1 /* SCTP kernel reference Implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel reference Implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * The SCTP reference implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * The SCTP reference implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
73 #include <net/route.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110 /* Get the sndbuf space available at the time on the association. */
111 static inline int sctp_wspace(struct sctp_association *asoc)
113 struct sock *sk = asoc->base.sk;
116 if (asoc->ep->sndbuf_policy) {
117 /* make sure that no association uses more than sk_sndbuf */
118 amt = sk->sk_sndbuf - asoc->sndbuf_used;
120 /* do socket level accounting */
121 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
130 /* Increment the used sndbuf space count of the corresponding association by
131 * the size of the outgoing data chunk.
132 * Also, set the skb destructor for sndbuf accounting later.
134 * Since it is always 1-1 between chunk and skb, and also a new skb is always
135 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
136 * destructor in the data chunk skb for the purpose of the sndbuf space
139 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
141 struct sctp_association *asoc = chunk->asoc;
142 struct sock *sk = asoc->base.sk;
144 /* The sndbuf space is tracked per association. */
145 sctp_association_hold(asoc);
147 skb_set_owner_w(chunk->skb, sk);
149 chunk->skb->destructor = sctp_wfree;
150 /* Save the chunk pointer in skb for sctp_wfree to use later. */
151 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
153 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
154 sizeof(struct sk_buff) +
155 sizeof(struct sctp_chunk);
157 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
160 /* Verify that this is a valid address. */
161 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
166 /* Verify basic sockaddr. */
167 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
171 /* Is this a valid SCTP address? */
172 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
175 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
181 /* Look up the association by its id. If this is not a UDP-style
182 * socket, the ID field is always ignored.
184 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
186 struct sctp_association *asoc = NULL;
188 /* If this is not a UDP-style socket, assoc id should be ignored. */
189 if (!sctp_style(sk, UDP)) {
190 /* Return NULL if the socket state is not ESTABLISHED. It
191 * could be a TCP-style listening socket or a socket which
192 * hasn't yet called connect() to establish an association.
194 if (!sctp_sstate(sk, ESTABLISHED))
197 /* Get the first and the only association from the list. */
198 if (!list_empty(&sctp_sk(sk)->ep->asocs))
199 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
200 struct sctp_association, asocs);
204 /* Otherwise this is a UDP-style socket. */
205 if (!id || (id == (sctp_assoc_t)-1))
208 spin_lock_bh(&sctp_assocs_id_lock);
209 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
210 spin_unlock_bh(&sctp_assocs_id_lock);
212 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
218 /* Look up the transport from an address and an assoc id. If both address and
219 * id are specified, the associations matching the address and the id should be
222 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
223 struct sockaddr_storage *addr,
226 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
227 struct sctp_transport *transport;
228 union sctp_addr *laddr = (union sctp_addr *)addr;
230 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
237 id_asoc = sctp_id2assoc(sk, id);
238 if (id_asoc && (id_asoc != addr_asoc))
241 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
242 (union sctp_addr *)addr);
247 /* API 3.1.2 bind() - UDP Style Syntax
248 * The syntax of bind() is,
250 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
252 * sd - the socket descriptor returned by socket().
253 * addr - the address structure (struct sockaddr_in or struct
254 * sockaddr_in6 [RFC 2553]),
255 * addr_len - the size of the address structure.
257 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
263 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
266 /* Disallow binding twice. */
267 if (!sctp_sk(sk)->ep->base.bind_addr.port)
268 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
273 sctp_release_sock(sk);
278 static long sctp_get_port_local(struct sock *, union sctp_addr *);
280 /* Verify this is a valid sockaddr. */
281 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
282 union sctp_addr *addr, int len)
286 /* Check minimum size. */
287 if (len < sizeof (struct sockaddr))
290 /* Does this PF support this AF? */
291 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
294 /* If we get this far, af is valid. */
295 af = sctp_get_af_specific(addr->sa.sa_family);
297 if (len < af->sockaddr_len)
303 /* Bind a local address either to an endpoint or to an association. */
304 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
306 struct sctp_sock *sp = sctp_sk(sk);
307 struct sctp_endpoint *ep = sp->ep;
308 struct sctp_bind_addr *bp = &ep->base.bind_addr;
313 /* Common sockaddr verification. */
314 af = sctp_sockaddr_af(sp, addr, len);
316 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
321 snum = ntohs(addr->v4.sin_port);
323 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
324 ", port: %d, new port: %d, len: %d)\n",
330 /* PF specific bind() address verification. */
331 if (!sp->pf->bind_verify(sp, addr))
332 return -EADDRNOTAVAIL;
334 /* We must either be unbound, or bind to the same port.
335 * It's OK to allow 0 ports if we are already bound.
336 * We'll just inhert an already bound port in this case
341 else if (snum != bp->port) {
342 SCTP_DEBUG_PRINTK("sctp_do_bind:"
343 " New port %d does not match existing port "
344 "%d.\n", snum, bp->port);
349 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
352 /* Make sure we are allowed to bind here.
353 * The function sctp_get_port_local() does duplicate address
356 if ((ret = sctp_get_port_local(sk, addr))) {
357 if (ret == (long) sk) {
358 /* This endpoint has a conflicting address. */
365 /* Refresh ephemeral port. */
367 bp->port = inet_sk(sk)->num;
369 /* Add the address to the bind address list. */
370 sctp_local_bh_disable();
371 sctp_write_lock(&ep->base.addr_lock);
373 /* Use GFP_ATOMIC since BHs are disabled. */
374 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
375 sctp_write_unlock(&ep->base.addr_lock);
376 sctp_local_bh_enable();
378 /* Copy back into socket for getsockname() use. */
380 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
381 af->to_sk_saddr(addr, sk);
387 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
389 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
390 * at any one time. If a sender, after sending an ASCONF chunk, decides
391 * it needs to transfer another ASCONF Chunk, it MUST wait until the
392 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
393 * subsequent ASCONF. Note this restriction binds each side, so at any
394 * time two ASCONF may be in-transit on any given association (one sent
395 * from each endpoint).
397 static int sctp_send_asconf(struct sctp_association *asoc,
398 struct sctp_chunk *chunk)
402 /* If there is an outstanding ASCONF chunk, queue it for later
405 if (asoc->addip_last_asconf) {
406 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
410 /* Hold the chunk until an ASCONF_ACK is received. */
411 sctp_chunk_hold(chunk);
412 retval = sctp_primitive_ASCONF(asoc, chunk);
414 sctp_chunk_free(chunk);
416 asoc->addip_last_asconf = chunk;
422 /* Add a list of addresses as bind addresses to local endpoint or
425 * Basically run through each address specified in the addrs/addrcnt
426 * array/length pair, determine if it is IPv6 or IPv4 and call
427 * sctp_do_bind() on it.
429 * If any of them fails, then the operation will be reversed and the
430 * ones that were added will be removed.
432 * Only sctp_setsockopt_bindx() is supposed to call this function.
434 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
439 struct sockaddr *sa_addr;
442 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
446 for (cnt = 0; cnt < addrcnt; cnt++) {
447 /* The list may contain either IPv4 or IPv6 address;
448 * determine the address length for walking thru the list.
450 sa_addr = (struct sockaddr *)addr_buf;
451 af = sctp_get_af_specific(sa_addr->sa_family);
457 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
460 addr_buf += af->sockaddr_len;
464 /* Failed. Cleanup the ones that have been added */
466 sctp_bindx_rem(sk, addrs, cnt);
474 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
475 * associations that are part of the endpoint indicating that a list of local
476 * addresses are added to the endpoint.
478 * If any of the addresses is already in the bind address list of the
479 * association, we do not send the chunk for that association. But it will not
480 * affect other associations.
482 * Only sctp_setsockopt_bindx() is supposed to call this function.
484 static int sctp_send_asconf_add_ip(struct sock *sk,
485 struct sockaddr *addrs,
488 struct sctp_sock *sp;
489 struct sctp_endpoint *ep;
490 struct sctp_association *asoc;
491 struct sctp_bind_addr *bp;
492 struct sctp_chunk *chunk;
493 struct sctp_sockaddr_entry *laddr;
494 union sctp_addr *addr;
495 union sctp_addr saveaddr;
498 struct list_head *pos;
503 if (!sctp_addip_enable)
509 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
510 __FUNCTION__, sk, addrs, addrcnt);
512 list_for_each(pos, &ep->asocs) {
513 asoc = list_entry(pos, struct sctp_association, asocs);
515 if (!asoc->peer.asconf_capable)
518 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
521 if (!sctp_state(asoc, ESTABLISHED))
524 /* Check if any address in the packed array of addresses is
525 * in the bind address list of the association. If so,
526 * do not send the asconf chunk to its peer, but continue with
527 * other associations.
530 for (i = 0; i < addrcnt; i++) {
531 addr = (union sctp_addr *)addr_buf;
532 af = sctp_get_af_specific(addr->v4.sin_family);
538 if (sctp_assoc_lookup_laddr(asoc, addr))
541 addr_buf += af->sockaddr_len;
546 /* Use the first address in bind addr list of association as
547 * Address Parameter of ASCONF CHUNK.
549 sctp_read_lock(&asoc->base.addr_lock);
550 bp = &asoc->base.bind_addr;
551 p = bp->address_list.next;
552 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
553 sctp_read_unlock(&asoc->base.addr_lock);
555 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
556 addrcnt, SCTP_PARAM_ADD_IP);
562 retval = sctp_send_asconf(asoc, chunk);
566 /* Add the new addresses to the bind address list with
567 * use_as_src set to 0.
569 sctp_local_bh_disable();
570 sctp_write_lock(&asoc->base.addr_lock);
572 for (i = 0; i < addrcnt; i++) {
573 addr = (union sctp_addr *)addr_buf;
574 af = sctp_get_af_specific(addr->v4.sin_family);
575 memcpy(&saveaddr, addr, af->sockaddr_len);
576 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
578 addr_buf += af->sockaddr_len;
580 sctp_write_unlock(&asoc->base.addr_lock);
581 sctp_local_bh_enable();
588 /* Remove a list of addresses from bind addresses list. Do not remove the
591 * Basically run through each address specified in the addrs/addrcnt
592 * array/length pair, determine if it is IPv6 or IPv4 and call
593 * sctp_del_bind() on it.
595 * If any of them fails, then the operation will be reversed and the
596 * ones that were removed will be added back.
598 * At least one address has to be left; if only one address is
599 * available, the operation will return -EBUSY.
601 * Only sctp_setsockopt_bindx() is supposed to call this function.
603 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
605 struct sctp_sock *sp = sctp_sk(sk);
606 struct sctp_endpoint *ep = sp->ep;
608 struct sctp_bind_addr *bp = &ep->base.bind_addr;
611 union sctp_addr *sa_addr;
614 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
618 for (cnt = 0; cnt < addrcnt; cnt++) {
619 /* If the bind address list is empty or if there is only one
620 * bind address, there is nothing more to be removed (we need
621 * at least one address here).
623 if (list_empty(&bp->address_list) ||
624 (sctp_list_single_entry(&bp->address_list))) {
629 sa_addr = (union sctp_addr *)addr_buf;
630 af = sctp_get_af_specific(sa_addr->sa.sa_family);
636 if (!af->addr_valid(sa_addr, sp, NULL)) {
637 retval = -EADDRNOTAVAIL;
641 if (sa_addr->v4.sin_port != htons(bp->port)) {
646 /* FIXME - There is probably a need to check if sk->sk_saddr and
647 * sk->sk_rcv_addr are currently set to one of the addresses to
648 * be removed. This is something which needs to be looked into
649 * when we are fixing the outstanding issues with multi-homing
650 * socket routing and failover schemes. Refer to comments in
651 * sctp_do_bind(). -daisy
653 sctp_local_bh_disable();
654 sctp_write_lock(&ep->base.addr_lock);
656 retval = sctp_del_bind_addr(bp, sa_addr);
658 sctp_write_unlock(&ep->base.addr_lock);
659 sctp_local_bh_enable();
661 addr_buf += af->sockaddr_len;
664 /* Failed. Add the ones that has been removed back */
666 sctp_bindx_add(sk, addrs, cnt);
674 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
675 * the associations that are part of the endpoint indicating that a list of
676 * local addresses are removed from the endpoint.
678 * If any of the addresses is already in the bind address list of the
679 * association, we do not send the chunk for that association. But it will not
680 * affect other associations.
682 * Only sctp_setsockopt_bindx() is supposed to call this function.
684 static int sctp_send_asconf_del_ip(struct sock *sk,
685 struct sockaddr *addrs,
688 struct sctp_sock *sp;
689 struct sctp_endpoint *ep;
690 struct sctp_association *asoc;
691 struct sctp_transport *transport;
692 struct sctp_bind_addr *bp;
693 struct sctp_chunk *chunk;
694 union sctp_addr *laddr;
697 struct list_head *pos, *pos1;
698 struct sctp_sockaddr_entry *saddr;
702 if (!sctp_addip_enable)
708 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
709 __FUNCTION__, sk, addrs, addrcnt);
711 list_for_each(pos, &ep->asocs) {
712 asoc = list_entry(pos, struct sctp_association, asocs);
714 if (!asoc->peer.asconf_capable)
717 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
720 if (!sctp_state(asoc, ESTABLISHED))
723 /* Check if any address in the packed array of addresses is
724 * not present in the bind address list of the association.
725 * If so, do not send the asconf chunk to its peer, but
726 * continue with other associations.
729 for (i = 0; i < addrcnt; i++) {
730 laddr = (union sctp_addr *)addr_buf;
731 af = sctp_get_af_specific(laddr->v4.sin_family);
737 if (!sctp_assoc_lookup_laddr(asoc, laddr))
740 addr_buf += af->sockaddr_len;
745 /* Find one address in the association's bind address list
746 * that is not in the packed array of addresses. This is to
747 * make sure that we do not delete all the addresses in the
750 sctp_read_lock(&asoc->base.addr_lock);
751 bp = &asoc->base.bind_addr;
752 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
754 sctp_read_unlock(&asoc->base.addr_lock);
758 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
765 /* Reset use_as_src flag for the addresses in the bind address
766 * list that are to be deleted.
768 sctp_local_bh_disable();
769 sctp_write_lock(&asoc->base.addr_lock);
771 for (i = 0; i < addrcnt; i++) {
772 laddr = (union sctp_addr *)addr_buf;
773 af = sctp_get_af_specific(laddr->v4.sin_family);
774 list_for_each(pos1, &bp->address_list) {
775 saddr = list_entry(pos1,
776 struct sctp_sockaddr_entry,
778 if (sctp_cmp_addr_exact(&saddr->a, laddr))
779 saddr->use_as_src = 0;
781 addr_buf += af->sockaddr_len;
783 sctp_write_unlock(&asoc->base.addr_lock);
784 sctp_local_bh_enable();
786 /* Update the route and saddr entries for all the transports
787 * as some of the addresses in the bind address list are
788 * about to be deleted and cannot be used as source addresses.
790 list_for_each(pos1, &asoc->peer.transport_addr_list) {
791 transport = list_entry(pos1, struct sctp_transport,
793 dst_release(transport->dst);
794 sctp_transport_route(transport, NULL,
795 sctp_sk(asoc->base.sk));
798 retval = sctp_send_asconf(asoc, chunk);
804 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
807 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
810 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
811 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
814 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
815 * Section 3.1.2 for this usage.
817 * addrs is a pointer to an array of one or more socket addresses. Each
818 * address is contained in its appropriate structure (i.e. struct
819 * sockaddr_in or struct sockaddr_in6) the family of the address type
820 * must be used to distinguish the address length (note that this
821 * representation is termed a "packed array" of addresses). The caller
822 * specifies the number of addresses in the array with addrcnt.
824 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
825 * -1, and sets errno to the appropriate error code.
827 * For SCTP, the port given in each socket address must be the same, or
828 * sctp_bindx() will fail, setting errno to EINVAL.
830 * The flags parameter is formed from the bitwise OR of zero or more of
831 * the following currently defined flags:
833 * SCTP_BINDX_ADD_ADDR
835 * SCTP_BINDX_REM_ADDR
837 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
838 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
839 * addresses from the association. The two flags are mutually exclusive;
840 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
841 * not remove all addresses from an association; sctp_bindx() will
842 * reject such an attempt with EINVAL.
844 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
845 * additional addresses with an endpoint after calling bind(). Or use
846 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
847 * socket is associated with so that no new association accepted will be
848 * associated with those addresses. If the endpoint supports dynamic
849 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
850 * endpoint to send the appropriate message to the peer to change the
851 * peers address lists.
853 * Adding and removing addresses from a connected association is
854 * optional functionality. Implementations that do not support this
855 * functionality should return EOPNOTSUPP.
857 * Basically do nothing but copying the addresses from user to kernel
858 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
859 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
862 * We don't use copy_from_user() for optimization: we first do the
863 * sanity checks (buffer size -fast- and access check-healthy
864 * pointer); if all of those succeed, then we can alloc the memory
865 * (expensive operation) needed to copy the data to kernel. Then we do
866 * the copying without checking the user space area
867 * (__copy_from_user()).
869 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
872 * sk The sk of the socket
873 * addrs The pointer to the addresses in user land
874 * addrssize Size of the addrs buffer
875 * op Operation to perform (add or remove, see the flags of
878 * Returns 0 if ok, <0 errno code on error.
880 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
881 struct sockaddr __user *addrs,
882 int addrs_size, int op)
884 struct sockaddr *kaddrs;
888 struct sockaddr *sa_addr;
892 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
893 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
895 if (unlikely(addrs_size <= 0))
898 /* Check the user passed a healthy pointer. */
899 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
902 /* Alloc space for the address array in kernel memory. */
903 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
904 if (unlikely(!kaddrs))
907 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
912 /* Walk through the addrs buffer and count the number of addresses. */
914 while (walk_size < addrs_size) {
915 sa_addr = (struct sockaddr *)addr_buf;
916 af = sctp_get_af_specific(sa_addr->sa_family);
918 /* If the address family is not supported or if this address
919 * causes the address buffer to overflow return EINVAL.
921 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
926 addr_buf += af->sockaddr_len;
927 walk_size += af->sockaddr_len;
932 case SCTP_BINDX_ADD_ADDR:
933 err = sctp_bindx_add(sk, kaddrs, addrcnt);
936 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
939 case SCTP_BINDX_REM_ADDR:
940 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
943 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
957 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
959 * Common routine for handling connect() and sctp_connectx().
960 * Connect will come in with just a single address.
962 static int __sctp_connect(struct sock* sk,
963 struct sockaddr *kaddrs,
966 struct sctp_sock *sp;
967 struct sctp_endpoint *ep;
968 struct sctp_association *asoc = NULL;
969 struct sctp_association *asoc2;
970 struct sctp_transport *transport;
978 union sctp_addr *sa_addr = NULL;
981 unsigned int f_flags = 0;
986 /* connect() cannot be done on a socket that is already in ESTABLISHED
987 * state - UDP-style peeled off socket or a TCP-style socket that
988 * is already connected.
989 * It cannot be done even on a TCP-style listening socket.
991 if (sctp_sstate(sk, ESTABLISHED) ||
992 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
997 /* Walk through the addrs buffer and count the number of addresses. */
999 while (walk_size < addrs_size) {
1000 sa_addr = (union sctp_addr *)addr_buf;
1001 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1002 port = ntohs(sa_addr->v4.sin_port);
1004 /* If the address family is not supported or if this address
1005 * causes the address buffer to overflow return EINVAL.
1007 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012 /* Save current address so we can work with it */
1013 memcpy(&to, sa_addr, af->sockaddr_len);
1015 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1019 /* Make sure the destination port is correctly set
1022 if (asoc && asoc->peer.port && asoc->peer.port != port)
1026 /* Check if there already is a matching association on the
1027 * endpoint (other than the one created here).
1029 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1030 if (asoc2 && asoc2 != asoc) {
1031 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1038 /* If we could not find a matching association on the endpoint,
1039 * make sure that there is no peeled-off association matching
1040 * the peer address even on another socket.
1042 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1043 err = -EADDRNOTAVAIL;
1048 /* If a bind() or sctp_bindx() is not called prior to
1049 * an sctp_connectx() call, the system picks an
1050 * ephemeral port and will choose an address set
1051 * equivalent to binding with a wildcard address.
1053 if (!ep->base.bind_addr.port) {
1054 if (sctp_autobind(sk)) {
1060 * If an unprivileged user inherits a 1-many
1061 * style socket with open associations on a
1062 * privileged port, it MAY be permitted to
1063 * accept new associations, but it SHOULD NOT
1064 * be permitted to open new associations.
1066 if (ep->base.bind_addr.port < PROT_SOCK &&
1067 !capable(CAP_NET_BIND_SERVICE)) {
1073 scope = sctp_scope(&to);
1074 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1081 /* Prime the peer's transport structures. */
1082 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1090 addr_buf += af->sockaddr_len;
1091 walk_size += af->sockaddr_len;
1094 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1099 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1104 /* Initialize sk's dport and daddr for getpeername() */
1105 inet_sk(sk)->dport = htons(asoc->peer.port);
1106 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1107 af->to_sk_daddr(sa_addr, sk);
1110 /* in-kernel sockets don't generally have a file allocated to them
1111 * if all they do is call sock_create_kern().
1113 if (sk->sk_socket->file)
1114 f_flags = sk->sk_socket->file->f_flags;
1116 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1118 err = sctp_wait_for_connect(asoc, &timeo);
1120 /* Don't free association on exit. */
1125 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1126 " kaddrs: %p err: %d\n",
1129 sctp_association_free(asoc);
1133 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1136 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1138 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1139 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1140 * or IPv6 addresses.
1142 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1143 * Section 3.1.2 for this usage.
1145 * addrs is a pointer to an array of one or more socket addresses. Each
1146 * address is contained in its appropriate structure (i.e. struct
1147 * sockaddr_in or struct sockaddr_in6) the family of the address type
1148 * must be used to distengish the address length (note that this
1149 * representation is termed a "packed array" of addresses). The caller
1150 * specifies the number of addresses in the array with addrcnt.
1152 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1153 * -1, and sets errno to the appropriate error code.
1155 * For SCTP, the port given in each socket address must be the same, or
1156 * sctp_connectx() will fail, setting errno to EINVAL.
1158 * An application can use sctp_connectx to initiate an association with
1159 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1160 * allows a caller to specify multiple addresses at which a peer can be
1161 * reached. The way the SCTP stack uses the list of addresses to set up
1162 * the association is implementation dependant. This function only
1163 * specifies that the stack will try to make use of all the addresses in
1164 * the list when needed.
1166 * Note that the list of addresses passed in is only used for setting up
1167 * the association. It does not necessarily equal the set of addresses
1168 * the peer uses for the resulting association. If the caller wants to
1169 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1170 * retrieve them after the association has been set up.
1172 * Basically do nothing but copying the addresses from user to kernel
1173 * land and invoking either sctp_connectx(). This is used for tunneling
1174 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1176 * We don't use copy_from_user() for optimization: we first do the
1177 * sanity checks (buffer size -fast- and access check-healthy
1178 * pointer); if all of those succeed, then we can alloc the memory
1179 * (expensive operation) needed to copy the data to kernel. Then we do
1180 * the copying without checking the user space area
1181 * (__copy_from_user()).
1183 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1186 * sk The sk of the socket
1187 * addrs The pointer to the addresses in user land
1188 * addrssize Size of the addrs buffer
1190 * Returns 0 if ok, <0 errno code on error.
1192 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1193 struct sockaddr __user *addrs,
1197 struct sockaddr *kaddrs;
1199 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1200 __FUNCTION__, sk, addrs, addrs_size);
1202 if (unlikely(addrs_size <= 0))
1205 /* Check the user passed a healthy pointer. */
1206 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1209 /* Alloc space for the address array in kernel memory. */
1210 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1211 if (unlikely(!kaddrs))
1214 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1217 err = __sctp_connect(sk, kaddrs, addrs_size);
1224 /* API 3.1.4 close() - UDP Style Syntax
1225 * Applications use close() to perform graceful shutdown (as described in
1226 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1227 * by a UDP-style socket.
1231 * ret = close(int sd);
1233 * sd - the socket descriptor of the associations to be closed.
1235 * To gracefully shutdown a specific association represented by the
1236 * UDP-style socket, an application should use the sendmsg() call,
1237 * passing no user data, but including the appropriate flag in the
1238 * ancillary data (see Section xxxx).
1240 * If sd in the close() call is a branched-off socket representing only
1241 * one association, the shutdown is performed on that association only.
1243 * 4.1.6 close() - TCP Style Syntax
1245 * Applications use close() to gracefully close down an association.
1249 * int close(int sd);
1251 * sd - the socket descriptor of the association to be closed.
1253 * After an application calls close() on a socket descriptor, no further
1254 * socket operations will succeed on that descriptor.
1256 * API 7.1.4 SO_LINGER
1258 * An application using the TCP-style socket can use this option to
1259 * perform the SCTP ABORT primitive. The linger option structure is:
1262 * int l_onoff; // option on/off
1263 * int l_linger; // linger time
1266 * To enable the option, set l_onoff to 1. If the l_linger value is set
1267 * to 0, calling close() is the same as the ABORT primitive. If the
1268 * value is set to a negative value, the setsockopt() call will return
1269 * an error. If the value is set to a positive value linger_time, the
1270 * close() can be blocked for at most linger_time ms. If the graceful
1271 * shutdown phase does not finish during this period, close() will
1272 * return but the graceful shutdown phase continues in the system.
1274 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1276 struct sctp_endpoint *ep;
1277 struct sctp_association *asoc;
1278 struct list_head *pos, *temp;
1280 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1283 sk->sk_shutdown = SHUTDOWN_MASK;
1285 ep = sctp_sk(sk)->ep;
1287 /* Walk all associations on an endpoint. */
1288 list_for_each_safe(pos, temp, &ep->asocs) {
1289 asoc = list_entry(pos, struct sctp_association, asocs);
1291 if (sctp_style(sk, TCP)) {
1292 /* A closed association can still be in the list if
1293 * it belongs to a TCP-style listening socket that is
1294 * not yet accepted. If so, free it. If not, send an
1295 * ABORT or SHUTDOWN based on the linger options.
1297 if (sctp_state(asoc, CLOSED)) {
1298 sctp_unhash_established(asoc);
1299 sctp_association_free(asoc);
1304 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1305 struct sctp_chunk *chunk;
1307 chunk = sctp_make_abort_user(asoc, NULL, 0);
1309 sctp_primitive_ABORT(asoc, chunk);
1311 sctp_primitive_SHUTDOWN(asoc, NULL);
1314 /* Clean up any skbs sitting on the receive queue. */
1315 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1316 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1318 /* On a TCP-style socket, block for at most linger_time if set. */
1319 if (sctp_style(sk, TCP) && timeout)
1320 sctp_wait_for_close(sk, timeout);
1322 /* This will run the backlog queue. */
1323 sctp_release_sock(sk);
1325 /* Supposedly, no process has access to the socket, but
1326 * the net layers still may.
1328 sctp_local_bh_disable();
1329 sctp_bh_lock_sock(sk);
1331 /* Hold the sock, since sk_common_release() will put sock_put()
1332 * and we have just a little more cleanup.
1335 sk_common_release(sk);
1337 sctp_bh_unlock_sock(sk);
1338 sctp_local_bh_enable();
1342 SCTP_DBG_OBJCNT_DEC(sock);
1345 /* Handle EPIPE error. */
1346 static int sctp_error(struct sock *sk, int flags, int err)
1349 err = sock_error(sk) ? : -EPIPE;
1350 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1351 send_sig(SIGPIPE, current, 0);
1355 /* API 3.1.3 sendmsg() - UDP Style Syntax
1357 * An application uses sendmsg() and recvmsg() calls to transmit data to
1358 * and receive data from its peer.
1360 * ssize_t sendmsg(int socket, const struct msghdr *message,
1363 * socket - the socket descriptor of the endpoint.
1364 * message - pointer to the msghdr structure which contains a single
1365 * user message and possibly some ancillary data.
1367 * See Section 5 for complete description of the data
1370 * flags - flags sent or received with the user message, see Section
1371 * 5 for complete description of the flags.
1373 * Note: This function could use a rewrite especially when explicit
1374 * connect support comes in.
1376 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1378 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1380 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1381 struct msghdr *msg, size_t msg_len)
1383 struct sctp_sock *sp;
1384 struct sctp_endpoint *ep;
1385 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1386 struct sctp_transport *transport, *chunk_tp;
1387 struct sctp_chunk *chunk;
1389 struct sockaddr *msg_name = NULL;
1390 struct sctp_sndrcvinfo default_sinfo = { 0 };
1391 struct sctp_sndrcvinfo *sinfo;
1392 struct sctp_initmsg *sinit;
1393 sctp_assoc_t associd = 0;
1394 sctp_cmsgs_t cmsgs = { NULL };
1398 __u16 sinfo_flags = 0;
1399 struct sctp_datamsg *datamsg;
1400 struct list_head *pos;
1401 int msg_flags = msg->msg_flags;
1403 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1410 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1412 /* We cannot send a message over a TCP-style listening socket. */
1413 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1418 /* Parse out the SCTP CMSGs. */
1419 err = sctp_msghdr_parse(msg, &cmsgs);
1422 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1426 /* Fetch the destination address for this packet. This
1427 * address only selects the association--it is not necessarily
1428 * the address we will send to.
1429 * For a peeled-off socket, msg_name is ignored.
1431 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1432 int msg_namelen = msg->msg_namelen;
1434 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1439 if (msg_namelen > sizeof(to))
1440 msg_namelen = sizeof(to);
1441 memcpy(&to, msg->msg_name, msg_namelen);
1442 msg_name = msg->msg_name;
1448 /* Did the user specify SNDRCVINFO? */
1450 sinfo_flags = sinfo->sinfo_flags;
1451 associd = sinfo->sinfo_assoc_id;
1454 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1455 msg_len, sinfo_flags);
1457 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1458 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1463 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1464 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1465 * If SCTP_ABORT is set, the message length could be non zero with
1466 * the msg_iov set to the user abort reason.
1468 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1469 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1474 /* If SCTP_ADDR_OVER is set, there must be an address
1475 * specified in msg_name.
1477 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1484 SCTP_DEBUG_PRINTK("About to look up association.\n");
1488 /* If a msg_name has been specified, assume this is to be used. */
1490 /* Look for a matching association on the endpoint. */
1491 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1493 /* If we could not find a matching association on the
1494 * endpoint, make sure that it is not a TCP-style
1495 * socket that already has an association or there is
1496 * no peeled-off association on another socket.
1498 if ((sctp_style(sk, TCP) &&
1499 sctp_sstate(sk, ESTABLISHED)) ||
1500 sctp_endpoint_is_peeled_off(ep, &to)) {
1501 err = -EADDRNOTAVAIL;
1506 asoc = sctp_id2assoc(sk, associd);
1514 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1516 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1517 * socket that has an association in CLOSED state. This can
1518 * happen when an accepted socket has an association that is
1521 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1526 if (sinfo_flags & SCTP_EOF) {
1527 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1529 sctp_primitive_SHUTDOWN(asoc, NULL);
1533 if (sinfo_flags & SCTP_ABORT) {
1535 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1541 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1542 sctp_primitive_ABORT(asoc, chunk);
1548 /* Do we need to create the association? */
1550 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1552 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1557 /* Check for invalid stream against the stream counts,
1558 * either the default or the user specified stream counts.
1561 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1562 /* Check against the defaults. */
1563 if (sinfo->sinfo_stream >=
1564 sp->initmsg.sinit_num_ostreams) {
1569 /* Check against the requested. */
1570 if (sinfo->sinfo_stream >=
1571 sinit->sinit_num_ostreams) {
1579 * API 3.1.2 bind() - UDP Style Syntax
1580 * If a bind() or sctp_bindx() is not called prior to a
1581 * sendmsg() call that initiates a new association, the
1582 * system picks an ephemeral port and will choose an address
1583 * set equivalent to binding with a wildcard address.
1585 if (!ep->base.bind_addr.port) {
1586 if (sctp_autobind(sk)) {
1592 * If an unprivileged user inherits a one-to-many
1593 * style socket with open associations on a privileged
1594 * port, it MAY be permitted to accept new associations,
1595 * but it SHOULD NOT be permitted to open new
1598 if (ep->base.bind_addr.port < PROT_SOCK &&
1599 !capable(CAP_NET_BIND_SERVICE)) {
1605 scope = sctp_scope(&to);
1606 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1613 /* If the SCTP_INIT ancillary data is specified, set all
1614 * the association init values accordingly.
1617 if (sinit->sinit_num_ostreams) {
1618 asoc->c.sinit_num_ostreams =
1619 sinit->sinit_num_ostreams;
1621 if (sinit->sinit_max_instreams) {
1622 asoc->c.sinit_max_instreams =
1623 sinit->sinit_max_instreams;
1625 if (sinit->sinit_max_attempts) {
1626 asoc->max_init_attempts
1627 = sinit->sinit_max_attempts;
1629 if (sinit->sinit_max_init_timeo) {
1630 asoc->max_init_timeo =
1631 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1635 /* Prime the peer's transport structures. */
1636 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1641 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1648 /* ASSERT: we have a valid association at this point. */
1649 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1652 /* If the user didn't specify SNDRCVINFO, make up one with
1655 default_sinfo.sinfo_stream = asoc->default_stream;
1656 default_sinfo.sinfo_flags = asoc->default_flags;
1657 default_sinfo.sinfo_ppid = asoc->default_ppid;
1658 default_sinfo.sinfo_context = asoc->default_context;
1659 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1660 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1661 sinfo = &default_sinfo;
1664 /* API 7.1.7, the sndbuf size per association bounds the
1665 * maximum size of data that can be sent in a single send call.
1667 if (msg_len > sk->sk_sndbuf) {
1672 if (asoc->pmtu_pending)
1673 sctp_assoc_pending_pmtu(asoc);
1675 /* If fragmentation is disabled and the message length exceeds the
1676 * association fragmentation point, return EMSGSIZE. The I-D
1677 * does not specify what this error is, but this looks like
1680 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1686 /* Check for invalid stream. */
1687 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1693 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1694 if (!sctp_wspace(asoc)) {
1695 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1700 /* If an address is passed with the sendto/sendmsg call, it is used
1701 * to override the primary destination address in the TCP model, or
1702 * when SCTP_ADDR_OVER flag is set in the UDP model.
1704 if ((sctp_style(sk, TCP) && msg_name) ||
1705 (sinfo_flags & SCTP_ADDR_OVER)) {
1706 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1714 /* Auto-connect, if we aren't connected already. */
1715 if (sctp_state(asoc, CLOSED)) {
1716 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1719 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1722 /* Break the message into multiple chunks of maximum size. */
1723 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1729 /* Now send the (possibly) fragmented message. */
1730 list_for_each(pos, &datamsg->chunks) {
1731 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1732 sctp_datamsg_track(chunk);
1734 /* Do accounting for the write space. */
1735 sctp_set_owner_w(chunk);
1737 chunk->transport = chunk_tp;
1739 /* Send it to the lower layers. Note: all chunks
1740 * must either fail or succeed. The lower layer
1741 * works that way today. Keep it that way or this
1744 err = sctp_primitive_SEND(asoc, chunk);
1745 /* Did the lower layer accept the chunk? */
1747 sctp_chunk_free(chunk);
1748 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1751 sctp_datamsg_free(datamsg);
1757 /* If we are already past ASSOCIATE, the lower
1758 * layers are responsible for association cleanup.
1764 sctp_association_free(asoc);
1766 sctp_release_sock(sk);
1769 return sctp_error(sk, msg_flags, err);
1776 err = sock_error(sk);
1786 /* This is an extended version of skb_pull() that removes the data from the
1787 * start of a skb even when data is spread across the list of skb's in the
1788 * frag_list. len specifies the total amount of data that needs to be removed.
1789 * when 'len' bytes could be removed from the skb, it returns 0.
1790 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1791 * could not be removed.
1793 static int sctp_skb_pull(struct sk_buff *skb, int len)
1795 struct sk_buff *list;
1796 int skb_len = skb_headlen(skb);
1799 if (len <= skb_len) {
1800 __skb_pull(skb, len);
1804 __skb_pull(skb, skb_len);
1806 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1807 rlen = sctp_skb_pull(list, len);
1808 skb->len -= (len-rlen);
1809 skb->data_len -= (len-rlen);
1820 /* API 3.1.3 recvmsg() - UDP Style Syntax
1822 * ssize_t recvmsg(int socket, struct msghdr *message,
1825 * socket - the socket descriptor of the endpoint.
1826 * message - pointer to the msghdr structure which contains a single
1827 * user message and possibly some ancillary data.
1829 * See Section 5 for complete description of the data
1832 * flags - flags sent or received with the user message, see Section
1833 * 5 for complete description of the flags.
1835 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1837 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1838 struct msghdr *msg, size_t len, int noblock,
1839 int flags, int *addr_len)
1841 struct sctp_ulpevent *event = NULL;
1842 struct sctp_sock *sp = sctp_sk(sk);
1843 struct sk_buff *skb;
1848 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1849 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1850 "len", len, "knoblauch", noblock,
1851 "flags", flags, "addr_len", addr_len);
1855 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1860 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1864 /* Get the total length of the skb including any skb's in the
1873 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1875 event = sctp_skb2event(skb);
1880 sock_recv_timestamp(msg, sk, skb);
1881 if (sctp_ulpevent_is_notification(event)) {
1882 msg->msg_flags |= MSG_NOTIFICATION;
1883 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1885 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1888 /* Check if we allow SCTP_SNDRCVINFO. */
1889 if (sp->subscribe.sctp_data_io_event)
1890 sctp_ulpevent_read_sndrcvinfo(event, msg);
1892 /* FIXME: we should be calling IP/IPv6 layers. */
1893 if (sk->sk_protinfo.af_inet.cmsg_flags)
1894 ip_cmsg_recv(msg, skb);
1899 /* If skb's length exceeds the user's buffer, update the skb and
1900 * push it back to the receive_queue so that the next call to
1901 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1903 if (skb_len > copied) {
1904 msg->msg_flags &= ~MSG_EOR;
1905 if (flags & MSG_PEEK)
1907 sctp_skb_pull(skb, copied);
1908 skb_queue_head(&sk->sk_receive_queue, skb);
1910 /* When only partial message is copied to the user, increase
1911 * rwnd by that amount. If all the data in the skb is read,
1912 * rwnd is updated when the event is freed.
1914 sctp_assoc_rwnd_increase(event->asoc, copied);
1916 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1917 (event->msg_flags & MSG_EOR))
1918 msg->msg_flags |= MSG_EOR;
1920 msg->msg_flags &= ~MSG_EOR;
1923 if (flags & MSG_PEEK) {
1924 /* Release the skb reference acquired after peeking the skb in
1925 * sctp_skb_recv_datagram().
1929 /* Free the event which includes releasing the reference to
1930 * the owner of the skb, freeing the skb and updating the
1933 sctp_ulpevent_free(event);
1936 sctp_release_sock(sk);
1940 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1942 * This option is a on/off flag. If enabled no SCTP message
1943 * fragmentation will be performed. Instead if a message being sent
1944 * exceeds the current PMTU size, the message will NOT be sent and
1945 * instead a error will be indicated to the user.
1947 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1948 char __user *optval, int optlen)
1952 if (optlen < sizeof(int))
1955 if (get_user(val, (int __user *)optval))
1958 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1963 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1966 if (optlen != sizeof(struct sctp_event_subscribe))
1968 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1973 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1975 * This socket option is applicable to the UDP-style socket only. When
1976 * set it will cause associations that are idle for more than the
1977 * specified number of seconds to automatically close. An association
1978 * being idle is defined an association that has NOT sent or received
1979 * user data. The special value of '0' indicates that no automatic
1980 * close of any associations should be performed. The option expects an
1981 * integer defining the number of seconds of idle time before an
1982 * association is closed.
1984 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1987 struct sctp_sock *sp = sctp_sk(sk);
1989 /* Applicable to UDP-style socket only */
1990 if (sctp_style(sk, TCP))
1992 if (optlen != sizeof(int))
1994 if (copy_from_user(&sp->autoclose, optval, optlen))
2000 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2002 * Applications can enable or disable heartbeats for any peer address of
2003 * an association, modify an address's heartbeat interval, force a
2004 * heartbeat to be sent immediately, and adjust the address's maximum
2005 * number of retransmissions sent before an address is considered
2006 * unreachable. The following structure is used to access and modify an
2007 * address's parameters:
2009 * struct sctp_paddrparams {
2010 * sctp_assoc_t spp_assoc_id;
2011 * struct sockaddr_storage spp_address;
2012 * uint32_t spp_hbinterval;
2013 * uint16_t spp_pathmaxrxt;
2014 * uint32_t spp_pathmtu;
2015 * uint32_t spp_sackdelay;
2016 * uint32_t spp_flags;
2019 * spp_assoc_id - (one-to-many style socket) This is filled in the
2020 * application, and identifies the association for
2022 * spp_address - This specifies which address is of interest.
2023 * spp_hbinterval - This contains the value of the heartbeat interval,
2024 * in milliseconds. If a value of zero
2025 * is present in this field then no changes are to
2026 * be made to this parameter.
2027 * spp_pathmaxrxt - This contains the maximum number of
2028 * retransmissions before this address shall be
2029 * considered unreachable. If a value of zero
2030 * is present in this field then no changes are to
2031 * be made to this parameter.
2032 * spp_pathmtu - When Path MTU discovery is disabled the value
2033 * specified here will be the "fixed" path mtu.
2034 * Note that if the spp_address field is empty
2035 * then all associations on this address will
2036 * have this fixed path mtu set upon them.
2038 * spp_sackdelay - When delayed sack is enabled, this value specifies
2039 * the number of milliseconds that sacks will be delayed
2040 * for. This value will apply to all addresses of an
2041 * association if the spp_address field is empty. Note
2042 * also, that if delayed sack is enabled and this
2043 * value is set to 0, no change is made to the last
2044 * recorded delayed sack timer value.
2046 * spp_flags - These flags are used to control various features
2047 * on an association. The flag field may contain
2048 * zero or more of the following options.
2050 * SPP_HB_ENABLE - Enable heartbeats on the
2051 * specified address. Note that if the address
2052 * field is empty all addresses for the association
2053 * have heartbeats enabled upon them.
2055 * SPP_HB_DISABLE - Disable heartbeats on the
2056 * speicifed address. Note that if the address
2057 * field is empty all addresses for the association
2058 * will have their heartbeats disabled. Note also
2059 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2060 * mutually exclusive, only one of these two should
2061 * be specified. Enabling both fields will have
2062 * undetermined results.
2064 * SPP_HB_DEMAND - Request a user initiated heartbeat
2065 * to be made immediately.
2067 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2068 * heartbeat delayis to be set to the value of 0
2071 * SPP_PMTUD_ENABLE - This field will enable PMTU
2072 * discovery upon the specified address. Note that
2073 * if the address feild is empty then all addresses
2074 * on the association are effected.
2076 * SPP_PMTUD_DISABLE - This field will disable PMTU
2077 * discovery upon the specified address. Note that
2078 * if the address feild is empty then all addresses
2079 * on the association are effected. Not also that
2080 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2081 * exclusive. Enabling both will have undetermined
2084 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2085 * on delayed sack. The time specified in spp_sackdelay
2086 * is used to specify the sack delay for this address. Note
2087 * that if spp_address is empty then all addresses will
2088 * enable delayed sack and take on the sack delay
2089 * value specified in spp_sackdelay.
2090 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2091 * off delayed sack. If the spp_address field is blank then
2092 * delayed sack is disabled for the entire association. Note
2093 * also that this field is mutually exclusive to
2094 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2097 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2098 struct sctp_transport *trans,
2099 struct sctp_association *asoc,
2100 struct sctp_sock *sp,
2103 int sackdelay_change)
2107 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2108 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2113 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2114 * this field is ignored. Note also that a value of zero indicates
2115 * the current setting should be left unchanged.
2117 if (params->spp_flags & SPP_HB_ENABLE) {
2119 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2120 * set. This lets us use 0 value when this flag
2123 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2124 params->spp_hbinterval = 0;
2126 if (params->spp_hbinterval ||
2127 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2130 msecs_to_jiffies(params->spp_hbinterval);
2133 msecs_to_jiffies(params->spp_hbinterval);
2135 sp->hbinterval = params->spp_hbinterval;
2142 trans->param_flags =
2143 (trans->param_flags & ~SPP_HB) | hb_change;
2146 (asoc->param_flags & ~SPP_HB) | hb_change;
2149 (sp->param_flags & ~SPP_HB) | hb_change;
2153 /* When Path MTU discovery is disabled the value specified here will
2154 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2155 * include the flag SPP_PMTUD_DISABLE for this field to have any
2158 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2160 trans->pathmtu = params->spp_pathmtu;
2161 sctp_assoc_sync_pmtu(asoc);
2163 asoc->pathmtu = params->spp_pathmtu;
2164 sctp_frag_point(sp, params->spp_pathmtu);
2166 sp->pathmtu = params->spp_pathmtu;
2172 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2173 (params->spp_flags & SPP_PMTUD_ENABLE);
2174 trans->param_flags =
2175 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2177 sctp_transport_pmtu(trans);
2178 sctp_assoc_sync_pmtu(asoc);
2182 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2185 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2189 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2190 * value of this field is ignored. Note also that a value of zero
2191 * indicates the current setting should be left unchanged.
2193 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2196 msecs_to_jiffies(params->spp_sackdelay);
2199 msecs_to_jiffies(params->spp_sackdelay);
2201 sp->sackdelay = params->spp_sackdelay;
2205 if (sackdelay_change) {
2207 trans->param_flags =
2208 (trans->param_flags & ~SPP_SACKDELAY) |
2212 (asoc->param_flags & ~SPP_SACKDELAY) |
2216 (sp->param_flags & ~SPP_SACKDELAY) |
2221 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2222 * of this field is ignored. Note also that a value of zero
2223 * indicates the current setting should be left unchanged.
2225 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2227 trans->pathmaxrxt = params->spp_pathmaxrxt;
2229 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2231 sp->pathmaxrxt = params->spp_pathmaxrxt;
2238 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2239 char __user *optval, int optlen)
2241 struct sctp_paddrparams params;
2242 struct sctp_transport *trans = NULL;
2243 struct sctp_association *asoc = NULL;
2244 struct sctp_sock *sp = sctp_sk(sk);
2246 int hb_change, pmtud_change, sackdelay_change;
2248 if (optlen != sizeof(struct sctp_paddrparams))
2251 if (copy_from_user(¶ms, optval, optlen))
2254 /* Validate flags and value parameters. */
2255 hb_change = params.spp_flags & SPP_HB;
2256 pmtud_change = params.spp_flags & SPP_PMTUD;
2257 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2259 if (hb_change == SPP_HB ||
2260 pmtud_change == SPP_PMTUD ||
2261 sackdelay_change == SPP_SACKDELAY ||
2262 params.spp_sackdelay > 500 ||
2264 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2267 /* If an address other than INADDR_ANY is specified, and
2268 * no transport is found, then the request is invalid.
2270 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
2271 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2272 params.spp_assoc_id);
2277 /* Get association, if assoc_id != 0 and the socket is a one
2278 * to many style socket, and an association was not found, then
2279 * the id was invalid.
2281 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2282 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2285 /* Heartbeat demand can only be sent on a transport or
2286 * association, but not a socket.
2288 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2291 /* Process parameters. */
2292 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2293 hb_change, pmtud_change,
2299 /* If changes are for association, also apply parameters to each
2302 if (!trans && asoc) {
2303 struct list_head *pos;
2305 list_for_each(pos, &asoc->peer.transport_addr_list) {
2306 trans = list_entry(pos, struct sctp_transport,
2308 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2309 hb_change, pmtud_change,
2317 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2319 * This options will get or set the delayed ack timer. The time is set
2320 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2321 * endpoints default delayed ack timer value. If the assoc_id field is
2322 * non-zero, then the set or get effects the specified association.
2324 * struct sctp_assoc_value {
2325 * sctp_assoc_t assoc_id;
2326 * uint32_t assoc_value;
2329 * assoc_id - This parameter, indicates which association the
2330 * user is preforming an action upon. Note that if
2331 * this field's value is zero then the endpoints
2332 * default value is changed (effecting future
2333 * associations only).
2335 * assoc_value - This parameter contains the number of milliseconds
2336 * that the user is requesting the delayed ACK timer
2337 * be set to. Note that this value is defined in
2338 * the standard to be between 200 and 500 milliseconds.
2340 * Note: a value of zero will leave the value alone,
2341 * but disable SACK delay. A non-zero value will also
2342 * enable SACK delay.
2345 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2346 char __user *optval, int optlen)
2348 struct sctp_assoc_value params;
2349 struct sctp_transport *trans = NULL;
2350 struct sctp_association *asoc = NULL;
2351 struct sctp_sock *sp = sctp_sk(sk);
2353 if (optlen != sizeof(struct sctp_assoc_value))
2356 if (copy_from_user(¶ms, optval, optlen))
2359 /* Validate value parameter. */
2360 if (params.assoc_value > 500)
2363 /* Get association, if assoc_id != 0 and the socket is a one
2364 * to many style socket, and an association was not found, then
2365 * the id was invalid.
2367 asoc = sctp_id2assoc(sk, params.assoc_id);
2368 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2371 if (params.assoc_value) {
2374 msecs_to_jiffies(params.assoc_value);
2376 (asoc->param_flags & ~SPP_SACKDELAY) |
2377 SPP_SACKDELAY_ENABLE;
2379 sp->sackdelay = params.assoc_value;
2381 (sp->param_flags & ~SPP_SACKDELAY) |
2382 SPP_SACKDELAY_ENABLE;
2387 (asoc->param_flags & ~SPP_SACKDELAY) |
2388 SPP_SACKDELAY_DISABLE;
2391 (sp->param_flags & ~SPP_SACKDELAY) |
2392 SPP_SACKDELAY_DISABLE;
2396 /* If change is for association, also apply to each transport. */
2398 struct list_head *pos;
2400 list_for_each(pos, &asoc->peer.transport_addr_list) {
2401 trans = list_entry(pos, struct sctp_transport,
2403 if (params.assoc_value) {
2405 msecs_to_jiffies(params.assoc_value);
2406 trans->param_flags =
2407 (trans->param_flags & ~SPP_SACKDELAY) |
2408 SPP_SACKDELAY_ENABLE;
2410 trans->param_flags =
2411 (trans->param_flags & ~SPP_SACKDELAY) |
2412 SPP_SACKDELAY_DISABLE;
2420 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2422 * Applications can specify protocol parameters for the default association
2423 * initialization. The option name argument to setsockopt() and getsockopt()
2426 * Setting initialization parameters is effective only on an unconnected
2427 * socket (for UDP-style sockets only future associations are effected
2428 * by the change). With TCP-style sockets, this option is inherited by
2429 * sockets derived from a listener socket.
2431 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2433 struct sctp_initmsg sinit;
2434 struct sctp_sock *sp = sctp_sk(sk);
2436 if (optlen != sizeof(struct sctp_initmsg))
2438 if (copy_from_user(&sinit, optval, optlen))
2441 if (sinit.sinit_num_ostreams)
2442 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2443 if (sinit.sinit_max_instreams)
2444 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2445 if (sinit.sinit_max_attempts)
2446 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2447 if (sinit.sinit_max_init_timeo)
2448 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2454 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2456 * Applications that wish to use the sendto() system call may wish to
2457 * specify a default set of parameters that would normally be supplied
2458 * through the inclusion of ancillary data. This socket option allows
2459 * such an application to set the default sctp_sndrcvinfo structure.
2460 * The application that wishes to use this socket option simply passes
2461 * in to this call the sctp_sndrcvinfo structure defined in Section
2462 * 5.2.2) The input parameters accepted by this call include
2463 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2464 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2465 * to this call if the caller is using the UDP model.
2467 static int sctp_setsockopt_default_send_param(struct sock *sk,
2468 char __user *optval, int optlen)
2470 struct sctp_sndrcvinfo info;
2471 struct sctp_association *asoc;
2472 struct sctp_sock *sp = sctp_sk(sk);
2474 if (optlen != sizeof(struct sctp_sndrcvinfo))
2476 if (copy_from_user(&info, optval, optlen))
2479 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2480 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2484 asoc->default_stream = info.sinfo_stream;
2485 asoc->default_flags = info.sinfo_flags;
2486 asoc->default_ppid = info.sinfo_ppid;
2487 asoc->default_context = info.sinfo_context;
2488 asoc->default_timetolive = info.sinfo_timetolive;
2490 sp->default_stream = info.sinfo_stream;
2491 sp->default_flags = info.sinfo_flags;
2492 sp->default_ppid = info.sinfo_ppid;
2493 sp->default_context = info.sinfo_context;
2494 sp->default_timetolive = info.sinfo_timetolive;
2500 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2502 * Requests that the local SCTP stack use the enclosed peer address as
2503 * the association primary. The enclosed address must be one of the
2504 * association peer's addresses.
2506 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2509 struct sctp_prim prim;
2510 struct sctp_transport *trans;
2512 if (optlen != sizeof(struct sctp_prim))
2515 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2518 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2522 sctp_assoc_set_primary(trans->asoc, trans);
2528 * 7.1.5 SCTP_NODELAY
2530 * Turn on/off any Nagle-like algorithm. This means that packets are
2531 * generally sent as soon as possible and no unnecessary delays are
2532 * introduced, at the cost of more packets in the network. Expects an
2533 * integer boolean flag.
2535 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2540 if (optlen < sizeof(int))
2542 if (get_user(val, (int __user *)optval))
2545 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2551 * 7.1.1 SCTP_RTOINFO
2553 * The protocol parameters used to initialize and bound retransmission
2554 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2555 * and modify these parameters.
2556 * All parameters are time values, in milliseconds. A value of 0, when
2557 * modifying the parameters, indicates that the current value should not
2561 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2562 struct sctp_rtoinfo rtoinfo;
2563 struct sctp_association *asoc;
2565 if (optlen != sizeof (struct sctp_rtoinfo))
2568 if (copy_from_user(&rtoinfo, optval, optlen))
2571 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2573 /* Set the values to the specific association */
2574 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2578 if (rtoinfo.srto_initial != 0)
2580 msecs_to_jiffies(rtoinfo.srto_initial);
2581 if (rtoinfo.srto_max != 0)
2582 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2583 if (rtoinfo.srto_min != 0)
2584 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2586 /* If there is no association or the association-id = 0
2587 * set the values to the endpoint.
2589 struct sctp_sock *sp = sctp_sk(sk);
2591 if (rtoinfo.srto_initial != 0)
2592 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2593 if (rtoinfo.srto_max != 0)
2594 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2595 if (rtoinfo.srto_min != 0)
2596 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2604 * 7.1.2 SCTP_ASSOCINFO
2606 * This option is used to tune the maximum retransmission attempts
2607 * of the association.
2608 * Returns an error if the new association retransmission value is
2609 * greater than the sum of the retransmission value of the peer.
2610 * See [SCTP] for more information.
2613 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2616 struct sctp_assocparams assocparams;
2617 struct sctp_association *asoc;
2619 if (optlen != sizeof(struct sctp_assocparams))
2621 if (copy_from_user(&assocparams, optval, optlen))
2624 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2626 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2629 /* Set the values to the specific association */
2631 if (assocparams.sasoc_asocmaxrxt != 0) {
2634 struct list_head *pos;
2635 struct sctp_transport *peer_addr;
2637 list_for_each(pos, &asoc->peer.transport_addr_list) {
2638 peer_addr = list_entry(pos,
2639 struct sctp_transport,
2641 path_sum += peer_addr->pathmaxrxt;
2645 /* Only validate asocmaxrxt if we have more then
2646 * one path/transport. We do this because path
2647 * retransmissions are only counted when we have more
2651 assocparams.sasoc_asocmaxrxt > path_sum)
2654 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2657 if (assocparams.sasoc_cookie_life != 0) {
2658 asoc->cookie_life.tv_sec =
2659 assocparams.sasoc_cookie_life / 1000;
2660 asoc->cookie_life.tv_usec =
2661 (assocparams.sasoc_cookie_life % 1000)
2665 /* Set the values to the endpoint */
2666 struct sctp_sock *sp = sctp_sk(sk);
2668 if (assocparams.sasoc_asocmaxrxt != 0)
2669 sp->assocparams.sasoc_asocmaxrxt =
2670 assocparams.sasoc_asocmaxrxt;
2671 if (assocparams.sasoc_cookie_life != 0)
2672 sp->assocparams.sasoc_cookie_life =
2673 assocparams.sasoc_cookie_life;
2679 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2681 * This socket option is a boolean flag which turns on or off mapped V4
2682 * addresses. If this option is turned on and the socket is type
2683 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2684 * If this option is turned off, then no mapping will be done of V4
2685 * addresses and a user will receive both PF_INET6 and PF_INET type
2686 * addresses on the socket.
2688 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2691 struct sctp_sock *sp = sctp_sk(sk);
2693 if (optlen < sizeof(int))
2695 if (get_user(val, (int __user *)optval))
2706 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2708 * This socket option specifies the maximum size to put in any outgoing
2709 * SCTP chunk. If a message is larger than this size it will be
2710 * fragmented by SCTP into the specified size. Note that the underlying
2711 * SCTP implementation may fragment into smaller sized chunks when the
2712 * PMTU of the underlying association is smaller than the value set by
2715 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2717 struct sctp_association *asoc;
2718 struct list_head *pos;
2719 struct sctp_sock *sp = sctp_sk(sk);
2722 if (optlen < sizeof(int))
2724 if (get_user(val, (int __user *)optval))
2726 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2728 sp->user_frag = val;
2730 /* Update the frag_point of the existing associations. */
2731 list_for_each(pos, &(sp->ep->asocs)) {
2732 asoc = list_entry(pos, struct sctp_association, asocs);
2733 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2741 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2743 * Requests that the peer mark the enclosed address as the association
2744 * primary. The enclosed address must be one of the association's
2745 * locally bound addresses. The following structure is used to make a
2746 * set primary request:
2748 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2751 struct sctp_sock *sp;
2752 struct sctp_endpoint *ep;
2753 struct sctp_association *asoc = NULL;
2754 struct sctp_setpeerprim prim;
2755 struct sctp_chunk *chunk;
2761 if (!sctp_addip_enable)
2764 if (optlen != sizeof(struct sctp_setpeerprim))
2767 if (copy_from_user(&prim, optval, optlen))
2770 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2774 if (!asoc->peer.asconf_capable)
2777 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2780 if (!sctp_state(asoc, ESTABLISHED))
2783 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2784 return -EADDRNOTAVAIL;
2786 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2787 chunk = sctp_make_asconf_set_prim(asoc,
2788 (union sctp_addr *)&prim.sspp_addr);
2792 err = sctp_send_asconf(asoc, chunk);
2794 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2799 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2802 struct sctp_setadaptation adaptation;
2804 if (optlen != sizeof(struct sctp_setadaptation))
2806 if (copy_from_user(&adaptation, optval, optlen))
2809 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2815 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2817 * The context field in the sctp_sndrcvinfo structure is normally only
2818 * used when a failed message is retrieved holding the value that was
2819 * sent down on the actual send call. This option allows the setting of
2820 * a default context on an association basis that will be received on
2821 * reading messages from the peer. This is especially helpful in the
2822 * one-2-many model for an application to keep some reference to an
2823 * internal state machine that is processing messages on the
2824 * association. Note that the setting of this value only effects
2825 * received messages from the peer and does not effect the value that is
2826 * saved with outbound messages.
2828 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2831 struct sctp_assoc_value params;
2832 struct sctp_sock *sp;
2833 struct sctp_association *asoc;
2835 if (optlen != sizeof(struct sctp_assoc_value))
2837 if (copy_from_user(¶ms, optval, optlen))
2842 if (params.assoc_id != 0) {
2843 asoc = sctp_id2assoc(sk, params.assoc_id);
2846 asoc->default_rcv_context = params.assoc_value;
2848 sp->default_rcv_context = params.assoc_value;
2855 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2857 * This options will at a minimum specify if the implementation is doing
2858 * fragmented interleave. Fragmented interleave, for a one to many
2859 * socket, is when subsequent calls to receive a message may return
2860 * parts of messages from different associations. Some implementations
2861 * may allow you to turn this value on or off. If so, when turned off,
2862 * no fragment interleave will occur (which will cause a head of line
2863 * blocking amongst multiple associations sharing the same one to many
2864 * socket). When this option is turned on, then each receive call may
2865 * come from a different association (thus the user must receive data
2866 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2867 * association each receive belongs to.
2869 * This option takes a boolean value. A non-zero value indicates that
2870 * fragmented interleave is on. A value of zero indicates that
2871 * fragmented interleave is off.
2873 * Note that it is important that an implementation that allows this
2874 * option to be turned on, have it off by default. Otherwise an unaware
2875 * application using the one to many model may become confused and act
2878 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2879 char __user *optval,
2884 if (optlen != sizeof(int))
2886 if (get_user(val, (int __user *)optval))
2889 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2895 * 7.1.25. Set or Get the sctp partial delivery point
2896 * (SCTP_PARTIAL_DELIVERY_POINT)
2897 * This option will set or get the SCTP partial delivery point. This
2898 * point is the size of a message where the partial delivery API will be
2899 * invoked to help free up rwnd space for the peer. Setting this to a
2900 * lower value will cause partial delivery's to happen more often. The
2901 * calls argument is an integer that sets or gets the partial delivery
2904 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2905 char __user *optval,
2910 if (optlen != sizeof(u32))
2912 if (get_user(val, (int __user *)optval))
2915 sctp_sk(sk)->pd_point = val;
2917 return 0; /* is this the right error code? */
2921 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2923 * This option will allow a user to change the maximum burst of packets
2924 * that can be emitted by this association. Note that the default value
2925 * is 4, and some implementations may restrict this setting so that it
2926 * can only be lowered.
2928 * NOTE: This text doesn't seem right. Do this on a socket basis with
2929 * future associations inheriting the socket value.
2931 static int sctp_setsockopt_maxburst(struct sock *sk,
2932 char __user *optval,
2937 if (optlen != sizeof(int))
2939 if (get_user(val, (int __user *)optval))
2945 sctp_sk(sk)->max_burst = val;
2950 /* API 6.2 setsockopt(), getsockopt()
2952 * Applications use setsockopt() and getsockopt() to set or retrieve
2953 * socket options. Socket options are used to change the default
2954 * behavior of sockets calls. They are described in Section 7.
2958 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2959 * int __user *optlen);
2960 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2963 * sd - the socket descript.
2964 * level - set to IPPROTO_SCTP for all SCTP options.
2965 * optname - the option name.
2966 * optval - the buffer to store the value of the option.
2967 * optlen - the size of the buffer.
2969 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2970 char __user *optval, int optlen)
2974 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2977 /* I can hardly begin to describe how wrong this is. This is
2978 * so broken as to be worse than useless. The API draft
2979 * REALLY is NOT helpful here... I am not convinced that the
2980 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2981 * are at all well-founded.
2983 if (level != SOL_SCTP) {
2984 struct sctp_af *af = sctp_sk(sk)->pf->af;
2985 retval = af->setsockopt(sk, level, optname, optval, optlen);
2992 case SCTP_SOCKOPT_BINDX_ADD:
2993 /* 'optlen' is the size of the addresses buffer. */
2994 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2995 optlen, SCTP_BINDX_ADD_ADDR);
2998 case SCTP_SOCKOPT_BINDX_REM:
2999 /* 'optlen' is the size of the addresses buffer. */
3000 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3001 optlen, SCTP_BINDX_REM_ADDR);
3004 case SCTP_SOCKOPT_CONNECTX:
3005 /* 'optlen' is the size of the addresses buffer. */
3006 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3010 case SCTP_DISABLE_FRAGMENTS:
3011 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3015 retval = sctp_setsockopt_events(sk, optval, optlen);
3018 case SCTP_AUTOCLOSE:
3019 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3022 case SCTP_PEER_ADDR_PARAMS:
3023 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3026 case SCTP_DELAYED_ACK_TIME:
3027 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3029 case SCTP_PARTIAL_DELIVERY_POINT:
3030 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3034 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3036 case SCTP_DEFAULT_SEND_PARAM:
3037 retval = sctp_setsockopt_default_send_param(sk, optval,
3040 case SCTP_PRIMARY_ADDR:
3041 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3043 case SCTP_SET_PEER_PRIMARY_ADDR:
3044 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3047 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3050 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3052 case SCTP_ASSOCINFO:
3053 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3055 case SCTP_I_WANT_MAPPED_V4_ADDR:
3056 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3059 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3061 case SCTP_ADAPTATION_LAYER:
3062 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3065 retval = sctp_setsockopt_context(sk, optval, optlen);
3067 case SCTP_FRAGMENT_INTERLEAVE:
3068 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3070 case SCTP_MAX_BURST:
3071 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3074 retval = -ENOPROTOOPT;
3078 sctp_release_sock(sk);
3084 /* API 3.1.6 connect() - UDP Style Syntax
3086 * An application may use the connect() call in the UDP model to initiate an
3087 * association without sending data.
3091 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3093 * sd: the socket descriptor to have a new association added to.
3095 * nam: the address structure (either struct sockaddr_in or struct
3096 * sockaddr_in6 defined in RFC2553 [7]).
3098 * len: the size of the address.
3100 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3108 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3109 __FUNCTION__, sk, addr, addr_len);
3111 /* Validate addr_len before calling common connect/connectx routine. */
3112 af = sctp_get_af_specific(addr->sa_family);
3113 if (!af || addr_len < af->sockaddr_len) {
3116 /* Pass correct addr len to common routine (so it knows there
3117 * is only one address being passed.
3119 err = __sctp_connect(sk, addr, af->sockaddr_len);
3122 sctp_release_sock(sk);
3126 /* FIXME: Write comments. */
3127 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3129 return -EOPNOTSUPP; /* STUB */
3132 /* 4.1.4 accept() - TCP Style Syntax
3134 * Applications use accept() call to remove an established SCTP
3135 * association from the accept queue of the endpoint. A new socket
3136 * descriptor will be returned from accept() to represent the newly
3137 * formed association.
3139 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3141 struct sctp_sock *sp;
3142 struct sctp_endpoint *ep;
3143 struct sock *newsk = NULL;
3144 struct sctp_association *asoc;
3153 if (!sctp_style(sk, TCP)) {
3154 error = -EOPNOTSUPP;
3158 if (!sctp_sstate(sk, LISTENING)) {
3163 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3165 error = sctp_wait_for_accept(sk, timeo);
3169 /* We treat the list of associations on the endpoint as the accept
3170 * queue and pick the first association on the list.
3172 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3174 newsk = sp->pf->create_accept_sk(sk, asoc);
3180 /* Populate the fields of the newsk from the oldsk and migrate the
3181 * asoc to the newsk.
3183 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3186 sctp_release_sock(sk);
3191 /* The SCTP ioctl handler. */
3192 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3194 return -ENOIOCTLCMD;
3197 /* This is the function which gets called during socket creation to
3198 * initialized the SCTP-specific portion of the sock.
3199 * The sock structure should already be zero-filled memory.
3201 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3203 struct sctp_endpoint *ep;
3204 struct sctp_sock *sp;
3206 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3210 /* Initialize the SCTP per socket area. */
3211 switch (sk->sk_type) {
3212 case SOCK_SEQPACKET:
3213 sp->type = SCTP_SOCKET_UDP;
3216 sp->type = SCTP_SOCKET_TCP;
3219 return -ESOCKTNOSUPPORT;
3222 /* Initialize default send parameters. These parameters can be
3223 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3225 sp->default_stream = 0;
3226 sp->default_ppid = 0;
3227 sp->default_flags = 0;
3228 sp->default_context = 0;
3229 sp->default_timetolive = 0;
3231 sp->default_rcv_context = 0;
3232 sp->max_burst = sctp_max_burst;
3234 /* Initialize default setup parameters. These parameters
3235 * can be modified with the SCTP_INITMSG socket option or
3236 * overridden by the SCTP_INIT CMSG.
3238 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3239 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3240 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3241 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3243 /* Initialize default RTO related parameters. These parameters can
3244 * be modified for with the SCTP_RTOINFO socket option.
3246 sp->rtoinfo.srto_initial = sctp_rto_initial;
3247 sp->rtoinfo.srto_max = sctp_rto_max;
3248 sp->rtoinfo.srto_min = sctp_rto_min;
3250 /* Initialize default association related parameters. These parameters
3251 * can be modified with the SCTP_ASSOCINFO socket option.
3253 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3254 sp->assocparams.sasoc_number_peer_destinations = 0;
3255 sp->assocparams.sasoc_peer_rwnd = 0;
3256 sp->assocparams.sasoc_local_rwnd = 0;
3257 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3259 /* Initialize default event subscriptions. By default, all the
3262 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3264 /* Default Peer Address Parameters. These defaults can
3265 * be modified via SCTP_PEER_ADDR_PARAMS
3267 sp->hbinterval = sctp_hb_interval;
3268 sp->pathmaxrxt = sctp_max_retrans_path;
3269 sp->pathmtu = 0; // allow default discovery
3270 sp->sackdelay = sctp_sack_timeout;
3271 sp->param_flags = SPP_HB_ENABLE |
3273 SPP_SACKDELAY_ENABLE;
3275 /* If enabled no SCTP message fragmentation will be performed.
3276 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3278 sp->disable_fragments = 0;
3280 /* Enable Nagle algorithm by default. */
3283 /* Enable by default. */
3286 /* Auto-close idle associations after the configured
3287 * number of seconds. A value of 0 disables this
3288 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3289 * for UDP-style sockets only.
3293 /* User specified fragmentation limit. */
3296 sp->adaptation_ind = 0;
3298 sp->pf = sctp_get_pf_specific(sk->sk_family);
3300 /* Control variables for partial data delivery. */
3301 atomic_set(&sp->pd_mode, 0);
3302 skb_queue_head_init(&sp->pd_lobby);
3303 sp->frag_interleave = 0;
3305 /* Create a per socket endpoint structure. Even if we
3306 * change the data structure relationships, this may still
3307 * be useful for storing pre-connect address information.
3309 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3316 SCTP_DBG_OBJCNT_INC(sock);
3320 /* Cleanup any SCTP per socket resources. */
3321 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3323 struct sctp_endpoint *ep;
3325 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3327 /* Release our hold on the endpoint. */
3328 ep = sctp_sk(sk)->ep;
3329 sctp_endpoint_free(ep);
3334 /* API 4.1.7 shutdown() - TCP Style Syntax
3335 * int shutdown(int socket, int how);
3337 * sd - the socket descriptor of the association to be closed.
3338 * how - Specifies the type of shutdown. The values are
3341 * Disables further receive operations. No SCTP
3342 * protocol action is taken.
3344 * Disables further send operations, and initiates
3345 * the SCTP shutdown sequence.
3347 * Disables further send and receive operations
3348 * and initiates the SCTP shutdown sequence.
3350 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3352 struct sctp_endpoint *ep;
3353 struct sctp_association *asoc;
3355 if (!sctp_style(sk, TCP))
3358 if (how & SEND_SHUTDOWN) {
3359 ep = sctp_sk(sk)->ep;
3360 if (!list_empty(&ep->asocs)) {
3361 asoc = list_entry(ep->asocs.next,
3362 struct sctp_association, asocs);
3363 sctp_primitive_SHUTDOWN(asoc, NULL);
3368 /* 7.2.1 Association Status (SCTP_STATUS)
3370 * Applications can retrieve current status information about an
3371 * association, including association state, peer receiver window size,
3372 * number of unacked data chunks, and number of data chunks pending
3373 * receipt. This information is read-only.
3375 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3376 char __user *optval,
3379 struct sctp_status status;
3380 struct sctp_association *asoc = NULL;
3381 struct sctp_transport *transport;
3382 sctp_assoc_t associd;
3385 if (len < sizeof(status)) {
3390 len = sizeof(status);
3391 if (copy_from_user(&status, optval, len)) {
3396 associd = status.sstat_assoc_id;
3397 asoc = sctp_id2assoc(sk, associd);
3403 transport = asoc->peer.primary_path;
3405 status.sstat_assoc_id = sctp_assoc2id(asoc);
3406 status.sstat_state = asoc->state;
3407 status.sstat_rwnd = asoc->peer.rwnd;
3408 status.sstat_unackdata = asoc->unack_data;
3410 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3411 status.sstat_instrms = asoc->c.sinit_max_instreams;
3412 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3413 status.sstat_fragmentation_point = asoc->frag_point;
3414 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3415 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3416 transport->af_specific->sockaddr_len);
3417 /* Map ipv4 address into v4-mapped-on-v6 address. */
3418 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3419 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3420 status.sstat_primary.spinfo_state = transport->state;
3421 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3422 status.sstat_primary.spinfo_srtt = transport->srtt;
3423 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3424 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3426 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3427 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3429 if (put_user(len, optlen)) {
3434 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3435 len, status.sstat_state, status.sstat_rwnd,
3436 status.sstat_assoc_id);
3438 if (copy_to_user(optval, &status, len)) {
3448 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3450 * Applications can retrieve information about a specific peer address
3451 * of an association, including its reachability state, congestion
3452 * window, and retransmission timer values. This information is
3455 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3456 char __user *optval,
3459 struct sctp_paddrinfo pinfo;
3460 struct sctp_transport *transport;
3463 if (len < sizeof(pinfo)) {
3468 len = sizeof(pinfo);
3469 if (copy_from_user(&pinfo, optval, len)) {
3474 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3475 pinfo.spinfo_assoc_id);
3479 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3480 pinfo.spinfo_state = transport->state;
3481 pinfo.spinfo_cwnd = transport->cwnd;
3482 pinfo.spinfo_srtt = transport->srtt;
3483 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3484 pinfo.spinfo_mtu = transport->pathmtu;
3486 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3487 pinfo.spinfo_state = SCTP_ACTIVE;
3489 if (put_user(len, optlen)) {
3494 if (copy_to_user(optval, &pinfo, len)) {
3503 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3505 * This option is a on/off flag. If enabled no SCTP message
3506 * fragmentation will be performed. Instead if a message being sent
3507 * exceeds the current PMTU size, the message will NOT be sent and
3508 * instead a error will be indicated to the user.
3510 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3511 char __user *optval, int __user *optlen)
3515 if (len < sizeof(int))
3519 val = (sctp_sk(sk)->disable_fragments == 1);
3520 if (put_user(len, optlen))
3522 if (copy_to_user(optval, &val, len))
3527 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3529 * This socket option is used to specify various notifications and
3530 * ancillary data the user wishes to receive.
3532 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3535 if (len < sizeof(struct sctp_event_subscribe))
3537 len = sizeof(struct sctp_event_subscribe);
3538 if (put_user(len, optlen))
3540 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3545 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3547 * This socket option is applicable to the UDP-style socket only. When
3548 * set it will cause associations that are idle for more than the
3549 * specified number of seconds to automatically close. An association
3550 * being idle is defined an association that has NOT sent or received
3551 * user data. The special value of '0' indicates that no automatic
3552 * close of any associations should be performed. The option expects an
3553 * integer defining the number of seconds of idle time before an
3554 * association is closed.
3556 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3558 /* Applicable to UDP-style socket only */
3559 if (sctp_style(sk, TCP))
3561 if (len < sizeof(int))
3564 if (put_user(len, optlen))
3566 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3571 /* Helper routine to branch off an association to a new socket. */
3572 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3573 struct socket **sockp)
3575 struct sock *sk = asoc->base.sk;
3576 struct socket *sock;
3577 struct inet_sock *inetsk;
3581 /* An association cannot be branched off from an already peeled-off
3582 * socket, nor is this supported for tcp style sockets.
3584 if (!sctp_style(sk, UDP))
3587 /* Create a new socket. */
3588 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3592 /* Populate the fields of the newsk from the oldsk and migrate the
3593 * asoc to the newsk.
3595 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3597 /* Make peeled-off sockets more like 1-1 accepted sockets.
3598 * Set the daddr and initialize id to something more random
3600 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3601 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3602 inetsk = inet_sk(sock->sk);
3603 inetsk->id = asoc->next_tsn ^ jiffies;
3610 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3612 sctp_peeloff_arg_t peeloff;
3613 struct socket *newsock;
3615 struct sctp_association *asoc;
3617 if (len < sizeof(sctp_peeloff_arg_t))
3619 len = sizeof(sctp_peeloff_arg_t);
3620 if (copy_from_user(&peeloff, optval, len))
3623 asoc = sctp_id2assoc(sk, peeloff.associd);
3629 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3631 retval = sctp_do_peeloff(asoc, &newsock);
3635 /* Map the socket to an unused fd that can be returned to the user. */
3636 retval = sock_map_fd(newsock);
3638 sock_release(newsock);
3642 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3643 __FUNCTION__, sk, asoc, newsock->sk, retval);
3645 /* Return the fd mapped to the new socket. */
3646 peeloff.sd = retval;
3647 if (put_user(len, optlen))
3649 if (copy_to_user(optval, &peeloff, len))
3656 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3658 * Applications can enable or disable heartbeats for any peer address of
3659 * an association, modify an address's heartbeat interval, force a
3660 * heartbeat to be sent immediately, and adjust the address's maximum
3661 * number of retransmissions sent before an address is considered
3662 * unreachable. The following structure is used to access and modify an
3663 * address's parameters:
3665 * struct sctp_paddrparams {
3666 * sctp_assoc_t spp_assoc_id;
3667 * struct sockaddr_storage spp_address;
3668 * uint32_t spp_hbinterval;
3669 * uint16_t spp_pathmaxrxt;
3670 * uint32_t spp_pathmtu;
3671 * uint32_t spp_sackdelay;
3672 * uint32_t spp_flags;
3675 * spp_assoc_id - (one-to-many style socket) This is filled in the
3676 * application, and identifies the association for
3678 * spp_address - This specifies which address is of interest.
3679 * spp_hbinterval - This contains the value of the heartbeat interval,
3680 * in milliseconds. If a value of zero
3681 * is present in this field then no changes are to
3682 * be made to this parameter.
3683 * spp_pathmaxrxt - This contains the maximum number of
3684 * retransmissions before this address shall be
3685 * considered unreachable. If a value of zero
3686 * is present in this field then no changes are to
3687 * be made to this parameter.
3688 * spp_pathmtu - When Path MTU discovery is disabled the value
3689 * specified here will be the "fixed" path mtu.
3690 * Note that if the spp_address field is empty
3691 * then all associations on this address will
3692 * have this fixed path mtu set upon them.
3694 * spp_sackdelay - When delayed sack is enabled, this value specifies
3695 * the number of milliseconds that sacks will be delayed
3696 * for. This value will apply to all addresses of an
3697 * association if the spp_address field is empty. Note
3698 * also, that if delayed sack is enabled and this
3699 * value is set to 0, no change is made to the last
3700 * recorded delayed sack timer value.
3702 * spp_flags - These flags are used to control various features
3703 * on an association. The flag field may contain
3704 * zero or more of the following options.
3706 * SPP_HB_ENABLE - Enable heartbeats on the
3707 * specified address. Note that if the address
3708 * field is empty all addresses for the association
3709 * have heartbeats enabled upon them.
3711 * SPP_HB_DISABLE - Disable heartbeats on the
3712 * speicifed address. Note that if the address
3713 * field is empty all addresses for the association
3714 * will have their heartbeats disabled. Note also
3715 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3716 * mutually exclusive, only one of these two should
3717 * be specified. Enabling both fields will have
3718 * undetermined results.
3720 * SPP_HB_DEMAND - Request a user initiated heartbeat
3721 * to be made immediately.
3723 * SPP_PMTUD_ENABLE - This field will enable PMTU
3724 * discovery upon the specified address. Note that
3725 * if the address feild is empty then all addresses
3726 * on the association are effected.
3728 * SPP_PMTUD_DISABLE - This field will disable PMTU
3729 * discovery upon the specified address. Note that
3730 * if the address feild is empty then all addresses
3731 * on the association are effected. Not also that
3732 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3733 * exclusive. Enabling both will have undetermined
3736 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3737 * on delayed sack. The time specified in spp_sackdelay
3738 * is used to specify the sack delay for this address. Note
3739 * that if spp_address is empty then all addresses will
3740 * enable delayed sack and take on the sack delay
3741 * value specified in spp_sackdelay.
3742 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3743 * off delayed sack. If the spp_address field is blank then
3744 * delayed sack is disabled for the entire association. Note
3745 * also that this field is mutually exclusive to
3746 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3749 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3750 char __user *optval, int __user *optlen)
3752 struct sctp_paddrparams params;
3753 struct sctp_transport *trans = NULL;
3754 struct sctp_association *asoc = NULL;
3755 struct sctp_sock *sp = sctp_sk(sk);
3757 if (len < sizeof(struct sctp_paddrparams))
3759 len = sizeof(struct sctp_paddrparams);
3760 if (copy_from_user(¶ms, optval, len))
3763 /* If an address other than INADDR_ANY is specified, and
3764 * no transport is found, then the request is invalid.
3766 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3767 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3768 params.spp_assoc_id);
3770 SCTP_DEBUG_PRINTK("Failed no transport\n");
3775 /* Get association, if assoc_id != 0 and the socket is a one
3776 * to many style socket, and an association was not found, then
3777 * the id was invalid.
3779 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3780 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3781 SCTP_DEBUG_PRINTK("Failed no association\n");
3786 /* Fetch transport values. */
3787 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3788 params.spp_pathmtu = trans->pathmtu;
3789 params.spp_pathmaxrxt = trans->pathmaxrxt;
3790 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3792 /*draft-11 doesn't say what to return in spp_flags*/
3793 params.spp_flags = trans->param_flags;
3795 /* Fetch association values. */
3796 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3797 params.spp_pathmtu = asoc->pathmtu;
3798 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3799 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3801 /*draft-11 doesn't say what to return in spp_flags*/
3802 params.spp_flags = asoc->param_flags;
3804 /* Fetch socket values. */
3805 params.spp_hbinterval = sp->hbinterval;
3806 params.spp_pathmtu = sp->pathmtu;
3807 params.spp_sackdelay = sp->sackdelay;
3808 params.spp_pathmaxrxt = sp->pathmaxrxt;
3810 /*draft-11 doesn't say what to return in spp_flags*/
3811 params.spp_flags = sp->param_flags;
3814 if (copy_to_user(optval, ¶ms, len))
3817 if (put_user(len, optlen))
3823 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3825 * This options will get or set the delayed ack timer. The time is set
3826 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3827 * endpoints default delayed ack timer value. If the assoc_id field is
3828 * non-zero, then the set or get effects the specified association.
3830 * struct sctp_assoc_value {
3831 * sctp_assoc_t assoc_id;
3832 * uint32_t assoc_value;
3835 * assoc_id - This parameter, indicates which association the
3836 * user is preforming an action upon. Note that if
3837 * this field's value is zero then the endpoints
3838 * default value is changed (effecting future
3839 * associations only).
3841 * assoc_value - This parameter contains the number of milliseconds
3842 * that the user is requesting the delayed ACK timer
3843 * be set to. Note that this value is defined in
3844 * the standard to be between 200 and 500 milliseconds.
3846 * Note: a value of zero will leave the value alone,
3847 * but disable SACK delay. A non-zero value will also
3848 * enable SACK delay.
3850 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3851 char __user *optval,
3854 struct sctp_assoc_value params;
3855 struct sctp_association *asoc = NULL;
3856 struct sctp_sock *sp = sctp_sk(sk);
3858 if (len < sizeof(struct sctp_assoc_value))
3861 len = sizeof(struct sctp_assoc_value);
3863 if (copy_from_user(¶ms, optval, len))
3866 /* Get association, if assoc_id != 0 and the socket is a one
3867 * to many style socket, and an association was not found, then
3868 * the id was invalid.
3870 asoc = sctp_id2assoc(sk, params.assoc_id);
3871 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3875 /* Fetch association values. */
3876 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3877 params.assoc_value = jiffies_to_msecs(
3880 params.assoc_value = 0;
3882 /* Fetch socket values. */
3883 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3884 params.assoc_value = sp->sackdelay;
3886 params.assoc_value = 0;
3889 if (copy_to_user(optval, ¶ms, len))
3892 if (put_user(len, optlen))
3898 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3900 * Applications can specify protocol parameters for the default association
3901 * initialization. The option name argument to setsockopt() and getsockopt()
3904 * Setting initialization parameters is effective only on an unconnected
3905 * socket (for UDP-style sockets only future associations are effected
3906 * by the change). With TCP-style sockets, this option is inherited by
3907 * sockets derived from a listener socket.
3909 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3911 if (len < sizeof(struct sctp_initmsg))
3913 len = sizeof(struct sctp_initmsg);
3914 if (put_user(len, optlen))
3916 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3921 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3922 char __user *optval,
3926 struct sctp_association *asoc;
3927 struct list_head *pos;
3930 if (len < sizeof(sctp_assoc_t))
3933 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3936 /* For UDP-style sockets, id specifies the association to query. */
3937 asoc = sctp_id2assoc(sk, id);
3941 list_for_each(pos, &asoc->peer.transport_addr_list) {
3949 * Old API for getting list of peer addresses. Does not work for 32-bit
3950 * programs running on a 64-bit kernel
3952 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3953 char __user *optval,
3956 struct sctp_association *asoc;
3957 struct list_head *pos;
3959 struct sctp_getaddrs_old getaddrs;
3960 struct sctp_transport *from;
3962 union sctp_addr temp;
3963 struct sctp_sock *sp = sctp_sk(sk);
3966 if (len < sizeof(struct sctp_getaddrs_old))
3969 len = sizeof(struct sctp_getaddrs_old);
3971 if (copy_from_user(&getaddrs, optval, len))
3974 if (getaddrs.addr_num <= 0) return -EINVAL;
3976 /* For UDP-style sockets, id specifies the association to query. */
3977 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3981 to = (void __user *)getaddrs.addrs;
3982 list_for_each(pos, &asoc->peer.transport_addr_list) {
3983 from = list_entry(pos, struct sctp_transport, transports);
3984 memcpy(&temp, &from->ipaddr, sizeof(temp));
3985 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3986 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3987 if (copy_to_user(to, &temp, addrlen))
3991 if (cnt >= getaddrs.addr_num) break;
3993 getaddrs.addr_num = cnt;
3994 if (put_user(len, optlen))
3996 if (copy_to_user(optval, &getaddrs, len))
4002 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4003 char __user *optval, int __user *optlen)
4005 struct sctp_association *asoc;
4006 struct list_head *pos;
4008 struct sctp_getaddrs getaddrs;
4009 struct sctp_transport *from;
4011 union sctp_addr temp;
4012 struct sctp_sock *sp = sctp_sk(sk);
4017 if (len < sizeof(struct sctp_getaddrs))
4020 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4023 /* For UDP-style sockets, id specifies the association to query. */
4024 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4028 to = optval + offsetof(struct sctp_getaddrs,addrs);
4029 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4031 list_for_each(pos, &asoc->peer.transport_addr_list) {
4032 from = list_entry(pos, struct sctp_transport, transports);
4033 memcpy(&temp, &from->ipaddr, sizeof(temp));
4034 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4035 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4036 if (space_left < addrlen)
4038 if (copy_to_user(to, &temp, addrlen))
4042 space_left -= addrlen;
4045 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4047 bytes_copied = ((char __user *)to) - optval;
4048 if (put_user(bytes_copied, optlen))
4054 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4055 char __user *optval,
4059 struct sctp_bind_addr *bp;
4060 struct sctp_association *asoc;
4061 struct list_head *pos, *temp;
4062 struct sctp_sockaddr_entry *addr;
4063 rwlock_t *addr_lock;
4066 if (len < sizeof(sctp_assoc_t))
4069 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4073 * For UDP-style sockets, id specifies the association to query.
4074 * If the id field is set to the value '0' then the locally bound
4075 * addresses are returned without regard to any particular
4079 bp = &sctp_sk(sk)->ep->base.bind_addr;
4080 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4082 asoc = sctp_id2assoc(sk, id);
4085 bp = &asoc->base.bind_addr;
4086 addr_lock = &asoc->base.addr_lock;
4089 sctp_read_lock(addr_lock);
4091 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4092 * addresses from the global local address list.
4094 if (sctp_list_single_entry(&bp->address_list)) {
4095 addr = list_entry(bp->address_list.next,
4096 struct sctp_sockaddr_entry, list);
4097 if (sctp_is_any(&addr->a)) {
4098 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
4099 addr = list_entry(pos,
4100 struct sctp_sockaddr_entry,
4102 if ((PF_INET == sk->sk_family) &&
4103 (AF_INET6 == addr->a.sa.sa_family))
4113 list_for_each(pos, &bp->address_list) {
4118 sctp_read_unlock(addr_lock);
4122 /* Helper function that copies local addresses to user and returns the number
4123 * of addresses copied.
4125 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4126 int max_addrs, void *to,
4129 struct list_head *pos, *next;
4130 struct sctp_sockaddr_entry *addr;
4131 union sctp_addr temp;
4135 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4136 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4137 if ((PF_INET == sk->sk_family) &&
4138 (AF_INET6 == addr->a.sa.sa_family))
4140 memcpy(&temp, &addr->a, sizeof(temp));
4141 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4143 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4144 memcpy(to, &temp, addrlen);
4147 *bytes_copied += addrlen;
4149 if (cnt >= max_addrs) break;
4155 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4156 size_t space_left, int *bytes_copied)
4158 struct list_head *pos, *next;
4159 struct sctp_sockaddr_entry *addr;
4160 union sctp_addr temp;
4164 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4165 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4166 if ((PF_INET == sk->sk_family) &&
4167 (AF_INET6 == addr->a.sa.sa_family))
4169 memcpy(&temp, &addr->a, sizeof(temp));
4170 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4172 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4173 if (space_left < addrlen)
4175 memcpy(to, &temp, addrlen);
4179 space_left -= addrlen;
4180 *bytes_copied += addrlen;
4186 /* Old API for getting list of local addresses. Does not work for 32-bit
4187 * programs running on a 64-bit kernel
4189 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4190 char __user *optval, int __user *optlen)
4192 struct sctp_bind_addr *bp;
4193 struct sctp_association *asoc;
4194 struct list_head *pos;
4196 struct sctp_getaddrs_old getaddrs;
4197 struct sctp_sockaddr_entry *addr;
4199 union sctp_addr temp;
4200 struct sctp_sock *sp = sctp_sk(sk);
4202 rwlock_t *addr_lock;
4206 int bytes_copied = 0;
4208 if (len < sizeof(struct sctp_getaddrs_old))
4211 len = sizeof(struct sctp_getaddrs_old);
4212 if (copy_from_user(&getaddrs, optval, len))
4215 if (getaddrs.addr_num <= 0) return -EINVAL;
4217 * For UDP-style sockets, id specifies the association to query.
4218 * If the id field is set to the value '0' then the locally bound
4219 * addresses are returned without regard to any particular
4222 if (0 == getaddrs.assoc_id) {
4223 bp = &sctp_sk(sk)->ep->base.bind_addr;
4224 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4226 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4229 bp = &asoc->base.bind_addr;
4230 addr_lock = &asoc->base.addr_lock;
4233 to = getaddrs.addrs;
4235 /* Allocate space for a local instance of packed array to hold all
4236 * the data. We store addresses here first and then put write them
4237 * to the user in one shot.
4239 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4244 sctp_read_lock(addr_lock);
4246 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4247 * addresses from the global local address list.
4249 if (sctp_list_single_entry(&bp->address_list)) {
4250 addr = list_entry(bp->address_list.next,
4251 struct sctp_sockaddr_entry, list);
4252 if (sctp_is_any(&addr->a)) {
4253 cnt = sctp_copy_laddrs_old(sk, bp->port,
4255 addrs, &bytes_copied);
4261 list_for_each(pos, &bp->address_list) {
4262 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4263 memcpy(&temp, &addr->a, sizeof(temp));
4264 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4265 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4266 memcpy(buf, &temp, addrlen);
4268 bytes_copied += addrlen;
4270 if (cnt >= getaddrs.addr_num) break;
4274 sctp_read_unlock(addr_lock);
4276 /* copy the entire address list into the user provided space */
4277 if (copy_to_user(to, addrs, bytes_copied)) {
4282 /* copy the leading structure back to user */
4283 getaddrs.addr_num = cnt;
4284 if (copy_to_user(optval, &getaddrs, len))
4292 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4293 char __user *optval, int __user *optlen)
4295 struct sctp_bind_addr *bp;
4296 struct sctp_association *asoc;
4297 struct list_head *pos;
4299 struct sctp_getaddrs getaddrs;
4300 struct sctp_sockaddr_entry *addr;
4302 union sctp_addr temp;
4303 struct sctp_sock *sp = sctp_sk(sk);
4305 rwlock_t *addr_lock;
4308 int bytes_copied = 0;
4312 if (len < sizeof(struct sctp_getaddrs))
4315 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4319 * For UDP-style sockets, id specifies the association to query.
4320 * If the id field is set to the value '0' then the locally bound
4321 * addresses are returned without regard to any particular
4324 if (0 == getaddrs.assoc_id) {
4325 bp = &sctp_sk(sk)->ep->base.bind_addr;
4326 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4328 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4331 bp = &asoc->base.bind_addr;
4332 addr_lock = &asoc->base.addr_lock;
4335 to = optval + offsetof(struct sctp_getaddrs,addrs);
4336 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4338 addrs = kmalloc(space_left, GFP_KERNEL);
4342 sctp_read_lock(addr_lock);
4344 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4345 * addresses from the global local address list.
4347 if (sctp_list_single_entry(&bp->address_list)) {
4348 addr = list_entry(bp->address_list.next,
4349 struct sctp_sockaddr_entry, list);
4350 if (sctp_is_any(&addr->a)) {
4351 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4352 space_left, &bytes_copied);
4362 list_for_each(pos, &bp->address_list) {
4363 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4364 memcpy(&temp, &addr->a, sizeof(temp));
4365 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4366 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4367 if (space_left < addrlen) {
4368 err = -ENOMEM; /*fixme: right error?*/
4371 memcpy(buf, &temp, addrlen);
4373 bytes_copied += addrlen;
4375 space_left -= addrlen;
4379 sctp_read_unlock(addr_lock);
4381 if (copy_to_user(to, addrs, bytes_copied)) {
4385 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4389 if (put_user(bytes_copied, optlen))
4395 sctp_read_unlock(addr_lock);
4402 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4404 * Requests that the local SCTP stack use the enclosed peer address as
4405 * the association primary. The enclosed address must be one of the
4406 * association peer's addresses.
4408 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4409 char __user *optval, int __user *optlen)
4411 struct sctp_prim prim;
4412 struct sctp_association *asoc;
4413 struct sctp_sock *sp = sctp_sk(sk);
4415 if (len < sizeof(struct sctp_prim))
4418 len = sizeof(struct sctp_prim);
4420 if (copy_from_user(&prim, optval, len))
4423 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4427 if (!asoc->peer.primary_path)
4430 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4431 asoc->peer.primary_path->af_specific->sockaddr_len);
4433 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4434 (union sctp_addr *)&prim.ssp_addr);
4436 if (put_user(len, optlen))
4438 if (copy_to_user(optval, &prim, len))
4445 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4447 * Requests that the local endpoint set the specified Adaptation Layer
4448 * Indication parameter for all future INIT and INIT-ACK exchanges.
4450 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4451 char __user *optval, int __user *optlen)
4453 struct sctp_setadaptation adaptation;
4455 if (len < sizeof(struct sctp_setadaptation))
4458 len = sizeof(struct sctp_setadaptation);
4460 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4462 if (put_user(len, optlen))
4464 if (copy_to_user(optval, &adaptation, len))
4472 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4474 * Applications that wish to use the sendto() system call may wish to
4475 * specify a default set of parameters that would normally be supplied
4476 * through the inclusion of ancillary data. This socket option allows
4477 * such an application to set the default sctp_sndrcvinfo structure.
4480 * The application that wishes to use this socket option simply passes
4481 * in to this call the sctp_sndrcvinfo structure defined in Section
4482 * 5.2.2) The input parameters accepted by this call include
4483 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4484 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4485 * to this call if the caller is using the UDP model.
4487 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4489 static int sctp_getsockopt_default_send_param(struct sock *sk,
4490 int len, char __user *optval,
4493 struct sctp_sndrcvinfo info;
4494 struct sctp_association *asoc;
4495 struct sctp_sock *sp = sctp_sk(sk);
4497 if (len < sizeof(struct sctp_sndrcvinfo))
4500 len = sizeof(struct sctp_sndrcvinfo);
4502 if (copy_from_user(&info, optval, len))
4505 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4506 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4510 info.sinfo_stream = asoc->default_stream;
4511 info.sinfo_flags = asoc->default_flags;
4512 info.sinfo_ppid = asoc->default_ppid;
4513 info.sinfo_context = asoc->default_context;
4514 info.sinfo_timetolive = asoc->default_timetolive;
4516 info.sinfo_stream = sp->default_stream;
4517 info.sinfo_flags = sp->default_flags;
4518 info.sinfo_ppid = sp->default_ppid;
4519 info.sinfo_context = sp->default_context;
4520 info.sinfo_timetolive = sp->default_timetolive;
4523 if (put_user(len, optlen))
4525 if (copy_to_user(optval, &info, len))
4533 * 7.1.5 SCTP_NODELAY
4535 * Turn on/off any Nagle-like algorithm. This means that packets are
4536 * generally sent as soon as possible and no unnecessary delays are
4537 * introduced, at the cost of more packets in the network. Expects an
4538 * integer boolean flag.
4541 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4542 char __user *optval, int __user *optlen)
4546 if (len < sizeof(int))
4550 val = (sctp_sk(sk)->nodelay == 1);
4551 if (put_user(len, optlen))
4553 if (copy_to_user(optval, &val, len))
4560 * 7.1.1 SCTP_RTOINFO
4562 * The protocol parameters used to initialize and bound retransmission
4563 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4564 * and modify these parameters.
4565 * All parameters are time values, in milliseconds. A value of 0, when
4566 * modifying the parameters, indicates that the current value should not
4570 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4571 char __user *optval,
4572 int __user *optlen) {
4573 struct sctp_rtoinfo rtoinfo;
4574 struct sctp_association *asoc;
4576 if (len < sizeof (struct sctp_rtoinfo))
4579 len = sizeof(struct sctp_rtoinfo);
4581 if (copy_from_user(&rtoinfo, optval, len))
4584 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4586 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4589 /* Values corresponding to the specific association. */
4591 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4592 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4593 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4595 /* Values corresponding to the endpoint. */
4596 struct sctp_sock *sp = sctp_sk(sk);
4598 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4599 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4600 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4603 if (put_user(len, optlen))
4606 if (copy_to_user(optval, &rtoinfo, len))
4614 * 7.1.2 SCTP_ASSOCINFO
4616 * This option is used to tune the maximum retransmission attempts
4617 * of the association.
4618 * Returns an error if the new association retransmission value is
4619 * greater than the sum of the retransmission value of the peer.
4620 * See [SCTP] for more information.
4623 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4624 char __user *optval,
4628 struct sctp_assocparams assocparams;
4629 struct sctp_association *asoc;
4630 struct list_head *pos;
4633 if (len < sizeof (struct sctp_assocparams))
4636 len = sizeof(struct sctp_assocparams);
4638 if (copy_from_user(&assocparams, optval, len))
4641 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4643 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4646 /* Values correspoinding to the specific association */
4648 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4649 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4650 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4651 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4653 (asoc->cookie_life.tv_usec
4656 list_for_each(pos, &asoc->peer.transport_addr_list) {
4660 assocparams.sasoc_number_peer_destinations = cnt;
4662 /* Values corresponding to the endpoint */
4663 struct sctp_sock *sp = sctp_sk(sk);
4665 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4666 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4667 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4668 assocparams.sasoc_cookie_life =
4669 sp->assocparams.sasoc_cookie_life;
4670 assocparams.sasoc_number_peer_destinations =
4672 sasoc_number_peer_destinations;
4675 if (put_user(len, optlen))
4678 if (copy_to_user(optval, &assocparams, len))
4685 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4687 * This socket option is a boolean flag which turns on or off mapped V4
4688 * addresses. If this option is turned on and the socket is type
4689 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4690 * If this option is turned off, then no mapping will be done of V4
4691 * addresses and a user will receive both PF_INET6 and PF_INET type
4692 * addresses on the socket.
4694 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4695 char __user *optval, int __user *optlen)
4698 struct sctp_sock *sp = sctp_sk(sk);
4700 if (len < sizeof(int))
4705 if (put_user(len, optlen))
4707 if (copy_to_user(optval, &val, len))
4714 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4715 * (chapter and verse is quoted at sctp_setsockopt_context())
4717 static int sctp_getsockopt_context(struct sock *sk, int len,
4718 char __user *optval, int __user *optlen)
4720 struct sctp_assoc_value params;
4721 struct sctp_sock *sp;
4722 struct sctp_association *asoc;
4724 if (len < sizeof(struct sctp_assoc_value))
4727 len = sizeof(struct sctp_assoc_value);
4729 if (copy_from_user(¶ms, optval, len))
4734 if (params.assoc_id != 0) {
4735 asoc = sctp_id2assoc(sk, params.assoc_id);
4738 params.assoc_value = asoc->default_rcv_context;
4740 params.assoc_value = sp->default_rcv_context;
4743 if (put_user(len, optlen))
4745 if (copy_to_user(optval, ¶ms, len))
4752 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4754 * This socket option specifies the maximum size to put in any outgoing
4755 * SCTP chunk. If a message is larger than this size it will be
4756 * fragmented by SCTP into the specified size. Note that the underlying
4757 * SCTP implementation may fragment into smaller sized chunks when the
4758 * PMTU of the underlying association is smaller than the value set by
4761 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4762 char __user *optval, int __user *optlen)
4766 if (len < sizeof(int))
4771 val = sctp_sk(sk)->user_frag;
4772 if (put_user(len, optlen))
4774 if (copy_to_user(optval, &val, len))
4781 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4782 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4784 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4785 char __user *optval, int __user *optlen)
4789 if (len < sizeof(int))
4794 val = sctp_sk(sk)->frag_interleave;
4795 if (put_user(len, optlen))
4797 if (copy_to_user(optval, &val, len))
4804 * 7.1.25. Set or Get the sctp partial delivery point
4805 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4807 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4808 char __user *optval,
4813 if (len < sizeof(u32))
4818 val = sctp_sk(sk)->pd_point;
4819 if (put_user(len, optlen))
4821 if (copy_to_user(optval, &val, len))
4828 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
4829 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
4831 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
4832 char __user *optval,
4837 if (len < sizeof(int))
4842 val = sctp_sk(sk)->max_burst;
4843 if (put_user(len, optlen))
4845 if (copy_to_user(optval, &val, len))
4851 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4852 char __user *optval, int __user *optlen)
4857 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4860 /* I can hardly begin to describe how wrong this is. This is
4861 * so broken as to be worse than useless. The API draft
4862 * REALLY is NOT helpful here... I am not convinced that the
4863 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4864 * are at all well-founded.
4866 if (level != SOL_SCTP) {
4867 struct sctp_af *af = sctp_sk(sk)->pf->af;
4869 retval = af->getsockopt(sk, level, optname, optval, optlen);
4873 if (get_user(len, optlen))
4880 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4882 case SCTP_DISABLE_FRAGMENTS:
4883 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4887 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4889 case SCTP_AUTOCLOSE:
4890 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4892 case SCTP_SOCKOPT_PEELOFF:
4893 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4895 case SCTP_PEER_ADDR_PARAMS:
4896 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4899 case SCTP_DELAYED_ACK_TIME:
4900 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4904 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4906 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4907 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4910 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4911 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4914 case SCTP_GET_PEER_ADDRS_OLD:
4915 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4918 case SCTP_GET_LOCAL_ADDRS_OLD:
4919 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4922 case SCTP_GET_PEER_ADDRS:
4923 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4926 case SCTP_GET_LOCAL_ADDRS:
4927 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4930 case SCTP_DEFAULT_SEND_PARAM:
4931 retval = sctp_getsockopt_default_send_param(sk, len,
4934 case SCTP_PRIMARY_ADDR:
4935 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4938 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4941 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4943 case SCTP_ASSOCINFO:
4944 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4946 case SCTP_I_WANT_MAPPED_V4_ADDR:
4947 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4950 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4952 case SCTP_GET_PEER_ADDR_INFO:
4953 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4956 case SCTP_ADAPTATION_LAYER:
4957 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4961 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4963 case SCTP_FRAGMENT_INTERLEAVE:
4964 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4967 case SCTP_PARTIAL_DELIVERY_POINT:
4968 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
4971 case SCTP_MAX_BURST:
4972 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
4975 retval = -ENOPROTOOPT;
4979 sctp_release_sock(sk);
4983 static void sctp_hash(struct sock *sk)
4988 static void sctp_unhash(struct sock *sk)
4993 /* Check if port is acceptable. Possibly find first available port.
4995 * The port hash table (contained in the 'global' SCTP protocol storage
4996 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4997 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4998 * list (the list number is the port number hashed out, so as you
4999 * would expect from a hash function, all the ports in a given list have
5000 * such a number that hashes out to the same list number; you were
5001 * expecting that, right?); so each list has a set of ports, with a
5002 * link to the socket (struct sock) that uses it, the port number and
5003 * a fastreuse flag (FIXME: NPI ipg).
5005 static struct sctp_bind_bucket *sctp_bucket_create(
5006 struct sctp_bind_hashbucket *head, unsigned short snum);
5008 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5010 struct sctp_bind_hashbucket *head; /* hash list */
5011 struct sctp_bind_bucket *pp; /* hash list port iterator */
5012 unsigned short snum;
5015 snum = ntohs(addr->v4.sin_port);
5017 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5018 sctp_local_bh_disable();
5021 /* Search for an available port.
5023 * 'sctp_port_rover' was the last port assigned, so
5024 * we start to search from 'sctp_port_rover +
5025 * 1'. What we do is first check if port 'rover' is
5026 * already in the hash table; if not, we use that; if
5027 * it is, we try next.
5029 int low = sysctl_local_port_range[0];
5030 int high = sysctl_local_port_range[1];
5031 int remaining = (high - low) + 1;
5035 sctp_spin_lock(&sctp_port_alloc_lock);
5036 rover = sctp_port_rover;
5039 if ((rover < low) || (rover > high))
5041 index = sctp_phashfn(rover);
5042 head = &sctp_port_hashtable[index];
5043 sctp_spin_lock(&head->lock);
5044 for (pp = head->chain; pp; pp = pp->next)
5045 if (pp->port == rover)
5049 sctp_spin_unlock(&head->lock);
5050 } while (--remaining > 0);
5051 sctp_port_rover = rover;
5052 sctp_spin_unlock(&sctp_port_alloc_lock);
5054 /* Exhausted local port range during search? */
5059 /* OK, here is the one we will use. HEAD (the port
5060 * hash table list entry) is non-NULL and we hold it's
5065 /* We are given an specific port number; we verify
5066 * that it is not being used. If it is used, we will
5067 * exahust the search in the hash list corresponding
5068 * to the port number (snum) - we detect that with the
5069 * port iterator, pp being NULL.
5071 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5072 sctp_spin_lock(&head->lock);
5073 for (pp = head->chain; pp; pp = pp->next) {
5074 if (pp->port == snum)
5081 if (!hlist_empty(&pp->owner)) {
5082 /* We had a port hash table hit - there is an
5083 * available port (pp != NULL) and it is being
5084 * used by other socket (pp->owner not empty); that other
5085 * socket is going to be sk2.
5087 int reuse = sk->sk_reuse;
5089 struct hlist_node *node;
5091 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5092 if (pp->fastreuse && sk->sk_reuse &&
5093 sk->sk_state != SCTP_SS_LISTENING)
5096 /* Run through the list of sockets bound to the port
5097 * (pp->port) [via the pointers bind_next and
5098 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5099 * we get the endpoint they describe and run through
5100 * the endpoint's list of IP (v4 or v6) addresses,
5101 * comparing each of the addresses with the address of
5102 * the socket sk. If we find a match, then that means
5103 * that this port/socket (sk) combination are already
5106 sk_for_each_bound(sk2, node, &pp->owner) {
5107 struct sctp_endpoint *ep2;
5108 ep2 = sctp_sk(sk2)->ep;
5110 if (reuse && sk2->sk_reuse &&
5111 sk2->sk_state != SCTP_SS_LISTENING)
5114 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5120 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5123 /* If there was a hash table miss, create a new port. */
5125 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5128 /* In either case (hit or miss), make sure fastreuse is 1 only
5129 * if sk->sk_reuse is too (that is, if the caller requested
5130 * SO_REUSEADDR on this socket -sk-).
5132 if (hlist_empty(&pp->owner)) {
5133 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5137 } else if (pp->fastreuse &&
5138 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5141 /* We are set, so fill up all the data in the hash table
5142 * entry, tie the socket list information with the rest of the
5143 * sockets FIXME: Blurry, NPI (ipg).
5146 if (!sctp_sk(sk)->bind_hash) {
5147 inet_sk(sk)->num = snum;
5148 sk_add_bind_node(sk, &pp->owner);
5149 sctp_sk(sk)->bind_hash = pp;
5154 sctp_spin_unlock(&head->lock);
5157 sctp_local_bh_enable();
5161 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5162 * port is requested.
5164 static int sctp_get_port(struct sock *sk, unsigned short snum)
5167 union sctp_addr addr;
5168 struct sctp_af *af = sctp_sk(sk)->pf->af;
5170 /* Set up a dummy address struct from the sk. */
5171 af->from_sk(&addr, sk);
5172 addr.v4.sin_port = htons(snum);
5174 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5175 ret = sctp_get_port_local(sk, &addr);
5177 return (ret ? 1 : 0);
5181 * 3.1.3 listen() - UDP Style Syntax
5183 * By default, new associations are not accepted for UDP style sockets.
5184 * An application uses listen() to mark a socket as being able to
5185 * accept new associations.
5187 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5189 struct sctp_sock *sp = sctp_sk(sk);
5190 struct sctp_endpoint *ep = sp->ep;
5192 /* Only UDP style sockets that are not peeled off are allowed to
5195 if (!sctp_style(sk, UDP))
5198 /* If backlog is zero, disable listening. */
5200 if (sctp_sstate(sk, CLOSED))
5203 sctp_unhash_endpoint(ep);
5204 sk->sk_state = SCTP_SS_CLOSED;
5207 /* Return if we are already listening. */
5208 if (sctp_sstate(sk, LISTENING))
5212 * If a bind() or sctp_bindx() is not called prior to a listen()
5213 * call that allows new associations to be accepted, the system
5214 * picks an ephemeral port and will choose an address set equivalent
5215 * to binding with a wildcard address.
5217 * This is not currently spelled out in the SCTP sockets
5218 * extensions draft, but follows the practice as seen in TCP
5221 * Additionally, turn off fastreuse flag since we are not listening
5223 sk->sk_state = SCTP_SS_LISTENING;
5224 if (!ep->base.bind_addr.port) {
5225 if (sctp_autobind(sk))
5228 sctp_sk(sk)->bind_hash->fastreuse = 0;
5230 sctp_hash_endpoint(ep);
5235 * 4.1.3 listen() - TCP Style Syntax
5237 * Applications uses listen() to ready the SCTP endpoint for accepting
5238 * inbound associations.
5240 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5242 struct sctp_sock *sp = sctp_sk(sk);
5243 struct sctp_endpoint *ep = sp->ep;
5245 /* If backlog is zero, disable listening. */
5247 if (sctp_sstate(sk, CLOSED))
5250 sctp_unhash_endpoint(ep);
5251 sk->sk_state = SCTP_SS_CLOSED;
5254 if (sctp_sstate(sk, LISTENING))
5258 * If a bind() or sctp_bindx() is not called prior to a listen()
5259 * call that allows new associations to be accepted, the system
5260 * picks an ephemeral port and will choose an address set equivalent
5261 * to binding with a wildcard address.
5263 * This is not currently spelled out in the SCTP sockets
5264 * extensions draft, but follows the practice as seen in TCP
5267 sk->sk_state = SCTP_SS_LISTENING;
5268 if (!ep->base.bind_addr.port) {
5269 if (sctp_autobind(sk))
5272 sctp_sk(sk)->bind_hash->fastreuse = 0;
5274 sk->sk_max_ack_backlog = backlog;
5275 sctp_hash_endpoint(ep);
5280 * Move a socket to LISTENING state.
5282 int sctp_inet_listen(struct socket *sock, int backlog)
5284 struct sock *sk = sock->sk;
5285 struct crypto_hash *tfm = NULL;
5288 if (unlikely(backlog < 0))
5293 if (sock->state != SS_UNCONNECTED)
5296 /* Allocate HMAC for generating cookie. */
5297 if (sctp_hmac_alg) {
5298 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5300 if (net_ratelimit()) {
5302 "SCTP: failed to load transform for %s: %ld\n",
5303 sctp_hmac_alg, PTR_ERR(tfm));
5310 switch (sock->type) {
5311 case SOCK_SEQPACKET:
5312 err = sctp_seqpacket_listen(sk, backlog);
5315 err = sctp_stream_listen(sk, backlog);
5324 /* Store away the transform reference. */
5325 sctp_sk(sk)->hmac = tfm;
5327 sctp_release_sock(sk);
5330 crypto_free_hash(tfm);
5335 * This function is done by modeling the current datagram_poll() and the
5336 * tcp_poll(). Note that, based on these implementations, we don't
5337 * lock the socket in this function, even though it seems that,
5338 * ideally, locking or some other mechanisms can be used to ensure
5339 * the integrity of the counters (sndbuf and wmem_alloc) used
5340 * in this place. We assume that we don't need locks either until proven
5343 * Another thing to note is that we include the Async I/O support
5344 * here, again, by modeling the current TCP/UDP code. We don't have
5345 * a good way to test with it yet.
5347 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5349 struct sock *sk = sock->sk;
5350 struct sctp_sock *sp = sctp_sk(sk);
5353 poll_wait(file, sk->sk_sleep, wait);
5355 /* A TCP-style listening socket becomes readable when the accept queue
5358 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5359 return (!list_empty(&sp->ep->asocs)) ?
5360 (POLLIN | POLLRDNORM) : 0;
5364 /* Is there any exceptional events? */
5365 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5367 if (sk->sk_shutdown & RCV_SHUTDOWN)
5369 if (sk->sk_shutdown == SHUTDOWN_MASK)
5372 /* Is it readable? Reconsider this code with TCP-style support. */
5373 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5374 (sk->sk_shutdown & RCV_SHUTDOWN))
5375 mask |= POLLIN | POLLRDNORM;
5377 /* The association is either gone or not ready. */
5378 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5381 /* Is it writable? */
5382 if (sctp_writeable(sk)) {
5383 mask |= POLLOUT | POLLWRNORM;
5385 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5387 * Since the socket is not locked, the buffer
5388 * might be made available after the writeable check and
5389 * before the bit is set. This could cause a lost I/O
5390 * signal. tcp_poll() has a race breaker for this race
5391 * condition. Based on their implementation, we put
5392 * in the following code to cover it as well.
5394 if (sctp_writeable(sk))
5395 mask |= POLLOUT | POLLWRNORM;
5400 /********************************************************************
5401 * 2nd Level Abstractions
5402 ********************************************************************/
5404 static struct sctp_bind_bucket *sctp_bucket_create(
5405 struct sctp_bind_hashbucket *head, unsigned short snum)
5407 struct sctp_bind_bucket *pp;
5409 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5410 SCTP_DBG_OBJCNT_INC(bind_bucket);
5414 INIT_HLIST_HEAD(&pp->owner);
5415 if ((pp->next = head->chain) != NULL)
5416 pp->next->pprev = &pp->next;
5418 pp->pprev = &head->chain;
5423 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5424 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5426 if (pp && hlist_empty(&pp->owner)) {
5428 pp->next->pprev = pp->pprev;
5429 *(pp->pprev) = pp->next;
5430 kmem_cache_free(sctp_bucket_cachep, pp);
5431 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5435 /* Release this socket's reference to a local port. */
5436 static inline void __sctp_put_port(struct sock *sk)
5438 struct sctp_bind_hashbucket *head =
5439 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5440 struct sctp_bind_bucket *pp;
5442 sctp_spin_lock(&head->lock);
5443 pp = sctp_sk(sk)->bind_hash;
5444 __sk_del_bind_node(sk);
5445 sctp_sk(sk)->bind_hash = NULL;
5446 inet_sk(sk)->num = 0;
5447 sctp_bucket_destroy(pp);
5448 sctp_spin_unlock(&head->lock);
5451 void sctp_put_port(struct sock *sk)
5453 sctp_local_bh_disable();
5454 __sctp_put_port(sk);
5455 sctp_local_bh_enable();
5459 * The system picks an ephemeral port and choose an address set equivalent
5460 * to binding with a wildcard address.
5461 * One of those addresses will be the primary address for the association.
5462 * This automatically enables the multihoming capability of SCTP.
5464 static int sctp_autobind(struct sock *sk)
5466 union sctp_addr autoaddr;
5470 /* Initialize a local sockaddr structure to INADDR_ANY. */
5471 af = sctp_sk(sk)->pf->af;
5473 port = htons(inet_sk(sk)->num);
5474 af->inaddr_any(&autoaddr, port);
5476 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5479 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5482 * 4.2 The cmsghdr Structure *
5484 * When ancillary data is sent or received, any number of ancillary data
5485 * objects can be specified by the msg_control and msg_controllen members of
5486 * the msghdr structure, because each object is preceded by
5487 * a cmsghdr structure defining the object's length (the cmsg_len member).
5488 * Historically Berkeley-derived implementations have passed only one object
5489 * at a time, but this API allows multiple objects to be
5490 * passed in a single call to sendmsg() or recvmsg(). The following example
5491 * shows two ancillary data objects in a control buffer.
5493 * |<--------------------------- msg_controllen -------------------------->|
5496 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5498 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5501 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5503 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5506 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5507 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5509 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5511 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5518 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5519 sctp_cmsgs_t *cmsgs)
5521 struct cmsghdr *cmsg;
5523 for (cmsg = CMSG_FIRSTHDR(msg);
5525 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5526 if (!CMSG_OK(msg, cmsg))
5529 /* Should we parse this header or ignore? */
5530 if (cmsg->cmsg_level != IPPROTO_SCTP)
5533 /* Strictly check lengths following example in SCM code. */
5534 switch (cmsg->cmsg_type) {
5536 /* SCTP Socket API Extension
5537 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5539 * This cmsghdr structure provides information for
5540 * initializing new SCTP associations with sendmsg().
5541 * The SCTP_INITMSG socket option uses this same data
5542 * structure. This structure is not used for
5545 * cmsg_level cmsg_type cmsg_data[]
5546 * ------------ ------------ ----------------------
5547 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5549 if (cmsg->cmsg_len !=
5550 CMSG_LEN(sizeof(struct sctp_initmsg)))
5552 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5556 /* SCTP Socket API Extension
5557 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5559 * This cmsghdr structure specifies SCTP options for
5560 * sendmsg() and describes SCTP header information
5561 * about a received message through recvmsg().
5563 * cmsg_level cmsg_type cmsg_data[]
5564 * ------------ ------------ ----------------------
5565 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5567 if (cmsg->cmsg_len !=
5568 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5572 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5574 /* Minimally, validate the sinfo_flags. */
5575 if (cmsgs->info->sinfo_flags &
5576 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5577 SCTP_ABORT | SCTP_EOF))
5589 * Wait for a packet..
5590 * Note: This function is the same function as in core/datagram.c
5591 * with a few modifications to make lksctp work.
5593 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5598 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5600 /* Socket errors? */
5601 error = sock_error(sk);
5605 if (!skb_queue_empty(&sk->sk_receive_queue))
5608 /* Socket shut down? */
5609 if (sk->sk_shutdown & RCV_SHUTDOWN)
5612 /* Sequenced packets can come disconnected. If so we report the
5617 /* Is there a good reason to think that we may receive some data? */
5618 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5621 /* Handle signals. */
5622 if (signal_pending(current))
5625 /* Let another process have a go. Since we are going to sleep
5626 * anyway. Note: This may cause odd behaviors if the message
5627 * does not fit in the user's buffer, but this seems to be the
5628 * only way to honor MSG_DONTWAIT realistically.
5630 sctp_release_sock(sk);
5631 *timeo_p = schedule_timeout(*timeo_p);
5635 finish_wait(sk->sk_sleep, &wait);
5639 error = sock_intr_errno(*timeo_p);
5642 finish_wait(sk->sk_sleep, &wait);
5647 /* Receive a datagram.
5648 * Note: This is pretty much the same routine as in core/datagram.c
5649 * with a few changes to make lksctp work.
5651 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5652 int noblock, int *err)
5655 struct sk_buff *skb;
5658 timeo = sock_rcvtimeo(sk, noblock);
5660 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5661 timeo, MAX_SCHEDULE_TIMEOUT);
5664 /* Again only user level code calls this function,
5665 * so nothing interrupt level
5666 * will suddenly eat the receive_queue.
5668 * Look at current nfs client by the way...
5669 * However, this function was corrent in any case. 8)
5671 if (flags & MSG_PEEK) {
5672 spin_lock_bh(&sk->sk_receive_queue.lock);
5673 skb = skb_peek(&sk->sk_receive_queue);
5675 atomic_inc(&skb->users);
5676 spin_unlock_bh(&sk->sk_receive_queue.lock);
5678 skb = skb_dequeue(&sk->sk_receive_queue);
5684 /* Caller is allowed not to check sk->sk_err before calling. */
5685 error = sock_error(sk);
5689 if (sk->sk_shutdown & RCV_SHUTDOWN)
5692 /* User doesn't want to wait. */
5696 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5705 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5706 static void __sctp_write_space(struct sctp_association *asoc)
5708 struct sock *sk = asoc->base.sk;
5709 struct socket *sock = sk->sk_socket;
5711 if ((sctp_wspace(asoc) > 0) && sock) {
5712 if (waitqueue_active(&asoc->wait))
5713 wake_up_interruptible(&asoc->wait);
5715 if (sctp_writeable(sk)) {
5716 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5717 wake_up_interruptible(sk->sk_sleep);
5719 /* Note that we try to include the Async I/O support
5720 * here by modeling from the current TCP/UDP code.
5721 * We have not tested with it yet.
5723 if (sock->fasync_list &&
5724 !(sk->sk_shutdown & SEND_SHUTDOWN))
5725 sock_wake_async(sock, 2, POLL_OUT);
5730 /* Do accounting for the sndbuf space.
5731 * Decrement the used sndbuf space of the corresponding association by the
5732 * data size which was just transmitted(freed).
5734 static void sctp_wfree(struct sk_buff *skb)
5736 struct sctp_association *asoc;
5737 struct sctp_chunk *chunk;
5740 /* Get the saved chunk pointer. */
5741 chunk = *((struct sctp_chunk **)(skb->cb));
5744 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5745 sizeof(struct sk_buff) +
5746 sizeof(struct sctp_chunk);
5748 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5751 __sctp_write_space(asoc);
5753 sctp_association_put(asoc);
5756 /* Do accounting for the receive space on the socket.
5757 * Accounting for the association is done in ulpevent.c
5758 * We set this as a destructor for the cloned data skbs so that
5759 * accounting is done at the correct time.
5761 void sctp_sock_rfree(struct sk_buff *skb)
5763 struct sock *sk = skb->sk;
5764 struct sctp_ulpevent *event = sctp_skb2event(skb);
5766 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5770 /* Helper function to wait for space in the sndbuf. */
5771 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5774 struct sock *sk = asoc->base.sk;
5776 long current_timeo = *timeo_p;
5779 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5780 asoc, (long)(*timeo_p), msg_len);
5782 /* Increment the association's refcnt. */
5783 sctp_association_hold(asoc);
5785 /* Wait on the association specific sndbuf space. */
5787 prepare_to_wait_exclusive(&asoc->wait, &wait,
5788 TASK_INTERRUPTIBLE);
5791 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5794 if (signal_pending(current))
5795 goto do_interrupted;
5796 if (msg_len <= sctp_wspace(asoc))
5799 /* Let another process have a go. Since we are going
5802 sctp_release_sock(sk);
5803 current_timeo = schedule_timeout(current_timeo);
5804 BUG_ON(sk != asoc->base.sk);
5807 *timeo_p = current_timeo;
5811 finish_wait(&asoc->wait, &wait);
5813 /* Release the association's refcnt. */
5814 sctp_association_put(asoc);
5823 err = sock_intr_errno(*timeo_p);
5831 /* If socket sndbuf has changed, wake up all per association waiters. */
5832 void sctp_write_space(struct sock *sk)
5834 struct sctp_association *asoc;
5835 struct list_head *pos;
5837 /* Wake up the tasks in each wait queue. */
5838 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5839 asoc = list_entry(pos, struct sctp_association, asocs);
5840 __sctp_write_space(asoc);
5844 /* Is there any sndbuf space available on the socket?
5846 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5847 * associations on the same socket. For a UDP-style socket with
5848 * multiple associations, it is possible for it to be "unwriteable"
5849 * prematurely. I assume that this is acceptable because
5850 * a premature "unwriteable" is better than an accidental "writeable" which
5851 * would cause an unwanted block under certain circumstances. For the 1-1
5852 * UDP-style sockets or TCP-style sockets, this code should work.
5855 static int sctp_writeable(struct sock *sk)
5859 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5865 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5866 * returns immediately with EINPROGRESS.
5868 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5870 struct sock *sk = asoc->base.sk;
5872 long current_timeo = *timeo_p;
5875 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5878 /* Increment the association's refcnt. */
5879 sctp_association_hold(asoc);
5882 prepare_to_wait_exclusive(&asoc->wait, &wait,
5883 TASK_INTERRUPTIBLE);
5886 if (sk->sk_shutdown & RCV_SHUTDOWN)
5888 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5891 if (signal_pending(current))
5892 goto do_interrupted;
5894 if (sctp_state(asoc, ESTABLISHED))
5897 /* Let another process have a go. Since we are going
5900 sctp_release_sock(sk);
5901 current_timeo = schedule_timeout(current_timeo);
5904 *timeo_p = current_timeo;
5908 finish_wait(&asoc->wait, &wait);
5910 /* Release the association's refcnt. */
5911 sctp_association_put(asoc);
5916 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5919 err = -ECONNREFUSED;
5923 err = sock_intr_errno(*timeo_p);
5931 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5933 struct sctp_endpoint *ep;
5937 ep = sctp_sk(sk)->ep;
5941 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5942 TASK_INTERRUPTIBLE);
5944 if (list_empty(&ep->asocs)) {
5945 sctp_release_sock(sk);
5946 timeo = schedule_timeout(timeo);
5951 if (!sctp_sstate(sk, LISTENING))
5955 if (!list_empty(&ep->asocs))
5958 err = sock_intr_errno(timeo);
5959 if (signal_pending(current))
5967 finish_wait(sk->sk_sleep, &wait);
5972 static void sctp_wait_for_close(struct sock *sk, long timeout)
5977 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5978 if (list_empty(&sctp_sk(sk)->ep->asocs))
5980 sctp_release_sock(sk);
5981 timeout = schedule_timeout(timeout);
5983 } while (!signal_pending(current) && timeout);
5985 finish_wait(sk->sk_sleep, &wait);
5988 static void sctp_sock_rfree_frag(struct sk_buff *skb)
5990 struct sk_buff *frag;
5995 /* Don't forget the fragments. */
5996 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5997 sctp_sock_rfree_frag(frag);
6000 sctp_sock_rfree(skb);
6003 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6005 struct sk_buff *frag;
6010 /* Don't forget the fragments. */
6011 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6012 sctp_skb_set_owner_r_frag(frag, sk);
6015 sctp_skb_set_owner_r(skb, sk);
6018 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6019 * and its messages to the newsk.
6021 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6022 struct sctp_association *assoc,
6023 sctp_socket_type_t type)
6025 struct sctp_sock *oldsp = sctp_sk(oldsk);
6026 struct sctp_sock *newsp = sctp_sk(newsk);
6027 struct sctp_bind_bucket *pp; /* hash list port iterator */
6028 struct sctp_endpoint *newep = newsp->ep;
6029 struct sk_buff *skb, *tmp;
6030 struct sctp_ulpevent *event;
6033 /* Migrate socket buffer sizes and all the socket level options to the
6036 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6037 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6038 /* Brute force copy old sctp opt. */
6039 inet_sk_copy_descendant(newsk, oldsk);
6041 /* Restore the ep value that was overwritten with the above structure
6047 /* Hook this new socket in to the bind_hash list. */
6048 pp = sctp_sk(oldsk)->bind_hash;
6049 sk_add_bind_node(newsk, &pp->owner);
6050 sctp_sk(newsk)->bind_hash = pp;
6051 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6053 /* Copy the bind_addr list from the original endpoint to the new
6054 * endpoint so that we can handle restarts properly
6056 if (PF_INET6 == assoc->base.sk->sk_family)
6057 flags = SCTP_ADDR6_ALLOWED;
6058 if (assoc->peer.ipv4_address)
6059 flags |= SCTP_ADDR4_PEERSUPP;
6060 if (assoc->peer.ipv6_address)
6061 flags |= SCTP_ADDR6_PEERSUPP;
6062 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
6063 &oldsp->ep->base.bind_addr,
6064 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
6066 /* Move any messages in the old socket's receive queue that are for the
6067 * peeled off association to the new socket's receive queue.
6069 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6070 event = sctp_skb2event(skb);
6071 if (event->asoc == assoc) {
6072 sctp_sock_rfree_frag(skb);
6073 __skb_unlink(skb, &oldsk->sk_receive_queue);
6074 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6075 sctp_skb_set_owner_r_frag(skb, newsk);
6079 /* Clean up any messages pending delivery due to partial
6080 * delivery. Three cases:
6081 * 1) No partial deliver; no work.
6082 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6083 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6085 skb_queue_head_init(&newsp->pd_lobby);
6086 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6088 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6089 struct sk_buff_head *queue;
6091 /* Decide which queue to move pd_lobby skbs to. */
6092 if (assoc->ulpq.pd_mode) {
6093 queue = &newsp->pd_lobby;
6095 queue = &newsk->sk_receive_queue;
6097 /* Walk through the pd_lobby, looking for skbs that
6098 * need moved to the new socket.
6100 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6101 event = sctp_skb2event(skb);
6102 if (event->asoc == assoc) {
6103 sctp_sock_rfree_frag(skb);
6104 __skb_unlink(skb, &oldsp->pd_lobby);
6105 __skb_queue_tail(queue, skb);
6106 sctp_skb_set_owner_r_frag(skb, newsk);
6110 /* Clear up any skbs waiting for the partial
6111 * delivery to finish.
6113 if (assoc->ulpq.pd_mode)
6114 sctp_clear_pd(oldsk, NULL);
6118 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6119 sctp_sock_rfree_frag(skb);
6120 sctp_skb_set_owner_r_frag(skb, newsk);
6123 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6124 sctp_sock_rfree_frag(skb);
6125 sctp_skb_set_owner_r_frag(skb, newsk);
6128 /* Set the type of socket to indicate that it is peeled off from the
6129 * original UDP-style socket or created with the accept() call on a
6130 * TCP-style socket..
6134 /* Mark the new socket "in-use" by the user so that any packets
6135 * that may arrive on the association after we've moved it are
6136 * queued to the backlog. This prevents a potential race between
6137 * backlog processing on the old socket and new-packet processing
6138 * on the new socket.
6140 * The caller has just allocated newsk so we can guarantee that other
6141 * paths won't try to lock it and then oldsk.
6143 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6144 sctp_assoc_migrate(assoc, newsk);
6146 /* If the association on the newsk is already closed before accept()
6147 * is called, set RCV_SHUTDOWN flag.
6149 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6150 newsk->sk_shutdown |= RCV_SHUTDOWN;
6152 newsk->sk_state = SCTP_SS_ESTABLISHED;
6153 sctp_release_sock(newsk);
6156 /* This proto struct describes the ULP interface for SCTP. */
6157 struct proto sctp_prot = {
6159 .owner = THIS_MODULE,
6160 .close = sctp_close,
6161 .connect = sctp_connect,
6162 .disconnect = sctp_disconnect,
6163 .accept = sctp_accept,
6164 .ioctl = sctp_ioctl,
6165 .init = sctp_init_sock,
6166 .destroy = sctp_destroy_sock,
6167 .shutdown = sctp_shutdown,
6168 .setsockopt = sctp_setsockopt,
6169 .getsockopt = sctp_getsockopt,
6170 .sendmsg = sctp_sendmsg,
6171 .recvmsg = sctp_recvmsg,
6173 .backlog_rcv = sctp_backlog_rcv,
6175 .unhash = sctp_unhash,
6176 .get_port = sctp_get_port,
6177 .obj_size = sizeof(struct sctp_sock),
6180 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6181 struct proto sctpv6_prot = {
6183 .owner = THIS_MODULE,
6184 .close = sctp_close,
6185 .connect = sctp_connect,
6186 .disconnect = sctp_disconnect,
6187 .accept = sctp_accept,
6188 .ioctl = sctp_ioctl,
6189 .init = sctp_init_sock,
6190 .destroy = sctp_destroy_sock,
6191 .shutdown = sctp_shutdown,
6192 .setsockopt = sctp_setsockopt,
6193 .getsockopt = sctp_getsockopt,
6194 .sendmsg = sctp_sendmsg,
6195 .recvmsg = sctp_recvmsg,
6197 .backlog_rcv = sctp_backlog_rcv,
6199 .unhash = sctp_unhash,
6200 .get_port = sctp_get_port,
6201 .obj_size = sizeof(struct sctp6_sock),
6203 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */