_t assoc_value;
3814 * assoc_id - This parameter, indicates which association the
3815 * user is preforming an action upon. Note that if
3816 * this field's value is zero then the endpoints
3817 * default value is changed (effecting future
3818 * associations only).
3820 * assoc_value - This parameter contains the number of milliseconds
3821 * that the user is requesting the delayed ACK timer
3822 * be set to. Note that this value is defined in
3823 * the standard to be between 200 and 500 milliseconds.
3825 * Note: a value of zero will leave the value alone,
3826 * but disable SACK delay. A non-zero value will also
3827 * enable SACK delay.
3829 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3830 char __user *optval,
3833 struct sctp_assoc_value params;
3834 struct sctp_association *asoc = NULL;
3835 struct sctp_sock *sp = sctp_sk(sk);
3837 if (len != sizeof(struct sctp_assoc_value))
3840 if (copy_from_user(¶ms, optval, len))
3843 /* Get association, if assoc_id != 0 and the socket is a one
3844 * to many style socket, and an association was not found, then
3845 * the id was invalid.
3847 asoc = sctp_id2assoc(sk, params.assoc_id);
3848 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3852 /* Fetch association values. */
3853 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3854 params.assoc_value = jiffies_to_msecs(
3857 params.assoc_value = 0;
3859 /* Fetch socket values. */
3860 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3861 params.assoc_value = sp->sackdelay;
3863 params.assoc_value = 0;
3866 if (copy_to_user(optval, ¶ms, len))
3869 if (put_user(len, optlen))
3875 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3877 * Applications can specify protocol parameters for the default association
3878 * initialization. The option name argument to setsockopt() and getsockopt()
3881 * Setting initialization parameters is effective only on an unconnected
3882 * socket (for UDP-style sockets only future associations are effected
3883 * by the change). With TCP-style sockets, this option is inherited by
3884 * sockets derived from a listener socket.
3886 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3888 if (len != sizeof(struct sctp_initmsg))
3890 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3895 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3896 char __user *optval,
3900 struct sctp_association *asoc;
3901 struct list_head *pos;
3904 if (len != sizeof(sctp_assoc_t))
3907 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3910 /* For UDP-style sockets, id specifies the association to query. */
3911 asoc = sctp_id2assoc(sk, id);
3915 list_for_each(pos, &asoc->peer.transport_addr_list) {
3923 * Old API for getting list of peer addresses. Does not work for 32-bit
3924 * programs running on a 64-bit kernel
3926 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3927 char __user *optval,
3930 struct sctp_association *asoc;
3931 struct list_head *pos;
3933 struct sctp_getaddrs_old getaddrs;
3934 struct sctp_transport *from;
3936 union sctp_addr temp;
3937 struct sctp_sock *sp = sctp_sk(sk);
3940 if (len != sizeof(struct sctp_getaddrs_old))
3943 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3946 if (getaddrs.addr_num <= 0) return -EINVAL;
3948 /* For UDP-style sockets, id specifies the association to query. */
3949 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3953 to = (void __user *)getaddrs.addrs;
3954 list_for_each(pos, &asoc->peer.transport_addr_list) {
3955 from = list_entry(pos, struct sctp_transport, transports);
3956 memcpy(&temp, &from->ipaddr, sizeof(temp));
3957 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3958 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3959 if (copy_to_user(to, &temp, addrlen))
3963 if (cnt >= getaddrs.addr_num) break;
3965 getaddrs.addr_num = cnt;
3966 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3972 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3973 char __user *optval, int __user *optlen)
3975 struct sctp_association *asoc;
3976 struct list_head *pos;
3978 struct sctp_getaddrs getaddrs;
3979 struct sctp_transport *from;
3981 union sctp_addr temp;
3982 struct sctp_sock *sp = sctp_sk(sk);
3987 if (len < sizeof(struct sctp_getaddrs))
3990 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3993 /* For UDP-style sockets, id specifies the association to query. */
3994 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3998 to = optval + offsetof(struct sctp_getaddrs,addrs);
3999 space_left = len - sizeof(struct sctp_getaddrs) -
4000 offsetof(struct sctp_getaddrs,addrs);
4002 list_for_each(pos, &asoc->peer.transport_addr_list) {
4003 from = list_entry(pos, struct sctp_transport, transports);
4004 memcpy(&temp, &from->ipaddr, sizeof(temp));
4005 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4006 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4007 if (space_left < addrlen)
4009 if (copy_to_user(to, &temp, addrlen))
4013 space_left -= addrlen;
4016 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4018 bytes_copied = ((char __user *)to) - optval;
4019 if (put_user(bytes_copied, optlen))
4025 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4026 char __user *optval,
4030 struct sctp_bind_addr *bp;
4031 struct sctp_association *asoc;
4032 struct list_head *pos, *temp;
4033 struct sctp_sockaddr_entry *addr;
4034 rwlock_t *addr_lock;
4037 if (len != sizeof(sctp_assoc_t))
4040 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4044 * For UDP-style sockets, id specifies the association to query.
4045 * If the id field is set to the value '0' then the locally bound
4046 * addresses are returned without regard to any particular
4050 bp = &sctp_sk(sk)->ep->base.bind_addr;
4051 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4053 asoc = sctp_id2assoc(sk, id);
4056 bp = &asoc->base.bind_addr;
4057 addr_lock = &asoc->base.addr_lock;
4060 sctp_read_lock(addr_lock);
4062 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4063 * addresses from the global local address list.
4065 if (sctp_list_single_entry(&bp->address_list)) {
4066 addr = list_entry(bp->address_list.next,
4067 struct sctp_sockaddr_entry, list);
4068 if (sctp_is_any(&addr->a)) {
4069 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
4070 addr = list_entry(pos,
4071 struct sctp_sockaddr_entry,
4073 if ((PF_INET == sk->sk_family) &&
4074 (AF_INET6 == addr->a.sa.sa_family))
4084 list_for_each(pos, &bp->address_list) {
4089 sctp_read_unlock(addr_lock);
4093 /* Helper function that copies local addresses to user and returns the number
4094 * of addresses copied.
4096 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4097 int max_addrs, void *to,
4100 struct list_head *pos, *next;
4101 struct sctp_sockaddr_entry *addr;
4102 union sctp_addr temp;
4106 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4107 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4108 if ((PF_INET == sk->sk_family) &&
4109 (AF_INET6 == addr->a.sa.sa_family))
4111 memcpy(&temp, &addr->a, sizeof(temp));
4112 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4114 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4115 memcpy(to, &temp, addrlen);
4118 *bytes_copied += addrlen;
4120 if (cnt >= max_addrs) break;
4126 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4127 size_t space_left, int *bytes_copied)
4129 struct list_head *pos, *next;
4130 struct sctp_sockaddr_entry *addr;
4131 union sctp_addr temp;
4135 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4136 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4137 if ((PF_INET == sk->sk_family) &&
4138 (AF_INET6 == addr->a.sa.sa_family))
4140 memcpy(&temp, &addr->a, sizeof(temp));
4141 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4143 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4144 if (space_left < addrlen)
4146 memcpy(to, &temp, addrlen);
4150 space_left -= addrlen;
4151 bytes_copied += addrlen;
4157 /* Old API for getting list of local addresses. Does not work for 32-bit
4158 * programs running on a 64-bit kernel
4160 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4161 char __user *optval, int __user *optlen)
4163 struct sctp_bind_addr *bp;
4164 struct sctp_association *asoc;
4165 struct list_head *pos;
4167 struct sctp_getaddrs_old getaddrs;
4168 struct sctp_sockaddr_entry *addr;
4170 union sctp_addr temp;
4171 struct sctp_sock *sp = sctp_sk(sk);
4173 rwlock_t *addr_lock;
4177 int bytes_copied = 0;
4179 if (len != sizeof(struct sctp_getaddrs_old))
4182 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
4185 if (getaddrs.addr_num <= 0) return -EINVAL;
4187 * For UDP-style sockets, id specifies the association to query.
4188 * If the id field is set to the value '0' then the locally bound
4189 * addresses are returned without regard to any particular
4192 if (0 == getaddrs.assoc_id) {
4193 bp = &sctp_sk(sk)->ep->base.bind_addr;
4194 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4196 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4199 bp = &asoc->base.bind_addr;
4200 addr_lock = &asoc->base.addr_lock;
4203 to = getaddrs.addrs;
4205 /* Allocate space for a local instance of packed array to hold all
4206 * the data. We store addresses here first and then put write them
4207 * to the user in one shot.
4209 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4214 sctp_read_lock(addr_lock);
4216 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4217 * addresses from the global local address list.
4219 if (sctp_list_single_entry(&bp->address_list)) {
4220 addr = list_entry(bp->address_list.next,
4221 struct sctp_sockaddr_entry, list);
4222 if (sctp_is_any(&addr->a)) {
4223 cnt = sctp_copy_laddrs_old(sk, bp->port,
4225 addrs, &bytes_copied);
4231 list_for_each(pos, &bp->address_list) {
4232 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4233 memcpy(&temp, &addr->a, sizeof(temp));
4234 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4235 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4236 memcpy(buf, &temp, addrlen);
4238 bytes_copied += addrlen;
4240 if (cnt >= getaddrs.addr_num) break;
4244 sctp_read_unlock(addr_lock);
4246 /* copy the entire address list into the user provided space */
4247 if (copy_to_user(to, addrs, bytes_copied)) {
4252 /* copy the leading structure back to user */
4253 getaddrs.addr_num = cnt;
4254 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4262 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4263 char __user *optval, int __user *optlen)
4265 struct sctp_bind_addr *bp;
4266 struct sctp_association *asoc;
4267 struct list_head *pos;
4269 struct sctp_getaddrs getaddrs;
4270 struct sctp_sockaddr_entry *addr;
4272 union sctp_addr temp;
4273 struct sctp_sock *sp = sctp_sk(sk);
4275 rwlock_t *addr_lock;
4278 int bytes_copied = 0;
4282 if (len <= sizeof(struct sctp_getaddrs))
4285 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4289 * For UDP-style sockets, id specifies the association to query.
4290 * If the id field is set to the value '0' then the locally bound
4291 * addresses are returned without regard to any particular
4294 if (0 == getaddrs.assoc_id) {
4295 bp = &sctp_sk(sk)->ep->base.bind_addr;
4296 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4298 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4301 bp = &asoc->base.bind_addr;
4302 addr_lock = &asoc->base.addr_lock;
4305 to = optval + offsetof(struct sctp_getaddrs,addrs);
4306 space_left = len - sizeof(struct sctp_getaddrs) -
4307 offsetof(struct sctp_getaddrs,addrs);
4308 addrs = kmalloc(space_left, GFP_KERNEL);
4312 sctp_read_lock(addr_lock);
4314 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4315 * addresses from the global local address list.
4317 if (sctp_list_single_entry(&bp->address_list)) {
4318 addr = list_entry(bp->address_list.next,
4319 struct sctp_sockaddr_entry, list);
4320 if (sctp_is_any(&addr->a)) {
4321 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4322 space_left, &bytes_copied);
4332 list_for_each(pos, &bp->address_list) {
4333 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4334 memcpy(&temp, &addr->a, sizeof(temp));
4335 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4336 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4337 if (space_left < addrlen) {
4338 err = -ENOMEM; /*fixme: right error?*/
4341 memcpy(buf, &temp, addrlen);
4343 bytes_copied += addrlen;
4345 space_left -= addrlen;
4349 sctp_read_unlock(addr_lock);
4351 if (copy_to_user(to, addrs, bytes_copied)) {
4355 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4359 if (put_user(bytes_copied, optlen))
4366 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4368 * Requests that the local SCTP stack use the enclosed peer address as
4369 * the association primary. The enclosed address must be one of the
4370 * association peer's addresses.
4372 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4373 char __user *optval, int __user *optlen)
4375 struct sctp_prim prim;
4376 struct sctp_association *asoc;
4377 struct sctp_sock *sp = sctp_sk(sk);
4379 if (len != sizeof(struct sctp_prim))
4382 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4385 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4389 if (!asoc->peer.primary_path)
4392 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4393 asoc->peer.primary_path->af_specific->sockaddr_len);
4395 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4396 (union sctp_addr *)&prim.ssp_addr);
4398 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4405 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4407 * Requests that the local endpoint set the specified Adaptation Layer
4408 * Indication parameter for all future INIT and INIT-ACK exchanges.
4410 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4411 char __user *optval, int __user *optlen)
4413 struct sctp_setadaptation adaptation;
4415 if (len != sizeof(struct sctp_setadaptation))
4418 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4419 if (copy_to_user(optval, &adaptation, len))
4427 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4429 * Applications that wish to use the sendto() system call may wish to
4430 * specify a default set of parameters that would normally be supplied
4431 * through the inclusion of ancillary data. This socket option allows
4432 * such an application to set the default sctp_sndrcvinfo structure.
4435 * The application that wishes to use this socket option simply passes
4436 * in to this call the sctp_sndrcvinfo structure defined in Section
4437 * 5.2.2) The input parameters accepted by this call include
4438 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4439 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4440 * to this call if the caller is using the UDP model.
4442 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4444 static int sctp_getsockopt_default_send_param(struct sock *sk,
4445 int len, char __user *optval,
4448 struct sctp_sndrcvinfo info;
4449 struct sctp_association *asoc;
4450 struct sctp_sock *sp = sctp_sk(sk);
4452 if (len != sizeof(struct sctp_sndrcvinfo))
4454 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4457 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4458 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4462 info.sinfo_stream = asoc->default_stream;
4463 info.sinfo_flags = asoc->default_flags;
4464 info.sinfo_ppid = asoc->default_ppid;
4465 info.sinfo_context = asoc->default_context;
4466 info.sinfo_timetolive = asoc->default_timetolive;
4468 info.sinfo_stream = sp->default_stream;
4469 info.sinfo_flags = sp->default_flags;
4470 info.sinfo_ppid = sp->default_ppid;
4471 info.sinfo_context = sp->default_context;
4472 info.sinfo_timetolive = sp->default_timetolive;
4475 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4483 * 7.1.5 SCTP_NODELAY
4485 * Turn on/off any Nagle-like algorithm. This means that packets are
4486 * generally sent as soon as possible and no unnecessary delays are
4487 * introduced, at the cost of more packets in the network. Expects an
4488 * integer boolean flag.
4491 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4492 char __user *optval, int __user *optlen)
4496 if (len < sizeof(int))
4500 val = (sctp_sk(sk)->nodelay == 1);
4501 if (put_user(len, optlen))
4503 if (copy_to_user(optval, &val, len))
4510 * 7.1.1 SCTP_RTOINFO
4512 * The protocol parameters used to initialize and bound retransmission
4513 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4514 * and modify these parameters.
4515 * All parameters are time values, in milliseconds. A value of 0, when
4516 * modifying the parameters, indicates that the current value should not
4520 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4521 char __user *optval,
4522 int __user *optlen) {
4523 struct sctp_rtoinfo rtoinfo;
4524 struct sctp_association *asoc;
4526 if (len != sizeof (struct sctp_rtoinfo))
4529 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4532 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4534 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4537 /* Values corresponding to the specific association. */
4539 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4540 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4541 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4543 /* Values corresponding to the endpoint. */
4544 struct sctp_sock *sp = sctp_sk(sk);
4546 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4547 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4548 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4551 if (put_user(len, optlen))
4554 if (copy_to_user(optval, &rtoinfo, len))
4562 * 7.1.2 SCTP_ASSOCINFO
4564 * This option is used to tune the maximum retransmission attempts
4565 * of the association.
4566 * Returns an error if the new association retransmission value is
4567 * greater than the sum of the retransmission value of the peer.
4568 * See [SCTP] for more information.
4571 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4572 char __user *optval,
4576 struct sctp_assocparams assocparams;
4577 struct sctp_association *asoc;
4578 struct list_head *pos;
4581 if (len != sizeof (struct sctp_assocparams))
4584 if (copy_from_user(&assocparams, optval,
4585 sizeof (struct sctp_assocparams)))
4588 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4590 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4593 /* Values correspoinding to the specific association */
4595 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4596 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4597 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4598 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4600 (asoc->cookie_life.tv_usec
4603 list_for_each(pos, &asoc->peer.transport_addr_list) {
4607 assocparams.sasoc_number_peer_destinations = cnt;
4609 /* Values corresponding to the endpoint */
4610 struct sctp_sock *sp = sctp_sk(sk);
4612 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4613 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4614 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4615 assocparams.sasoc_cookie_life =
4616 sp->assocparams.sasoc_cookie_life;
4617 assocparams.sasoc_number_peer_destinations =
4619 sasoc_number_peer_destinations;
4622 if (put_user(len, optlen))
4625 if (copy_to_user(optval, &assocparams, len))
4632 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4634 * This socket option is a boolean flag which turns on or off mapped V4
4635 * addresses. If this option is turned on and the socket is type
4636 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4637 * If this option is turned off, then no mapping will be done of V4
4638 * addresses and a user will receive both PF_INET6 and PF_INET type
4639 * addresses on the socket.
4641 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4642 char __user *optval, int __user *optlen)
4645 struct sctp_sock *sp = sctp_sk(sk);
4647 if (len < sizeof(int))
4652 if (put_user(len, optlen))
4654 if (copy_to_user(optval, &val, len))
4661 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4662 * (chapter and verse is quoted at sctp_setsockopt_context())
4664 static int sctp_getsockopt_context(struct sock *sk, int len,
4665 char __user *optval, int __user *optlen)
4667 struct sctp_assoc_value params;
4668 struct sctp_sock *sp;
4669 struct sctp_association *asoc;
4671 if (len != sizeof(struct sctp_assoc_value))
4674 if (copy_from_user(¶ms, optval, len))
4679 if (params.assoc_id != 0) {
4680 asoc = sctp_id2assoc(sk, params.assoc_id);
4683 params.assoc_value = asoc->default_rcv_context;
4685 params.assoc_value = sp->default_rcv_context;
4688 if (put_user(len, optlen))
4690 if (copy_to_user(optval, ¶ms, len))
4697 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4699 * This socket option specifies the maximum size to put in any outgoing
4700 * SCTP chunk. If a message is larger than this size it will be
4701 * fragmented by SCTP into the specified size. Note that the underlying
4702 * SCTP implementation may fragment into smaller sized chunks when the
4703 * PMTU of the underlying association is smaller than the value set by
4706 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4707 char __user *optval, int __user *optlen)
4711 if (len < sizeof(int))
4716 val = sctp_sk(sk)->user_frag;
4717 if (put_user(len, optlen))
4719 if (copy_to_user(optval, &val, len))
4726 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4727 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4729 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4730 char __user *optval, int __user *optlen)
4734 if (len < sizeof(int))
4739 val = sctp_sk(sk)->frag_interleave;
4740 if (put_user(len, optlen))
4742 if (copy_to_user(optval, &val, len))
4749 * 7.1.25. Set or Get the sctp partial delivery point
4750 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4752 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4753 char __user *optval,
4758 if (len < sizeof(u32))
4763 val = sctp_sk(sk)->pd_point;
4764 if (put_user(len, optlen))
4766 if (copy_to_user(optval, &val, len))
4773 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
4774 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
4776 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
4777 char __user *optval,
4782 if (len < sizeof(int))
4787 val = sctp_sk(sk)->max_burst;
4788 if (put_user(len, optlen))
4790 if (copy_to_user(optval, &val, len))
4796 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4797 char __user *optval, int __user *optlen)
4802 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4805 /* I can hardly begin to describe how wrong this is. This is
4806 * so broken as to be worse than useless. The API draft
4807 * REALLY is NOT helpful here... I am not convinced that the
4808 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4809 * are at all well-founded.
4811 if (level != SOL_SCTP) {
4812 struct sctp_af *af = sctp_sk(sk)->pf->af;
4814 retval = af->getsockopt(sk, level, optname, optval, optlen);
4818 if (get_user(len, optlen))
4825 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4827 case SCTP_DISABLE_FRAGMENTS:
4828 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4832 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4834 case SCTP_AUTOCLOSE:
4835 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4837 case SCTP_SOCKOPT_PEELOFF:
4838 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4840 case SCTP_PEER_ADDR_PARAMS:
4841 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4844 case SCTP_DELAYED_ACK_TIME:
4845 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4849 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4851 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4852 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4855 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4856 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4859 case SCTP_GET_PEER_ADDRS_OLD:
4860 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4863 case SCTP_GET_LOCAL_ADDRS_OLD:
4864 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4867 case SCTP_GET_PEER_ADDRS:
4868 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4871 case SCTP_GET_LOCAL_ADDRS:
4872 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4875 case SCTP_DEFAULT_SEND_PARAM:
4876 retval = sctp_getsockopt_default_send_param(sk, len,
4879 case SCTP_PRIMARY_ADDR:
4880 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4883 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4886 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4888 case SCTP_ASSOCINFO:
4889 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4891 case SCTP_I_WANT_MAPPED_V4_ADDR:
4892 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4895 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4897 case SCTP_GET_PEER_ADDR_INFO:
4898 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4901 case SCTP_ADAPTATION_LAYER:
4902 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4906 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4908 case SCTP_FRAGMENT_INTERLEAVE:
4909 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4912 case SCTP_PARTIAL_DELIVERY_POINT:
4913 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
4916 case SCTP_MAX_BURST:
4917 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
4920 retval = -ENOPROTOOPT;
4924 sctp_release_sock(sk);
4928 static void sctp_hash(struct sock *sk)
4933 static void sctp_unhash(struct sock *sk)
4938 /* Check if port is acceptable. Possibly find first available port.
4940 * The port hash table (contained in the 'global' SCTP protocol storage
4941 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4942 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4943 * list (the list number is the port number hashed out, so as you
4944 * would expect from a hash function, all the ports in a given list have
4945 * such a number that hashes out to the same list number; you were
4946 * expecting that, right?); so each list has a set of ports, with a
4947 * link to the socket (struct sock) that uses it, the port number and
4948 * a fastreuse flag (FIXME: NPI ipg).
4950 static struct sctp_bind_bucket *sctp_bucket_create(
4951 struct sctp_bind_hashbucket *head, unsigned short snum);
4953 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4955 struct sctp_bind_hashbucket *head; /* hash list */
4956 struct sctp_bind_bucket *pp; /* hash list port iterator */
4957 unsigned short snum;
4960 snum = ntohs(addr->v4.sin_port);
4962 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4963 sctp_local_bh_disable();
4966 /* Search for an available port.
4968 * 'sctp_port_rover' was the last port assigned, so
4969 * we start to search from 'sctp_port_rover +
4970 * 1'. What we do is first check if port 'rover' is
4971 * already in the hash table; if not, we use that; if
4972 * it is, we try next.
4974 int low = sysctl_local_port_range[0];
4975 int high = sysctl_local_port_range[1];
4976 int remaining = (high - low) + 1;
4980 sctp_spin_lock(&sctp_port_alloc_lock);
4981 rover = sctp_port_rover;
4984 if ((rover < low) || (rover > high))
4986 index = sctp_phashfn(rover);
4987 head = &sctp_port_hashtable[index];
4988 sctp_spin_lock(&head->lock);
4989 for (pp = head->chain; pp; pp = pp->next)
4990 if (pp->port == rover)
4994 sctp_spin_unlock(&head->lock);
4995 } while (--remaining > 0);
4996 sctp_port_rover = rover;
4997 sctp_spin_unlock(&sctp_port_alloc_lock);
4999 /* Exhausted local port range during search? */
5004 /* OK, here is the one we will use. HEAD (the port
5005 * hash table list entry) is non-NULL and we hold it's
5010 /* We are given an specific port number; we verify
5011 * that it is not being used. If it is used, we will
5012 * exahust the search in the hash list corresponding
5013 * to the port number (snum) - we detect that with the
5014 * port iterator, pp being NULL.
5016 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5017 sctp_spin_lock(&head->lock);
5018 for (pp = head->chain; pp; pp = pp->next) {
5019 if (pp->port == snum)
5026 if (!hlist_empty(&pp->owner)) {
5027 /* We had a port hash table hit - there is an
5028 * available port (pp != NULL) and it is being
5029 * used by other socket (pp->owner not empty); that other
5030 * socket is going to be sk2.
5032 int reuse = sk->sk_reuse;
5034 struct hlist_node *node;
5036 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5037 if (pp->fastreuse && sk->sk_reuse &&
5038 sk->sk_state != SCTP_SS_LISTENING)
5041 /* Run through the list of sockets bound to the port
5042 * (pp->port) [via the pointers bind_next and
5043 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5044 * we get the endpoint they describe and run through
5045 * the endpoint's list of IP (v4 or v6) addresses,
5046 * comparing each of the addresses with the address of
5047 * the socket sk. If we find a match, then that means
5048 * that this port/socket (sk) combination are already
5051 sk_for_each_bound(sk2, node, &pp->owner) {
5052 struct sctp_endpoint *ep2;
5053 ep2 = sctp_sk(sk2)->ep;
5055 if (reuse && sk2->sk_reuse &&
5056 sk2->sk_state != SCTP_SS_LISTENING)
5059 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5065 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5068 /* If there was a hash table miss, create a new port. */
5070 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5073 /* In either case (hit or miss), make sure fastreuse is 1 only
5074 * if sk->sk_reuse is too (that is, if the caller requested
5075 * SO_REUSEADDR on this socket -sk-).
5077 if (hlist_empty(&pp->owner)) {
5078 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5082 } else if (pp->fastreuse &&
5083 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5086 /* We are set, so fill up all the data in the hash table
5087 * entry, tie the socket list information with the rest of the
5088 * sockets FIXME: Blurry, NPI (ipg).
5091 if (!sctp_sk(sk)->bind_hash) {
5092 inet_sk(sk)->num = snum;
5093 sk_add_bind_node(sk, &pp->owner);
5094 sctp_sk(sk)->bind_hash = pp;
5099 sctp_spin_unlock(&head->lock);
5102 sctp_local_bh_enable();
5106 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5107 * port is requested.
5109 static int sctp_get_port(struct sock *sk, unsigned short snum)
5112 union sctp_addr addr;
5113 struct sctp_af *af = sctp_sk(sk)->pf->af;
5115 /* Set up a dummy address struct from the sk. */
5116 af->from_sk(&addr, sk);
5117 addr.v4.sin_port = htons(snum);
5119 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5120 ret = sctp_get_port_local(sk, &addr);
5122 return (ret ? 1 : 0);
5126 * 3.1.3 listen() - UDP Style Syntax
5128 * By default, new associations are not accepted for UDP style sockets.
5129 * An application uses listen() to mark a socket as being able to
5130 * accept new associations.
5132 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5134 struct sctp_sock *sp = sctp_sk(sk);
5135 struct sctp_endpoint *ep = sp->ep;
5137 /* Only UDP style sockets that are not peeled off are allowed to
5140 if (!sctp_style(sk, UDP))
5143 /* If backlog is zero, disable listening. */
5145 if (sctp_sstate(sk, CLOSED))
5148 sctp_unhash_endpoint(ep);
5149 sk->sk_state = SCTP_SS_CLOSED;
5152 /* Return if we are already listening. */
5153 if (sctp_sstate(sk, LISTENING))
5157 * If a bind() or sctp_bindx() is not called prior to a listen()
5158 * call that allows new associations to be accepted, the system
5159 * picks an ephemeral port and will choose an address set equivalent
5160 * to binding with a wildcard address.
5162 * This is not currently spelled out in the SCTP sockets
5163 * extensions draft, but follows the practice as seen in TCP
5166 * Additionally, turn off fastreuse flag since we are not listening
5168 sk->sk_state = SCTP_SS_LISTENING;
5169 if (!ep->base.bind_addr.port) {
5170 if (sctp_autobind(sk))
5173 sctp_sk(sk)->bind_hash->fastreuse = 0;
5175 sctp_hash_endpoint(ep);
5180 * 4.1.3 listen() - TCP Style Syntax
5182 * Applications uses listen() to ready the SCTP endpoint for accepting
5183 * inbound associations.
5185 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5187 struct sctp_sock *sp = sctp_sk(sk);
5188 struct sctp_endpoint *ep = sp->ep;
5190 /* If backlog is zero, disable listening. */
5192 if (sctp_sstate(sk, CLOSED))
5195 sctp_unhash_endpoint(ep);
5196 sk->sk_state = SCTP_SS_CLOSED;
5199 if (sctp_sstate(sk, LISTENING))
5203 * If a bind() or sctp_bindx() is not called prior to a listen()
5204 * call that allows new associations to be accepted, the system
5205 * picks an ephemeral port and will choose an address set equivalent
5206 * to binding with a wildcard address.
5208 * This is not currently spelled out in the SCTP sockets
5209 * extensions draft, but follows the practice as seen in TCP
5212 sk->sk_state = SCTP_SS_LISTENING;
5213 if (!ep->base.bind_addr.port) {
5214 if (sctp_autobind(sk))
5217 sctp_sk(sk)->bind_hash->fastreuse = 0;
5219 sk->sk_max_ack_backlog = backlog;
5220 sctp_hash_endpoint(ep);
5225 * Move a socket to LISTENING state.
5227 int sctp_inet_listen(struct socket *sock, int backlog)
5229 struct sock *sk = sock->sk;
5230 struct crypto_hash *tfm = NULL;
5233 if (unlikely(backlog < 0))
5238 if (sock->state != SS_UNCONNECTED)
5241 /* Allocate HMAC for generating cookie. */
5242 if (sctp_hmac_alg) {
5243 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5245 if (net_ratelimit()) {
5247 "SCTP: failed to load transform for %s: %ld\n",
5248 sctp_hmac_alg, PTR_ERR(tfm));
5255 switch (sock->type) {
5256 case SOCK_SEQPACKET:
5257 err = sctp_seqpacket_listen(sk, backlog);
5260 err = sctp_stream_listen(sk, backlog);
5269 /* Store away the transform reference. */
5270 sctp_sk(sk)->hmac = tfm;
5272 sctp_release_sock(sk);
5275 crypto_free_hash(tfm);
5280 * This function is done by modeling the current datagram_poll() and the
5281 * tcp_poll(). Note that, based on these implementations, we don't
5282 * lock the socket in this function, even though it seems that,
5283 * ideally, locking or some other mechanisms can be used to ensure
5284 * the integrity of the counters (sndbuf and wmem_alloc) used
5285 * in this place. We assume that we don't need locks either until proven
5288 * Another thing to note is that we include the Async I/O support
5289 * here, again, by modeling the current TCP/UDP code. We don't have
5290 * a good way to test with it yet.
5292 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5294 struct sock *sk = sock->sk;
5295 struct sctp_sock *sp = sctp_sk(sk);
5298 poll_wait(file, sk->sk_sleep, wait);
5300 /* A TCP-style listening socket becomes readable when the accept queue
5303 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5304 return (!list_empty(&sp->ep->asocs)) ?
5305 (POLLIN | POLLRDNORM) : 0;
5309 /* Is there any exceptional events? */
5310 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5312 if (sk->sk_shutdown & RCV_SHUTDOWN)
5314 if (sk->sk_shutdown == SHUTDOWN_MASK)
5317 /* Is it readable? Reconsider this code with TCP-style support. */
5318 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5319 (sk->sk_shutdown & RCV_SHUTDOWN))
5320 mask |= POLLIN | POLLRDNORM;
5322 /* The association is either gone or not ready. */
5323 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5326 /* Is it writable? */
5327 if (sctp_writeable(sk)) {
5328 mask |= POLLOUT | POLLWRNORM;
5330 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5332 * Since the socket is not locked, the buffer
5333 * might be made available after the writeable check and
5334 * before the bit is set. This could cause a lost I/O
5335 * signal. tcp_poll() has a race breaker for this race
5336 * condition. Based on their implementation, we put
5337 * in the following code to cover it as well.
5339 if (sctp_writeable(sk))
5340 mask |= POLLOUT | POLLWRNORM;
5345 /********************************************************************
5346 * 2nd Level Abstractions
5347 ********************************************************************/
5349 static struct sctp_bind_bucket *sctp_bucket_create(
5350 struct sctp_bind_hashbucket *head, unsigned short snum)
5352 struct sctp_bind_bucket *pp;
5354 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5355 SCTP_DBG_OBJCNT_INC(bind_bucket);
5359 INIT_HLIST_HEAD(&pp->owner);
5360 if ((pp->next = head->chain) != NULL)
5361 pp->next->pprev = &pp->next;
5363 pp->pprev = &head->chain;
5368 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5369 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5371 if (pp && hlist_empty(&pp->owner)) {
5373 pp->next->pprev = pp->pprev;
5374 *(pp->pprev) = pp->next;
5375 kmem_cache_free(sctp_bucket_cachep, pp);
5376 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5380 /* Release this socket's reference to a local port. */
5381 static inline void __sctp_put_port(struct sock *sk)
5383 struct sctp_bind_hashbucket *head =
5384 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5385 struct sctp_bind_bucket *pp;
5387 sctp_spin_lock(&head->lock);
5388 pp = sctp_sk(sk)->bind_hash;
5389 __sk_del_bind_node(sk);
5390 sctp_sk(sk)->bind_hash = NULL;
5391 inet_sk(sk)->num = 0;
5392 sctp_bucket_destroy(pp);
5393 sctp_spin_unlock(&head->lock);
5396 void sctp_put_port(struct sock *sk)
5398 sctp_local_bh_disable();
5399 __sctp_put_port(sk);
5400 sctp_local_bh_enable();
5404 * The system picks an ephemeral port and choose an address set equivalent
5405 * to binding with a wildcard address.
5406 * One of those addresses will be the primary address for the association.
5407 * This automatically enables the multihoming capability of SCTP.
5409 static int sctp_autobind(struct sock *sk)
5411 union sctp_addr autoaddr;
5415 /* Initialize a local sockaddr structure to INADDR_ANY. */
5416 af = sctp_sk(sk)->pf->af;
5418 port = htons(inet_sk(sk)->num);
5419 af->inaddr_any(&autoaddr, port);
5421 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5424 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5427 * 4.2 The cmsghdr Structure *
5429 * When ancillary data is sent or received, any number of ancillary data
5430 * objects can be specified by the msg_control and msg_controllen members of
5431 * the msghdr structure, because each object is preceded by
5432 * a cmsghdr structure defining the object's length (the cmsg_len member).
5433 * Historically Berkeley-derived implementations have passed only one object
5434 * at a time, but this API allows multiple objects to be
5435 * passed in a single call to sendmsg() or recvmsg(). The following example
5436 * shows two ancillary data objects in a control buffer.
5438 * |<--------------------------- msg_controllen -------------------------->|
5441 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5443 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5446 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5448 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5451 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5452 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5454 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5456 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5463 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5464 sctp_cmsgs_t *cmsgs)
5466 struct cmsghdr *cmsg;
5468 for (cmsg = CMSG_FIRSTHDR(msg);
5470 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5471 if (!CMSG_OK(msg, cmsg))
5474 /* Should we parse this header or ignore? */
5475 if (cmsg->cmsg_level != IPPROTO_SCTP)
5478 /* Strictly check lengths following example in SCM code. */
5479 switch (cmsg->cmsg_type) {
5481 /* SCTP Socket API Extension
5482 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5484 * This cmsghdr structure provides information for
5485 * initializing new SCTP associations with sendmsg().
5486 * The SCTP_INITMSG socket option uses this same data
5487 * structure. This structure is not used for
5490 * cmsg_level cmsg_type cmsg_data[]
5491 * ------------ ------------ ----------------------
5492 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5494 if (cmsg->cmsg_len !=
5495 CMSG_LEN(sizeof(struct sctp_initmsg)))
5497 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5501 /* SCTP Socket API Extension
5502 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5504 * This cmsghdr structure specifies SCTP options for
5505 * sendmsg() and describes SCTP header information
5506 * about a received message through recvmsg().
5508 * cmsg_level cmsg_type cmsg_data[]
5509 * ------------ ------------ ----------------------
5510 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5512 if (cmsg->cmsg_len !=
5513 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5517 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5519 /* Minimally, validate the sinfo_flags. */
5520 if (cmsgs->info->sinfo_flags &
5521 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5522 SCTP_ABORT | SCTP_EOF))
5534 * Wait for a packet..
5535 * Note: This function is the same function as in core/datagram.c
5536 * with a few modifications to make lksctp work.
5538 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5543 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5545 /* Socket errors? */
5546 error = sock_error(sk);
5550 if (!skb_queue_empty(&sk->sk_receive_queue))
5553 /* Socket shut down? */
5554 if (sk->sk_shutdown & RCV_SHUTDOWN)
5557 /* Sequenced packets can come disconnected. If so we report the
5562 /* Is there a good reason to think that we may receive some data? */
5563 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5566 /* Handle signals. */
5567 if (signal_pending(current))
5570 /* Let another process have a go. Since we are going to sleep
5571 * anyway. Note: This may cause odd behaviors if the message
5572 * does not fit in the user's buffer, but this seems to be the
5573 * only way to honor MSG_DONTWAIT realistically.
5575 sctp_release_sock(sk);
5576 *timeo_p = schedule_timeout(*timeo_p);
5580 finish_wait(sk->sk_sleep, &wait);
5584 error = sock_intr_errno(*timeo_p);
5587 finish_wait(sk->sk_sleep, &wait);
5592 /* Receive a datagram.
5593 * Note: This is pretty much the same routine as in core/datagram.c
5594 * with a few changes to make lksctp work.
5596 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5597 int noblock, int *err)
5600 struct sk_buff *skb;
5603 timeo = sock_rcvtimeo(sk, noblock);
5605 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5606 timeo, MAX_SCHEDULE_TIMEOUT);
5609 /* Again only user level code calls this function,
5610 * so nothing interrupt level
5611 * will suddenly eat the receive_queue.
5613 * Look at current nfs client by the way...
5614 * However, this function was corrent in any case. 8)
5616 if (flags & MSG_PEEK) {
5617 spin_lock_bh(&sk->sk_receive_queue.lock);
5618 skb = skb_peek(&sk->sk_receive_queue);
5620 atomic_inc(&skb->users);
5621 spin_unlock_bh(&sk->sk_receive_queue.lock);
5623 skb = skb_dequeue(&sk->sk_receive_queue);
5629 /* Caller is allowed not to check sk->sk_err before calling. */
5630 error = sock_error(sk);
5634 if (sk->sk_shutdown & RCV_SHUTDOWN)
5637 /* User doesn't want to wait. */
5641 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5650 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5651 static void __sctp_write_space(struct sctp_association *asoc)
5653 struct sock *sk = asoc->base.sk;
5654 struct socket *sock = sk->sk_socket;
5656 if ((sctp_wspace(asoc) > 0) && sock) {
5657 if (waitqueue_active(&asoc->wait))
5658 wake_up_interruptible(&asoc->wait);
5660 if (sctp_writeable(sk)) {
5661 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5662 wake_up_interruptible(sk->sk_sleep);
5664 /* Note that we try to include the Async I/O support
5665 * here by modeling from the current TCP/UDP code.
5666 * We have not tested with it yet.
5668 if (sock->fasync_list &&
5669 !(sk->sk_shutdown & SEND_SHUTDOWN))
5670 sock_wake_async(sock, 2, POLL_OUT);
5675 /* Do accounting for the sndbuf space.
5676 * Decrement the used sndbuf space of the corresponding association by the
5677 * data size which was just transmitted(freed).
5679 static void sctp_wfree(struct sk_buff *skb)
5681 struct sctp_association *asoc;
5682 struct sctp_chunk *chunk;
5685 /* Get the saved chunk pointer. */
5686 chunk = *((struct sctp_chunk **)(skb->cb));
5689 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5690 sizeof(struct sk_buff) +
5691 sizeof(struct sctp_chunk);
5693 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5696 __sctp_write_space(asoc);
5698 sctp_association_put(asoc);
5701 /* Do accounting for the receive space on the socket.
5702 * Accounting for the association is done in ulpevent.c
5703 * We set this as a destructor for the cloned data skbs so that
5704 * accounting is done at the correct time.
5706 void sctp_sock_rfree(struct sk_buff *skb)
5708 struct sock *sk = skb->sk;
5709 struct sctp_ulpevent *event = sctp_skb2event(skb);
5711 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5715 /* Helper function to wait for space in the sndbuf. */
5716 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5719 struct sock *sk = asoc->base.sk;
5721 long current_timeo = *timeo_p;
5724 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5725 asoc, (long)(*timeo_p), msg_len);
5727 /* Increment the association's refcnt. */
5728 sctp_association_hold(asoc);
5730 /* Wait on the association specific sndbuf space. */
5732 prepare_to_wait_exclusive(&asoc->wait, &wait,
5733 TASK_INTERRUPTIBLE);
5736 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5739 if (signal_pending(current))
5740 goto do_interrupted;
5741 if (msg_len <= sctp_wspace(asoc))
5744 /* Let another process have a go. Since we are going
5747 sctp_release_sock(sk);
5748 current_timeo = schedule_timeout(current_timeo);
5749 BUG_ON(sk != asoc->base.sk);
5752 *timeo_p = current_timeo;
5756 finish_wait(&asoc->wait, &wait);
5758 /* Release the association's refcnt. */
5759 sctp_association_put(asoc);
5768 err = sock_intr_errno(*timeo_p);
5776 /* If socket sndbuf has changed, wake up all per association waiters. */
5777 void sctp_write_space(struct sock *sk)
5779 struct sctp_association *asoc;
5780 struct list_head *pos;
5782 /* Wake up the tasks in each wait queue. */
5783 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5784 asoc = list_entry(pos, struct sctp_association, asocs);
5785 __sctp_write_space(asoc);
5789 /* Is there any sndbuf space available on the socket?
5791 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5792 * associations on the same socket. For a UDP-style socket with
5793 * multiple associations, it is possible for it to be "unwriteable"
5794 * prematurely. I assume that this is acceptable because
5795 * a premature "unwriteable" is better than an accidental "writeable" which
5796 * would cause an unwanted block under certain circumstances. For the 1-1
5797 * UDP-style sockets or TCP-style sockets, this code should work.
5800 static int sctp_writeable(struct sock *sk)
5804 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5810 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5811 * returns immediately with EINPROGRESS.
5813 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5815 struct sock *sk = asoc->base.sk;
5817 long current_timeo = *timeo_p;
5820 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5823 /* Increment the association's refcnt. */
5824 sctp_association_hold(asoc);
5827 prepare_to_wait_exclusive(&asoc->wait, &wait,
5828 TASK_INTERRUPTIBLE);
5831 if (sk->sk_shutdown & RCV_SHUTDOWN)
5833 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5836 if (signal_pending(current))
5837 goto do_interrupted;
5839 if (sctp_state(asoc, ESTABLISHED))
5842 /* Let another process have a go. Since we are going
5845 sctp_release_sock(sk);
5846 current_timeo = schedule_timeout(current_timeo);
5849 *timeo_p = current_timeo;
5853 finish_wait(&asoc->wait, &wait);
5855 /* Release the association's refcnt. */
5856 sctp_association_put(asoc);
5861 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5864 err = -ECONNREFUSED;
5868 err = sock_intr_errno(*timeo_p);
5876 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5878 struct sctp_endpoint *ep;
5882 ep = sctp_sk(sk)->ep;
5886 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5887 TASK_INTERRUPTIBLE);
5889 if (list_empty(&ep->asocs)) {
5890 sctp_release_sock(sk);
5891 timeo = schedule_timeout(timeo);
5896 if (!sctp_sstate(sk, LISTENING))
5900 if (!list_empty(&ep->asocs))
5903 err = sock_intr_errno(timeo);
5904 if (signal_pending(current))
5912 finish_wait(sk->sk_sleep, &wait);
5917 void sctp_wait_for_close(struct sock *sk, long timeout)
5922 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5923 if (list_empty(&sctp_sk(sk)->ep->asocs))
5925 sctp_release_sock(sk);
5926 timeout = schedule_timeout(timeout);
5928 } while (!signal_pending(current) && timeout);
5930 finish_wait(sk->sk_sleep, &wait);
5933 static void sctp_sock_rfree_frag(struct sk_buff *skb)
5935 struct sk_buff *frag;
5940 /* Don't forget the fragments. */
5941 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5942 sctp_sock_rfree_frag(frag);
5945 sctp_sock_rfree(skb);
5948 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
5950 struct sk_buff *frag;
5955 /* Don't forget the fragments. */
5956 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5957 sctp_skb_set_owner_r_frag(frag, sk);
5960 sctp_skb_set_owner_r(skb, sk);
5963 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5964 * and its messages to the newsk.
5966 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5967 struct sctp_association *assoc,
5968 sctp_socket_type_t type)
5970 struct sctp_sock *oldsp = sctp_sk(oldsk);
5971 struct sctp_sock *newsp = sctp_sk(newsk);
5972 struct sctp_bind_bucket *pp; /* hash list port iterator */
5973 struct sctp_endpoint *newep = newsp->ep;
5974 struct sk_buff *skb, *tmp;
5975 struct sctp_ulpevent *event;
5978 /* Migrate socket buffer sizes and all the socket level options to the
5981 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5982 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5983 /* Brute force copy old sctp opt. */
5984 inet_sk_copy_descendant(newsk, oldsk);
5986 /* Restore the ep value that was overwritten with the above structure
5992 /* Hook this new socket in to the bind_hash list. */
5993 pp = sctp_sk(oldsk)->bind_hash;
5994 sk_add_bind_node(newsk, &pp->owner);
5995 sctp_sk(newsk)->bind_hash = pp;
5996 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5998 /* Copy the bind_addr list from the original endpoint to the new
5999 * endpoint so that we can handle restarts properly
6001 if (PF_INET6 == assoc->base.sk->sk_family)
6002 flags = SCTP_ADDR6_ALLOWED;
6003 if (assoc->peer.ipv4_address)
6004 flags |= SCTP_ADDR4_PEERSUPP;
6005 if (assoc->peer.ipv6_address)
6006 flags |= SCTP_ADDR6_PEERSUPP;
6007 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
6008 &oldsp->ep->base.bind_addr,
6009 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
6011 /* Move any messages in the old socket's receive queue that are for the
6012 * peeled off association to the new socket's receive queue.
6014 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6015 event = sctp_skb2event(skb);
6016 if (event->asoc == assoc) {
6017 sctp_sock_rfree_frag(skb);
6018 __skb_unlink(skb, &oldsk->sk_receive_queue);
6019 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6020 sctp_skb_set_owner_r_frag(skb, newsk);
6024 /* Clean up any messages pending delivery due to partial
6025 * delivery. Three cases:
6026 * 1) No partial deliver; no work.
6027 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6028 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6030 skb_queue_head_init(&newsp->pd_lobby);
6031 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6033 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6034 struct sk_buff_head *queue;
6036 /* Decide which queue to move pd_lobby skbs to. */
6037 if (assoc->ulpq.pd_mode) {
6038 queue = &newsp->pd_lobby;
6040 queue = &newsk->sk_receive_queue;
6042 /* Walk through the pd_lobby, looking for skbs that
6043 * need moved to the new socket.
6045 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6046 event = sctp_skb2event(skb);
6047 if (event->asoc == assoc) {
6048 sctp_sock_rfree_frag(skb);
6049 __skb_unlink(skb, &oldsp->pd_lobby);
6050 __skb_queue_tail(queue, skb);
6051 sctp_skb_set_owner_r_frag(skb, newsk);
6055 /* Clear up any skbs waiting for the partial
6056 * delivery to finish.
6058 if (assoc->ulpq.pd_mode)
6059 sctp_clear_pd(oldsk, NULL);
6063 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6064 sctp_sock_rfree_frag(skb);
6065 sctp_skb_set_owner_r_frag(skb, newsk);
6068 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6069 sctp_sock_rfree_frag(skb);
6070 sctp_skb_set_owner_r_frag(skb, newsk);
6073 /* Set the type of socket to indicate that it is peeled off from the
6074 * original UDP-style socket or created with the accept() call on a
6075 * TCP-style socket..
6079 /* Mark the new socket "in-use" by the user so that any packets
6080 * that may arrive on the association after we've moved it are
6081 * queued to the backlog. This prevents a potential race between
6082 * backlog processing on the old socket and new-packet processing
6083 * on the new socket.
6085 sctp_lock_sock(newsk);
6086 sctp_assoc_migrate(assoc, newsk);
6088 /* If the association on the newsk is already closed before accept()
6089 * is called, set RCV_SHUTDOWN flag.
6091 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6092 newsk->sk_shutdown |= RCV_SHUTDOWN;
6094 newsk->sk_state = SCTP_SS_ESTABLISHED;
6095 sctp_release_sock(newsk);
6098 /* This proto struct describes the ULP interface for SCTP. */
6099 struct proto sctp_prot = {
6101 .owner = THIS_MODULE,
6102 .close = sctp_close,
6103 .connect = sctp_connect,
6104 .disconnect = sctp_disconnect,
6105 .accept = sctp_accept,
6106 .ioctl = sctp_ioctl,
6107 .init = sctp_init_sock,
6108 .destroy = sctp_destroy_sock,
6109 .shutdown = sctp_shutdown,
6110 .setsockopt = sctp_setsockopt,
6111 .getsockopt = sctp_getsockopt,
6112 .sendmsg = sctp_sendmsg,
6113 .recvmsg = sctp_recvmsg,
6115 .backlog_rcv = sctp_backlog_rcv,
6117 .unhash = sctp_unhash,
6118 .get_port = sctp_get_port,
6119 .obj_size = sizeof(struct sctp_sock),
6122 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6123 struct proto sctpv6_prot = {
6125 .owner = THIS_MODULE,
6126 .close = sctp_close,
6127 .connect = sctp_connect,
6128 .disconnect = sctp_disconnect,
6129 .accept = sctp_accept,
6130 .ioctl = sctp_ioctl,
6131 .init = sctp_init_sock,
6132 .destroy = sctp_destroy_sock,
6133 .shutdown = sctp_shutdown,
6134 .setsockopt = sctp_setsockopt,
6135 .getsockopt = sctp_getsockopt,
6136 .sendmsg = sctp_sendmsg,
6137 .recvmsg = sctp_recvmsg,
6139 .backlog_rcv = sctp_backlog_rcv,
6141 .unhash = sctp_unhash,
6142 .get_port = sctp_get_port,
6143 .obj_size = sizeof(struct sctp6_sock),
6145 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */