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/config.h>
61 #include <linux/types.h>
62 #include <linux/kernel.h>
63 #include <linux/wait.h>
64 #include <linux/time.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 kmem_cache_t *sctp_bucket_cachep;
112 /* Get the sndbuf space available at the time on the association. */
113 static inline int sctp_wspace(struct sctp_association *asoc)
115 struct sock *sk = asoc->base.sk;
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 sk->sk_wmem_queued += SCTP_DATA_SNDSIZE(chunk) +
160 sizeof(struct sk_buff) +
161 sizeof(struct sctp_chunk);
163 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
166 /* Verify that this is a valid address. */
167 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
172 /* Verify basic sockaddr. */
173 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
177 /* Is this a valid SCTP address? */
178 if (!af->addr_valid(addr, sctp_sk(sk)))
181 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
187 /* Look up the association by its id. If this is not a UDP-style
188 * socket, the ID field is always ignored.
190 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
192 struct sctp_association *asoc = NULL;
194 /* If this is not a UDP-style socket, assoc id should be ignored. */
195 if (!sctp_style(sk, UDP)) {
196 /* Return NULL if the socket state is not ESTABLISHED. It
197 * could be a TCP-style listening socket or a socket which
198 * hasn't yet called connect() to establish an association.
200 if (!sctp_sstate(sk, ESTABLISHED))
203 /* Get the first and the only association from the list. */
204 if (!list_empty(&sctp_sk(sk)->ep->asocs))
205 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
206 struct sctp_association, asocs);
210 /* Otherwise this is a UDP-style socket. */
211 if (!id || (id == (sctp_assoc_t)-1))
214 spin_lock_bh(&sctp_assocs_id_lock);
215 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
216 spin_unlock_bh(&sctp_assocs_id_lock);
218 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
224 /* Look up the transport from an address and an assoc id. If both address and
225 * id are specified, the associations matching the address and the id should be
228 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
229 struct sockaddr_storage *addr,
232 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
233 struct sctp_transport *transport;
234 union sctp_addr *laddr = (union sctp_addr *)addr;
236 laddr->v4.sin_port = ntohs(laddr->v4.sin_port);
237 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
238 (union sctp_addr *)addr,
240 laddr->v4.sin_port = htons(laddr->v4.sin_port);
245 id_asoc = sctp_id2assoc(sk, id);
246 if (id_asoc && (id_asoc != addr_asoc))
249 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
250 (union sctp_addr *)addr);
255 /* API 3.1.2 bind() - UDP Style Syntax
256 * The syntax of bind() is,
258 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
260 * sd - the socket descriptor returned by socket().
261 * addr - the address structure (struct sockaddr_in or struct
262 * sockaddr_in6 [RFC 2553]),
263 * addr_len - the size of the address structure.
265 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
271 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
274 /* Disallow binding twice. */
275 if (!sctp_sk(sk)->ep->base.bind_addr.port)
276 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
281 sctp_release_sock(sk);
286 static long sctp_get_port_local(struct sock *, union sctp_addr *);
288 /* Verify this is a valid sockaddr. */
289 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
290 union sctp_addr *addr, int len)
294 /* Check minimum size. */
295 if (len < sizeof (struct sockaddr))
298 /* Does this PF support this AF? */
299 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
302 /* If we get this far, af is valid. */
303 af = sctp_get_af_specific(addr->sa.sa_family);
305 if (len < af->sockaddr_len)
311 /* Bind a local address either to an endpoint or to an association. */
312 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
314 struct sctp_sock *sp = sctp_sk(sk);
315 struct sctp_endpoint *ep = sp->ep;
316 struct sctp_bind_addr *bp = &ep->base.bind_addr;
321 /* Common sockaddr verification. */
322 af = sctp_sockaddr_af(sp, addr, len);
324 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
329 snum = ntohs(addr->v4.sin_port);
331 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
332 ", port: %d, new port: %d, len: %d)\n",
338 /* PF specific bind() address verification. */
339 if (!sp->pf->bind_verify(sp, addr))
340 return -EADDRNOTAVAIL;
342 /* We must either be unbound, or bind to the same port. */
343 if (bp->port && (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);
350 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
353 /* Make sure we are allowed to bind here.
354 * The function sctp_get_port_local() does duplicate address
357 if ((ret = sctp_get_port_local(sk, addr))) {
358 if (ret == (long) sk) {
359 /* This endpoint has a conflicting address. */
366 /* Refresh ephemeral port. */
368 bp->port = inet_sk(sk)->num;
370 /* Add the address to the bind address list. */
371 sctp_local_bh_disable();
372 sctp_write_lock(&ep->base.addr_lock);
374 /* Use GFP_ATOMIC since BHs are disabled. */
375 addr->v4.sin_port = ntohs(addr->v4.sin_port);
376 ret = sctp_add_bind_addr(bp, addr, GFP_ATOMIC);
377 addr->v4.sin_port = htons(addr->v4.sin_port);
378 sctp_write_unlock(&ep->base.addr_lock);
379 sctp_local_bh_enable();
381 /* Copy back into socket for getsockname() use. */
383 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
384 af->to_sk_saddr(addr, sk);
390 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
392 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
393 * at any one time. If a sender, after sending an ASCONF chunk, decides
394 * it needs to transfer another ASCONF Chunk, it MUST wait until the
395 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
396 * subsequent ASCONF. Note this restriction binds each side, so at any
397 * time two ASCONF may be in-transit on any given association (one sent
398 * from each endpoint).
400 static int sctp_send_asconf(struct sctp_association *asoc,
401 struct sctp_chunk *chunk)
405 /* If there is an outstanding ASCONF chunk, queue it for later
408 if (asoc->addip_last_asconf) {
409 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
413 /* Hold the chunk until an ASCONF_ACK is received. */
414 sctp_chunk_hold(chunk);
415 retval = sctp_primitive_ASCONF(asoc, chunk);
417 sctp_chunk_free(chunk);
419 asoc->addip_last_asconf = chunk;
425 /* Add a list of addresses as bind addresses to local endpoint or
428 * Basically run through each address specified in the addrs/addrcnt
429 * array/length pair, determine if it is IPv6 or IPv4 and call
430 * sctp_do_bind() on it.
432 * If any of them fails, then the operation will be reversed and the
433 * ones that were added will be removed.
435 * Only sctp_setsockopt_bindx() is supposed to call this function.
437 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
442 struct sockaddr *sa_addr;
445 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
449 for (cnt = 0; cnt < addrcnt; cnt++) {
450 /* The list may contain either IPv4 or IPv6 address;
451 * determine the address length for walking thru the list.
453 sa_addr = (struct sockaddr *)addr_buf;
454 af = sctp_get_af_specific(sa_addr->sa_family);
460 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
463 addr_buf += af->sockaddr_len;
467 /* Failed. Cleanup the ones that have been added */
469 sctp_bindx_rem(sk, addrs, cnt);
477 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
478 * associations that are part of the endpoint indicating that a list of local
479 * addresses are added to the endpoint.
481 * If any of the addresses is already in the bind address list of the
482 * association, we do not send the chunk for that association. But it will not
483 * affect other associations.
485 * Only sctp_setsockopt_bindx() is supposed to call this function.
487 static int sctp_send_asconf_add_ip(struct sock *sk,
488 struct sockaddr *addrs,
491 struct sctp_sock *sp;
492 struct sctp_endpoint *ep;
493 struct sctp_association *asoc;
494 struct sctp_bind_addr *bp;
495 struct sctp_chunk *chunk;
496 struct sctp_sockaddr_entry *laddr;
497 union sctp_addr *addr;
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);
566 /* FIXME: After sending the add address ASCONF chunk, we
567 * cannot append the address to the association's binding
568 * address list, because the new address may be used as the
569 * source of a message sent to the peer before the ASCONF
570 * chunk is received by the peer. So we should wait until
571 * ASCONF_ACK is received.
579 /* Remove a list of addresses from bind addresses list. Do not remove the
582 * Basically run through each address specified in the addrs/addrcnt
583 * array/length pair, determine if it is IPv6 or IPv4 and call
584 * sctp_del_bind() on it.
586 * If any of them fails, then the operation will be reversed and the
587 * ones that were removed will be added back.
589 * At least one address has to be left; if only one address is
590 * available, the operation will return -EBUSY.
592 * Only sctp_setsockopt_bindx() is supposed to call this function.
594 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
596 struct sctp_sock *sp = sctp_sk(sk);
597 struct sctp_endpoint *ep = sp->ep;
599 struct sctp_bind_addr *bp = &ep->base.bind_addr;
601 union sctp_addr saveaddr;
603 struct sockaddr *sa_addr;
606 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
610 for (cnt = 0; cnt < addrcnt; cnt++) {
611 /* If the bind address list is empty or if there is only one
612 * bind address, there is nothing more to be removed (we need
613 * at least one address here).
615 if (list_empty(&bp->address_list) ||
616 (sctp_list_single_entry(&bp->address_list))) {
621 /* The list may contain either IPv4 or IPv6 address;
622 * determine the address length to copy the address to
625 sa_addr = (struct sockaddr *)addr_buf;
626 af = sctp_get_af_specific(sa_addr->sa_family);
631 memcpy(&saveaddr, sa_addr, af->sockaddr_len);
632 saveaddr.v4.sin_port = ntohs(saveaddr.v4.sin_port);
633 if (saveaddr.v4.sin_port != bp->port) {
638 /* FIXME - There is probably a need to check if sk->sk_saddr and
639 * sk->sk_rcv_addr are currently set to one of the addresses to
640 * be removed. This is something which needs to be looked into
641 * when we are fixing the outstanding issues with multi-homing
642 * socket routing and failover schemes. Refer to comments in
643 * sctp_do_bind(). -daisy
645 sctp_local_bh_disable();
646 sctp_write_lock(&ep->base.addr_lock);
648 retval = sctp_del_bind_addr(bp, &saveaddr);
650 sctp_write_unlock(&ep->base.addr_lock);
651 sctp_local_bh_enable();
653 addr_buf += af->sockaddr_len;
656 /* Failed. Add the ones that has been removed back */
658 sctp_bindx_add(sk, addrs, cnt);
666 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
667 * the associations that are part of the endpoint indicating that a list of
668 * local addresses are removed from the endpoint.
670 * If any of the addresses is already in the bind address list of the
671 * association, we do not send the chunk for that association. But it will not
672 * affect other associations.
674 * Only sctp_setsockopt_bindx() is supposed to call this function.
676 static int sctp_send_asconf_del_ip(struct sock *sk,
677 struct sockaddr *addrs,
680 struct sctp_sock *sp;
681 struct sctp_endpoint *ep;
682 struct sctp_association *asoc;
683 struct sctp_bind_addr *bp;
684 struct sctp_chunk *chunk;
685 union sctp_addr *laddr;
688 struct list_head *pos;
692 if (!sctp_addip_enable)
698 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
699 __FUNCTION__, sk, addrs, addrcnt);
701 list_for_each(pos, &ep->asocs) {
702 asoc = list_entry(pos, struct sctp_association, asocs);
704 if (!asoc->peer.asconf_capable)
707 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
710 if (!sctp_state(asoc, ESTABLISHED))
713 /* Check if any address in the packed array of addresses is
714 * not present in the bind address list of the association.
715 * If so, do not send the asconf chunk to its peer, but
716 * continue with other associations.
719 for (i = 0; i < addrcnt; i++) {
720 laddr = (union sctp_addr *)addr_buf;
721 af = sctp_get_af_specific(laddr->v4.sin_family);
727 if (!sctp_assoc_lookup_laddr(asoc, laddr))
730 addr_buf += af->sockaddr_len;
735 /* Find one address in the association's bind address list
736 * that is not in the packed array of addresses. This is to
737 * make sure that we do not delete all the addresses in the
740 sctp_read_lock(&asoc->base.addr_lock);
741 bp = &asoc->base.bind_addr;
742 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
744 sctp_read_unlock(&asoc->base.addr_lock);
748 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
755 retval = sctp_send_asconf(asoc, chunk);
757 /* FIXME: After sending the delete address ASCONF chunk, we
758 * cannot remove the addresses from the association's bind
759 * address list, because there maybe some packet send to
760 * the delete addresses, so we should wait until ASCONF_ACK
761 * packet is received.
768 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
771 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
774 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
775 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
778 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
779 * Section 3.1.2 for this usage.
781 * addrs is a pointer to an array of one or more socket addresses. Each
782 * address is contained in its appropriate structure (i.e. struct
783 * sockaddr_in or struct sockaddr_in6) the family of the address type
784 * must be used to distengish the address length (note that this
785 * representation is termed a "packed array" of addresses). The caller
786 * specifies the number of addresses in the array with addrcnt.
788 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
789 * -1, and sets errno to the appropriate error code.
791 * For SCTP, the port given in each socket address must be the same, or
792 * sctp_bindx() will fail, setting errno to EINVAL.
794 * The flags parameter is formed from the bitwise OR of zero or more of
795 * the following currently defined flags:
797 * SCTP_BINDX_ADD_ADDR
799 * SCTP_BINDX_REM_ADDR
801 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
802 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
803 * addresses from the association. The two flags are mutually exclusive;
804 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
805 * not remove all addresses from an association; sctp_bindx() will
806 * reject such an attempt with EINVAL.
808 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
809 * additional addresses with an endpoint after calling bind(). Or use
810 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
811 * socket is associated with so that no new association accepted will be
812 * associated with those addresses. If the endpoint supports dynamic
813 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
814 * endpoint to send the appropriate message to the peer to change the
815 * peers address lists.
817 * Adding and removing addresses from a connected association is
818 * optional functionality. Implementations that do not support this
819 * functionality should return EOPNOTSUPP.
821 * Basically do nothing but copying the addresses from user to kernel
822 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
823 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
826 * We don't use copy_from_user() for optimization: we first do the
827 * sanity checks (buffer size -fast- and access check-healthy
828 * pointer); if all of those succeed, then we can alloc the memory
829 * (expensive operation) needed to copy the data to kernel. Then we do
830 * the copying without checking the user space area
831 * (__copy_from_user()).
833 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
836 * sk The sk of the socket
837 * addrs The pointer to the addresses in user land
838 * addrssize Size of the addrs buffer
839 * op Operation to perform (add or remove, see the flags of
842 * Returns 0 if ok, <0 errno code on error.
844 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
845 struct sockaddr __user *addrs,
846 int addrs_size, int op)
848 struct sockaddr *kaddrs;
852 struct sockaddr *sa_addr;
856 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
857 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
859 if (unlikely(addrs_size <= 0))
862 /* Check the user passed a healthy pointer. */
863 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
866 /* Alloc space for the address array in kernel memory. */
867 kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL);
868 if (unlikely(!kaddrs))
871 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
876 /* Walk through the addrs buffer and count the number of addresses. */
878 while (walk_size < addrs_size) {
879 sa_addr = (struct sockaddr *)addr_buf;
880 af = sctp_get_af_specific(sa_addr->sa_family);
882 /* If the address family is not supported or if this address
883 * causes the address buffer to overflow return EINVAL.
885 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
890 addr_buf += af->sockaddr_len;
891 walk_size += af->sockaddr_len;
896 case SCTP_BINDX_ADD_ADDR:
897 err = sctp_bindx_add(sk, kaddrs, addrcnt);
900 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
903 case SCTP_BINDX_REM_ADDR:
904 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
907 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
921 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
923 * Common routine for handling connect() and sctp_connectx().
924 * Connect will come in with just a single address.
926 static int __sctp_connect(struct sock* sk,
927 struct sockaddr *kaddrs,
930 struct sctp_sock *sp;
931 struct sctp_endpoint *ep;
932 struct sctp_association *asoc = NULL;
933 struct sctp_association *asoc2;
934 struct sctp_transport *transport;
942 struct sockaddr *sa_addr;
948 /* connect() cannot be done on a socket that is already in ESTABLISHED
949 * state - UDP-style peeled off socket or a TCP-style socket that
950 * is already connected.
951 * It cannot be done even on a TCP-style listening socket.
953 if (sctp_sstate(sk, ESTABLISHED) ||
954 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
959 /* Walk through the addrs buffer and count the number of addresses. */
961 while (walk_size < addrs_size) {
962 sa_addr = (struct sockaddr *)addr_buf;
963 af = sctp_get_af_specific(sa_addr->sa_family);
965 /* If the address family is not supported or if this address
966 * causes the address buffer to overflow return EINVAL.
968 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
973 err = sctp_verify_addr(sk, (union sctp_addr *)sa_addr,
978 memcpy(&to, sa_addr, af->sockaddr_len);
979 to.v4.sin_port = ntohs(to.v4.sin_port);
981 /* Check if there already is a matching association on the
982 * endpoint (other than the one created here).
984 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
985 if (asoc2 && asoc2 != asoc) {
986 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
993 /* If we could not find a matching association on the endpoint,
994 * make sure that there is no peeled-off association matching
995 * the peer address even on another socket.
997 if (sctp_endpoint_is_peeled_off(ep, &to)) {
998 err = -EADDRNOTAVAIL;
1003 /* If a bind() or sctp_bindx() is not called prior to
1004 * an sctp_connectx() call, the system picks an
1005 * ephemeral port and will choose an address set
1006 * equivalent to binding with a wildcard address.
1008 if (!ep->base.bind_addr.port) {
1009 if (sctp_autobind(sk)) {
1015 scope = sctp_scope(&to);
1016 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1023 /* Prime the peer's transport structures. */
1024 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1032 addr_buf += af->sockaddr_len;
1033 walk_size += af->sockaddr_len;
1036 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1041 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1046 /* Initialize sk's dport and daddr for getpeername() */
1047 inet_sk(sk)->dport = htons(asoc->peer.port);
1048 af = sctp_get_af_specific(to.sa.sa_family);
1049 af->to_sk_daddr(&to, sk);
1051 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1052 err = sctp_wait_for_connect(asoc, &timeo);
1054 /* Don't free association on exit. */
1059 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1060 " kaddrs: %p err: %d\n",
1063 sctp_association_free(asoc);
1067 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1070 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1072 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1073 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1074 * or IPv6 addresses.
1076 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1077 * Section 3.1.2 for this usage.
1079 * addrs is a pointer to an array of one or more socket addresses. Each
1080 * address is contained in its appropriate structure (i.e. struct
1081 * sockaddr_in or struct sockaddr_in6) the family of the address type
1082 * must be used to distengish the address length (note that this
1083 * representation is termed a "packed array" of addresses). The caller
1084 * specifies the number of addresses in the array with addrcnt.
1086 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1087 * -1, and sets errno to the appropriate error code.
1089 * For SCTP, the port given in each socket address must be the same, or
1090 * sctp_connectx() will fail, setting errno to EINVAL.
1092 * An application can use sctp_connectx to initiate an association with
1093 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1094 * allows a caller to specify multiple addresses at which a peer can be
1095 * reached. The way the SCTP stack uses the list of addresses to set up
1096 * the association is implementation dependant. This function only
1097 * specifies that the stack will try to make use of all the addresses in
1098 * the list when needed.
1100 * Note that the list of addresses passed in is only used for setting up
1101 * the association. It does not necessarily equal the set of addresses
1102 * the peer uses for the resulting association. If the caller wants to
1103 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1104 * retrieve them after the association has been set up.
1106 * Basically do nothing but copying the addresses from user to kernel
1107 * land and invoking either sctp_connectx(). This is used for tunneling
1108 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1110 * We don't use copy_from_user() for optimization: we first do the
1111 * sanity checks (buffer size -fast- and access check-healthy
1112 * pointer); if all of those succeed, then we can alloc the memory
1113 * (expensive operation) needed to copy the data to kernel. Then we do
1114 * the copying without checking the user space area
1115 * (__copy_from_user()).
1117 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1120 * sk The sk of the socket
1121 * addrs The pointer to the addresses in user land
1122 * addrssize Size of the addrs buffer
1124 * Returns 0 if ok, <0 errno code on error.
1126 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1127 struct sockaddr __user *addrs,
1131 struct sockaddr *kaddrs;
1133 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1134 __FUNCTION__, sk, addrs, addrs_size);
1136 if (unlikely(addrs_size <= 0))
1139 /* Check the user passed a healthy pointer. */
1140 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1143 /* Alloc space for the address array in kernel memory. */
1144 kaddrs = (struct sockaddr *)kmalloc(addrs_size, GFP_KERNEL);
1145 if (unlikely(!kaddrs))
1148 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1151 err = __sctp_connect(sk, kaddrs, addrs_size);
1158 /* API 3.1.4 close() - UDP Style Syntax
1159 * Applications use close() to perform graceful shutdown (as described in
1160 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1161 * by a UDP-style socket.
1165 * ret = close(int sd);
1167 * sd - the socket descriptor of the associations to be closed.
1169 * To gracefully shutdown a specific association represented by the
1170 * UDP-style socket, an application should use the sendmsg() call,
1171 * passing no user data, but including the appropriate flag in the
1172 * ancillary data (see Section xxxx).
1174 * If sd in the close() call is a branched-off socket representing only
1175 * one association, the shutdown is performed on that association only.
1177 * 4.1.6 close() - TCP Style Syntax
1179 * Applications use close() to gracefully close down an association.
1183 * int close(int sd);
1185 * sd - the socket descriptor of the association to be closed.
1187 * After an application calls close() on a socket descriptor, no further
1188 * socket operations will succeed on that descriptor.
1190 * API 7.1.4 SO_LINGER
1192 * An application using the TCP-style socket can use this option to
1193 * perform the SCTP ABORT primitive. The linger option structure is:
1196 * int l_onoff; // option on/off
1197 * int l_linger; // linger time
1200 * To enable the option, set l_onoff to 1. If the l_linger value is set
1201 * to 0, calling close() is the same as the ABORT primitive. If the
1202 * value is set to a negative value, the setsockopt() call will return
1203 * an error. If the value is set to a positive value linger_time, the
1204 * close() can be blocked for at most linger_time ms. If the graceful
1205 * shutdown phase does not finish during this period, close() will
1206 * return but the graceful shutdown phase continues in the system.
1208 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1210 struct sctp_endpoint *ep;
1211 struct sctp_association *asoc;
1212 struct list_head *pos, *temp;
1214 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1217 sk->sk_shutdown = SHUTDOWN_MASK;
1219 ep = sctp_sk(sk)->ep;
1221 /* Walk all associations on a socket, not on an endpoint. */
1222 list_for_each_safe(pos, temp, &ep->asocs) {
1223 asoc = list_entry(pos, struct sctp_association, asocs);
1225 if (sctp_style(sk, TCP)) {
1226 /* A closed association can still be in the list if
1227 * it belongs to a TCP-style listening socket that is
1228 * not yet accepted. If so, free it. If not, send an
1229 * ABORT or SHUTDOWN based on the linger options.
1231 if (sctp_state(asoc, CLOSED)) {
1232 sctp_unhash_established(asoc);
1233 sctp_association_free(asoc);
1235 } else if (sock_flag(sk, SOCK_LINGER) &&
1237 sctp_primitive_ABORT(asoc, NULL);
1239 sctp_primitive_SHUTDOWN(asoc, NULL);
1241 sctp_primitive_SHUTDOWN(asoc, NULL);
1244 /* Clean up any skbs sitting on the receive queue. */
1245 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1246 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1248 /* On a TCP-style socket, block for at most linger_time if set. */
1249 if (sctp_style(sk, TCP) && timeout)
1250 sctp_wait_for_close(sk, timeout);
1252 /* This will run the backlog queue. */
1253 sctp_release_sock(sk);
1255 /* Supposedly, no process has access to the socket, but
1256 * the net layers still may.
1258 sctp_local_bh_disable();
1259 sctp_bh_lock_sock(sk);
1261 /* Hold the sock, since sk_common_release() will put sock_put()
1262 * and we have just a little more cleanup.
1265 sk_common_release(sk);
1267 sctp_bh_unlock_sock(sk);
1268 sctp_local_bh_enable();
1272 SCTP_DBG_OBJCNT_DEC(sock);
1275 /* Handle EPIPE error. */
1276 static int sctp_error(struct sock *sk, int flags, int err)
1279 err = sock_error(sk) ? : -EPIPE;
1280 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1281 send_sig(SIGPIPE, current, 0);
1285 /* API 3.1.3 sendmsg() - UDP Style Syntax
1287 * An application uses sendmsg() and recvmsg() calls to transmit data to
1288 * and receive data from its peer.
1290 * ssize_t sendmsg(int socket, const struct msghdr *message,
1293 * socket - the socket descriptor of the endpoint.
1294 * message - pointer to the msghdr structure which contains a single
1295 * user message and possibly some ancillary data.
1297 * See Section 5 for complete description of the data
1300 * flags - flags sent or received with the user message, see Section
1301 * 5 for complete description of the flags.
1303 * Note: This function could use a rewrite especially when explicit
1304 * connect support comes in.
1306 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1308 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1310 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1311 struct msghdr *msg, size_t msg_len)
1313 struct sctp_sock *sp;
1314 struct sctp_endpoint *ep;
1315 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1316 struct sctp_transport *transport, *chunk_tp;
1317 struct sctp_chunk *chunk;
1319 struct sockaddr *msg_name = NULL;
1320 struct sctp_sndrcvinfo default_sinfo = { 0 };
1321 struct sctp_sndrcvinfo *sinfo;
1322 struct sctp_initmsg *sinit;
1323 sctp_assoc_t associd = 0;
1324 sctp_cmsgs_t cmsgs = { NULL };
1328 __u16 sinfo_flags = 0;
1329 struct sctp_datamsg *datamsg;
1330 struct list_head *pos;
1331 int msg_flags = msg->msg_flags;
1333 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1340 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1342 /* We cannot send a message over a TCP-style listening socket. */
1343 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1348 /* Parse out the SCTP CMSGs. */
1349 err = sctp_msghdr_parse(msg, &cmsgs);
1352 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1356 /* Fetch the destination address for this packet. This
1357 * address only selects the association--it is not necessarily
1358 * the address we will send to.
1359 * For a peeled-off socket, msg_name is ignored.
1361 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1362 int msg_namelen = msg->msg_namelen;
1364 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1369 if (msg_namelen > sizeof(to))
1370 msg_namelen = sizeof(to);
1371 memcpy(&to, msg->msg_name, msg_namelen);
1372 SCTP_DEBUG_PRINTK("Just memcpy'd. msg_name is "
1374 to.v4.sin_addr.s_addr, to.v4.sin_port);
1376 to.v4.sin_port = ntohs(to.v4.sin_port);
1377 msg_name = msg->msg_name;
1383 /* Did the user specify SNDRCVINFO? */
1385 sinfo_flags = sinfo->sinfo_flags;
1386 associd = sinfo->sinfo_assoc_id;
1389 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1390 msg_len, sinfo_flags);
1392 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1393 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1398 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1399 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1400 * If SCTP_ABORT is set, the message length could be non zero with
1401 * the msg_iov set to the user abort reason.
1403 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1404 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1409 /* If SCTP_ADDR_OVER is set, there must be an address
1410 * specified in msg_name.
1412 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1419 SCTP_DEBUG_PRINTK("About to look up association.\n");
1423 /* If a msg_name has been specified, assume this is to be used. */
1425 /* Look for a matching association on the endpoint. */
1426 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1428 /* If we could not find a matching association on the
1429 * endpoint, make sure that it is not a TCP-style
1430 * socket that already has an association or there is
1431 * no peeled-off association on another socket.
1433 if ((sctp_style(sk, TCP) &&
1434 sctp_sstate(sk, ESTABLISHED)) ||
1435 sctp_endpoint_is_peeled_off(ep, &to)) {
1436 err = -EADDRNOTAVAIL;
1441 asoc = sctp_id2assoc(sk, associd);
1449 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1451 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1452 * socket that has an association in CLOSED state. This can
1453 * happen when an accepted socket has an association that is
1456 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1461 if (sinfo_flags & SCTP_EOF) {
1462 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1464 sctp_primitive_SHUTDOWN(asoc, NULL);
1468 if (sinfo_flags & SCTP_ABORT) {
1469 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1470 sctp_primitive_ABORT(asoc, msg);
1476 /* Do we need to create the association? */
1478 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1480 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1485 /* Check for invalid stream against the stream counts,
1486 * either the default or the user specified stream counts.
1489 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1490 /* Check against the defaults. */
1491 if (sinfo->sinfo_stream >=
1492 sp->initmsg.sinit_num_ostreams) {
1497 /* Check against the requested. */
1498 if (sinfo->sinfo_stream >=
1499 sinit->sinit_num_ostreams) {
1507 * API 3.1.2 bind() - UDP Style Syntax
1508 * If a bind() or sctp_bindx() is not called prior to a
1509 * sendmsg() call that initiates a new association, the
1510 * system picks an ephemeral port and will choose an address
1511 * set equivalent to binding with a wildcard address.
1513 if (!ep->base.bind_addr.port) {
1514 if (sctp_autobind(sk)) {
1520 scope = sctp_scope(&to);
1521 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1528 /* If the SCTP_INIT ancillary data is specified, set all
1529 * the association init values accordingly.
1532 if (sinit->sinit_num_ostreams) {
1533 asoc->c.sinit_num_ostreams =
1534 sinit->sinit_num_ostreams;
1536 if (sinit->sinit_max_instreams) {
1537 asoc->c.sinit_max_instreams =
1538 sinit->sinit_max_instreams;
1540 if (sinit->sinit_max_attempts) {
1541 asoc->max_init_attempts
1542 = sinit->sinit_max_attempts;
1544 if (sinit->sinit_max_init_timeo) {
1545 asoc->max_init_timeo =
1546 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1550 /* Prime the peer's transport structures. */
1551 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1556 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1563 /* ASSERT: we have a valid association at this point. */
1564 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1567 /* If the user didn't specify SNDRCVINFO, make up one with
1570 default_sinfo.sinfo_stream = asoc->default_stream;
1571 default_sinfo.sinfo_flags = asoc->default_flags;
1572 default_sinfo.sinfo_ppid = asoc->default_ppid;
1573 default_sinfo.sinfo_context = asoc->default_context;
1574 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1575 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1576 sinfo = &default_sinfo;
1579 /* API 7.1.7, the sndbuf size per association bounds the
1580 * maximum size of data that can be sent in a single send call.
1582 if (msg_len > sk->sk_sndbuf) {
1587 /* If fragmentation is disabled and the message length exceeds the
1588 * association fragmentation point, return EMSGSIZE. The I-D
1589 * does not specify what this error is, but this looks like
1592 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1598 /* Check for invalid stream. */
1599 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1605 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1606 if (!sctp_wspace(asoc)) {
1607 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1612 /* If an address is passed with the sendto/sendmsg call, it is used
1613 * to override the primary destination address in the TCP model, or
1614 * when SCTP_ADDR_OVER flag is set in the UDP model.
1616 if ((sctp_style(sk, TCP) && msg_name) ||
1617 (sinfo_flags & SCTP_ADDR_OVER)) {
1618 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1626 /* Auto-connect, if we aren't connected already. */
1627 if (sctp_state(asoc, CLOSED)) {
1628 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1631 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1634 /* Break the message into multiple chunks of maximum size. */
1635 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1641 /* Now send the (possibly) fragmented message. */
1642 list_for_each(pos, &datamsg->chunks) {
1643 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1644 sctp_datamsg_track(chunk);
1646 /* Do accounting for the write space. */
1647 sctp_set_owner_w(chunk);
1649 chunk->transport = chunk_tp;
1651 /* Send it to the lower layers. Note: all chunks
1652 * must either fail or succeed. The lower layer
1653 * works that way today. Keep it that way or this
1656 err = sctp_primitive_SEND(asoc, chunk);
1657 /* Did the lower layer accept the chunk? */
1659 sctp_chunk_free(chunk);
1660 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1663 sctp_datamsg_free(datamsg);
1669 /* If we are already past ASSOCIATE, the lower
1670 * layers are responsible for association cleanup.
1676 sctp_association_free(asoc);
1678 sctp_release_sock(sk);
1681 return sctp_error(sk, msg_flags, err);
1688 err = sock_error(sk);
1698 /* This is an extended version of skb_pull() that removes the data from the
1699 * start of a skb even when data is spread across the list of skb's in the
1700 * frag_list. len specifies the total amount of data that needs to be removed.
1701 * when 'len' bytes could be removed from the skb, it returns 0.
1702 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1703 * could not be removed.
1705 static int sctp_skb_pull(struct sk_buff *skb, int len)
1707 struct sk_buff *list;
1708 int skb_len = skb_headlen(skb);
1711 if (len <= skb_len) {
1712 __skb_pull(skb, len);
1716 __skb_pull(skb, skb_len);
1718 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1719 rlen = sctp_skb_pull(list, len);
1720 skb->len -= (len-rlen);
1721 skb->data_len -= (len-rlen);
1732 /* API 3.1.3 recvmsg() - UDP Style Syntax
1734 * ssize_t recvmsg(int socket, struct msghdr *message,
1737 * socket - the socket descriptor of the endpoint.
1738 * message - pointer to the msghdr structure which contains a single
1739 * user message and possibly some ancillary data.
1741 * See Section 5 for complete description of the data
1744 * flags - flags sent or received with the user message, see Section
1745 * 5 for complete description of the flags.
1747 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1749 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1750 struct msghdr *msg, size_t len, int noblock,
1751 int flags, int *addr_len)
1753 struct sctp_ulpevent *event = NULL;
1754 struct sctp_sock *sp = sctp_sk(sk);
1755 struct sk_buff *skb;
1760 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1761 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1762 "len", len, "knoblauch", noblock,
1763 "flags", flags, "addr_len", addr_len);
1767 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1772 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1776 /* Get the total length of the skb including any skb's in the
1785 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1787 event = sctp_skb2event(skb);
1792 sock_recv_timestamp(msg, sk, skb);
1793 if (sctp_ulpevent_is_notification(event)) {
1794 msg->msg_flags |= MSG_NOTIFICATION;
1795 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1797 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1800 /* Check if we allow SCTP_SNDRCVINFO. */
1801 if (sp->subscribe.sctp_data_io_event)
1802 sctp_ulpevent_read_sndrcvinfo(event, msg);
1804 /* FIXME: we should be calling IP/IPv6 layers. */
1805 if (sk->sk_protinfo.af_inet.cmsg_flags)
1806 ip_cmsg_recv(msg, skb);
1811 /* If skb's length exceeds the user's buffer, update the skb and
1812 * push it back to the receive_queue so that the next call to
1813 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1815 if (skb_len > copied) {
1816 msg->msg_flags &= ~MSG_EOR;
1817 if (flags & MSG_PEEK)
1819 sctp_skb_pull(skb, copied);
1820 skb_queue_head(&sk->sk_receive_queue, skb);
1822 /* When only partial message is copied to the user, increase
1823 * rwnd by that amount. If all the data in the skb is read,
1824 * rwnd is updated when the event is freed.
1826 sctp_assoc_rwnd_increase(event->asoc, copied);
1828 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1829 (event->msg_flags & MSG_EOR))
1830 msg->msg_flags |= MSG_EOR;
1832 msg->msg_flags &= ~MSG_EOR;
1835 if (flags & MSG_PEEK) {
1836 /* Release the skb reference acquired after peeking the skb in
1837 * sctp_skb_recv_datagram().
1841 /* Free the event which includes releasing the reference to
1842 * the owner of the skb, freeing the skb and updating the
1845 sctp_ulpevent_free(event);
1848 sctp_release_sock(sk);
1852 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1854 * This option is a on/off flag. If enabled no SCTP message
1855 * fragmentation will be performed. Instead if a message being sent
1856 * exceeds the current PMTU size, the message will NOT be sent and
1857 * instead a error will be indicated to the user.
1859 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1860 char __user *optval, int optlen)
1864 if (optlen < sizeof(int))
1867 if (get_user(val, (int __user *)optval))
1870 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1875 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1878 if (optlen != sizeof(struct sctp_event_subscribe))
1880 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1885 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1887 * This socket option is applicable to the UDP-style socket only. When
1888 * set it will cause associations that are idle for more than the
1889 * specified number of seconds to automatically close. An association
1890 * being idle is defined an association that has NOT sent or received
1891 * user data. The special value of '0' indicates that no automatic
1892 * close of any associations should be performed. The option expects an
1893 * integer defining the number of seconds of idle time before an
1894 * association is closed.
1896 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1899 struct sctp_sock *sp = sctp_sk(sk);
1901 /* Applicable to UDP-style socket only */
1902 if (sctp_style(sk, TCP))
1904 if (optlen != sizeof(int))
1906 if (copy_from_user(&sp->autoclose, optval, optlen))
1909 sp->ep->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE] = sp->autoclose * HZ;
1913 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1915 * Applications can enable or disable heartbeats for any peer address of
1916 * an association, modify an address's heartbeat interval, force a
1917 * heartbeat to be sent immediately, and adjust the address's maximum
1918 * number of retransmissions sent before an address is considered
1919 * unreachable. The following structure is used to access and modify an
1920 * address's parameters:
1922 * struct sctp_paddrparams {
1923 * sctp_assoc_t spp_assoc_id;
1924 * struct sockaddr_storage spp_address;
1925 * uint32_t spp_hbinterval;
1926 * uint16_t spp_pathmaxrxt;
1929 * spp_assoc_id - (UDP style socket) This is filled in the application,
1930 * and identifies the association for this query.
1931 * spp_address - This specifies which address is of interest.
1932 * spp_hbinterval - This contains the value of the heartbeat interval,
1933 * in milliseconds. A value of 0, when modifying the
1934 * parameter, specifies that the heartbeat on this
1935 * address should be disabled. A value of UINT32_MAX
1936 * (4294967295), when modifying the parameter,
1937 * specifies that a heartbeat should be sent
1938 * immediately to the peer address, and the current
1939 * interval should remain unchanged.
1940 * spp_pathmaxrxt - This contains the maximum number of
1941 * retransmissions before this address shall be
1942 * considered unreachable.
1944 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
1945 char __user *optval, int optlen)
1947 struct sctp_paddrparams params;
1948 struct sctp_transport *trans;
1951 if (optlen != sizeof(struct sctp_paddrparams))
1953 if (copy_from_user(¶ms, optval, optlen))
1957 * API 7. Socket Options (setting the default value for the endpoint)
1958 * All options that support specific settings on an association by
1959 * filling in either an association id variable or a sockaddr_storage
1960 * SHOULD also support setting of the same value for the entire endpoint
1961 * (i.e. future associations). To accomplish this the following logic is
1962 * used when setting one of these options:
1964 * c) If neither the sockaddr_storage or association identification is
1965 * set i.e. the sockaddr_storage is set to all 0's (INADDR_ANY) and
1966 * the association identification is 0, the settings are a default
1967 * and to be applied to the endpoint (all future associations).
1970 /* update default value for endpoint (all future associations) */
1971 if (!params.spp_assoc_id &&
1972 sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
1973 /* Manual heartbeat on an endpoint is invalid. */
1974 if (0xffffffff == params.spp_hbinterval)
1976 else if (params.spp_hbinterval)
1977 sctp_sk(sk)->paddrparam.spp_hbinterval =
1978 params.spp_hbinterval;
1979 if (params.spp_pathmaxrxt)
1980 sctp_sk(sk)->paddrparam.spp_pathmaxrxt =
1981 params.spp_pathmaxrxt;
1985 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
1986 params.spp_assoc_id);
1990 /* Applications can enable or disable heartbeats for any peer address
1991 * of an association, modify an address's heartbeat interval, force a
1992 * heartbeat to be sent immediately, and adjust the address's maximum
1993 * number of retransmissions sent before an address is considered
1996 * The value of the heartbeat interval, in milliseconds. A value of
1997 * UINT32_MAX (4294967295), when modifying the parameter, specifies
1998 * that a heartbeat should be sent immediately to the peer address,
1999 * and the current interval should remain unchanged.
2001 if (0xffffffff == params.spp_hbinterval) {
2002 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2006 /* The value of the heartbeat interval, in milliseconds. A value of 0,
2007 * when modifying the parameter, specifies that the heartbeat on this
2008 * address should be disabled.
2010 if (params.spp_hbinterval) {
2011 trans->hb_allowed = 1;
2012 trans->hb_interval =
2013 msecs_to_jiffies(params.spp_hbinterval);
2015 trans->hb_allowed = 0;
2018 /* spp_pathmaxrxt contains the maximum number of retransmissions
2019 * before this address shall be considered unreachable.
2021 if (params.spp_pathmaxrxt)
2022 trans->max_retrans = params.spp_pathmaxrxt;
2027 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2029 * Applications can specify protocol parameters for the default association
2030 * initialization. The option name argument to setsockopt() and getsockopt()
2033 * Setting initialization parameters is effective only on an unconnected
2034 * socket (for UDP-style sockets only future associations are effected
2035 * by the change). With TCP-style sockets, this option is inherited by
2036 * sockets derived from a listener socket.
2038 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2040 struct sctp_initmsg sinit;
2041 struct sctp_sock *sp = sctp_sk(sk);
2043 if (optlen != sizeof(struct sctp_initmsg))
2045 if (copy_from_user(&sinit, optval, optlen))
2048 if (sinit.sinit_num_ostreams)
2049 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2050 if (sinit.sinit_max_instreams)
2051 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2052 if (sinit.sinit_max_attempts)
2053 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2054 if (sinit.sinit_max_init_timeo)
2055 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2061 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2063 * Applications that wish to use the sendto() system call may wish to
2064 * specify a default set of parameters that would normally be supplied
2065 * through the inclusion of ancillary data. This socket option allows
2066 * such an application to set the default sctp_sndrcvinfo structure.
2067 * The application that wishes to use this socket option simply passes
2068 * in to this call the sctp_sndrcvinfo structure defined in Section
2069 * 5.2.2) The input parameters accepted by this call include
2070 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2071 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2072 * to this call if the caller is using the UDP model.
2074 static int sctp_setsockopt_default_send_param(struct sock *sk,
2075 char __user *optval, int optlen)
2077 struct sctp_sndrcvinfo info;
2078 struct sctp_association *asoc;
2079 struct sctp_sock *sp = sctp_sk(sk);
2081 if (optlen != sizeof(struct sctp_sndrcvinfo))
2083 if (copy_from_user(&info, optval, optlen))
2086 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2087 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2091 asoc->default_stream = info.sinfo_stream;
2092 asoc->default_flags = info.sinfo_flags;
2093 asoc->default_ppid = info.sinfo_ppid;
2094 asoc->default_context = info.sinfo_context;
2095 asoc->default_timetolive = info.sinfo_timetolive;
2097 sp->default_stream = info.sinfo_stream;
2098 sp->default_flags = info.sinfo_flags;
2099 sp->default_ppid = info.sinfo_ppid;
2100 sp->default_context = info.sinfo_context;
2101 sp->default_timetolive = info.sinfo_timetolive;
2107 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2109 * Requests that the local SCTP stack use the enclosed peer address as
2110 * the association primary. The enclosed address must be one of the
2111 * association peer's addresses.
2113 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2116 struct sctp_prim prim;
2117 struct sctp_transport *trans;
2119 if (optlen != sizeof(struct sctp_prim))
2122 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2125 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2129 sctp_assoc_set_primary(trans->asoc, trans);
2135 * 7.1.5 SCTP_NODELAY
2137 * Turn on/off any Nagle-like algorithm. This means that packets are
2138 * generally sent as soon as possible and no unnecessary delays are
2139 * introduced, at the cost of more packets in the network. Expects an
2140 * integer boolean flag.
2142 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2147 if (optlen < sizeof(int))
2149 if (get_user(val, (int __user *)optval))
2152 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2158 * 7.1.1 SCTP_RTOINFO
2160 * The protocol parameters used to initialize and bound retransmission
2161 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2162 * and modify these parameters.
2163 * All parameters are time values, in milliseconds. A value of 0, when
2164 * modifying the parameters, indicates that the current value should not
2168 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2169 struct sctp_rtoinfo rtoinfo;
2170 struct sctp_association *asoc;
2172 if (optlen != sizeof (struct sctp_rtoinfo))
2175 if (copy_from_user(&rtoinfo, optval, optlen))
2178 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2180 /* Set the values to the specific association */
2181 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2185 if (rtoinfo.srto_initial != 0)
2187 msecs_to_jiffies(rtoinfo.srto_initial);
2188 if (rtoinfo.srto_max != 0)
2189 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2190 if (rtoinfo.srto_min != 0)
2191 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2193 /* If there is no association or the association-id = 0
2194 * set the values to the endpoint.
2196 struct sctp_sock *sp = sctp_sk(sk);
2198 if (rtoinfo.srto_initial != 0)
2199 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2200 if (rtoinfo.srto_max != 0)
2201 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2202 if (rtoinfo.srto_min != 0)
2203 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2211 * 7.1.2 SCTP_ASSOCINFO
2213 * This option is used to tune the the maximum retransmission attempts
2214 * of the association.
2215 * Returns an error if the new association retransmission value is
2216 * greater than the sum of the retransmission value of the peer.
2217 * See [SCTP] for more information.
2220 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2223 struct sctp_assocparams assocparams;
2224 struct sctp_association *asoc;
2226 if (optlen != sizeof(struct sctp_assocparams))
2228 if (copy_from_user(&assocparams, optval, optlen))
2231 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2233 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2236 /* Set the values to the specific association */
2238 if (assocparams.sasoc_asocmaxrxt != 0)
2239 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2240 if (assocparams.sasoc_cookie_life != 0) {
2241 asoc->cookie_life.tv_sec =
2242 assocparams.sasoc_cookie_life / 1000;
2243 asoc->cookie_life.tv_usec =
2244 (assocparams.sasoc_cookie_life % 1000)
2248 /* Set the values to the endpoint */
2249 struct sctp_sock *sp = sctp_sk(sk);
2251 if (assocparams.sasoc_asocmaxrxt != 0)
2252 sp->assocparams.sasoc_asocmaxrxt =
2253 assocparams.sasoc_asocmaxrxt;
2254 if (assocparams.sasoc_cookie_life != 0)
2255 sp->assocparams.sasoc_cookie_life =
2256 assocparams.sasoc_cookie_life;
2262 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2264 * This socket option is a boolean flag which turns on or off mapped V4
2265 * addresses. If this option is turned on and the socket is type
2266 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2267 * If this option is turned off, then no mapping will be done of V4
2268 * addresses and a user will receive both PF_INET6 and PF_INET type
2269 * addresses on the socket.
2271 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2274 struct sctp_sock *sp = sctp_sk(sk);
2276 if (optlen < sizeof(int))
2278 if (get_user(val, (int __user *)optval))
2289 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2291 * This socket option specifies the maximum size to put in any outgoing
2292 * SCTP chunk. If a message is larger than this size it will be
2293 * fragmented by SCTP into the specified size. Note that the underlying
2294 * SCTP implementation may fragment into smaller sized chunks when the
2295 * PMTU of the underlying association is smaller than the value set by
2298 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2300 struct sctp_association *asoc;
2301 struct list_head *pos;
2302 struct sctp_sock *sp = sctp_sk(sk);
2305 if (optlen < sizeof(int))
2307 if (get_user(val, (int __user *)optval))
2309 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2311 sp->user_frag = val;
2313 /* Update the frag_point of the existing associations. */
2314 list_for_each(pos, &(sp->ep->asocs)) {
2315 asoc = list_entry(pos, struct sctp_association, asocs);
2316 asoc->frag_point = sctp_frag_point(sp, asoc->pmtu);
2324 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2326 * Requests that the peer mark the enclosed address as the association
2327 * primary. The enclosed address must be one of the association's
2328 * locally bound addresses. The following structure is used to make a
2329 * set primary request:
2331 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2334 struct sctp_sock *sp;
2335 struct sctp_endpoint *ep;
2336 struct sctp_association *asoc = NULL;
2337 struct sctp_setpeerprim prim;
2338 struct sctp_chunk *chunk;
2344 if (!sctp_addip_enable)
2347 if (optlen != sizeof(struct sctp_setpeerprim))
2350 if (copy_from_user(&prim, optval, optlen))
2353 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2357 if (!asoc->peer.asconf_capable)
2360 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2363 if (!sctp_state(asoc, ESTABLISHED))
2366 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2367 return -EADDRNOTAVAIL;
2369 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2370 chunk = sctp_make_asconf_set_prim(asoc,
2371 (union sctp_addr *)&prim.sspp_addr);
2375 err = sctp_send_asconf(asoc, chunk);
2377 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2382 static int sctp_setsockopt_adaption_layer(struct sock *sk, char __user *optval,
2385 struct sctp_setadaption adaption;
2387 if (optlen != sizeof(struct sctp_setadaption))
2389 if (copy_from_user(&adaption, optval, optlen))
2392 sctp_sk(sk)->adaption_ind = adaption.ssb_adaption_ind;
2397 /* API 6.2 setsockopt(), getsockopt()
2399 * Applications use setsockopt() and getsockopt() to set or retrieve
2400 * socket options. Socket options are used to change the default
2401 * behavior of sockets calls. They are described in Section 7.
2405 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2406 * int __user *optlen);
2407 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2410 * sd - the socket descript.
2411 * level - set to IPPROTO_SCTP for all SCTP options.
2412 * optname - the option name.
2413 * optval - the buffer to store the value of the option.
2414 * optlen - the size of the buffer.
2416 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2417 char __user *optval, int optlen)
2421 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2424 /* I can hardly begin to describe how wrong this is. This is
2425 * so broken as to be worse than useless. The API draft
2426 * REALLY is NOT helpful here... I am not convinced that the
2427 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2428 * are at all well-founded.
2430 if (level != SOL_SCTP) {
2431 struct sctp_af *af = sctp_sk(sk)->pf->af;
2432 retval = af->setsockopt(sk, level, optname, optval, optlen);
2439 case SCTP_SOCKOPT_BINDX_ADD:
2440 /* 'optlen' is the size of the addresses buffer. */
2441 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2442 optlen, SCTP_BINDX_ADD_ADDR);
2445 case SCTP_SOCKOPT_BINDX_REM:
2446 /* 'optlen' is the size of the addresses buffer. */
2447 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2448 optlen, SCTP_BINDX_REM_ADDR);
2451 case SCTP_SOCKOPT_CONNECTX:
2452 /* 'optlen' is the size of the addresses buffer. */
2453 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2457 case SCTP_DISABLE_FRAGMENTS:
2458 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2462 retval = sctp_setsockopt_events(sk, optval, optlen);
2465 case SCTP_AUTOCLOSE:
2466 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2469 case SCTP_PEER_ADDR_PARAMS:
2470 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2474 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2476 case SCTP_DEFAULT_SEND_PARAM:
2477 retval = sctp_setsockopt_default_send_param(sk, optval,
2480 case SCTP_PRIMARY_ADDR:
2481 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2483 case SCTP_SET_PEER_PRIMARY_ADDR:
2484 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2487 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2490 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2492 case SCTP_ASSOCINFO:
2493 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2495 case SCTP_I_WANT_MAPPED_V4_ADDR:
2496 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2499 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2501 case SCTP_ADAPTION_LAYER:
2502 retval = sctp_setsockopt_adaption_layer(sk, optval, optlen);
2506 retval = -ENOPROTOOPT;
2510 sctp_release_sock(sk);
2516 /* API 3.1.6 connect() - UDP Style Syntax
2518 * An application may use the connect() call in the UDP model to initiate an
2519 * association without sending data.
2523 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2525 * sd: the socket descriptor to have a new association added to.
2527 * nam: the address structure (either struct sockaddr_in or struct
2528 * sockaddr_in6 defined in RFC2553 [7]).
2530 * len: the size of the address.
2532 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2540 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2541 __FUNCTION__, sk, addr, addr_len);
2543 /* Validate addr_len before calling common connect/connectx routine. */
2544 af = sctp_get_af_specific(addr->sa_family);
2545 if (!af || addr_len < af->sockaddr_len) {
2548 /* Pass correct addr len to common routine (so it knows there
2549 * is only one address being passed.
2551 err = __sctp_connect(sk, addr, af->sockaddr_len);
2554 sctp_release_sock(sk);
2558 /* FIXME: Write comments. */
2559 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2561 return -EOPNOTSUPP; /* STUB */
2564 /* 4.1.4 accept() - TCP Style Syntax
2566 * Applications use accept() call to remove an established SCTP
2567 * association from the accept queue of the endpoint. A new socket
2568 * descriptor will be returned from accept() to represent the newly
2569 * formed association.
2571 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2573 struct sctp_sock *sp;
2574 struct sctp_endpoint *ep;
2575 struct sock *newsk = NULL;
2576 struct sctp_association *asoc;
2585 if (!sctp_style(sk, TCP)) {
2586 error = -EOPNOTSUPP;
2590 if (!sctp_sstate(sk, LISTENING)) {
2595 timeo = sock_rcvtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
2597 error = sctp_wait_for_accept(sk, timeo);
2601 /* We treat the list of associations on the endpoint as the accept
2602 * queue and pick the first association on the list.
2604 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
2606 newsk = sp->pf->create_accept_sk(sk, asoc);
2612 /* Populate the fields of the newsk from the oldsk and migrate the
2613 * asoc to the newsk.
2615 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
2618 sctp_release_sock(sk);
2623 /* The SCTP ioctl handler. */
2624 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
2626 return -ENOIOCTLCMD;
2629 /* This is the function which gets called during socket creation to
2630 * initialized the SCTP-specific portion of the sock.
2631 * The sock structure should already be zero-filled memory.
2633 SCTP_STATIC int sctp_init_sock(struct sock *sk)
2635 struct sctp_endpoint *ep;
2636 struct sctp_sock *sp;
2638 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
2642 /* Initialize the SCTP per socket area. */
2643 switch (sk->sk_type) {
2644 case SOCK_SEQPACKET:
2645 sp->type = SCTP_SOCKET_UDP;
2648 sp->type = SCTP_SOCKET_TCP;
2651 return -ESOCKTNOSUPPORT;
2654 /* Initialize default send parameters. These parameters can be
2655 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
2657 sp->default_stream = 0;
2658 sp->default_ppid = 0;
2659 sp->default_flags = 0;
2660 sp->default_context = 0;
2661 sp->default_timetolive = 0;
2663 /* Initialize default setup parameters. These parameters
2664 * can be modified with the SCTP_INITMSG socket option or
2665 * overridden by the SCTP_INIT CMSG.
2667 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
2668 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
2669 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
2670 sp->initmsg.sinit_max_init_timeo = jiffies_to_msecs(sctp_rto_max);
2672 /* Initialize default RTO related parameters. These parameters can
2673 * be modified for with the SCTP_RTOINFO socket option.
2675 sp->rtoinfo.srto_initial = jiffies_to_msecs(sctp_rto_initial);
2676 sp->rtoinfo.srto_max = jiffies_to_msecs(sctp_rto_max);
2677 sp->rtoinfo.srto_min = jiffies_to_msecs(sctp_rto_min);
2679 /* Initialize default association related parameters. These parameters
2680 * can be modified with the SCTP_ASSOCINFO socket option.
2682 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
2683 sp->assocparams.sasoc_number_peer_destinations = 0;
2684 sp->assocparams.sasoc_peer_rwnd = 0;
2685 sp->assocparams.sasoc_local_rwnd = 0;
2686 sp->assocparams.sasoc_cookie_life =
2687 jiffies_to_msecs(sctp_valid_cookie_life);
2689 /* Initialize default event subscriptions. By default, all the
2692 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
2694 /* Default Peer Address Parameters. These defaults can
2695 * be modified via SCTP_PEER_ADDR_PARAMS
2697 sp->paddrparam.spp_hbinterval = jiffies_to_msecs(sctp_hb_interval);
2698 sp->paddrparam.spp_pathmaxrxt = sctp_max_retrans_path;
2700 /* If enabled no SCTP message fragmentation will be performed.
2701 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
2703 sp->disable_fragments = 0;
2705 /* Turn on/off any Nagle-like algorithm. */
2708 /* Enable by default. */
2711 /* Auto-close idle associations after the configured
2712 * number of seconds. A value of 0 disables this
2713 * feature. Configure through the SCTP_AUTOCLOSE socket option,
2714 * for UDP-style sockets only.
2718 /* User specified fragmentation limit. */
2721 sp->adaption_ind = 0;
2723 sp->pf = sctp_get_pf_specific(sk->sk_family);
2725 /* Control variables for partial data delivery. */
2727 skb_queue_head_init(&sp->pd_lobby);
2729 /* Create a per socket endpoint structure. Even if we
2730 * change the data structure relationships, this may still
2731 * be useful for storing pre-connect address information.
2733 ep = sctp_endpoint_new(sk, GFP_KERNEL);
2740 SCTP_DBG_OBJCNT_INC(sock);
2744 /* Cleanup any SCTP per socket resources. */
2745 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
2747 struct sctp_endpoint *ep;
2749 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
2751 /* Release our hold on the endpoint. */
2752 ep = sctp_sk(sk)->ep;
2753 sctp_endpoint_free(ep);
2758 /* API 4.1.7 shutdown() - TCP Style Syntax
2759 * int shutdown(int socket, int how);
2761 * sd - the socket descriptor of the association to be closed.
2762 * how - Specifies the type of shutdown. The values are
2765 * Disables further receive operations. No SCTP
2766 * protocol action is taken.
2768 * Disables further send operations, and initiates
2769 * the SCTP shutdown sequence.
2771 * Disables further send and receive operations
2772 * and initiates the SCTP shutdown sequence.
2774 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
2776 struct sctp_endpoint *ep;
2777 struct sctp_association *asoc;
2779 if (!sctp_style(sk, TCP))
2782 if (how & SEND_SHUTDOWN) {
2783 ep = sctp_sk(sk)->ep;
2784 if (!list_empty(&ep->asocs)) {
2785 asoc = list_entry(ep->asocs.next,
2786 struct sctp_association, asocs);
2787 sctp_primitive_SHUTDOWN(asoc, NULL);
2792 /* 7.2.1 Association Status (SCTP_STATUS)
2794 * Applications can retrieve current status information about an
2795 * association, including association state, peer receiver window size,
2796 * number of unacked data chunks, and number of data chunks pending
2797 * receipt. This information is read-only.
2799 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
2800 char __user *optval,
2803 struct sctp_status status;
2804 struct sctp_association *asoc = NULL;
2805 struct sctp_transport *transport;
2806 sctp_assoc_t associd;
2809 if (len != sizeof(status)) {
2814 if (copy_from_user(&status, optval, sizeof(status))) {
2819 associd = status.sstat_assoc_id;
2820 asoc = sctp_id2assoc(sk, associd);
2826 transport = asoc->peer.primary_path;
2828 status.sstat_assoc_id = sctp_assoc2id(asoc);
2829 status.sstat_state = asoc->state;
2830 status.sstat_rwnd = asoc->peer.rwnd;
2831 status.sstat_unackdata = asoc->unack_data;
2833 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
2834 status.sstat_instrms = asoc->c.sinit_max_instreams;
2835 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
2836 status.sstat_fragmentation_point = asoc->frag_point;
2837 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
2838 memcpy(&status.sstat_primary.spinfo_address,
2839 &(transport->ipaddr), sizeof(union sctp_addr));
2840 /* Map ipv4 address into v4-mapped-on-v6 address. */
2841 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
2842 (union sctp_addr *)&status.sstat_primary.spinfo_address);
2843 status.sstat_primary.spinfo_state = transport->state;
2844 status.sstat_primary.spinfo_cwnd = transport->cwnd;
2845 status.sstat_primary.spinfo_srtt = transport->srtt;
2846 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
2847 status.sstat_primary.spinfo_mtu = transport->pmtu;
2849 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
2850 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
2852 if (put_user(len, optlen)) {
2857 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
2858 len, status.sstat_state, status.sstat_rwnd,
2859 status.sstat_assoc_id);
2861 if (copy_to_user(optval, &status, len)) {
2871 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
2873 * Applications can retrieve information about a specific peer address
2874 * of an association, including its reachability state, congestion
2875 * window, and retransmission timer values. This information is
2878 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
2879 char __user *optval,
2882 struct sctp_paddrinfo pinfo;
2883 struct sctp_transport *transport;
2886 if (len != sizeof(pinfo)) {
2891 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
2896 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
2897 pinfo.spinfo_assoc_id);
2901 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
2902 pinfo.spinfo_state = transport->state;
2903 pinfo.spinfo_cwnd = transport->cwnd;
2904 pinfo.spinfo_srtt = transport->srtt;
2905 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
2906 pinfo.spinfo_mtu = transport->pmtu;
2908 if (pinfo.spinfo_state == SCTP_UNKNOWN)
2909 pinfo.spinfo_state = SCTP_ACTIVE;
2911 if (put_user(len, optlen)) {
2916 if (copy_to_user(optval, &pinfo, len)) {
2925 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2927 * This option is a on/off flag. If enabled no SCTP message
2928 * fragmentation will be performed. Instead if a message being sent
2929 * exceeds the current PMTU size, the message will NOT be sent and
2930 * instead a error will be indicated to the user.
2932 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
2933 char __user *optval, int __user *optlen)
2937 if (len < sizeof(int))
2941 val = (sctp_sk(sk)->disable_fragments == 1);
2942 if (put_user(len, optlen))
2944 if (copy_to_user(optval, &val, len))
2949 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
2951 * This socket option is used to specify various notifications and
2952 * ancillary data the user wishes to receive.
2954 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
2957 if (len != sizeof(struct sctp_event_subscribe))
2959 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
2964 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2966 * This socket option is applicable to the UDP-style socket only. When
2967 * set it will cause associations that are idle for more than the
2968 * specified number of seconds to automatically close. An association
2969 * being idle is defined an association that has NOT sent or received
2970 * user data. The special value of '0' indicates that no automatic
2971 * close of any associations should be performed. The option expects an
2972 * integer defining the number of seconds of idle time before an
2973 * association is closed.
2975 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
2977 /* Applicable to UDP-style socket only */
2978 if (sctp_style(sk, TCP))
2980 if (len != sizeof(int))
2982 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
2987 /* Helper routine to branch off an association to a new socket. */
2988 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
2989 struct socket **sockp)
2991 struct sock *sk = asoc->base.sk;
2992 struct socket *sock;
2995 /* An association cannot be branched off from an already peeled-off
2996 * socket, nor is this supported for tcp style sockets.
2998 if (!sctp_style(sk, UDP))
3001 /* Create a new socket. */
3002 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3006 /* Populate the fields of the newsk from the oldsk and migrate the
3007 * asoc to the newsk.
3009 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3015 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3017 sctp_peeloff_arg_t peeloff;
3018 struct socket *newsock;
3020 struct sctp_association *asoc;
3022 if (len != sizeof(sctp_peeloff_arg_t))
3024 if (copy_from_user(&peeloff, optval, len))
3027 asoc = sctp_id2assoc(sk, peeloff.associd);
3033 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3035 retval = sctp_do_peeloff(asoc, &newsock);
3039 /* Map the socket to an unused fd that can be returned to the user. */
3040 retval = sock_map_fd(newsock);
3042 sock_release(newsock);
3046 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3047 __FUNCTION__, sk, asoc, newsock->sk, retval);
3049 /* Return the fd mapped to the new socket. */
3050 peeloff.sd = retval;
3051 if (copy_to_user(optval, &peeloff, len))
3058 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3060 * Applications can enable or disable heartbeats for any peer address of
3061 * an association, modify an address's heartbeat interval, force a
3062 * heartbeat to be sent immediately, and adjust the address's maximum
3063 * number of retransmissions sent before an address is considered
3064 * unreachable. The following structure is used to access and modify an
3065 * address's parameters:
3067 * struct sctp_paddrparams {
3068 * sctp_assoc_t spp_assoc_id;
3069 * struct sockaddr_storage spp_address;
3070 * uint32_t spp_hbinterval;
3071 * uint16_t spp_pathmaxrxt;
3074 * spp_assoc_id - (UDP style socket) This is filled in the application,
3075 * and identifies the association for this query.
3076 * spp_address - This specifies which address is of interest.
3077 * spp_hbinterval - This contains the value of the heartbeat interval,
3078 * in milliseconds. A value of 0, when modifying the
3079 * parameter, specifies that the heartbeat on this
3080 * address should be disabled. A value of UINT32_MAX
3081 * (4294967295), when modifying the parameter,
3082 * specifies that a heartbeat should be sent
3083 * immediately to the peer address, and the current
3084 * interval should remain unchanged.
3085 * spp_pathmaxrxt - This contains the maximum number of
3086 * retransmissions before this address shall be
3087 * considered unreachable.
3089 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3090 char __user *optval, int __user *optlen)
3092 struct sctp_paddrparams params;
3093 struct sctp_transport *trans;
3095 if (len != sizeof(struct sctp_paddrparams))
3097 if (copy_from_user(¶ms, optval, len))
3100 /* If no association id is specified retrieve the default value
3101 * for the endpoint that will be used for all future associations
3103 if (!params.spp_assoc_id &&
3104 sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3105 params.spp_hbinterval = sctp_sk(sk)->paddrparam.spp_hbinterval;
3106 params.spp_pathmaxrxt = sctp_sk(sk)->paddrparam.spp_pathmaxrxt;
3111 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3112 params.spp_assoc_id);
3116 /* The value of the heartbeat interval, in milliseconds. A value of 0,
3117 * when modifying the parameter, specifies that the heartbeat on this
3118 * address should be disabled.
3120 if (!trans->hb_allowed)
3121 params.spp_hbinterval = 0;
3123 params.spp_hbinterval = jiffies_to_msecs(trans->hb_interval);
3125 /* spp_pathmaxrxt contains the maximum number of retransmissions
3126 * before this address shall be considered unreachable.
3128 params.spp_pathmaxrxt = trans->max_retrans;
3131 if (copy_to_user(optval, ¶ms, len))
3134 if (put_user(len, optlen))
3140 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3142 * Applications can specify protocol parameters for the default association
3143 * initialization. The option name argument to setsockopt() and getsockopt()
3146 * Setting initialization parameters is effective only on an unconnected
3147 * socket (for UDP-style sockets only future associations are effected
3148 * by the change). With TCP-style sockets, this option is inherited by
3149 * sockets derived from a listener socket.
3151 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3153 if (len != sizeof(struct sctp_initmsg))
3155 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3160 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3161 char __user *optval,
3165 struct sctp_association *asoc;
3166 struct list_head *pos;
3169 if (len != sizeof(sctp_assoc_t))
3172 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3175 /* For UDP-style sockets, id specifies the association to query. */
3176 asoc = sctp_id2assoc(sk, id);
3180 list_for_each(pos, &asoc->peer.transport_addr_list) {
3188 * Old API for getting list of peer addresses. Does not work for 32-bit
3189 * programs running on a 64-bit kernel
3191 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3192 char __user *optval,
3195 struct sctp_association *asoc;
3196 struct list_head *pos;
3198 struct sctp_getaddrs_old getaddrs;
3199 struct sctp_transport *from;
3201 union sctp_addr temp;
3202 struct sctp_sock *sp = sctp_sk(sk);
3205 if (len != sizeof(struct sctp_getaddrs_old))
3208 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3211 if (getaddrs.addr_num <= 0) return -EINVAL;
3213 /* For UDP-style sockets, id specifies the association to query. */
3214 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3218 to = (void __user *)getaddrs.addrs;
3219 list_for_each(pos, &asoc->peer.transport_addr_list) {
3220 from = list_entry(pos, struct sctp_transport, transports);
3221 memcpy(&temp, &from->ipaddr, sizeof(temp));
3222 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3223 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3224 temp.v4.sin_port = htons(temp.v4.sin_port);
3225 if (copy_to_user(to, &temp, addrlen))
3229 if (cnt >= getaddrs.addr_num) break;
3231 getaddrs.addr_num = cnt;
3232 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3238 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3239 char __user *optval, int __user *optlen)
3241 struct sctp_association *asoc;
3242 struct list_head *pos;
3244 struct sctp_getaddrs getaddrs;
3245 struct sctp_transport *from;
3247 union sctp_addr temp;
3248 struct sctp_sock *sp = sctp_sk(sk);
3253 if (len < sizeof(struct sctp_getaddrs))
3256 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3259 /* For UDP-style sockets, id specifies the association to query. */
3260 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3264 to = optval + offsetof(struct sctp_getaddrs,addrs);
3265 space_left = len - sizeof(struct sctp_getaddrs) -
3266 offsetof(struct sctp_getaddrs,addrs);
3268 list_for_each(pos, &asoc->peer.transport_addr_list) {
3269 from = list_entry(pos, struct sctp_transport, transports);
3270 memcpy(&temp, &from->ipaddr, sizeof(temp));
3271 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3272 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3273 if(space_left < addrlen)
3275 temp.v4.sin_port = htons(temp.v4.sin_port);
3276 if (copy_to_user(to, &temp, addrlen))
3280 space_left -= addrlen;
3283 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3285 bytes_copied = ((char __user *)to) - optval;
3286 if (put_user(bytes_copied, optlen))
3292 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3293 char __user *optval,
3297 struct sctp_bind_addr *bp;
3298 struct sctp_association *asoc;
3299 struct list_head *pos;
3300 struct sctp_sockaddr_entry *addr;
3301 rwlock_t *addr_lock;
3302 unsigned long flags;
3305 if (len != sizeof(sctp_assoc_t))
3308 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3312 * For UDP-style sockets, id specifies the association to query.
3313 * If the id field is set to the value '0' then the locally bound
3314 * addresses are returned without regard to any particular
3318 bp = &sctp_sk(sk)->ep->base.bind_addr;
3319 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3321 asoc = sctp_id2assoc(sk, id);
3324 bp = &asoc->base.bind_addr;
3325 addr_lock = &asoc->base.addr_lock;
3328 sctp_read_lock(addr_lock);
3330 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3331 * addresses from the global local address list.
3333 if (sctp_list_single_entry(&bp->address_list)) {
3334 addr = list_entry(bp->address_list.next,
3335 struct sctp_sockaddr_entry, list);
3336 if (sctp_is_any(&addr->a)) {
3337 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3338 list_for_each(pos, &sctp_local_addr_list) {
3339 addr = list_entry(pos,
3340 struct sctp_sockaddr_entry,
3342 if ((PF_INET == sk->sk_family) &&
3343 (AF_INET6 == addr->a.sa.sa_family))
3347 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3355 list_for_each(pos, &bp->address_list) {
3360 sctp_read_unlock(addr_lock);
3364 /* Helper function that copies local addresses to user and returns the number
3365 * of addresses copied.
3367 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3370 struct list_head *pos;
3371 struct sctp_sockaddr_entry *addr;
3372 unsigned long flags;
3373 union sctp_addr temp;
3377 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3378 list_for_each(pos, &sctp_local_addr_list) {
3379 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3380 if ((PF_INET == sk->sk_family) &&
3381 (AF_INET6 == addr->a.sa.sa_family))
3383 memcpy(&temp, &addr->a, sizeof(temp));
3384 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3386 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3387 temp.v4.sin_port = htons(port);
3388 if (copy_to_user(to, &temp, addrlen)) {
3389 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3395 if (cnt >= max_addrs) break;
3397 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3402 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
3403 void * __user *to, size_t space_left)
3405 struct list_head *pos;
3406 struct sctp_sockaddr_entry *addr;
3407 unsigned long flags;
3408 union sctp_addr temp;
3412 sctp_spin_lock_irqsave(&sctp_local_addr_lock, flags);
3413 list_for_each(pos, &sctp_local_addr_list) {
3414 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3415 if ((PF_INET == sk->sk_family) &&
3416 (AF_INET6 == addr->a.sa.sa_family))
3418 memcpy(&temp, &addr->a, sizeof(temp));
3419 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3421 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3422 if(space_left<addrlen)
3424 temp.v4.sin_port = htons(port);
3425 if (copy_to_user(*to, &temp, addrlen)) {
3426 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock,
3432 space_left -= addrlen;
3434 sctp_spin_unlock_irqrestore(&sctp_local_addr_lock, flags);
3439 /* Old API for getting list of local addresses. Does not work for 32-bit
3440 * programs running on a 64-bit kernel
3442 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3443 char __user *optval, int __user *optlen)
3445 struct sctp_bind_addr *bp;
3446 struct sctp_association *asoc;
3447 struct list_head *pos;
3449 struct sctp_getaddrs_old getaddrs;
3450 struct sctp_sockaddr_entry *addr;
3452 union sctp_addr temp;
3453 struct sctp_sock *sp = sctp_sk(sk);
3455 rwlock_t *addr_lock;
3458 if (len != sizeof(struct sctp_getaddrs_old))
3461 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3464 if (getaddrs.addr_num <= 0) return -EINVAL;
3466 * For UDP-style sockets, id specifies the association to query.
3467 * If the id field is set to the value '0' then the locally bound
3468 * addresses are returned without regard to any particular
3471 if (0 == getaddrs.assoc_id) {
3472 bp = &sctp_sk(sk)->ep->base.bind_addr;
3473 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3475 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3478 bp = &asoc->base.bind_addr;
3479 addr_lock = &asoc->base.addr_lock;
3482 to = getaddrs.addrs;
3484 sctp_read_lock(addr_lock);
3486 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3487 * addresses from the global local address list.
3489 if (sctp_list_single_entry(&bp->address_list)) {
3490 addr = list_entry(bp->address_list.next,
3491 struct sctp_sockaddr_entry, list);
3492 if (sctp_is_any(&addr->a)) {
3493 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
3504 list_for_each(pos, &bp->address_list) {
3505 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3506 memcpy(&temp, &addr->a, sizeof(temp));
3507 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3508 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3509 temp.v4.sin_port = htons(temp.v4.sin_port);
3510 if (copy_to_user(to, &temp, addrlen)) {
3516 if (cnt >= getaddrs.addr_num) break;
3520 getaddrs.addr_num = cnt;
3521 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3525 sctp_read_unlock(addr_lock);
3529 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
3530 char __user *optval, int __user *optlen)
3532 struct sctp_bind_addr *bp;
3533 struct sctp_association *asoc;
3534 struct list_head *pos;
3536 struct sctp_getaddrs getaddrs;
3537 struct sctp_sockaddr_entry *addr;
3539 union sctp_addr temp;
3540 struct sctp_sock *sp = sctp_sk(sk);
3542 rwlock_t *addr_lock;
3547 if (len <= sizeof(struct sctp_getaddrs))
3550 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3554 * For UDP-style sockets, id specifies the association to query.
3555 * If the id field is set to the value '0' then the locally bound
3556 * addresses are returned without regard to any particular
3559 if (0 == getaddrs.assoc_id) {
3560 bp = &sctp_sk(sk)->ep->base.bind_addr;
3561 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3563 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3566 bp = &asoc->base.bind_addr;
3567 addr_lock = &asoc->base.addr_lock;
3570 to = optval + offsetof(struct sctp_getaddrs,addrs);
3571 space_left = len - sizeof(struct sctp_getaddrs) -
3572 offsetof(struct sctp_getaddrs,addrs);
3574 sctp_read_lock(addr_lock);
3576 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
3577 * addresses from the global local address list.
3579 if (sctp_list_single_entry(&bp->address_list)) {
3580 addr = list_entry(bp->address_list.next,
3581 struct sctp_sockaddr_entry, list);
3582 if (sctp_is_any(&addr->a)) {
3583 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
3593 list_for_each(pos, &bp->address_list) {
3594 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3595 memcpy(&temp, &addr->a, sizeof(temp));
3596 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3597 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3598 if(space_left < addrlen)
3599 return -ENOMEM; /*fixme: right error?*/
3600 temp.v4.sin_port = htons(temp.v4.sin_port);
3601 if (copy_to_user(to, &temp, addrlen)) {
3607 space_left -= addrlen;
3611 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3613 bytes_copied = ((char __user *)to) - optval;
3614 if (put_user(bytes_copied, optlen))
3618 sctp_read_unlock(addr_lock);
3622 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3624 * Requests that the local SCTP stack use the enclosed peer address as
3625 * the association primary. The enclosed address must be one of the
3626 * association peer's addresses.
3628 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
3629 char __user *optval, int __user *optlen)
3631 struct sctp_prim prim;
3632 struct sctp_association *asoc;
3633 struct sctp_sock *sp = sctp_sk(sk);
3635 if (len != sizeof(struct sctp_prim))
3638 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3641 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
3645 if (!asoc->peer.primary_path)
3648 asoc->peer.primary_path->ipaddr.v4.sin_port =
3649 htons(asoc->peer.primary_path->ipaddr.v4.sin_port);
3650 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
3651 sizeof(union sctp_addr));
3652 asoc->peer.primary_path->ipaddr.v4.sin_port =
3653 ntohs(asoc->peer.primary_path->ipaddr.v4.sin_port);
3655 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
3656 (union sctp_addr *)&prim.ssp_addr);
3658 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
3665 * 7.1.11 Set Adaption Layer Indicator (SCTP_ADAPTION_LAYER)
3667 * Requests that the local endpoint set the specified Adaption Layer
3668 * Indication parameter for all future INIT and INIT-ACK exchanges.
3670 static int sctp_getsockopt_adaption_layer(struct sock *sk, int len,
3671 char __user *optval, int __user *optlen)
3673 struct sctp_setadaption adaption;
3675 if (len != sizeof(struct sctp_setadaption))
3678 adaption.ssb_adaption_ind = sctp_sk(sk)->adaption_ind;
3679 if (copy_to_user(optval, &adaption, len))
3687 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
3689 * Applications that wish to use the sendto() system call may wish to
3690 * specify a default set of parameters that would normally be supplied
3691 * through the inclusion of ancillary data. This socket option allows
3692 * such an application to set the default sctp_sndrcvinfo structure.
3695 * The application that wishes to use this socket option simply passes
3696 * in to this call the sctp_sndrcvinfo structure defined in Section
3697 * 5.2.2) The input parameters accepted by this call include
3698 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
3699 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
3700 * to this call if the caller is using the UDP model.
3702 * For getsockopt, it get the default sctp_sndrcvinfo structure.
3704 static int sctp_getsockopt_default_send_param(struct sock *sk,
3705 int len, char __user *optval,
3708 struct sctp_sndrcvinfo info;
3709 struct sctp_association *asoc;
3710 struct sctp_sock *sp = sctp_sk(sk);
3712 if (len != sizeof(struct sctp_sndrcvinfo))
3714 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
3717 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3718 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
3722 info.sinfo_stream = asoc->default_stream;
3723 info.sinfo_flags = asoc->default_flags;
3724 info.sinfo_ppid = asoc->default_ppid;
3725 info.sinfo_context = asoc->default_context;
3726 info.sinfo_timetolive = asoc->default_timetolive;
3728 info.sinfo_stream = sp->default_stream;
3729 info.sinfo_flags = sp->default_flags;
3730 info.sinfo_ppid = sp->default_ppid;
3731 info.sinfo_context = sp->default_context;
3732 info.sinfo_timetolive = sp->default_timetolive;
3735 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
3743 * 7.1.5 SCTP_NODELAY
3745 * Turn on/off any Nagle-like algorithm. This means that packets are
3746 * generally sent as soon as possible and no unnecessary delays are
3747 * introduced, at the cost of more packets in the network. Expects an
3748 * integer boolean flag.
3751 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
3752 char __user *optval, int __user *optlen)
3756 if (len < sizeof(int))
3760 val = (sctp_sk(sk)->nodelay == 1);
3761 if (put_user(len, optlen))
3763 if (copy_to_user(optval, &val, len))
3770 * 7.1.1 SCTP_RTOINFO
3772 * The protocol parameters used to initialize and bound retransmission
3773 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3774 * and modify these parameters.
3775 * All parameters are time values, in milliseconds. A value of 0, when
3776 * modifying the parameters, indicates that the current value should not
3780 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
3781 char __user *optval,
3782 int __user *optlen) {
3783 struct sctp_rtoinfo rtoinfo;
3784 struct sctp_association *asoc;
3786 if (len != sizeof (struct sctp_rtoinfo))
3789 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
3792 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3794 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3797 /* Values corresponding to the specific association. */
3799 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
3800 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
3801 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
3803 /* Values corresponding to the endpoint. */
3804 struct sctp_sock *sp = sctp_sk(sk);
3806 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
3807 rtoinfo.srto_max = sp->rtoinfo.srto_max;
3808 rtoinfo.srto_min = sp->rtoinfo.srto_min;
3811 if (put_user(len, optlen))
3814 if (copy_to_user(optval, &rtoinfo, len))
3822 * 7.1.2 SCTP_ASSOCINFO
3824 * This option is used to tune the the maximum retransmission attempts
3825 * of the association.
3826 * Returns an error if the new association retransmission value is
3827 * greater than the sum of the retransmission value of the peer.
3828 * See [SCTP] for more information.
3831 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
3832 char __user *optval,
3836 struct sctp_assocparams assocparams;
3837 struct sctp_association *asoc;
3838 struct list_head *pos;
3841 if (len != sizeof (struct sctp_assocparams))
3844 if (copy_from_user(&assocparams, optval,
3845 sizeof (struct sctp_assocparams)))
3848 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3850 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3853 /* Values correspoinding to the specific association */
3855 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
3856 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
3857 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
3858 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
3860 (asoc->cookie_life.tv_usec
3863 list_for_each(pos, &asoc->peer.transport_addr_list) {
3867 assocparams.sasoc_number_peer_destinations = cnt;
3869 /* Values corresponding to the endpoint */
3870 struct sctp_sock *sp = sctp_sk(sk);
3872 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
3873 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
3874 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
3875 assocparams.sasoc_cookie_life =
3876 sp->assocparams.sasoc_cookie_life;
3877 assocparams.sasoc_number_peer_destinations =
3879 sasoc_number_peer_destinations;
3882 if (put_user(len, optlen))
3885 if (copy_to_user(optval, &assocparams, len))
3892 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3894 * This socket option is a boolean flag which turns on or off mapped V4
3895 * addresses. If this option is turned on and the socket is type
3896 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3897 * If this option is turned off, then no mapping will be done of V4
3898 * addresses and a user will receive both PF_INET6 and PF_INET type
3899 * addresses on the socket.
3901 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
3902 char __user *optval, int __user *optlen)
3905 struct sctp_sock *sp = sctp_sk(sk);
3907 if (len < sizeof(int))
3912 if (put_user(len, optlen))
3914 if (copy_to_user(optval, &val, len))
3921 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
3923 * This socket option specifies the maximum size to put in any outgoing
3924 * SCTP chunk. If a message is larger than this size it will be
3925 * fragmented by SCTP into the specified size. Note that the underlying
3926 * SCTP implementation may fragment into smaller sized chunks when the
3927 * PMTU of the underlying association is smaller than the value set by
3930 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
3931 char __user *optval, int __user *optlen)
3935 if (len < sizeof(int))
3940 val = sctp_sk(sk)->user_frag;
3941 if (put_user(len, optlen))
3943 if (copy_to_user(optval, &val, len))
3949 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
3950 char __user *optval, int __user *optlen)
3955 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
3958 /* I can hardly begin to describe how wrong this is. This is
3959 * so broken as to be worse than useless. The API draft
3960 * REALLY is NOT helpful here... I am not convinced that the
3961 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
3962 * are at all well-founded.
3964 if (level != SOL_SCTP) {
3965 struct sctp_af *af = sctp_sk(sk)->pf->af;
3967 retval = af->getsockopt(sk, level, optname, optval, optlen);
3971 if (get_user(len, optlen))
3978 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
3980 case SCTP_DISABLE_FRAGMENTS:
3981 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
3985 retval = sctp_getsockopt_events(sk, len, optval, optlen);
3987 case SCTP_AUTOCLOSE:
3988 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
3990 case SCTP_SOCKOPT_PEELOFF:
3991 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
3993 case SCTP_PEER_ADDR_PARAMS:
3994 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
3998 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4000 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4001 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4004 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4005 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4008 case SCTP_GET_PEER_ADDRS_OLD:
4009 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4012 case SCTP_GET_LOCAL_ADDRS_OLD:
4013 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4016 case SCTP_GET_PEER_ADDRS:
4017 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4020 case SCTP_GET_LOCAL_ADDRS:
4021 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4024 case SCTP_DEFAULT_SEND_PARAM:
4025 retval = sctp_getsockopt_default_send_param(sk, len,
4028 case SCTP_PRIMARY_ADDR:
4029 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4032 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4035 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4037 case SCTP_ASSOCINFO:
4038 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4040 case SCTP_I_WANT_MAPPED_V4_ADDR:
4041 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4044 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4046 case SCTP_GET_PEER_ADDR_INFO:
4047 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4050 case SCTP_ADAPTION_LAYER:
4051 retval = sctp_getsockopt_adaption_layer(sk, len, optval,
4055 retval = -ENOPROTOOPT;
4059 sctp_release_sock(sk);
4063 static void sctp_hash(struct sock *sk)
4068 static void sctp_unhash(struct sock *sk)
4073 /* Check if port is acceptable. Possibly find first available port.
4075 * The port hash table (contained in the 'global' SCTP protocol storage
4076 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4077 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4078 * list (the list number is the port number hashed out, so as you
4079 * would expect from a hash function, all the ports in a given list have
4080 * such a number that hashes out to the same list number; you were
4081 * expecting that, right?); so each list has a set of ports, with a
4082 * link to the socket (struct sock) that uses it, the port number and
4083 * a fastreuse flag (FIXME: NPI ipg).
4085 static struct sctp_bind_bucket *sctp_bucket_create(
4086 struct sctp_bind_hashbucket *head, unsigned short snum);
4088 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4090 struct sctp_bind_hashbucket *head; /* hash list */
4091 struct sctp_bind_bucket *pp; /* hash list port iterator */
4092 unsigned short snum;
4095 /* NOTE: Remember to put this back to net order. */
4096 addr->v4.sin_port = ntohs(addr->v4.sin_port);
4097 snum = addr->v4.sin_port;
4099 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4100 sctp_local_bh_disable();
4103 /* Search for an available port.
4105 * 'sctp_port_rover' was the last port assigned, so
4106 * we start to search from 'sctp_port_rover +
4107 * 1'. What we do is first check if port 'rover' is
4108 * already in the hash table; if not, we use that; if
4109 * it is, we try next.
4111 int low = sysctl_local_port_range[0];
4112 int high = sysctl_local_port_range[1];
4113 int remaining = (high - low) + 1;
4117 sctp_spin_lock(&sctp_port_alloc_lock);
4118 rover = sctp_port_rover;
4121 if ((rover < low) || (rover > high))
4123 index = sctp_phashfn(rover);
4124 head = &sctp_port_hashtable[index];
4125 sctp_spin_lock(&head->lock);
4126 for (pp = head->chain; pp; pp = pp->next)
4127 if (pp->port == rover)
4131 sctp_spin_unlock(&head->lock);
4132 } while (--remaining > 0);
4133 sctp_port_rover = rover;
4134 sctp_spin_unlock(&sctp_port_alloc_lock);
4136 /* Exhausted local port range during search? */
4141 /* OK, here is the one we will use. HEAD (the port
4142 * hash table list entry) is non-NULL and we hold it's
4147 /* We are given an specific port number; we verify
4148 * that it is not being used. If it is used, we will
4149 * exahust the search in the hash list corresponding
4150 * to the port number (snum) - we detect that with the
4151 * port iterator, pp being NULL.
4153 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4154 sctp_spin_lock(&head->lock);
4155 for (pp = head->chain; pp; pp = pp->next) {
4156 if (pp->port == snum)
4163 if (!hlist_empty(&pp->owner)) {
4164 /* We had a port hash table hit - there is an
4165 * available port (pp != NULL) and it is being
4166 * used by other socket (pp->owner not empty); that other
4167 * socket is going to be sk2.
4169 int reuse = sk->sk_reuse;
4171 struct hlist_node *node;
4173 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4174 if (pp->fastreuse && sk->sk_reuse)
4177 /* Run through the list of sockets bound to the port
4178 * (pp->port) [via the pointers bind_next and
4179 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4180 * we get the endpoint they describe and run through
4181 * the endpoint's list of IP (v4 or v6) addresses,
4182 * comparing each of the addresses with the address of
4183 * the socket sk. If we find a match, then that means
4184 * that this port/socket (sk) combination are already
4187 sk_for_each_bound(sk2, node, &pp->owner) {
4188 struct sctp_endpoint *ep2;
4189 ep2 = sctp_sk(sk2)->ep;
4191 if (reuse && sk2->sk_reuse)
4194 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4200 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4203 /* If there was a hash table miss, create a new port. */
4205 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4208 /* In either case (hit or miss), make sure fastreuse is 1 only
4209 * if sk->sk_reuse is too (that is, if the caller requested
4210 * SO_REUSEADDR on this socket -sk-).
4212 if (hlist_empty(&pp->owner))
4213 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4214 else if (pp->fastreuse && !sk->sk_reuse)
4217 /* We are set, so fill up all the data in the hash table
4218 * entry, tie the socket list information with the rest of the
4219 * sockets FIXME: Blurry, NPI (ipg).
4222 inet_sk(sk)->num = snum;
4223 if (!sctp_sk(sk)->bind_hash) {
4224 sk_add_bind_node(sk, &pp->owner);
4225 sctp_sk(sk)->bind_hash = pp;
4230 sctp_spin_unlock(&head->lock);
4233 sctp_local_bh_enable();
4234 addr->v4.sin_port = htons(addr->v4.sin_port);
4238 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4239 * port is requested.
4241 static int sctp_get_port(struct sock *sk, unsigned short snum)
4244 union sctp_addr addr;
4245 struct sctp_af *af = sctp_sk(sk)->pf->af;
4247 /* Set up a dummy address struct from the sk. */
4248 af->from_sk(&addr, sk);
4249 addr.v4.sin_port = htons(snum);
4251 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4252 ret = sctp_get_port_local(sk, &addr);
4254 return (ret ? 1 : 0);
4258 * 3.1.3 listen() - UDP Style Syntax
4260 * By default, new associations are not accepted for UDP style sockets.
4261 * An application uses listen() to mark a socket as being able to
4262 * accept new associations.
4264 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4266 struct sctp_sock *sp = sctp_sk(sk);
4267 struct sctp_endpoint *ep = sp->ep;
4269 /* Only UDP style sockets that are not peeled off are allowed to
4272 if (!sctp_style(sk, UDP))
4275 /* If backlog is zero, disable listening. */
4277 if (sctp_sstate(sk, CLOSED))
4280 sctp_unhash_endpoint(ep);
4281 sk->sk_state = SCTP_SS_CLOSED;
4284 /* Return if we are already listening. */
4285 if (sctp_sstate(sk, LISTENING))
4289 * If a bind() or sctp_bindx() is not called prior to a listen()
4290 * call that allows new associations to be accepted, the system
4291 * picks an ephemeral port and will choose an address set equivalent
4292 * to binding with a wildcard address.
4294 * This is not currently spelled out in the SCTP sockets
4295 * extensions draft, but follows the practice as seen in TCP
4298 if (!ep->base.bind_addr.port) {
4299 if (sctp_autobind(sk))
4302 sk->sk_state = SCTP_SS_LISTENING;
4303 sctp_hash_endpoint(ep);
4308 * 4.1.3 listen() - TCP Style Syntax
4310 * Applications uses listen() to ready the SCTP endpoint for accepting
4311 * inbound associations.
4313 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4315 struct sctp_sock *sp = sctp_sk(sk);
4316 struct sctp_endpoint *ep = sp->ep;
4318 /* If backlog is zero, disable listening. */
4320 if (sctp_sstate(sk, CLOSED))
4323 sctp_unhash_endpoint(ep);
4324 sk->sk_state = SCTP_SS_CLOSED;
4327 if (sctp_sstate(sk, LISTENING))
4331 * If a bind() or sctp_bindx() is not called prior to a listen()
4332 * call that allows new associations to be accepted, the system
4333 * picks an ephemeral port and will choose an address set equivalent
4334 * to binding with a wildcard address.
4336 * This is not currently spelled out in the SCTP sockets
4337 * extensions draft, but follows the practice as seen in TCP
4340 if (!ep->base.bind_addr.port) {
4341 if (sctp_autobind(sk))
4344 sk->sk_state = SCTP_SS_LISTENING;
4345 sk->sk_max_ack_backlog = backlog;
4346 sctp_hash_endpoint(ep);
4351 * Move a socket to LISTENING state.
4353 int sctp_inet_listen(struct socket *sock, int backlog)
4355 struct sock *sk = sock->sk;
4356 struct crypto_tfm *tfm=NULL;
4359 if (unlikely(backlog < 0))
4364 if (sock->state != SS_UNCONNECTED)
4367 /* Allocate HMAC for generating cookie. */
4368 if (sctp_hmac_alg) {
4369 tfm = sctp_crypto_alloc_tfm(sctp_hmac_alg, 0);
4376 switch (sock->type) {
4377 case SOCK_SEQPACKET:
4378 err = sctp_seqpacket_listen(sk, backlog);
4381 err = sctp_stream_listen(sk, backlog);
4389 /* Store away the transform reference. */
4390 sctp_sk(sk)->hmac = tfm;
4392 sctp_release_sock(sk);
4395 sctp_crypto_free_tfm(tfm);
4400 * This function is done by modeling the current datagram_poll() and the
4401 * tcp_poll(). Note that, based on these implementations, we don't
4402 * lock the socket in this function, even though it seems that,
4403 * ideally, locking or some other mechanisms can be used to ensure
4404 * the integrity of the counters (sndbuf and wmem_queued) used
4405 * in this place. We assume that we don't need locks either until proven
4408 * Another thing to note is that we include the Async I/O support
4409 * here, again, by modeling the current TCP/UDP code. We don't have
4410 * a good way to test with it yet.
4412 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
4414 struct sock *sk = sock->sk;
4415 struct sctp_sock *sp = sctp_sk(sk);
4418 poll_wait(file, sk->sk_sleep, wait);
4420 /* A TCP-style listening socket becomes readable when the accept queue
4423 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4424 return (!list_empty(&sp->ep->asocs)) ?
4425 (POLLIN | POLLRDNORM) : 0;
4429 /* Is there any exceptional events? */
4430 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
4432 if (sk->sk_shutdown == SHUTDOWN_MASK)
4435 /* Is it readable? Reconsider this code with TCP-style support. */
4436 if (!skb_queue_empty(&sk->sk_receive_queue) ||
4437 (sk->sk_shutdown & RCV_SHUTDOWN))
4438 mask |= POLLIN | POLLRDNORM;
4440 /* The association is either gone or not ready. */
4441 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
4444 /* Is it writable? */
4445 if (sctp_writeable(sk)) {
4446 mask |= POLLOUT | POLLWRNORM;
4448 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
4450 * Since the socket is not locked, the buffer
4451 * might be made available after the writeable check and
4452 * before the bit is set. This could cause a lost I/O
4453 * signal. tcp_poll() has a race breaker for this race
4454 * condition. Based on their implementation, we put
4455 * in the following code to cover it as well.
4457 if (sctp_writeable(sk))
4458 mask |= POLLOUT | POLLWRNORM;
4463 /********************************************************************
4464 * 2nd Level Abstractions
4465 ********************************************************************/
4467 static struct sctp_bind_bucket *sctp_bucket_create(
4468 struct sctp_bind_hashbucket *head, unsigned short snum)
4470 struct sctp_bind_bucket *pp;
4472 pp = kmem_cache_alloc(sctp_bucket_cachep, SLAB_ATOMIC);
4473 SCTP_DBG_OBJCNT_INC(bind_bucket);
4477 INIT_HLIST_HEAD(&pp->owner);
4478 if ((pp->next = head->chain) != NULL)
4479 pp->next->pprev = &pp->next;
4481 pp->pprev = &head->chain;
4486 /* Caller must hold hashbucket lock for this tb with local BH disabled */
4487 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
4489 if (hlist_empty(&pp->owner)) {
4491 pp->next->pprev = pp->pprev;
4492 *(pp->pprev) = pp->next;
4493 kmem_cache_free(sctp_bucket_cachep, pp);
4494 SCTP_DBG_OBJCNT_DEC(bind_bucket);
4498 /* Release this socket's reference to a local port. */
4499 static inline void __sctp_put_port(struct sock *sk)
4501 struct sctp_bind_hashbucket *head =
4502 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
4503 struct sctp_bind_bucket *pp;
4505 sctp_spin_lock(&head->lock);
4506 pp = sctp_sk(sk)->bind_hash;
4507 __sk_del_bind_node(sk);
4508 sctp_sk(sk)->bind_hash = NULL;
4509 inet_sk(sk)->num = 0;
4510 sctp_bucket_destroy(pp);
4511 sctp_spin_unlock(&head->lock);
4514 void sctp_put_port(struct sock *sk)
4516 sctp_local_bh_disable();
4517 __sctp_put_port(sk);
4518 sctp_local_bh_enable();
4522 * The system picks an ephemeral port and choose an address set equivalent
4523 * to binding with a wildcard address.
4524 * One of those addresses will be the primary address for the association.
4525 * This automatically enables the multihoming capability of SCTP.
4527 static int sctp_autobind(struct sock *sk)
4529 union sctp_addr autoaddr;
4531 unsigned short port;
4533 /* Initialize a local sockaddr structure to INADDR_ANY. */
4534 af = sctp_sk(sk)->pf->af;
4536 port = htons(inet_sk(sk)->num);
4537 af->inaddr_any(&autoaddr, port);
4539 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
4542 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
4545 * 4.2 The cmsghdr Structure *
4547 * When ancillary data is sent or received, any number of ancillary data
4548 * objects can be specified by the msg_control and msg_controllen members of
4549 * the msghdr structure, because each object is preceded by
4550 * a cmsghdr structure defining the object's length (the cmsg_len member).
4551 * Historically Berkeley-derived implementations have passed only one object
4552 * at a time, but this API allows multiple objects to be
4553 * passed in a single call to sendmsg() or recvmsg(). The following example
4554 * shows two ancillary data objects in a control buffer.
4556 * |<--------------------------- msg_controllen -------------------------->|
4559 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
4561 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
4564 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
4566 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
4569 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4570 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
4572 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
4574 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
4581 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
4582 sctp_cmsgs_t *cmsgs)
4584 struct cmsghdr *cmsg;
4586 for (cmsg = CMSG_FIRSTHDR(msg);
4588 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
4589 if (!CMSG_OK(msg, cmsg))
4592 /* Should we parse this header or ignore? */
4593 if (cmsg->cmsg_level != IPPROTO_SCTP)
4596 /* Strictly check lengths following example in SCM code. */
4597 switch (cmsg->cmsg_type) {
4599 /* SCTP Socket API Extension
4600 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
4602 * This cmsghdr structure provides information for
4603 * initializing new SCTP associations with sendmsg().
4604 * The SCTP_INITMSG socket option uses this same data
4605 * structure. This structure is not used for
4608 * cmsg_level cmsg_type cmsg_data[]
4609 * ------------ ------------ ----------------------
4610 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
4612 if (cmsg->cmsg_len !=
4613 CMSG_LEN(sizeof(struct sctp_initmsg)))
4615 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
4619 /* SCTP Socket API Extension
4620 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
4622 * This cmsghdr structure specifies SCTP options for
4623 * sendmsg() and describes SCTP header information
4624 * about a received message through recvmsg().
4626 * cmsg_level cmsg_type cmsg_data[]
4627 * ------------ ------------ ----------------------
4628 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
4630 if (cmsg->cmsg_len !=
4631 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
4635 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
4637 /* Minimally, validate the sinfo_flags. */
4638 if (cmsgs->info->sinfo_flags &
4639 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
4640 SCTP_ABORT | SCTP_EOF))
4652 * Wait for a packet..
4653 * Note: This function is the same function as in core/datagram.c
4654 * with a few modifications to make lksctp work.
4656 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
4661 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
4663 /* Socket errors? */
4664 error = sock_error(sk);
4668 if (!skb_queue_empty(&sk->sk_receive_queue))
4671 /* Socket shut down? */
4672 if (sk->sk_shutdown & RCV_SHUTDOWN)
4675 /* Sequenced packets can come disconnected. If so we report the
4680 /* Is there a good reason to think that we may receive some data? */
4681 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
4684 /* Handle signals. */
4685 if (signal_pending(current))
4688 /* Let another process have a go. Since we are going to sleep
4689 * anyway. Note: This may cause odd behaviors if the message
4690 * does not fit in the user's buffer, but this seems to be the
4691 * only way to honor MSG_DONTWAIT realistically.
4693 sctp_release_sock(sk);
4694 *timeo_p = schedule_timeout(*timeo_p);
4698 finish_wait(sk->sk_sleep, &wait);
4702 error = sock_intr_errno(*timeo_p);
4705 finish_wait(sk->sk_sleep, &wait);
4710 /* Receive a datagram.
4711 * Note: This is pretty much the same routine as in core/datagram.c
4712 * with a few changes to make lksctp work.
4714 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
4715 int noblock, int *err)
4718 struct sk_buff *skb;
4721 /* Caller is allowed not to check sk->sk_err before calling. */
4722 error = sock_error(sk);
4726 timeo = sock_rcvtimeo(sk, noblock);
4728 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
4729 timeo, MAX_SCHEDULE_TIMEOUT);
4732 /* Again only user level code calls this function,
4733 * so nothing interrupt level
4734 * will suddenly eat the receive_queue.
4736 * Look at current nfs client by the way...
4737 * However, this function was corrent in any case. 8)
4739 if (flags & MSG_PEEK) {
4740 spin_lock_bh(&sk->sk_receive_queue.lock);
4741 skb = skb_peek(&sk->sk_receive_queue);
4743 atomic_inc(&skb->users);
4744 spin_unlock_bh(&sk->sk_receive_queue.lock);
4746 skb = skb_dequeue(&sk->sk_receive_queue);
4752 if (sk->sk_shutdown & RCV_SHUTDOWN)
4755 /* User doesn't want to wait. */
4759 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
4768 /* If sndbuf has changed, wake up per association sndbuf waiters. */
4769 static void __sctp_write_space(struct sctp_association *asoc)
4771 struct sock *sk = asoc->base.sk;
4772 struct socket *sock = sk->sk_socket;
4774 if ((sctp_wspace(asoc) > 0) && sock) {
4775 if (waitqueue_active(&asoc->wait))
4776 wake_up_interruptible(&asoc->wait);
4778 if (sctp_writeable(sk)) {
4779 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
4780 wake_up_interruptible(sk->sk_sleep);
4782 /* Note that we try to include the Async I/O support
4783 * here by modeling from the current TCP/UDP code.
4784 * We have not tested with it yet.
4786 if (sock->fasync_list &&
4787 !(sk->sk_shutdown & SEND_SHUTDOWN))
4788 sock_wake_async(sock, 2, POLL_OUT);
4793 /* Do accounting for the sndbuf space.
4794 * Decrement the used sndbuf space of the corresponding association by the
4795 * data size which was just transmitted(freed).
4797 static void sctp_wfree(struct sk_buff *skb)
4799 struct sctp_association *asoc;
4800 struct sctp_chunk *chunk;
4803 /* Get the saved chunk pointer. */
4804 chunk = *((struct sctp_chunk **)(skb->cb));
4807 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
4808 sizeof(struct sk_buff) +
4809 sizeof(struct sctp_chunk);
4811 sk->sk_wmem_queued -= SCTP_DATA_SNDSIZE(chunk) +
4812 sizeof(struct sk_buff) +
4813 sizeof(struct sctp_chunk);
4815 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
4818 __sctp_write_space(asoc);
4820 sctp_association_put(asoc);
4823 /* Helper function to wait for space in the sndbuf. */
4824 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
4827 struct sock *sk = asoc->base.sk;
4829 long current_timeo = *timeo_p;
4832 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
4833 asoc, (long)(*timeo_p), msg_len);
4835 /* Increment the association's refcnt. */
4836 sctp_association_hold(asoc);
4838 /* Wait on the association specific sndbuf space. */
4840 prepare_to_wait_exclusive(&asoc->wait, &wait,
4841 TASK_INTERRUPTIBLE);
4844 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
4847 if (signal_pending(current))
4848 goto do_interrupted;
4849 if (msg_len <= sctp_wspace(asoc))
4852 /* Let another process have a go. Since we are going
4855 sctp_release_sock(sk);
4856 current_timeo = schedule_timeout(current_timeo);
4859 *timeo_p = current_timeo;
4863 finish_wait(&asoc->wait, &wait);
4865 /* Release the association's refcnt. */
4866 sctp_association_put(asoc);
4875 err = sock_intr_errno(*timeo_p);
4883 /* If socket sndbuf has changed, wake up all per association waiters. */
4884 void sctp_write_space(struct sock *sk)
4886 struct sctp_association *asoc;
4887 struct list_head *pos;
4889 /* Wake up the tasks in each wait queue. */
4890 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
4891 asoc = list_entry(pos, struct sctp_association, asocs);
4892 __sctp_write_space(asoc);
4896 /* Is there any sndbuf space available on the socket?
4898 * Note that wmem_queued is the sum of the send buffers on all of the
4899 * associations on the same socket. For a UDP-style socket with
4900 * multiple associations, it is possible for it to be "unwriteable"
4901 * prematurely. I assume that this is acceptable because
4902 * a premature "unwriteable" is better than an accidental "writeable" which
4903 * would cause an unwanted block under certain circumstances. For the 1-1
4904 * UDP-style sockets or TCP-style sockets, this code should work.
4907 static int sctp_writeable(struct sock *sk)
4911 amt = sk->sk_sndbuf - sk->sk_wmem_queued;
4917 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
4918 * returns immediately with EINPROGRESS.
4920 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
4922 struct sock *sk = asoc->base.sk;
4924 long current_timeo = *timeo_p;
4927 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
4930 /* Increment the association's refcnt. */
4931 sctp_association_hold(asoc);
4934 prepare_to_wait_exclusive(&asoc->wait, &wait,
4935 TASK_INTERRUPTIBLE);
4938 if (sk->sk_shutdown & RCV_SHUTDOWN)
4940 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
4943 if (signal_pending(current))
4944 goto do_interrupted;
4946 if (sctp_state(asoc, ESTABLISHED))
4949 /* Let another process have a go. Since we are going
4952 sctp_release_sock(sk);
4953 current_timeo = schedule_timeout(current_timeo);
4956 *timeo_p = current_timeo;
4960 finish_wait(&asoc->wait, &wait);
4962 /* Release the association's refcnt. */
4963 sctp_association_put(asoc);
4968 if (asoc->init_err_counter + 1 >= asoc->max_init_attempts)
4971 err = -ECONNREFUSED;
4975 err = sock_intr_errno(*timeo_p);
4983 static int sctp_wait_for_accept(struct sock *sk, long timeo)
4985 struct sctp_endpoint *ep;
4989 ep = sctp_sk(sk)->ep;
4993 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
4994 TASK_INTERRUPTIBLE);
4996 if (list_empty(&ep->asocs)) {
4997 sctp_release_sock(sk);
4998 timeo = schedule_timeout(timeo);
5003 if (!sctp_sstate(sk, LISTENING))
5007 if (!list_empty(&ep->asocs))
5010 err = sock_intr_errno(timeo);
5011 if (signal_pending(current))
5019 finish_wait(sk->sk_sleep, &wait);
5024 void sctp_wait_for_close(struct sock *sk, long timeout)
5029 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5030 if (list_empty(&sctp_sk(sk)->ep->asocs))
5032 sctp_release_sock(sk);
5033 timeout = schedule_timeout(timeout);
5035 } while (!signal_pending(current) && timeout);
5037 finish_wait(sk->sk_sleep, &wait);
5040 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5041 * and its messages to the newsk.
5043 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5044 struct sctp_association *assoc,
5045 sctp_socket_type_t type)
5047 struct sctp_sock *oldsp = sctp_sk(oldsk);
5048 struct sctp_sock *newsp = sctp_sk(newsk);
5049 struct sctp_bind_bucket *pp; /* hash list port iterator */
5050 struct sctp_endpoint *newep = newsp->ep;
5051 struct sk_buff *skb, *tmp;
5052 struct sctp_ulpevent *event;
5055 /* Migrate socket buffer sizes and all the socket level options to the
5058 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5059 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5060 /* Brute force copy old sctp opt. */
5061 inet_sk_copy_descendant(newsk, oldsk);
5063 /* Restore the ep value that was overwritten with the above structure
5069 /* Hook this new socket in to the bind_hash list. */
5070 pp = sctp_sk(oldsk)->bind_hash;
5071 sk_add_bind_node(newsk, &pp->owner);
5072 sctp_sk(newsk)->bind_hash = pp;
5073 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5075 /* Copy the bind_addr list from the original endpoint to the new
5076 * endpoint so that we can handle restarts properly
5078 if (assoc->peer.ipv4_address)
5079 flags |= SCTP_ADDR4_PEERSUPP;
5080 if (assoc->peer.ipv6_address)
5081 flags |= SCTP_ADDR6_PEERSUPP;
5082 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5083 &oldsp->ep->base.bind_addr,
5084 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5086 /* Move any messages in the old socket's receive queue that are for the
5087 * peeled off association to the new socket's receive queue.
5089 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5090 event = sctp_skb2event(skb);
5091 if (event->asoc == assoc) {
5092 __skb_unlink(skb, &oldsk->sk_receive_queue);
5093 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5097 /* Clean up any messages pending delivery due to partial
5098 * delivery. Three cases:
5099 * 1) No partial deliver; no work.
5100 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5101 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5103 skb_queue_head_init(&newsp->pd_lobby);
5104 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5106 if (sctp_sk(oldsk)->pd_mode) {
5107 struct sk_buff_head *queue;
5109 /* Decide which queue to move pd_lobby skbs to. */
5110 if (assoc->ulpq.pd_mode) {
5111 queue = &newsp->pd_lobby;
5113 queue = &newsk->sk_receive_queue;
5115 /* Walk through the pd_lobby, looking for skbs that
5116 * need moved to the new socket.
5118 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5119 event = sctp_skb2event(skb);
5120 if (event->asoc == assoc) {
5121 __skb_unlink(skb, &oldsp->pd_lobby);
5122 __skb_queue_tail(queue, skb);
5126 /* Clear up any skbs waiting for the partial
5127 * delivery to finish.
5129 if (assoc->ulpq.pd_mode)
5130 sctp_clear_pd(oldsk);
5134 /* Set the type of socket to indicate that it is peeled off from the
5135 * original UDP-style socket or created with the accept() call on a
5136 * TCP-style socket..
5140 /* Migrate the association to the new socket. */
5141 sctp_assoc_migrate(assoc, newsk);
5143 /* If the association on the newsk is already closed before accept()
5144 * is called, set RCV_SHUTDOWN flag.
5146 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5147 newsk->sk_shutdown |= RCV_SHUTDOWN;
5149 newsk->sk_state = SCTP_SS_ESTABLISHED;
5152 /* This proto struct describes the ULP interface for SCTP. */
5153 struct proto sctp_prot = {
5155 .owner = THIS_MODULE,
5156 .close = sctp_close,
5157 .connect = sctp_connect,
5158 .disconnect = sctp_disconnect,
5159 .accept = sctp_accept,
5160 .ioctl = sctp_ioctl,
5161 .init = sctp_init_sock,
5162 .destroy = sctp_destroy_sock,
5163 .shutdown = sctp_shutdown,
5164 .setsockopt = sctp_setsockopt,
5165 .getsockopt = sctp_getsockopt,
5166 .sendmsg = sctp_sendmsg,
5167 .recvmsg = sctp_recvmsg,
5169 .backlog_rcv = sctp_backlog_rcv,
5171 .unhash = sctp_unhash,
5172 .get_port = sctp_get_port,
5173 .obj_size = sizeof(struct sctp_sock),
5176 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5177 struct proto sctpv6_prot = {
5179 .owner = THIS_MODULE,
5180 .close = sctp_close,
5181 .connect = sctp_connect,
5182 .disconnect = sctp_disconnect,
5183 .accept = sctp_accept,
5184 .ioctl = sctp_ioctl,
5185 .init = sctp_init_sock,
5186 .destroy = sctp_destroy_sock,
5187 .shutdown = sctp_shutdown,
5188 .setsockopt = sctp_setsockopt,
5189 .getsockopt = sctp_getsockopt,
5190 .sendmsg = sctp_sendmsg,
5191 .recvmsg = sctp_recvmsg,
5193 .backlog_rcv = sctp_backlog_rcv,
5195 .unhash = sctp_unhash,
5196 .get_port = sctp_get_port,
5197 .obj_size = sizeof(struct sctp6_sock),
5199 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */