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;
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 err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
1016 /* Make sure the destination port is correctly set
1019 if (asoc && asoc->peer.port && asoc->peer.port != port)
1022 memcpy(&to, sa_addr, af->sockaddr_len);
1024 /* Check if there already is a matching association on the
1025 * endpoint (other than the one created here).
1027 asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1028 if (asoc2 && asoc2 != asoc) {
1029 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1036 /* If we could not find a matching association on the endpoint,
1037 * make sure that there is no peeled-off association matching
1038 * the peer address even on another socket.
1040 if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1041 err = -EADDRNOTAVAIL;
1046 /* If a bind() or sctp_bindx() is not called prior to
1047 * an sctp_connectx() call, the system picks an
1048 * ephemeral port and will choose an address set
1049 * equivalent to binding with a wildcard address.
1051 if (!ep->base.bind_addr.port) {
1052 if (sctp_autobind(sk)) {
1058 * If an unprivileged user inherits a 1-many
1059 * style socket with open associations on a
1060 * privileged port, it MAY be permitted to
1061 * accept new associations, but it SHOULD NOT
1062 * be permitted to open new associations.
1064 if (ep->base.bind_addr.port < PROT_SOCK &&
1065 !capable(CAP_NET_BIND_SERVICE)) {
1071 scope = sctp_scope(sa_addr);
1072 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1079 /* Prime the peer's transport structures. */
1080 transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1088 addr_buf += af->sockaddr_len;
1089 walk_size += af->sockaddr_len;
1092 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1097 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1102 /* Initialize sk's dport and daddr for getpeername() */
1103 inet_sk(sk)->dport = htons(asoc->peer.port);
1104 af = sctp_get_af_specific(to.sa.sa_family);
1105 af->to_sk_daddr(&to, sk);
1108 /* in-kernel sockets don't generally have a file allocated to them
1109 * if all they do is call sock_create_kern().
1111 if (sk->sk_socket->file)
1112 f_flags = sk->sk_socket->file->f_flags;
1114 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1116 err = sctp_wait_for_connect(asoc, &timeo);
1118 /* Don't free association on exit. */
1123 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1124 " kaddrs: %p err: %d\n",
1127 sctp_association_free(asoc);
1131 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1134 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1136 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1137 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1138 * or IPv6 addresses.
1140 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1141 * Section 3.1.2 for this usage.
1143 * addrs is a pointer to an array of one or more socket addresses. Each
1144 * address is contained in its appropriate structure (i.e. struct
1145 * sockaddr_in or struct sockaddr_in6) the family of the address type
1146 * must be used to distengish the address length (note that this
1147 * representation is termed a "packed array" of addresses). The caller
1148 * specifies the number of addresses in the array with addrcnt.
1150 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1151 * -1, and sets errno to the appropriate error code.
1153 * For SCTP, the port given in each socket address must be the same, or
1154 * sctp_connectx() will fail, setting errno to EINVAL.
1156 * An application can use sctp_connectx to initiate an association with
1157 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1158 * allows a caller to specify multiple addresses at which a peer can be
1159 * reached. The way the SCTP stack uses the list of addresses to set up
1160 * the association is implementation dependant. This function only
1161 * specifies that the stack will try to make use of all the addresses in
1162 * the list when needed.
1164 * Note that the list of addresses passed in is only used for setting up
1165 * the association. It does not necessarily equal the set of addresses
1166 * the peer uses for the resulting association. If the caller wants to
1167 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1168 * retrieve them after the association has been set up.
1170 * Basically do nothing but copying the addresses from user to kernel
1171 * land and invoking either sctp_connectx(). This is used for tunneling
1172 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1174 * We don't use copy_from_user() for optimization: we first do the
1175 * sanity checks (buffer size -fast- and access check-healthy
1176 * pointer); if all of those succeed, then we can alloc the memory
1177 * (expensive operation) needed to copy the data to kernel. Then we do
1178 * the copying without checking the user space area
1179 * (__copy_from_user()).
1181 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1184 * sk The sk of the socket
1185 * addrs The pointer to the addresses in user land
1186 * addrssize Size of the addrs buffer
1188 * Returns 0 if ok, <0 errno code on error.
1190 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1191 struct sockaddr __user *addrs,
1195 struct sockaddr *kaddrs;
1197 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1198 __FUNCTION__, sk, addrs, addrs_size);
1200 if (unlikely(addrs_size <= 0))
1203 /* Check the user passed a healthy pointer. */
1204 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1207 /* Alloc space for the address array in kernel memory. */
1208 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1209 if (unlikely(!kaddrs))
1212 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1215 err = __sctp_connect(sk, kaddrs, addrs_size);
1222 /* API 3.1.4 close() - UDP Style Syntax
1223 * Applications use close() to perform graceful shutdown (as described in
1224 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1225 * by a UDP-style socket.
1229 * ret = close(int sd);
1231 * sd - the socket descriptor of the associations to be closed.
1233 * To gracefully shutdown a specific association represented by the
1234 * UDP-style socket, an application should use the sendmsg() call,
1235 * passing no user data, but including the appropriate flag in the
1236 * ancillary data (see Section xxxx).
1238 * If sd in the close() call is a branched-off socket representing only
1239 * one association, the shutdown is performed on that association only.
1241 * 4.1.6 close() - TCP Style Syntax
1243 * Applications use close() to gracefully close down an association.
1247 * int close(int sd);
1249 * sd - the socket descriptor of the association to be closed.
1251 * After an application calls close() on a socket descriptor, no further
1252 * socket operations will succeed on that descriptor.
1254 * API 7.1.4 SO_LINGER
1256 * An application using the TCP-style socket can use this option to
1257 * perform the SCTP ABORT primitive. The linger option structure is:
1260 * int l_onoff; // option on/off
1261 * int l_linger; // linger time
1264 * To enable the option, set l_onoff to 1. If the l_linger value is set
1265 * to 0, calling close() is the same as the ABORT primitive. If the
1266 * value is set to a negative value, the setsockopt() call will return
1267 * an error. If the value is set to a positive value linger_time, the
1268 * close() can be blocked for at most linger_time ms. If the graceful
1269 * shutdown phase does not finish during this period, close() will
1270 * return but the graceful shutdown phase continues in the system.
1272 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1274 struct sctp_endpoint *ep;
1275 struct sctp_association *asoc;
1276 struct list_head *pos, *temp;
1278 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1281 sk->sk_shutdown = SHUTDOWN_MASK;
1283 ep = sctp_sk(sk)->ep;
1285 /* Walk all associations on an endpoint. */
1286 list_for_each_safe(pos, temp, &ep->asocs) {
1287 asoc = list_entry(pos, struct sctp_association, asocs);
1289 if (sctp_style(sk, TCP)) {
1290 /* A closed association can still be in the list if
1291 * it belongs to a TCP-style listening socket that is
1292 * not yet accepted. If so, free it. If not, send an
1293 * ABORT or SHUTDOWN based on the linger options.
1295 if (sctp_state(asoc, CLOSED)) {
1296 sctp_unhash_established(asoc);
1297 sctp_association_free(asoc);
1302 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1303 struct sctp_chunk *chunk;
1305 chunk = sctp_make_abort_user(asoc, NULL, 0);
1307 sctp_primitive_ABORT(asoc, chunk);
1309 sctp_primitive_SHUTDOWN(asoc, NULL);
1312 /* Clean up any skbs sitting on the receive queue. */
1313 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1314 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1316 /* On a TCP-style socket, block for at most linger_time if set. */
1317 if (sctp_style(sk, TCP) && timeout)
1318 sctp_wait_for_close(sk, timeout);
1320 /* This will run the backlog queue. */
1321 sctp_release_sock(sk);
1323 /* Supposedly, no process has access to the socket, but
1324 * the net layers still may.
1326 sctp_local_bh_disable();
1327 sctp_bh_lock_sock(sk);
1329 /* Hold the sock, since sk_common_release() will put sock_put()
1330 * and we have just a little more cleanup.
1333 sk_common_release(sk);
1335 sctp_bh_unlock_sock(sk);
1336 sctp_local_bh_enable();
1340 SCTP_DBG_OBJCNT_DEC(sock);
1343 /* Handle EPIPE error. */
1344 static int sctp_error(struct sock *sk, int flags, int err)
1347 err = sock_error(sk) ? : -EPIPE;
1348 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1349 send_sig(SIGPIPE, current, 0);
1353 /* API 3.1.3 sendmsg() - UDP Style Syntax
1355 * An application uses sendmsg() and recvmsg() calls to transmit data to
1356 * and receive data from its peer.
1358 * ssize_t sendmsg(int socket, const struct msghdr *message,
1361 * socket - the socket descriptor of the endpoint.
1362 * message - pointer to the msghdr structure which contains a single
1363 * user message and possibly some ancillary data.
1365 * See Section 5 for complete description of the data
1368 * flags - flags sent or received with the user message, see Section
1369 * 5 for complete description of the flags.
1371 * Note: This function could use a rewrite especially when explicit
1372 * connect support comes in.
1374 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1376 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1378 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1379 struct msghdr *msg, size_t msg_len)
1381 struct sctp_sock *sp;
1382 struct sctp_endpoint *ep;
1383 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1384 struct sctp_transport *transport, *chunk_tp;
1385 struct sctp_chunk *chunk;
1387 struct sockaddr *msg_name = NULL;
1388 struct sctp_sndrcvinfo default_sinfo = { 0 };
1389 struct sctp_sndrcvinfo *sinfo;
1390 struct sctp_initmsg *sinit;
1391 sctp_assoc_t associd = 0;
1392 sctp_cmsgs_t cmsgs = { NULL };
1396 __u16 sinfo_flags = 0;
1397 struct sctp_datamsg *datamsg;
1398 struct list_head *pos;
1399 int msg_flags = msg->msg_flags;
1401 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1408 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1410 /* We cannot send a message over a TCP-style listening socket. */
1411 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1416 /* Parse out the SCTP CMSGs. */
1417 err = sctp_msghdr_parse(msg, &cmsgs);
1420 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1424 /* Fetch the destination address for this packet. This
1425 * address only selects the association--it is not necessarily
1426 * the address we will send to.
1427 * For a peeled-off socket, msg_name is ignored.
1429 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1430 int msg_namelen = msg->msg_namelen;
1432 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1437 if (msg_namelen > sizeof(to))
1438 msg_namelen = sizeof(to);
1439 memcpy(&to, msg->msg_name, msg_namelen);
1440 msg_name = msg->msg_name;
1446 /* Did the user specify SNDRCVINFO? */
1448 sinfo_flags = sinfo->sinfo_flags;
1449 associd = sinfo->sinfo_assoc_id;
1452 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1453 msg_len, sinfo_flags);
1455 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1456 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1461 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1462 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1463 * If SCTP_ABORT is set, the message length could be non zero with
1464 * the msg_iov set to the user abort reason.
1466 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1467 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1472 /* If SCTP_ADDR_OVER is set, there must be an address
1473 * specified in msg_name.
1475 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1482 SCTP_DEBUG_PRINTK("About to look up association.\n");
1486 /* If a msg_name has been specified, assume this is to be used. */
1488 /* Look for a matching association on the endpoint. */
1489 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1491 /* If we could not find a matching association on the
1492 * endpoint, make sure that it is not a TCP-style
1493 * socket that already has an association or there is
1494 * no peeled-off association on another socket.
1496 if ((sctp_style(sk, TCP) &&
1497 sctp_sstate(sk, ESTABLISHED)) ||
1498 sctp_endpoint_is_peeled_off(ep, &to)) {
1499 err = -EADDRNOTAVAIL;
1504 asoc = sctp_id2assoc(sk, associd);
1512 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1514 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1515 * socket that has an association in CLOSED state. This can
1516 * happen when an accepted socket has an association that is
1519 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1524 if (sinfo_flags & SCTP_EOF) {
1525 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1527 sctp_primitive_SHUTDOWN(asoc, NULL);
1531 if (sinfo_flags & SCTP_ABORT) {
1533 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1539 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1540 sctp_primitive_ABORT(asoc, chunk);
1546 /* Do we need to create the association? */
1548 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1550 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1555 /* Check for invalid stream against the stream counts,
1556 * either the default or the user specified stream counts.
1559 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1560 /* Check against the defaults. */
1561 if (sinfo->sinfo_stream >=
1562 sp->initmsg.sinit_num_ostreams) {
1567 /* Check against the requested. */
1568 if (sinfo->sinfo_stream >=
1569 sinit->sinit_num_ostreams) {
1577 * API 3.1.2 bind() - UDP Style Syntax
1578 * If a bind() or sctp_bindx() is not called prior to a
1579 * sendmsg() call that initiates a new association, the
1580 * system picks an ephemeral port and will choose an address
1581 * set equivalent to binding with a wildcard address.
1583 if (!ep->base.bind_addr.port) {
1584 if (sctp_autobind(sk)) {
1590 * If an unprivileged user inherits a one-to-many
1591 * style socket with open associations on a privileged
1592 * port, it MAY be permitted to accept new associations,
1593 * but it SHOULD NOT be permitted to open new
1596 if (ep->base.bind_addr.port < PROT_SOCK &&
1597 !capable(CAP_NET_BIND_SERVICE)) {
1603 scope = sctp_scope(&to);
1604 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1611 /* If the SCTP_INIT ancillary data is specified, set all
1612 * the association init values accordingly.
1615 if (sinit->sinit_num_ostreams) {
1616 asoc->c.sinit_num_ostreams =
1617 sinit->sinit_num_ostreams;
1619 if (sinit->sinit_max_instreams) {
1620 asoc->c.sinit_max_instreams =
1621 sinit->sinit_max_instreams;
1623 if (sinit->sinit_max_attempts) {
1624 asoc->max_init_attempts
1625 = sinit->sinit_max_attempts;
1627 if (sinit->sinit_max_init_timeo) {
1628 asoc->max_init_timeo =
1629 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1633 /* Prime the peer's transport structures. */
1634 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1639 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1646 /* ASSERT: we have a valid association at this point. */
1647 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1650 /* If the user didn't specify SNDRCVINFO, make up one with
1653 default_sinfo.sinfo_stream = asoc->default_stream;
1654 default_sinfo.sinfo_flags = asoc->default_flags;
1655 default_sinfo.sinfo_ppid = asoc->default_ppid;
1656 default_sinfo.sinfo_context = asoc->default_context;
1657 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1658 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1659 sinfo = &default_sinfo;
1662 /* API 7.1.7, the sndbuf size per association bounds the
1663 * maximum size of data that can be sent in a single send call.
1665 if (msg_len > sk->sk_sndbuf) {
1670 if (asoc->pmtu_pending)
1671 sctp_assoc_pending_pmtu(asoc);
1673 /* If fragmentation is disabled and the message length exceeds the
1674 * association fragmentation point, return EMSGSIZE. The I-D
1675 * does not specify what this error is, but this looks like
1678 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1684 /* Check for invalid stream. */
1685 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1691 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1692 if (!sctp_wspace(asoc)) {
1693 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1698 /* If an address is passed with the sendto/sendmsg call, it is used
1699 * to override the primary destination address in the TCP model, or
1700 * when SCTP_ADDR_OVER flag is set in the UDP model.
1702 if ((sctp_style(sk, TCP) && msg_name) ||
1703 (sinfo_flags & SCTP_ADDR_OVER)) {
1704 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1712 /* Auto-connect, if we aren't connected already. */
1713 if (sctp_state(asoc, CLOSED)) {
1714 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1717 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1720 /* Break the message into multiple chunks of maximum size. */
1721 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1727 /* Now send the (possibly) fragmented message. */
1728 list_for_each(pos, &datamsg->chunks) {
1729 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1730 sctp_datamsg_track(chunk);
1732 /* Do accounting for the write space. */
1733 sctp_set_owner_w(chunk);
1735 chunk->transport = chunk_tp;
1737 /* Send it to the lower layers. Note: all chunks
1738 * must either fail or succeed. The lower layer
1739 * works that way today. Keep it that way or this
1742 err = sctp_primitive_SEND(asoc, chunk);
1743 /* Did the lower layer accept the chunk? */
1745 sctp_chunk_free(chunk);
1746 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1749 sctp_datamsg_free(datamsg);
1755 /* If we are already past ASSOCIATE, the lower
1756 * layers are responsible for association cleanup.
1762 sctp_association_free(asoc);
1764 sctp_release_sock(sk);
1767 return sctp_error(sk, msg_flags, err);
1774 err = sock_error(sk);
1784 /* This is an extended version of skb_pull() that removes the data from the
1785 * start of a skb even when data is spread across the list of skb's in the
1786 * frag_list. len specifies the total amount of data that needs to be removed.
1787 * when 'len' bytes could be removed from the skb, it returns 0.
1788 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1789 * could not be removed.
1791 static int sctp_skb_pull(struct sk_buff *skb, int len)
1793 struct sk_buff *list;
1794 int skb_len = skb_headlen(skb);
1797 if (len <= skb_len) {
1798 __skb_pull(skb, len);
1802 __skb_pull(skb, skb_len);
1804 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1805 rlen = sctp_skb_pull(list, len);
1806 skb->len -= (len-rlen);
1807 skb->data_len -= (len-rlen);
1818 /* API 3.1.3 recvmsg() - UDP Style Syntax
1820 * ssize_t recvmsg(int socket, struct msghdr *message,
1823 * socket - the socket descriptor of the endpoint.
1824 * message - pointer to the msghdr structure which contains a single
1825 * user message and possibly some ancillary data.
1827 * See Section 5 for complete description of the data
1830 * flags - flags sent or received with the user message, see Section
1831 * 5 for complete description of the flags.
1833 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1835 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1836 struct msghdr *msg, size_t len, int noblock,
1837 int flags, int *addr_len)
1839 struct sctp_ulpevent *event = NULL;
1840 struct sctp_sock *sp = sctp_sk(sk);
1841 struct sk_buff *skb;
1846 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1847 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1848 "len", len, "knoblauch", noblock,
1849 "flags", flags, "addr_len", addr_len);
1853 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1858 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1862 /* Get the total length of the skb including any skb's in the
1871 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1873 event = sctp_skb2event(skb);
1878 sock_recv_timestamp(msg, sk, skb);
1879 if (sctp_ulpevent_is_notification(event)) {
1880 msg->msg_flags |= MSG_NOTIFICATION;
1881 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1883 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1886 /* Check if we allow SCTP_SNDRCVINFO. */
1887 if (sp->subscribe.sctp_data_io_event)
1888 sctp_ulpevent_read_sndrcvinfo(event, msg);
1890 /* FIXME: we should be calling IP/IPv6 layers. */
1891 if (sk->sk_protinfo.af_inet.cmsg_flags)
1892 ip_cmsg_recv(msg, skb);
1897 /* If skb's length exceeds the user's buffer, update the skb and
1898 * push it back to the receive_queue so that the next call to
1899 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1901 if (skb_len > copied) {
1902 msg->msg_flags &= ~MSG_EOR;
1903 if (flags & MSG_PEEK)
1905 sctp_skb_pull(skb, copied);
1906 skb_queue_head(&sk->sk_receive_queue, skb);
1908 /* When only partial message is copied to the user, increase
1909 * rwnd by that amount. If all the data in the skb is read,
1910 * rwnd is updated when the event is freed.
1912 sctp_assoc_rwnd_increase(event->asoc, copied);
1914 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1915 (event->msg_flags & MSG_EOR))
1916 msg->msg_flags |= MSG_EOR;
1918 msg->msg_flags &= ~MSG_EOR;
1921 if (flags & MSG_PEEK) {
1922 /* Release the skb reference acquired after peeking the skb in
1923 * sctp_skb_recv_datagram().
1927 /* Free the event which includes releasing the reference to
1928 * the owner of the skb, freeing the skb and updating the
1931 sctp_ulpevent_free(event);
1934 sctp_release_sock(sk);
1938 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1940 * This option is a on/off flag. If enabled no SCTP message
1941 * fragmentation will be performed. Instead if a message being sent
1942 * exceeds the current PMTU size, the message will NOT be sent and
1943 * instead a error will be indicated to the user.
1945 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1946 char __user *optval, int optlen)
1950 if (optlen < sizeof(int))
1953 if (get_user(val, (int __user *)optval))
1956 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1961 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1964 if (optlen != sizeof(struct sctp_event_subscribe))
1966 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1971 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1973 * This socket option is applicable to the UDP-style socket only. When
1974 * set it will cause associations that are idle for more than the
1975 * specified number of seconds to automatically close. An association
1976 * being idle is defined an association that has NOT sent or received
1977 * user data. The special value of '0' indicates that no automatic
1978 * close of any associations should be performed. The option expects an
1979 * integer defining the number of seconds of idle time before an
1980 * association is closed.
1982 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1985 struct sctp_sock *sp = sctp_sk(sk);
1987 /* Applicable to UDP-style socket only */
1988 if (sctp_style(sk, TCP))
1990 if (optlen != sizeof(int))
1992 if (copy_from_user(&sp->autoclose, optval, optlen))
1998 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2000 * Applications can enable or disable heartbeats for any peer address of
2001 * an association, modify an address's heartbeat interval, force a
2002 * heartbeat to be sent immediately, and adjust the address's maximum
2003 * number of retransmissions sent before an address is considered
2004 * unreachable. The following structure is used to access and modify an
2005 * address's parameters:
2007 * struct sctp_paddrparams {
2008 * sctp_assoc_t spp_assoc_id;
2009 * struct sockaddr_storage spp_address;
2010 * uint32_t spp_hbinterval;
2011 * uint16_t spp_pathmaxrxt;
2012 * uint32_t spp_pathmtu;
2013 * uint32_t spp_sackdelay;
2014 * uint32_t spp_flags;
2017 * spp_assoc_id - (one-to-many style socket) This is filled in the
2018 * application, and identifies the association for
2020 * spp_address - This specifies which address is of interest.
2021 * spp_hbinterval - This contains the value of the heartbeat interval,
2022 * in milliseconds. If a value of zero
2023 * is present in this field then no changes are to
2024 * be made to this parameter.
2025 * spp_pathmaxrxt - This contains the maximum number of
2026 * retransmissions before this address shall be
2027 * considered unreachable. If a value of zero
2028 * is present in this field then no changes are to
2029 * be made to this parameter.
2030 * spp_pathmtu - When Path MTU discovery is disabled the value
2031 * specified here will be the "fixed" path mtu.
2032 * Note that if the spp_address field is empty
2033 * then all associations on this address will
2034 * have this fixed path mtu set upon them.
2036 * spp_sackdelay - When delayed sack is enabled, this value specifies
2037 * the number of milliseconds that sacks will be delayed
2038 * for. This value will apply to all addresses of an
2039 * association if the spp_address field is empty. Note
2040 * also, that if delayed sack is enabled and this
2041 * value is set to 0, no change is made to the last
2042 * recorded delayed sack timer value.
2044 * spp_flags - These flags are used to control various features
2045 * on an association. The flag field may contain
2046 * zero or more of the following options.
2048 * SPP_HB_ENABLE - Enable heartbeats on the
2049 * specified address. Note that if the address
2050 * field is empty all addresses for the association
2051 * have heartbeats enabled upon them.
2053 * SPP_HB_DISABLE - Disable heartbeats on the
2054 * speicifed address. Note that if the address
2055 * field is empty all addresses for the association
2056 * will have their heartbeats disabled. Note also
2057 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2058 * mutually exclusive, only one of these two should
2059 * be specified. Enabling both fields will have
2060 * undetermined results.
2062 * SPP_HB_DEMAND - Request a user initiated heartbeat
2063 * to be made immediately.
2065 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2066 * heartbeat delayis to be set to the value of 0
2069 * SPP_PMTUD_ENABLE - This field will enable PMTU
2070 * discovery upon the specified address. Note that
2071 * if the address feild is empty then all addresses
2072 * on the association are effected.
2074 * SPP_PMTUD_DISABLE - This field will disable PMTU
2075 * discovery upon the specified address. Note that
2076 * if the address feild is empty then all addresses
2077 * on the association are effected. Not also that
2078 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2079 * exclusive. Enabling both will have undetermined
2082 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2083 * on delayed sack. The time specified in spp_sackdelay
2084 * is used to specify the sack delay for this address. Note
2085 * that if spp_address is empty then all addresses will
2086 * enable delayed sack and take on the sack delay
2087 * value specified in spp_sackdelay.
2088 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2089 * off delayed sack. If the spp_address field is blank then
2090 * delayed sack is disabled for the entire association. Note
2091 * also that this field is mutually exclusive to
2092 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2095 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2096 struct sctp_transport *trans,
2097 struct sctp_association *asoc,
2098 struct sctp_sock *sp,
2101 int sackdelay_change)
2105 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2106 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2111 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2112 * this field is ignored. Note also that a value of zero indicates
2113 * the current setting should be left unchanged.
2115 if (params->spp_flags & SPP_HB_ENABLE) {
2117 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2118 * set. This lets us use 0 value when this flag
2121 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2122 params->spp_hbinterval = 0;
2124 if (params->spp_hbinterval ||
2125 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2128 msecs_to_jiffies(params->spp_hbinterval);
2131 msecs_to_jiffies(params->spp_hbinterval);
2133 sp->hbinterval = params->spp_hbinterval;
2140 trans->param_flags =
2141 (trans->param_flags & ~SPP_HB) | hb_change;
2144 (asoc->param_flags & ~SPP_HB) | hb_change;
2147 (sp->param_flags & ~SPP_HB) | hb_change;
2151 /* When Path MTU discovery is disabled the value specified here will
2152 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2153 * include the flag SPP_PMTUD_DISABLE for this field to have any
2156 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2158 trans->pathmtu = params->spp_pathmtu;
2159 sctp_assoc_sync_pmtu(asoc);
2161 asoc->pathmtu = params->spp_pathmtu;
2162 sctp_frag_point(sp, params->spp_pathmtu);
2164 sp->pathmtu = params->spp_pathmtu;
2170 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2171 (params->spp_flags & SPP_PMTUD_ENABLE);
2172 trans->param_flags =
2173 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2175 sctp_transport_pmtu(trans);
2176 sctp_assoc_sync_pmtu(asoc);
2180 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2183 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2187 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2188 * value of this field is ignored. Note also that a value of zero
2189 * indicates the current setting should be left unchanged.
2191 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2194 msecs_to_jiffies(params->spp_sackdelay);
2197 msecs_to_jiffies(params->spp_sackdelay);
2199 sp->sackdelay = params->spp_sackdelay;
2203 if (sackdelay_change) {
2205 trans->param_flags =
2206 (trans->param_flags & ~SPP_SACKDELAY) |
2210 (asoc->param_flags & ~SPP_SACKDELAY) |
2214 (sp->param_flags & ~SPP_SACKDELAY) |
2219 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2220 * of this field is ignored. Note also that a value of zero
2221 * indicates the current setting should be left unchanged.
2223 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2225 trans->pathmaxrxt = params->spp_pathmaxrxt;
2227 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2229 sp->pathmaxrxt = params->spp_pathmaxrxt;
2236 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2237 char __user *optval, int optlen)
2239 struct sctp_paddrparams params;
2240 struct sctp_transport *trans = NULL;
2241 struct sctp_association *asoc = NULL;
2242 struct sctp_sock *sp = sctp_sk(sk);
2244 int hb_change, pmtud_change, sackdelay_change;
2246 if (optlen != sizeof(struct sctp_paddrparams))
2249 if (copy_from_user(¶ms, optval, optlen))
2252 /* Validate flags and value parameters. */
2253 hb_change = params.spp_flags & SPP_HB;
2254 pmtud_change = params.spp_flags & SPP_PMTUD;
2255 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2257 if (hb_change == SPP_HB ||
2258 pmtud_change == SPP_PMTUD ||
2259 sackdelay_change == SPP_SACKDELAY ||
2260 params.spp_sackdelay > 500 ||
2262 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2265 /* If an address other than INADDR_ANY is specified, and
2266 * no transport is found, then the request is invalid.
2268 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
2269 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2270 params.spp_assoc_id);
2275 /* Get association, if assoc_id != 0 and the socket is a one
2276 * to many style socket, and an association was not found, then
2277 * the id was invalid.
2279 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2280 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2283 /* Heartbeat demand can only be sent on a transport or
2284 * association, but not a socket.
2286 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2289 /* Process parameters. */
2290 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2291 hb_change, pmtud_change,
2297 /* If changes are for association, also apply parameters to each
2300 if (!trans && asoc) {
2301 struct list_head *pos;
2303 list_for_each(pos, &asoc->peer.transport_addr_list) {
2304 trans = list_entry(pos, struct sctp_transport,
2306 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2307 hb_change, pmtud_change,
2315 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2317 * This options will get or set the delayed ack timer. The time is set
2318 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2319 * endpoints default delayed ack timer value. If the assoc_id field is
2320 * non-zero, then the set or get effects the specified association.
2322 * struct sctp_assoc_value {
2323 * sctp_assoc_t assoc_id;
2324 * uint32_t assoc_value;
2327 * assoc_id - This parameter, indicates which association the
2328 * user is preforming an action upon. Note that if
2329 * this field's value is zero then the endpoints
2330 * default value is changed (effecting future
2331 * associations only).
2333 * assoc_value - This parameter contains the number of milliseconds
2334 * that the user is requesting the delayed ACK timer
2335 * be set to. Note that this value is defined in
2336 * the standard to be between 200 and 500 milliseconds.
2338 * Note: a value of zero will leave the value alone,
2339 * but disable SACK delay. A non-zero value will also
2340 * enable SACK delay.
2343 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2344 char __user *optval, int optlen)
2346 struct sctp_assoc_value params;
2347 struct sctp_transport *trans = NULL;
2348 struct sctp_association *asoc = NULL;
2349 struct sctp_sock *sp = sctp_sk(sk);
2351 if (optlen != sizeof(struct sctp_assoc_value))
2354 if (copy_from_user(¶ms, optval, optlen))
2357 /* Validate value parameter. */
2358 if (params.assoc_value > 500)
2361 /* Get association, if assoc_id != 0 and the socket is a one
2362 * to many style socket, and an association was not found, then
2363 * the id was invalid.
2365 asoc = sctp_id2assoc(sk, params.assoc_id);
2366 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2369 if (params.assoc_value) {
2372 msecs_to_jiffies(params.assoc_value);
2374 (asoc->param_flags & ~SPP_SACKDELAY) |
2375 SPP_SACKDELAY_ENABLE;
2377 sp->sackdelay = params.assoc_value;
2379 (sp->param_flags & ~SPP_SACKDELAY) |
2380 SPP_SACKDELAY_ENABLE;
2385 (asoc->param_flags & ~SPP_SACKDELAY) |
2386 SPP_SACKDELAY_DISABLE;
2389 (sp->param_flags & ~SPP_SACKDELAY) |
2390 SPP_SACKDELAY_DISABLE;
2394 /* If change is for association, also apply to each transport. */
2396 struct list_head *pos;
2398 list_for_each(pos, &asoc->peer.transport_addr_list) {
2399 trans = list_entry(pos, struct sctp_transport,
2401 if (params.assoc_value) {
2403 msecs_to_jiffies(params.assoc_value);
2404 trans->param_flags =
2405 (trans->param_flags & ~SPP_SACKDELAY) |
2406 SPP_SACKDELAY_ENABLE;
2408 trans->param_flags =
2409 (trans->param_flags & ~SPP_SACKDELAY) |
2410 SPP_SACKDELAY_DISABLE;
2418 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2420 * Applications can specify protocol parameters for the default association
2421 * initialization. The option name argument to setsockopt() and getsockopt()
2424 * Setting initialization parameters is effective only on an unconnected
2425 * socket (for UDP-style sockets only future associations are effected
2426 * by the change). With TCP-style sockets, this option is inherited by
2427 * sockets derived from a listener socket.
2429 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2431 struct sctp_initmsg sinit;
2432 struct sctp_sock *sp = sctp_sk(sk);
2434 if (optlen != sizeof(struct sctp_initmsg))
2436 if (copy_from_user(&sinit, optval, optlen))
2439 if (sinit.sinit_num_ostreams)
2440 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2441 if (sinit.sinit_max_instreams)
2442 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2443 if (sinit.sinit_max_attempts)
2444 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2445 if (sinit.sinit_max_init_timeo)
2446 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2452 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2454 * Applications that wish to use the sendto() system call may wish to
2455 * specify a default set of parameters that would normally be supplied
2456 * through the inclusion of ancillary data. This socket option allows
2457 * such an application to set the default sctp_sndrcvinfo structure.
2458 * The application that wishes to use this socket option simply passes
2459 * in to this call the sctp_sndrcvinfo structure defined in Section
2460 * 5.2.2) The input parameters accepted by this call include
2461 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2462 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2463 * to this call if the caller is using the UDP model.
2465 static int sctp_setsockopt_default_send_param(struct sock *sk,
2466 char __user *optval, int optlen)
2468 struct sctp_sndrcvinfo info;
2469 struct sctp_association *asoc;
2470 struct sctp_sock *sp = sctp_sk(sk);
2472 if (optlen != sizeof(struct sctp_sndrcvinfo))
2474 if (copy_from_user(&info, optval, optlen))
2477 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2478 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2482 asoc->default_stream = info.sinfo_stream;
2483 asoc->default_flags = info.sinfo_flags;
2484 asoc->default_ppid = info.sinfo_ppid;
2485 asoc->default_context = info.sinfo_context;
2486 asoc->default_timetolive = info.sinfo_timetolive;
2488 sp->default_stream = info.sinfo_stream;
2489 sp->default_flags = info.sinfo_flags;
2490 sp->default_ppid = info.sinfo_ppid;
2491 sp->default_context = info.sinfo_context;
2492 sp->default_timetolive = info.sinfo_timetolive;
2498 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2500 * Requests that the local SCTP stack use the enclosed peer address as
2501 * the association primary. The enclosed address must be one of the
2502 * association peer's addresses.
2504 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2507 struct sctp_prim prim;
2508 struct sctp_transport *trans;
2510 if (optlen != sizeof(struct sctp_prim))
2513 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2516 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2520 sctp_assoc_set_primary(trans->asoc, trans);
2526 * 7.1.5 SCTP_NODELAY
2528 * Turn on/off any Nagle-like algorithm. This means that packets are
2529 * generally sent as soon as possible and no unnecessary delays are
2530 * introduced, at the cost of more packets in the network. Expects an
2531 * integer boolean flag.
2533 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2538 if (optlen < sizeof(int))
2540 if (get_user(val, (int __user *)optval))
2543 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2549 * 7.1.1 SCTP_RTOINFO
2551 * The protocol parameters used to initialize and bound retransmission
2552 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2553 * and modify these parameters.
2554 * All parameters are time values, in milliseconds. A value of 0, when
2555 * modifying the parameters, indicates that the current value should not
2559 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2560 struct sctp_rtoinfo rtoinfo;
2561 struct sctp_association *asoc;
2563 if (optlen != sizeof (struct sctp_rtoinfo))
2566 if (copy_from_user(&rtoinfo, optval, optlen))
2569 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2571 /* Set the values to the specific association */
2572 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2576 if (rtoinfo.srto_initial != 0)
2578 msecs_to_jiffies(rtoinfo.srto_initial);
2579 if (rtoinfo.srto_max != 0)
2580 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2581 if (rtoinfo.srto_min != 0)
2582 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2584 /* If there is no association or the association-id = 0
2585 * set the values to the endpoint.
2587 struct sctp_sock *sp = sctp_sk(sk);
2589 if (rtoinfo.srto_initial != 0)
2590 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2591 if (rtoinfo.srto_max != 0)
2592 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2593 if (rtoinfo.srto_min != 0)
2594 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2602 * 7.1.2 SCTP_ASSOCINFO
2604 * This option is used to tune the maximum retransmission attempts
2605 * of the association.
2606 * Returns an error if the new association retransmission value is
2607 * greater than the sum of the retransmission value of the peer.
2608 * See [SCTP] for more information.
2611 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2614 struct sctp_assocparams assocparams;
2615 struct sctp_association *asoc;
2617 if (optlen != sizeof(struct sctp_assocparams))
2619 if (copy_from_user(&assocparams, optval, optlen))
2622 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2624 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2627 /* Set the values to the specific association */
2629 if (assocparams.sasoc_asocmaxrxt != 0) {
2632 struct list_head *pos;
2633 struct sctp_transport *peer_addr;
2635 list_for_each(pos, &asoc->peer.transport_addr_list) {
2636 peer_addr = list_entry(pos,
2637 struct sctp_transport,
2639 path_sum += peer_addr->pathmaxrxt;
2643 /* Only validate asocmaxrxt if we have more then
2644 * one path/transport. We do this because path
2645 * retransmissions are only counted when we have more
2649 assocparams.sasoc_asocmaxrxt > path_sum)
2652 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2655 if (assocparams.sasoc_cookie_life != 0) {
2656 asoc->cookie_life.tv_sec =
2657 assocparams.sasoc_cookie_life / 1000;
2658 asoc->cookie_life.tv_usec =
2659 (assocparams.sasoc_cookie_life % 1000)
2663 /* Set the values to the endpoint */
2664 struct sctp_sock *sp = sctp_sk(sk);
2666 if (assocparams.sasoc_asocmaxrxt != 0)
2667 sp->assocparams.sasoc_asocmaxrxt =
2668 assocparams.sasoc_asocmaxrxt;
2669 if (assocparams.sasoc_cookie_life != 0)
2670 sp->assocparams.sasoc_cookie_life =
2671 assocparams.sasoc_cookie_life;
2677 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2679 * This socket option is a boolean flag which turns on or off mapped V4
2680 * addresses. If this option is turned on and the socket is type
2681 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2682 * If this option is turned off, then no mapping will be done of V4
2683 * addresses and a user will receive both PF_INET6 and PF_INET type
2684 * addresses on the socket.
2686 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2689 struct sctp_sock *sp = sctp_sk(sk);
2691 if (optlen < sizeof(int))
2693 if (get_user(val, (int __user *)optval))
2704 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2706 * This socket option specifies the maximum size to put in any outgoing
2707 * SCTP chunk. If a message is larger than this size it will be
2708 * fragmented by SCTP into the specified size. Note that the underlying
2709 * SCTP implementation may fragment into smaller sized chunks when the
2710 * PMTU of the underlying association is smaller than the value set by
2713 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2715 struct sctp_association *asoc;
2716 struct list_head *pos;
2717 struct sctp_sock *sp = sctp_sk(sk);
2720 if (optlen < sizeof(int))
2722 if (get_user(val, (int __user *)optval))
2724 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2726 sp->user_frag = val;
2728 /* Update the frag_point of the existing associations. */
2729 list_for_each(pos, &(sp->ep->asocs)) {
2730 asoc = list_entry(pos, struct sctp_association, asocs);
2731 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2739 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2741 * Requests that the peer mark the enclosed address as the association
2742 * primary. The enclosed address must be one of the association's
2743 * locally bound addresses. The following structure is used to make a
2744 * set primary request:
2746 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2749 struct sctp_sock *sp;
2750 struct sctp_endpoint *ep;
2751 struct sctp_association *asoc = NULL;
2752 struct sctp_setpeerprim prim;
2753 struct sctp_chunk *chunk;
2759 if (!sctp_addip_enable)
2762 if (optlen != sizeof(struct sctp_setpeerprim))
2765 if (copy_from_user(&prim, optval, optlen))
2768 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2772 if (!asoc->peer.asconf_capable)
2775 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2778 if (!sctp_state(asoc, ESTABLISHED))
2781 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2782 return -EADDRNOTAVAIL;
2784 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2785 chunk = sctp_make_asconf_set_prim(asoc,
2786 (union sctp_addr *)&prim.sspp_addr);
2790 err = sctp_send_asconf(asoc, chunk);
2792 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2797 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2800 struct sctp_setadaptation adaptation;
2802 if (optlen != sizeof(struct sctp_setadaptation))
2804 if (copy_from_user(&adaptation, optval, optlen))
2807 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2813 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2815 * The context field in the sctp_sndrcvinfo structure is normally only
2816 * used when a failed message is retrieved holding the value that was
2817 * sent down on the actual send call. This option allows the setting of
2818 * a default context on an association basis that will be received on
2819 * reading messages from the peer. This is especially helpful in the
2820 * one-2-many model for an application to keep some reference to an
2821 * internal state machine that is processing messages on the
2822 * association. Note that the setting of this value only effects
2823 * received messages from the peer and does not effect the value that is
2824 * saved with outbound messages.
2826 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2829 struct sctp_assoc_value params;
2830 struct sctp_sock *sp;
2831 struct sctp_association *asoc;
2833 if (optlen != sizeof(struct sctp_assoc_value))
2835 if (copy_from_user(¶ms, optval, optlen))
2840 if (params.assoc_id != 0) {
2841 asoc = sctp_id2assoc(sk, params.assoc_id);
2844 asoc->default_rcv_context = params.assoc_value;
2846 sp->default_rcv_context = params.assoc_value;
2853 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2855 * This options will at a minimum specify if the implementation is doing
2856 * fragmented interleave. Fragmented interleave, for a one to many
2857 * socket, is when subsequent calls to receive a message may return
2858 * parts of messages from different associations. Some implementations
2859 * may allow you to turn this value on or off. If so, when turned off,
2860 * no fragment interleave will occur (which will cause a head of line
2861 * blocking amongst multiple associations sharing the same one to many
2862 * socket). When this option is turned on, then each receive call may
2863 * come from a different association (thus the user must receive data
2864 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2865 * association each receive belongs to.
2867 * This option takes a boolean value. A non-zero value indicates that
2868 * fragmented interleave is on. A value of zero indicates that
2869 * fragmented interleave is off.
2871 * Note that it is important that an implementation that allows this
2872 * option to be turned on, have it off by default. Otherwise an unaware
2873 * application using the one to many model may become confused and act
2876 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2877 char __user *optval,
2882 if (optlen != sizeof(int))
2884 if (get_user(val, (int __user *)optval))
2887 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2893 * 7.1.25. Set or Get the sctp partial delivery point
2894 * (SCTP_PARTIAL_DELIVERY_POINT)
2895 * This option will set or get the SCTP partial delivery point. This
2896 * point is the size of a message where the partial delivery API will be
2897 * invoked to help free up rwnd space for the peer. Setting this to a
2898 * lower value will cause partial delivery's to happen more often. The
2899 * calls argument is an integer that sets or gets the partial delivery
2902 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2903 char __user *optval,
2908 if (optlen != sizeof(u32))
2910 if (get_user(val, (int __user *)optval))
2913 sctp_sk(sk)->pd_point = val;
2915 return 0; /* is this the right error code? */
2919 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2921 * This option will allow a user to change the maximum burst of packets
2922 * that can be emitted by this association. Note that the default value
2923 * is 4, and some implementations may restrict this setting so that it
2924 * can only be lowered.
2926 * NOTE: This text doesn't seem right. Do this on a socket basis with
2927 * future associations inheriting the socket value.
2929 static int sctp_setsockopt_maxburst(struct sock *sk,
2930 char __user *optval,
2935 if (optlen != sizeof(int))
2937 if (get_user(val, (int __user *)optval))
2943 sctp_sk(sk)->max_burst = val;
2948 /* API 6.2 setsockopt(), getsockopt()
2950 * Applications use setsockopt() and getsockopt() to set or retrieve
2951 * socket options. Socket options are used to change the default
2952 * behavior of sockets calls. They are described in Section 7.
2956 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2957 * int __user *optlen);
2958 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2961 * sd - the socket descript.
2962 * level - set to IPPROTO_SCTP for all SCTP options.
2963 * optname - the option name.
2964 * optval - the buffer to store the value of the option.
2965 * optlen - the size of the buffer.
2967 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2968 char __user *optval, int optlen)
2972 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2975 /* I can hardly begin to describe how wrong this is. This is
2976 * so broken as to be worse than useless. The API draft
2977 * REALLY is NOT helpful here... I am not convinced that the
2978 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2979 * are at all well-founded.
2981 if (level != SOL_SCTP) {
2982 struct sctp_af *af = sctp_sk(sk)->pf->af;
2983 retval = af->setsockopt(sk, level, optname, optval, optlen);
2990 case SCTP_SOCKOPT_BINDX_ADD:
2991 /* 'optlen' is the size of the addresses buffer. */
2992 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2993 optlen, SCTP_BINDX_ADD_ADDR);
2996 case SCTP_SOCKOPT_BINDX_REM:
2997 /* 'optlen' is the size of the addresses buffer. */
2998 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2999 optlen, SCTP_BINDX_REM_ADDR);
3002 case SCTP_SOCKOPT_CONNECTX:
3003 /* 'optlen' is the size of the addresses buffer. */
3004 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3008 case SCTP_DISABLE_FRAGMENTS:
3009 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3013 retval = sctp_setsockopt_events(sk, optval, optlen);
3016 case SCTP_AUTOCLOSE:
3017 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3020 case SCTP_PEER_ADDR_PARAMS:
3021 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3024 case SCTP_DELAYED_ACK_TIME:
3025 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3027 case SCTP_PARTIAL_DELIVERY_POINT:
3028 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3032 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3034 case SCTP_DEFAULT_SEND_PARAM:
3035 retval = sctp_setsockopt_default_send_param(sk, optval,
3038 case SCTP_PRIMARY_ADDR:
3039 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3041 case SCTP_SET_PEER_PRIMARY_ADDR:
3042 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3045 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3048 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3050 case SCTP_ASSOCINFO:
3051 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3053 case SCTP_I_WANT_MAPPED_V4_ADDR:
3054 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3057 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3059 case SCTP_ADAPTATION_LAYER:
3060 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3063 retval = sctp_setsockopt_context(sk, optval, optlen);
3065 case SCTP_FRAGMENT_INTERLEAVE:
3066 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3068 case SCTP_MAX_BURST:
3069 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3072 retval = -ENOPROTOOPT;
3076 sctp_release_sock(sk);
3082 /* API 3.1.6 connect() - UDP Style Syntax
3084 * An application may use the connect() call in the UDP model to initiate an
3085 * association without sending data.
3089 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3091 * sd: the socket descriptor to have a new association added to.
3093 * nam: the address structure (either struct sockaddr_in or struct
3094 * sockaddr_in6 defined in RFC2553 [7]).
3096 * len: the size of the address.
3098 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3106 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3107 __FUNCTION__, sk, addr, addr_len);
3109 /* Validate addr_len before calling common connect/connectx routine. */
3110 af = sctp_get_af_specific(addr->sa_family);
3111 if (!af || addr_len < af->sockaddr_len) {
3114 /* Pass correct addr len to common routine (so it knows there
3115 * is only one address being passed.
3117 err = __sctp_connect(sk, addr, af->sockaddr_len);
3120 sctp_release_sock(sk);
3124 /* FIXME: Write comments. */
3125 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3127 return -EOPNOTSUPP; /* STUB */
3130 /* 4.1.4 accept() - TCP Style Syntax
3132 * Applications use accept() call to remove an established SCTP
3133 * association from the accept queue of the endpoint. A new socket
3134 * descriptor will be returned from accept() to represent the newly
3135 * formed association.
3137 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3139 struct sctp_sock *sp;
3140 struct sctp_endpoint *ep;
3141 struct sock *newsk = NULL;
3142 struct sctp_association *asoc;
3151 if (!sctp_style(sk, TCP)) {
3152 error = -EOPNOTSUPP;
3156 if (!sctp_sstate(sk, LISTENING)) {
3161 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3163 error = sctp_wait_for_accept(sk, timeo);
3167 /* We treat the list of associations on the endpoint as the accept
3168 * queue and pick the first association on the list.
3170 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3172 newsk = sp->pf->create_accept_sk(sk, asoc);
3178 /* Populate the fields of the newsk from the oldsk and migrate the
3179 * asoc to the newsk.
3181 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3184 sctp_release_sock(sk);
3189 /* The SCTP ioctl handler. */
3190 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3192 return -ENOIOCTLCMD;
3195 /* This is the function which gets called during socket creation to
3196 * initialized the SCTP-specific portion of the sock.
3197 * The sock structure should already be zero-filled memory.
3199 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3201 struct sctp_endpoint *ep;
3202 struct sctp_sock *sp;
3204 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3208 /* Initialize the SCTP per socket area. */
3209 switch (sk->sk_type) {
3210 case SOCK_SEQPACKET:
3211 sp->type = SCTP_SOCKET_UDP;
3214 sp->type = SCTP_SOCKET_TCP;
3217 return -ESOCKTNOSUPPORT;
3220 /* Initialize default send parameters. These parameters can be
3221 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3223 sp->default_stream = 0;
3224 sp->default_ppid = 0;
3225 sp->default_flags = 0;
3226 sp->default_context = 0;
3227 sp->default_timetolive = 0;
3229 sp->default_rcv_context = 0;
3230 sp->max_burst = sctp_max_burst;
3232 /* Initialize default setup parameters. These parameters
3233 * can be modified with the SCTP_INITMSG socket option or
3234 * overridden by the SCTP_INIT CMSG.
3236 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3237 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3238 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3239 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3241 /* Initialize default RTO related parameters. These parameters can
3242 * be modified for with the SCTP_RTOINFO socket option.
3244 sp->rtoinfo.srto_initial = sctp_rto_initial;
3245 sp->rtoinfo.srto_max = sctp_rto_max;
3246 sp->rtoinfo.srto_min = sctp_rto_min;
3248 /* Initialize default association related parameters. These parameters
3249 * can be modified with the SCTP_ASSOCINFO socket option.
3251 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3252 sp->assocparams.sasoc_number_peer_destinations = 0;
3253 sp->assocparams.sasoc_peer_rwnd = 0;
3254 sp->assocparams.sasoc_local_rwnd = 0;
3255 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3257 /* Initialize default event subscriptions. By default, all the
3260 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3262 /* Default Peer Address Parameters. These defaults can
3263 * be modified via SCTP_PEER_ADDR_PARAMS
3265 sp->hbinterval = sctp_hb_interval;
3266 sp->pathmaxrxt = sctp_max_retrans_path;
3267 sp->pathmtu = 0; // allow default discovery
3268 sp->sackdelay = sctp_sack_timeout;
3269 sp->param_flags = SPP_HB_ENABLE |
3271 SPP_SACKDELAY_ENABLE;
3273 /* If enabled no SCTP message fragmentation will be performed.
3274 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3276 sp->disable_fragments = 0;
3278 /* Enable Nagle algorithm by default. */
3281 /* Enable by default. */
3284 /* Auto-close idle associations after the configured
3285 * number of seconds. A value of 0 disables this
3286 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3287 * for UDP-style sockets only.
3291 /* User specified fragmentation limit. */
3294 sp->adaptation_ind = 0;
3296 sp->pf = sctp_get_pf_specific(sk->sk_family);
3298 /* Control variables for partial data delivery. */
3299 atomic_set(&sp->pd_mode, 0);
3300 skb_queue_head_init(&sp->pd_lobby);
3301 sp->frag_interleave = 0;
3303 /* Create a per socket endpoint structure. Even if we
3304 * change the data structure relationships, this may still
3305 * be useful for storing pre-connect address information.
3307 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3314 SCTP_DBG_OBJCNT_INC(sock);
3318 /* Cleanup any SCTP per socket resources. */
3319 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3321 struct sctp_endpoint *ep;
3323 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3325 /* Release our hold on the endpoint. */
3326 ep = sctp_sk(sk)->ep;
3327 sctp_endpoint_free(ep);
3332 /* API 4.1.7 shutdown() - TCP Style Syntax
3333 * int shutdown(int socket, int how);
3335 * sd - the socket descriptor of the association to be closed.
3336 * how - Specifies the type of shutdown. The values are
3339 * Disables further receive operations. No SCTP
3340 * protocol action is taken.
3342 * Disables further send operations, and initiates
3343 * the SCTP shutdown sequence.
3345 * Disables further send and receive operations
3346 * and initiates the SCTP shutdown sequence.
3348 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3350 struct sctp_endpoint *ep;
3351 struct sctp_association *asoc;
3353 if (!sctp_style(sk, TCP))
3356 if (how & SEND_SHUTDOWN) {
3357 ep = sctp_sk(sk)->ep;
3358 if (!list_empty(&ep->asocs)) {
3359 asoc = list_entry(ep->asocs.next,
3360 struct sctp_association, asocs);
3361 sctp_primitive_SHUTDOWN(asoc, NULL);
3366 /* 7.2.1 Association Status (SCTP_STATUS)
3368 * Applications can retrieve current status information about an
3369 * association, including association state, peer receiver window size,
3370 * number of unacked data chunks, and number of data chunks pending
3371 * receipt. This information is read-only.
3373 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3374 char __user *optval,
3377 struct sctp_status status;
3378 struct sctp_association *asoc = NULL;
3379 struct sctp_transport *transport;
3380 sctp_assoc_t associd;
3383 if (len < sizeof(status)) {
3388 len = sizeof(status);
3389 if (copy_from_user(&status, optval, len)) {
3394 associd = status.sstat_assoc_id;
3395 asoc = sctp_id2assoc(sk, associd);
3401 transport = asoc->peer.primary_path;
3403 status.sstat_assoc_id = sctp_assoc2id(asoc);
3404 status.sstat_state = asoc->state;
3405 status.sstat_rwnd = asoc->peer.rwnd;
3406 status.sstat_unackdata = asoc->unack_data;
3408 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3409 status.sstat_instrms = asoc->c.sinit_max_instreams;
3410 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3411 status.sstat_fragmentation_point = asoc->frag_point;
3412 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3413 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3414 transport->af_specific->sockaddr_len);
3415 /* Map ipv4 address into v4-mapped-on-v6 address. */
3416 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3417 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3418 status.sstat_primary.spinfo_state = transport->state;
3419 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3420 status.sstat_primary.spinfo_srtt = transport->srtt;
3421 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3422 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3424 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3425 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3427 if (put_user(len, optlen)) {
3432 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3433 len, status.sstat_state, status.sstat_rwnd,
3434 status.sstat_assoc_id);
3436 if (copy_to_user(optval, &status, len)) {
3446 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3448 * Applications can retrieve information about a specific peer address
3449 * of an association, including its reachability state, congestion
3450 * window, and retransmission timer values. This information is
3453 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3454 char __user *optval,
3457 struct sctp_paddrinfo pinfo;
3458 struct sctp_transport *transport;
3461 if (len < sizeof(pinfo)) {
3466 len = sizeof(pinfo);
3467 if (copy_from_user(&pinfo, optval, len)) {
3472 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3473 pinfo.spinfo_assoc_id);
3477 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3478 pinfo.spinfo_state = transport->state;
3479 pinfo.spinfo_cwnd = transport->cwnd;
3480 pinfo.spinfo_srtt = transport->srtt;
3481 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3482 pinfo.spinfo_mtu = transport->pathmtu;
3484 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3485 pinfo.spinfo_state = SCTP_ACTIVE;
3487 if (put_user(len, optlen)) {
3492 if (copy_to_user(optval, &pinfo, len)) {
3501 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3503 * This option is a on/off flag. If enabled no SCTP message
3504 * fragmentation will be performed. Instead if a message being sent
3505 * exceeds the current PMTU size, the message will NOT be sent and
3506 * instead a error will be indicated to the user.
3508 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3509 char __user *optval, int __user *optlen)
3513 if (len < sizeof(int))
3517 val = (sctp_sk(sk)->disable_fragments == 1);
3518 if (put_user(len, optlen))
3520 if (copy_to_user(optval, &val, len))
3525 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3527 * This socket option is used to specify various notifications and
3528 * ancillary data the user wishes to receive.
3530 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3533 if (len < sizeof(struct sctp_event_subscribe))
3535 len = sizeof(struct sctp_event_subscribe);
3536 if (put_user(len, optlen))
3538 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3543 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3545 * This socket option is applicable to the UDP-style socket only. When
3546 * set it will cause associations that are idle for more than the
3547 * specified number of seconds to automatically close. An association
3548 * being idle is defined an association that has NOT sent or received
3549 * user data. The special value of '0' indicates that no automatic
3550 * close of any associations should be performed. The option expects an
3551 * integer defining the number of seconds of idle time before an
3552 * association is closed.
3554 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3556 /* Applicable to UDP-style socket only */
3557 if (sctp_style(sk, TCP))
3559 if (len < sizeof(int))
3562 if (put_user(len, optlen))
3564 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3569 /* Helper routine to branch off an association to a new socket. */
3570 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3571 struct socket **sockp)
3573 struct sock *sk = asoc->base.sk;
3574 struct socket *sock;
3575 struct inet_sock *inetsk;
3579 /* An association cannot be branched off from an already peeled-off
3580 * socket, nor is this supported for tcp style sockets.
3582 if (!sctp_style(sk, UDP))
3585 /* Create a new socket. */
3586 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3590 /* Populate the fields of the newsk from the oldsk and migrate the
3591 * asoc to the newsk.
3593 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3595 /* Make peeled-off sockets more like 1-1 accepted sockets.
3596 * Set the daddr and initialize id to something more random
3598 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3599 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3600 inetsk = inet_sk(sock->sk);
3601 inetsk->id = asoc->next_tsn ^ jiffies;
3608 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3610 sctp_peeloff_arg_t peeloff;
3611 struct socket *newsock;
3613 struct sctp_association *asoc;
3615 if (len < sizeof(sctp_peeloff_arg_t))
3617 len = sizeof(sctp_peeloff_arg_t);
3618 if (copy_from_user(&peeloff, optval, len))
3621 asoc = sctp_id2assoc(sk, peeloff.associd);
3627 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3629 retval = sctp_do_peeloff(asoc, &newsock);
3633 /* Map the socket to an unused fd that can be returned to the user. */
3634 retval = sock_map_fd(newsock);
3636 sock_release(newsock);
3640 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3641 __FUNCTION__, sk, asoc, newsock->sk, retval);
3643 /* Return the fd mapped to the new socket. */
3644 peeloff.sd = retval;
3645 if (put_user(len, optlen))
3647 if (copy_to_user(optval, &peeloff, len))
3654 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3656 * Applications can enable or disable heartbeats for any peer address of
3657 * an association, modify an address's heartbeat interval, force a
3658 * heartbeat to be sent immediately, and adjust the address's maximum
3659 * number of retransmissions sent before an address is considered
3660 * unreachable. The following structure is used to access and modify an
3661 * address's parameters:
3663 * struct sctp_paddrparams {
3664 * sctp_assoc_t spp_assoc_id;
3665 * struct sockaddr_storage spp_address;
3666 * uint32_t spp_hbinterval;
3667 * uint16_t spp_pathmaxrxt;
3668 * uint32_t spp_pathmtu;
3669 * uint32_t spp_sackdelay;
3670 * uint32_t spp_flags;
3673 * spp_assoc_id - (one-to-many style socket) This is filled in the
3674 * application, and identifies the association for
3676 * spp_address - This specifies which address is of interest.
3677 * spp_hbinterval - This contains the value of the heartbeat interval,
3678 * in milliseconds. If a value of zero
3679 * is present in this field then no changes are to
3680 * be made to this parameter.
3681 * spp_pathmaxrxt - This contains the maximum number of
3682 * retransmissions before this address shall be
3683 * considered unreachable. If a value of zero
3684 * is present in this field then no changes are to
3685 * be made to this parameter.
3686 * spp_pathmtu - When Path MTU discovery is disabled the value
3687 * specified here will be the "fixed" path mtu.
3688 * Note that if the spp_address field is empty
3689 * then all associations on this address will
3690 * have this fixed path mtu set upon them.
3692 * spp_sackdelay - When delayed sack is enabled, this value specifies
3693 * the number of milliseconds that sacks will be delayed
3694 * for. This value will apply to all addresses of an
3695 * association if the spp_address field is empty. Note
3696 * also, that if delayed sack is enabled and this
3697 * value is set to 0, no change is made to the last
3698 * recorded delayed sack timer value.
3700 * spp_flags - These flags are used to control various features
3701 * on an association. The flag field may contain
3702 * zero or more of the following options.
3704 * SPP_HB_ENABLE - Enable heartbeats on the
3705 * specified address. Note that if the address
3706 * field is empty all addresses for the association
3707 * have heartbeats enabled upon them.
3709 * SPP_HB_DISABLE - Disable heartbeats on the
3710 * speicifed address. Note that if the address
3711 * field is empty all addresses for the association
3712 * will have their heartbeats disabled. Note also
3713 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3714 * mutually exclusive, only one of these two should
3715 * be specified. Enabling both fields will have
3716 * undetermined results.
3718 * SPP_HB_DEMAND - Request a user initiated heartbeat
3719 * to be made immediately.
3721 * SPP_PMTUD_ENABLE - This field will enable PMTU
3722 * discovery upon the specified address. Note that
3723 * if the address feild is empty then all addresses
3724 * on the association are effected.
3726 * SPP_PMTUD_DISABLE - This field will disable PMTU
3727 * discovery upon the specified address. Note that
3728 * if the address feild is empty then all addresses
3729 * on the association are effected. Not also that
3730 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3731 * exclusive. Enabling both will have undetermined
3734 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3735 * on delayed sack. The time specified in spp_sackdelay
3736 * is used to specify the sack delay for this address. Note
3737 * that if spp_address is empty then all addresses will
3738 * enable delayed sack and take on the sack delay
3739 * value specified in spp_sackdelay.
3740 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3741 * off delayed sack. If the spp_address field is blank then
3742 * delayed sack is disabled for the entire association. Note
3743 * also that this field is mutually exclusive to
3744 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3747 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3748 char __user *optval, int __user *optlen)
3750 struct sctp_paddrparams params;
3751 struct sctp_transport *trans = NULL;
3752 struct sctp_association *asoc = NULL;
3753 struct sctp_sock *sp = sctp_sk(sk);
3755 if (len < sizeof(struct sctp_paddrparams))
3757 len = sizeof(struct sctp_paddrparams);
3758 if (copy_from_user(¶ms, optval, len))
3761 /* If an address other than INADDR_ANY is specified, and
3762 * no transport is found, then the request is invalid.
3764 if (!sctp_is_any(( union sctp_addr *)¶ms.spp_address)) {
3765 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
3766 params.spp_assoc_id);
3768 SCTP_DEBUG_PRINTK("Failed no transport\n");
3773 /* Get association, if assoc_id != 0 and the socket is a one
3774 * to many style socket, and an association was not found, then
3775 * the id was invalid.
3777 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3778 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3779 SCTP_DEBUG_PRINTK("Failed no association\n");
3784 /* Fetch transport values. */
3785 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3786 params.spp_pathmtu = trans->pathmtu;
3787 params.spp_pathmaxrxt = trans->pathmaxrxt;
3788 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3790 /*draft-11 doesn't say what to return in spp_flags*/
3791 params.spp_flags = trans->param_flags;
3793 /* Fetch association values. */
3794 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3795 params.spp_pathmtu = asoc->pathmtu;
3796 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3797 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3799 /*draft-11 doesn't say what to return in spp_flags*/
3800 params.spp_flags = asoc->param_flags;
3802 /* Fetch socket values. */
3803 params.spp_hbinterval = sp->hbinterval;
3804 params.spp_pathmtu = sp->pathmtu;
3805 params.spp_sackdelay = sp->sackdelay;
3806 params.spp_pathmaxrxt = sp->pathmaxrxt;
3808 /*draft-11 doesn't say what to return in spp_flags*/
3809 params.spp_flags = sp->param_flags;
3812 if (copy_to_user(optval, ¶ms, len))
3815 if (put_user(len, optlen))
3821 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3823 * This options will get or set the delayed ack timer. The time is set
3824 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3825 * endpoints default delayed ack timer value. If the assoc_id field is
3826 * non-zero, then the set or get effects the specified association.
3828 * struct sctp_assoc_value {
3829 * sctp_assoc_t assoc_id;
3830 * uint32_t assoc_value;
3833 * assoc_id - This parameter, indicates which association the
3834 * user is preforming an action upon. Note that if
3835 * this field's value is zero then the endpoints
3836 * default value is changed (effecting future
3837 * associations only).
3839 * assoc_value - This parameter contains the number of milliseconds
3840 * that the user is requesting the delayed ACK timer
3841 * be set to. Note that this value is defined in
3842 * the standard to be between 200 and 500 milliseconds.
3844 * Note: a value of zero will leave the value alone,
3845 * but disable SACK delay. A non-zero value will also
3846 * enable SACK delay.
3848 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3849 char __user *optval,
3852 struct sctp_assoc_value params;
3853 struct sctp_association *asoc = NULL;
3854 struct sctp_sock *sp = sctp_sk(sk);
3856 if (len < sizeof(struct sctp_assoc_value))
3859 len = sizeof(struct sctp_assoc_value);
3861 if (copy_from_user(¶ms, optval, len))
3864 /* Get association, if assoc_id != 0 and the socket is a one
3865 * to many style socket, and an association was not found, then
3866 * the id was invalid.
3868 asoc = sctp_id2assoc(sk, params.assoc_id);
3869 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3873 /* Fetch association values. */
3874 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3875 params.assoc_value = jiffies_to_msecs(
3878 params.assoc_value = 0;
3880 /* Fetch socket values. */
3881 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3882 params.assoc_value = sp->sackdelay;
3884 params.assoc_value = 0;
3887 if (copy_to_user(optval, ¶ms, len))
3890 if (put_user(len, optlen))
3896 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3898 * Applications can specify protocol parameters for the default association
3899 * initialization. The option name argument to setsockopt() and getsockopt()
3902 * Setting initialization parameters is effective only on an unconnected
3903 * socket (for UDP-style sockets only future associations are effected
3904 * by the change). With TCP-style sockets, this option is inherited by
3905 * sockets derived from a listener socket.
3907 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3909 if (len < sizeof(struct sctp_initmsg))
3911 len = sizeof(struct sctp_initmsg);
3912 if (put_user(len, optlen))
3914 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3919 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3920 char __user *optval,
3924 struct sctp_association *asoc;
3925 struct list_head *pos;
3928 if (len < sizeof(sctp_assoc_t))
3931 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3934 /* For UDP-style sockets, id specifies the association to query. */
3935 asoc = sctp_id2assoc(sk, id);
3939 list_for_each(pos, &asoc->peer.transport_addr_list) {
3947 * Old API for getting list of peer addresses. Does not work for 32-bit
3948 * programs running on a 64-bit kernel
3950 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3951 char __user *optval,
3954 struct sctp_association *asoc;
3955 struct list_head *pos;
3957 struct sctp_getaddrs_old getaddrs;
3958 struct sctp_transport *from;
3960 union sctp_addr temp;
3961 struct sctp_sock *sp = sctp_sk(sk);
3964 if (len < sizeof(struct sctp_getaddrs_old))
3967 len = sizeof(struct sctp_getaddrs_old);
3969 if (copy_from_user(&getaddrs, optval, len))
3972 if (getaddrs.addr_num <= 0) return -EINVAL;
3974 /* For UDP-style sockets, id specifies the association to query. */
3975 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3979 to = (void __user *)getaddrs.addrs;
3980 list_for_each(pos, &asoc->peer.transport_addr_list) {
3981 from = list_entry(pos, struct sctp_transport, transports);
3982 memcpy(&temp, &from->ipaddr, sizeof(temp));
3983 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3984 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3985 if (copy_to_user(to, &temp, addrlen))
3989 if (cnt >= getaddrs.addr_num) break;
3991 getaddrs.addr_num = cnt;
3992 if (put_user(len, optlen))
3994 if (copy_to_user(optval, &getaddrs, len))
4000 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4001 char __user *optval, int __user *optlen)
4003 struct sctp_association *asoc;
4004 struct list_head *pos;
4006 struct sctp_getaddrs getaddrs;
4007 struct sctp_transport *from;
4009 union sctp_addr temp;
4010 struct sctp_sock *sp = sctp_sk(sk);
4015 if (len < sizeof(struct sctp_getaddrs))
4018 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4021 /* For UDP-style sockets, id specifies the association to query. */
4022 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4026 to = optval + offsetof(struct sctp_getaddrs,addrs);
4027 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4029 list_for_each(pos, &asoc->peer.transport_addr_list) {
4030 from = list_entry(pos, struct sctp_transport, transports);
4031 memcpy(&temp, &from->ipaddr, sizeof(temp));
4032 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4033 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4034 if (space_left < addrlen)
4036 if (copy_to_user(to, &temp, addrlen))
4040 space_left -= addrlen;
4043 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4045 bytes_copied = ((char __user *)to) - optval;
4046 if (put_user(bytes_copied, optlen))
4052 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4053 char __user *optval,
4057 struct sctp_bind_addr *bp;
4058 struct sctp_association *asoc;
4059 struct list_head *pos, *temp;
4060 struct sctp_sockaddr_entry *addr;
4061 rwlock_t *addr_lock;
4064 if (len < sizeof(sctp_assoc_t))
4067 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4071 * For UDP-style sockets, id specifies the association to query.
4072 * If the id field is set to the value '0' then the locally bound
4073 * addresses are returned without regard to any particular
4077 bp = &sctp_sk(sk)->ep->base.bind_addr;
4078 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4080 asoc = sctp_id2assoc(sk, id);
4083 bp = &asoc->base.bind_addr;
4084 addr_lock = &asoc->base.addr_lock;
4087 sctp_read_lock(addr_lock);
4089 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4090 * addresses from the global local address list.
4092 if (sctp_list_single_entry(&bp->address_list)) {
4093 addr = list_entry(bp->address_list.next,
4094 struct sctp_sockaddr_entry, list);
4095 if (sctp_is_any(&addr->a)) {
4096 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
4097 addr = list_entry(pos,
4098 struct sctp_sockaddr_entry,
4100 if ((PF_INET == sk->sk_family) &&
4101 (AF_INET6 == addr->a.sa.sa_family))
4111 list_for_each(pos, &bp->address_list) {
4116 sctp_read_unlock(addr_lock);
4120 /* Helper function that copies local addresses to user and returns the number
4121 * of addresses copied.
4123 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4124 int max_addrs, void *to,
4127 struct list_head *pos, *next;
4128 struct sctp_sockaddr_entry *addr;
4129 union sctp_addr temp;
4133 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4134 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4135 if ((PF_INET == sk->sk_family) &&
4136 (AF_INET6 == addr->a.sa.sa_family))
4138 memcpy(&temp, &addr->a, sizeof(temp));
4139 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4141 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4142 memcpy(to, &temp, addrlen);
4145 *bytes_copied += addrlen;
4147 if (cnt >= max_addrs) break;
4153 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4154 size_t space_left, int *bytes_copied)
4156 struct list_head *pos, *next;
4157 struct sctp_sockaddr_entry *addr;
4158 union sctp_addr temp;
4162 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4163 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4164 if ((PF_INET == sk->sk_family) &&
4165 (AF_INET6 == addr->a.sa.sa_family))
4167 memcpy(&temp, &addr->a, sizeof(temp));
4168 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4170 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4171 if (space_left < addrlen)
4173 memcpy(to, &temp, addrlen);
4177 space_left -= addrlen;
4178 *bytes_copied += addrlen;
4184 /* Old API for getting list of local addresses. Does not work for 32-bit
4185 * programs running on a 64-bit kernel
4187 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4188 char __user *optval, int __user *optlen)
4190 struct sctp_bind_addr *bp;
4191 struct sctp_association *asoc;
4192 struct list_head *pos;
4194 struct sctp_getaddrs_old getaddrs;
4195 struct sctp_sockaddr_entry *addr;
4197 union sctp_addr temp;
4198 struct sctp_sock *sp = sctp_sk(sk);
4200 rwlock_t *addr_lock;
4204 int bytes_copied = 0;
4206 if (len < sizeof(struct sctp_getaddrs_old))
4209 len = sizeof(struct sctp_getaddrs_old);
4210 if (copy_from_user(&getaddrs, optval, len))
4213 if (getaddrs.addr_num <= 0) return -EINVAL;
4215 * For UDP-style sockets, id specifies the association to query.
4216 * If the id field is set to the value '0' then the locally bound
4217 * addresses are returned without regard to any particular
4220 if (0 == getaddrs.assoc_id) {
4221 bp = &sctp_sk(sk)->ep->base.bind_addr;
4222 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4224 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4227 bp = &asoc->base.bind_addr;
4228 addr_lock = &asoc->base.addr_lock;
4231 to = getaddrs.addrs;
4233 /* Allocate space for a local instance of packed array to hold all
4234 * the data. We store addresses here first and then put write them
4235 * to the user in one shot.
4237 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4242 sctp_read_lock(addr_lock);
4244 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4245 * addresses from the global local address list.
4247 if (sctp_list_single_entry(&bp->address_list)) {
4248 addr = list_entry(bp->address_list.next,
4249 struct sctp_sockaddr_entry, list);
4250 if (sctp_is_any(&addr->a)) {
4251 cnt = sctp_copy_laddrs_old(sk, bp->port,
4253 addrs, &bytes_copied);
4259 list_for_each(pos, &bp->address_list) {
4260 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4261 memcpy(&temp, &addr->a, sizeof(temp));
4262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4263 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4264 memcpy(buf, &temp, addrlen);
4266 bytes_copied += addrlen;
4268 if (cnt >= getaddrs.addr_num) break;
4272 sctp_read_unlock(addr_lock);
4274 /* copy the entire address list into the user provided space */
4275 if (copy_to_user(to, addrs, bytes_copied)) {
4280 /* copy the leading structure back to user */
4281 getaddrs.addr_num = cnt;
4282 if (copy_to_user(optval, &getaddrs, len))
4290 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4291 char __user *optval, int __user *optlen)
4293 struct sctp_bind_addr *bp;
4294 struct sctp_association *asoc;
4295 struct list_head *pos;
4297 struct sctp_getaddrs getaddrs;
4298 struct sctp_sockaddr_entry *addr;
4300 union sctp_addr temp;
4301 struct sctp_sock *sp = sctp_sk(sk);
4303 rwlock_t *addr_lock;
4306 int bytes_copied = 0;
4310 if (len < sizeof(struct sctp_getaddrs))
4313 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4317 * For UDP-style sockets, id specifies the association to query.
4318 * If the id field is set to the value '0' then the locally bound
4319 * addresses are returned without regard to any particular
4322 if (0 == getaddrs.assoc_id) {
4323 bp = &sctp_sk(sk)->ep->base.bind_addr;
4324 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4326 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4329 bp = &asoc->base.bind_addr;
4330 addr_lock = &asoc->base.addr_lock;
4333 to = optval + offsetof(struct sctp_getaddrs,addrs);
4334 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4336 addrs = kmalloc(space_left, GFP_KERNEL);
4340 sctp_read_lock(addr_lock);
4342 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4343 * addresses from the global local address list.
4345 if (sctp_list_single_entry(&bp->address_list)) {
4346 addr = list_entry(bp->address_list.next,
4347 struct sctp_sockaddr_entry, list);
4348 if (sctp_is_any(&addr->a)) {
4349 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4350 space_left, &bytes_copied);
4360 list_for_each(pos, &bp->address_list) {
4361 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4362 memcpy(&temp, &addr->a, sizeof(temp));
4363 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4364 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4365 if (space_left < addrlen) {
4366 err = -ENOMEM; /*fixme: right error?*/
4369 memcpy(buf, &temp, addrlen);
4371 bytes_copied += addrlen;
4373 space_left -= addrlen;
4377 sctp_read_unlock(addr_lock);
4379 if (copy_to_user(to, addrs, bytes_copied)) {
4383 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4387 if (put_user(bytes_copied, optlen))
4394 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4396 * Requests that the local SCTP stack use the enclosed peer address as
4397 * the association primary. The enclosed address must be one of the
4398 * association peer's addresses.
4400 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4401 char __user *optval, int __user *optlen)
4403 struct sctp_prim prim;
4404 struct sctp_association *asoc;
4405 struct sctp_sock *sp = sctp_sk(sk);
4407 if (len < sizeof(struct sctp_prim))
4410 len = sizeof(struct sctp_prim);
4412 if (copy_from_user(&prim, optval, len))
4415 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4419 if (!asoc->peer.primary_path)
4422 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4423 asoc->peer.primary_path->af_specific->sockaddr_len);
4425 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4426 (union sctp_addr *)&prim.ssp_addr);
4428 if (put_user(len, optlen))
4430 if (copy_to_user(optval, &prim, len))
4437 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4439 * Requests that the local endpoint set the specified Adaptation Layer
4440 * Indication parameter for all future INIT and INIT-ACK exchanges.
4442 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4443 char __user *optval, int __user *optlen)
4445 struct sctp_setadaptation adaptation;
4447 if (len < sizeof(struct sctp_setadaptation))
4450 len = sizeof(struct sctp_setadaptation);
4452 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4454 if (put_user(len, optlen))
4456 if (copy_to_user(optval, &adaptation, len))
4464 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4466 * Applications that wish to use the sendto() system call may wish to
4467 * specify a default set of parameters that would normally be supplied
4468 * through the inclusion of ancillary data. This socket option allows
4469 * such an application to set the default sctp_sndrcvinfo structure.
4472 * The application that wishes to use this socket option simply passes
4473 * in to this call the sctp_sndrcvinfo structure defined in Section
4474 * 5.2.2) The input parameters accepted by this call include
4475 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4476 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4477 * to this call if the caller is using the UDP model.
4479 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4481 static int sctp_getsockopt_default_send_param(struct sock *sk,
4482 int len, char __user *optval,
4485 struct sctp_sndrcvinfo info;
4486 struct sctp_association *asoc;
4487 struct sctp_sock *sp = sctp_sk(sk);
4489 if (len < sizeof(struct sctp_sndrcvinfo))
4492 len = sizeof(struct sctp_sndrcvinfo);
4494 if (copy_from_user(&info, optval, len))
4497 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4498 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4502 info.sinfo_stream = asoc->default_stream;
4503 info.sinfo_flags = asoc->default_flags;
4504 info.sinfo_ppid = asoc->default_ppid;
4505 info.sinfo_context = asoc->default_context;
4506 info.sinfo_timetolive = asoc->default_timetolive;
4508 info.sinfo_stream = sp->default_stream;
4509 info.sinfo_flags = sp->default_flags;
4510 info.sinfo_ppid = sp->default_ppid;
4511 info.sinfo_context = sp->default_context;
4512 info.sinfo_timetolive = sp->default_timetolive;
4515 if (put_user(len, optlen))
4517 if (copy_to_user(optval, &info, len))
4525 * 7.1.5 SCTP_NODELAY
4527 * Turn on/off any Nagle-like algorithm. This means that packets are
4528 * generally sent as soon as possible and no unnecessary delays are
4529 * introduced, at the cost of more packets in the network. Expects an
4530 * integer boolean flag.
4533 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4534 char __user *optval, int __user *optlen)
4538 if (len < sizeof(int))
4542 val = (sctp_sk(sk)->nodelay == 1);
4543 if (put_user(len, optlen))
4545 if (copy_to_user(optval, &val, len))
4552 * 7.1.1 SCTP_RTOINFO
4554 * The protocol parameters used to initialize and bound retransmission
4555 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4556 * and modify these parameters.
4557 * All parameters are time values, in milliseconds. A value of 0, when
4558 * modifying the parameters, indicates that the current value should not
4562 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4563 char __user *optval,
4564 int __user *optlen) {
4565 struct sctp_rtoinfo rtoinfo;
4566 struct sctp_association *asoc;
4568 if (len < sizeof (struct sctp_rtoinfo))
4571 len = sizeof(struct sctp_rtoinfo);
4573 if (copy_from_user(&rtoinfo, optval, len))
4576 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4578 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4581 /* Values corresponding to the specific association. */
4583 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4584 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4585 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4587 /* Values corresponding to the endpoint. */
4588 struct sctp_sock *sp = sctp_sk(sk);
4590 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4591 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4592 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4595 if (put_user(len, optlen))
4598 if (copy_to_user(optval, &rtoinfo, len))
4606 * 7.1.2 SCTP_ASSOCINFO
4608 * This option is used to tune the maximum retransmission attempts
4609 * of the association.
4610 * Returns an error if the new association retransmission value is
4611 * greater than the sum of the retransmission value of the peer.
4612 * See [SCTP] for more information.
4615 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4616 char __user *optval,
4620 struct sctp_assocparams assocparams;
4621 struct sctp_association *asoc;
4622 struct list_head *pos;
4625 if (len < sizeof (struct sctp_assocparams))
4628 len = sizeof(struct sctp_assocparams);
4630 if (copy_from_user(&assocparams, optval, len))
4633 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4635 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4638 /* Values correspoinding to the specific association */
4640 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4641 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4642 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4643 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4645 (asoc->cookie_life.tv_usec
4648 list_for_each(pos, &asoc->peer.transport_addr_list) {
4652 assocparams.sasoc_number_peer_destinations = cnt;
4654 /* Values corresponding to the endpoint */
4655 struct sctp_sock *sp = sctp_sk(sk);
4657 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4658 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4659 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4660 assocparams.sasoc_cookie_life =
4661 sp->assocparams.sasoc_cookie_life;
4662 assocparams.sasoc_number_peer_destinations =
4664 sasoc_number_peer_destinations;
4667 if (put_user(len, optlen))
4670 if (copy_to_user(optval, &assocparams, len))
4677 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4679 * This socket option is a boolean flag which turns on or off mapped V4
4680 * addresses. If this option is turned on and the socket is type
4681 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4682 * If this option is turned off, then no mapping will be done of V4
4683 * addresses and a user will receive both PF_INET6 and PF_INET type
4684 * addresses on the socket.
4686 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4687 char __user *optval, int __user *optlen)
4690 struct sctp_sock *sp = sctp_sk(sk);
4692 if (len < sizeof(int))
4697 if (put_user(len, optlen))
4699 if (copy_to_user(optval, &val, len))
4706 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4707 * (chapter and verse is quoted at sctp_setsockopt_context())
4709 static int sctp_getsockopt_context(struct sock *sk, int len,
4710 char __user *optval, int __user *optlen)
4712 struct sctp_assoc_value params;
4713 struct sctp_sock *sp;
4714 struct sctp_association *asoc;
4716 if (len < sizeof(struct sctp_assoc_value))
4719 len = sizeof(struct sctp_assoc_value);
4721 if (copy_from_user(¶ms, optval, len))
4726 if (params.assoc_id != 0) {
4727 asoc = sctp_id2assoc(sk, params.assoc_id);
4730 params.assoc_value = asoc->default_rcv_context;
4732 params.assoc_value = sp->default_rcv_context;
4735 if (put_user(len, optlen))
4737 if (copy_to_user(optval, ¶ms, len))
4744 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4746 * This socket option specifies the maximum size to put in any outgoing
4747 * SCTP chunk. If a message is larger than this size it will be
4748 * fragmented by SCTP into the specified size. Note that the underlying
4749 * SCTP implementation may fragment into smaller sized chunks when the
4750 * PMTU of the underlying association is smaller than the value set by
4753 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4754 char __user *optval, int __user *optlen)
4758 if (len < sizeof(int))
4763 val = sctp_sk(sk)->user_frag;
4764 if (put_user(len, optlen))
4766 if (copy_to_user(optval, &val, len))
4773 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4774 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4776 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4777 char __user *optval, int __user *optlen)
4781 if (len < sizeof(int))
4786 val = sctp_sk(sk)->frag_interleave;
4787 if (put_user(len, optlen))
4789 if (copy_to_user(optval, &val, len))
4796 * 7.1.25. Set or Get the sctp partial delivery point
4797 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4799 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4800 char __user *optval,
4805 if (len < sizeof(u32))
4810 val = sctp_sk(sk)->pd_point;
4811 if (put_user(len, optlen))
4813 if (copy_to_user(optval, &val, len))
4820 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
4821 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
4823 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
4824 char __user *optval,
4829 if (len < sizeof(int))
4834 val = sctp_sk(sk)->max_burst;
4835 if (put_user(len, optlen))
4837 if (copy_to_user(optval, &val, len))
4843 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4844 char __user *optval, int __user *optlen)
4849 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4852 /* I can hardly begin to describe how wrong this is. This is
4853 * so broken as to be worse than useless. The API draft
4854 * REALLY is NOT helpful here... I am not convinced that the
4855 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4856 * are at all well-founded.
4858 if (level != SOL_SCTP) {
4859 struct sctp_af *af = sctp_sk(sk)->pf->af;
4861 retval = af->getsockopt(sk, level, optname, optval, optlen);
4865 if (get_user(len, optlen))
4872 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4874 case SCTP_DISABLE_FRAGMENTS:
4875 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4879 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4881 case SCTP_AUTOCLOSE:
4882 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4884 case SCTP_SOCKOPT_PEELOFF:
4885 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4887 case SCTP_PEER_ADDR_PARAMS:
4888 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4891 case SCTP_DELAYED_ACK_TIME:
4892 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4896 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4898 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4899 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4902 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4903 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4906 case SCTP_GET_PEER_ADDRS_OLD:
4907 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4910 case SCTP_GET_LOCAL_ADDRS_OLD:
4911 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4914 case SCTP_GET_PEER_ADDRS:
4915 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4918 case SCTP_GET_LOCAL_ADDRS:
4919 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4922 case SCTP_DEFAULT_SEND_PARAM:
4923 retval = sctp_getsockopt_default_send_param(sk, len,
4926 case SCTP_PRIMARY_ADDR:
4927 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4930 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4933 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4935 case SCTP_ASSOCINFO:
4936 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4938 case SCTP_I_WANT_MAPPED_V4_ADDR:
4939 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4942 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4944 case SCTP_GET_PEER_ADDR_INFO:
4945 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4948 case SCTP_ADAPTATION_LAYER:
4949 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4953 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4955 case SCTP_FRAGMENT_INTERLEAVE:
4956 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4959 case SCTP_PARTIAL_DELIVERY_POINT:
4960 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
4963 case SCTP_MAX_BURST:
4964 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
4967 retval = -ENOPROTOOPT;
4971 sctp_release_sock(sk);
4975 static void sctp_hash(struct sock *sk)
4980 static void sctp_unhash(struct sock *sk)
4985 /* Check if port is acceptable. Possibly find first available port.
4987 * The port hash table (contained in the 'global' SCTP protocol storage
4988 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4989 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4990 * list (the list number is the port number hashed out, so as you
4991 * would expect from a hash function, all the ports in a given list have
4992 * such a number that hashes out to the same list number; you were
4993 * expecting that, right?); so each list has a set of ports, with a
4994 * link to the socket (struct sock) that uses it, the port number and
4995 * a fastreuse flag (FIXME: NPI ipg).
4997 static struct sctp_bind_bucket *sctp_bucket_create(
4998 struct sctp_bind_hashbucket *head, unsigned short snum);
5000 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5002 struct sctp_bind_hashbucket *head; /* hash list */
5003 struct sctp_bind_bucket *pp; /* hash list port iterator */
5004 unsigned short snum;
5007 snum = ntohs(addr->v4.sin_port);
5009 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5010 sctp_local_bh_disable();
5013 /* Search for an available port.
5015 * 'sctp_port_rover' was the last port assigned, so
5016 * we start to search from 'sctp_port_rover +
5017 * 1'. What we do is first check if port 'rover' is
5018 * already in the hash table; if not, we use that; if
5019 * it is, we try next.
5021 int low = sysctl_local_port_range[0];
5022 int high = sysctl_local_port_range[1];
5023 int remaining = (high - low) + 1;
5027 sctp_spin_lock(&sctp_port_alloc_lock);
5028 rover = sctp_port_rover;
5031 if ((rover < low) || (rover > high))
5033 index = sctp_phashfn(rover);
5034 head = &sctp_port_hashtable[index];
5035 sctp_spin_lock(&head->lock);
5036 for (pp = head->chain; pp; pp = pp->next)
5037 if (pp->port == rover)
5041 sctp_spin_unlock(&head->lock);
5042 } while (--remaining > 0);
5043 sctp_port_rover = rover;
5044 sctp_spin_unlock(&sctp_port_alloc_lock);
5046 /* Exhausted local port range during search? */
5051 /* OK, here is the one we will use. HEAD (the port
5052 * hash table list entry) is non-NULL and we hold it's
5057 /* We are given an specific port number; we verify
5058 * that it is not being used. If it is used, we will
5059 * exahust the search in the hash list corresponding
5060 * to the port number (snum) - we detect that with the
5061 * port iterator, pp being NULL.
5063 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5064 sctp_spin_lock(&head->lock);
5065 for (pp = head->chain; pp; pp = pp->next) {
5066 if (pp->port == snum)
5073 if (!hlist_empty(&pp->owner)) {
5074 /* We had a port hash table hit - there is an
5075 * available port (pp != NULL) and it is being
5076 * used by other socket (pp->owner not empty); that other
5077 * socket is going to be sk2.
5079 int reuse = sk->sk_reuse;
5081 struct hlist_node *node;
5083 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5084 if (pp->fastreuse && sk->sk_reuse &&
5085 sk->sk_state != SCTP_SS_LISTENING)
5088 /* Run through the list of sockets bound to the port
5089 * (pp->port) [via the pointers bind_next and
5090 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5091 * we get the endpoint they describe and run through
5092 * the endpoint's list of IP (v4 or v6) addresses,
5093 * comparing each of the addresses with the address of
5094 * the socket sk. If we find a match, then that means
5095 * that this port/socket (sk) combination are already
5098 sk_for_each_bound(sk2, node, &pp->owner) {
5099 struct sctp_endpoint *ep2;
5100 ep2 = sctp_sk(sk2)->ep;
5102 if (reuse && sk2->sk_reuse &&
5103 sk2->sk_state != SCTP_SS_LISTENING)
5106 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5112 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5115 /* If there was a hash table miss, create a new port. */
5117 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5120 /* In either case (hit or miss), make sure fastreuse is 1 only
5121 * if sk->sk_reuse is too (that is, if the caller requested
5122 * SO_REUSEADDR on this socket -sk-).
5124 if (hlist_empty(&pp->owner)) {
5125 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5129 } else if (pp->fastreuse &&
5130 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5133 /* We are set, so fill up all the data in the hash table
5134 * entry, tie the socket list information with the rest of the
5135 * sockets FIXME: Blurry, NPI (ipg).
5138 if (!sctp_sk(sk)->bind_hash) {
5139 inet_sk(sk)->num = snum;
5140 sk_add_bind_node(sk, &pp->owner);
5141 sctp_sk(sk)->bind_hash = pp;
5146 sctp_spin_unlock(&head->lock);
5149 sctp_local_bh_enable();
5153 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5154 * port is requested.
5156 static int sctp_get_port(struct sock *sk, unsigned short snum)
5159 union sctp_addr addr;
5160 struct sctp_af *af = sctp_sk(sk)->pf->af;
5162 /* Set up a dummy address struct from the sk. */
5163 af->from_sk(&addr, sk);
5164 addr.v4.sin_port = htons(snum);
5166 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5167 ret = sctp_get_port_local(sk, &addr);
5169 return (ret ? 1 : 0);
5173 * 3.1.3 listen() - UDP Style Syntax
5175 * By default, new associations are not accepted for UDP style sockets.
5176 * An application uses listen() to mark a socket as being able to
5177 * accept new associations.
5179 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5181 struct sctp_sock *sp = sctp_sk(sk);
5182 struct sctp_endpoint *ep = sp->ep;
5184 /* Only UDP style sockets that are not peeled off are allowed to
5187 if (!sctp_style(sk, UDP))
5190 /* If backlog is zero, disable listening. */
5192 if (sctp_sstate(sk, CLOSED))
5195 sctp_unhash_endpoint(ep);
5196 sk->sk_state = SCTP_SS_CLOSED;
5199 /* Return if we are already listening. */
5200 if (sctp_sstate(sk, LISTENING))
5204 * If a bind() or sctp_bindx() is not called prior to a listen()
5205 * call that allows new associations to be accepted, the system
5206 * picks an ephemeral port and will choose an address set equivalent
5207 * to binding with a wildcard address.
5209 * This is not currently spelled out in the SCTP sockets
5210 * extensions draft, but follows the practice as seen in TCP
5213 * Additionally, turn off fastreuse flag since we are not listening
5215 sk->sk_state = SCTP_SS_LISTENING;
5216 if (!ep->base.bind_addr.port) {
5217 if (sctp_autobind(sk))
5220 sctp_sk(sk)->bind_hash->fastreuse = 0;
5222 sctp_hash_endpoint(ep);
5227 * 4.1.3 listen() - TCP Style Syntax
5229 * Applications uses listen() to ready the SCTP endpoint for accepting
5230 * inbound associations.
5232 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5234 struct sctp_sock *sp = sctp_sk(sk);
5235 struct sctp_endpoint *ep = sp->ep;
5237 /* If backlog is zero, disable listening. */
5239 if (sctp_sstate(sk, CLOSED))
5242 sctp_unhash_endpoint(ep);
5243 sk->sk_state = SCTP_SS_CLOSED;
5246 if (sctp_sstate(sk, LISTENING))
5250 * If a bind() or sctp_bindx() is not called prior to a listen()
5251 * call that allows new associations to be accepted, the system
5252 * picks an ephemeral port and will choose an address set equivalent
5253 * to binding with a wildcard address.
5255 * This is not currently spelled out in the SCTP sockets
5256 * extensions draft, but follows the practice as seen in TCP
5259 sk->sk_state = SCTP_SS_LISTENING;
5260 if (!ep->base.bind_addr.port) {
5261 if (sctp_autobind(sk))
5264 sctp_sk(sk)->bind_hash->fastreuse = 0;
5266 sk->sk_max_ack_backlog = backlog;
5267 sctp_hash_endpoint(ep);
5272 * Move a socket to LISTENING state.
5274 int sctp_inet_listen(struct socket *sock, int backlog)
5276 struct sock *sk = sock->sk;
5277 struct crypto_hash *tfm = NULL;
5280 if (unlikely(backlog < 0))
5285 if (sock->state != SS_UNCONNECTED)
5288 /* Allocate HMAC for generating cookie. */
5289 if (sctp_hmac_alg) {
5290 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5292 if (net_ratelimit()) {
5294 "SCTP: failed to load transform for %s: %ld\n",
5295 sctp_hmac_alg, PTR_ERR(tfm));
5302 switch (sock->type) {
5303 case SOCK_SEQPACKET:
5304 err = sctp_seqpacket_listen(sk, backlog);
5307 err = sctp_stream_listen(sk, backlog);
5316 /* Store away the transform reference. */
5317 sctp_sk(sk)->hmac = tfm;
5319 sctp_release_sock(sk);
5322 crypto_free_hash(tfm);
5327 * This function is done by modeling the current datagram_poll() and the
5328 * tcp_poll(). Note that, based on these implementations, we don't
5329 * lock the socket in this function, even though it seems that,
5330 * ideally, locking or some other mechanisms can be used to ensure
5331 * the integrity of the counters (sndbuf and wmem_alloc) used
5332 * in this place. We assume that we don't need locks either until proven
5335 * Another thing to note is that we include the Async I/O support
5336 * here, again, by modeling the current TCP/UDP code. We don't have
5337 * a good way to test with it yet.
5339 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5341 struct sock *sk = sock->sk;
5342 struct sctp_sock *sp = sctp_sk(sk);
5345 poll_wait(file, sk->sk_sleep, wait);
5347 /* A TCP-style listening socket becomes readable when the accept queue
5350 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5351 return (!list_empty(&sp->ep->asocs)) ?
5352 (POLLIN | POLLRDNORM) : 0;
5356 /* Is there any exceptional events? */
5357 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5359 if (sk->sk_shutdown & RCV_SHUTDOWN)
5361 if (sk->sk_shutdown == SHUTDOWN_MASK)
5364 /* Is it readable? Reconsider this code with TCP-style support. */
5365 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5366 (sk->sk_shutdown & RCV_SHUTDOWN))
5367 mask |= POLLIN | POLLRDNORM;
5369 /* The association is either gone or not ready. */
5370 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5373 /* Is it writable? */
5374 if (sctp_writeable(sk)) {
5375 mask |= POLLOUT | POLLWRNORM;
5377 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5379 * Since the socket is not locked, the buffer
5380 * might be made available after the writeable check and
5381 * before the bit is set. This could cause a lost I/O
5382 * signal. tcp_poll() has a race breaker for this race
5383 * condition. Based on their implementation, we put
5384 * in the following code to cover it as well.
5386 if (sctp_writeable(sk))
5387 mask |= POLLOUT | POLLWRNORM;
5392 /********************************************************************
5393 * 2nd Level Abstractions
5394 ********************************************************************/
5396 static struct sctp_bind_bucket *sctp_bucket_create(
5397 struct sctp_bind_hashbucket *head, unsigned short snum)
5399 struct sctp_bind_bucket *pp;
5401 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5402 SCTP_DBG_OBJCNT_INC(bind_bucket);
5406 INIT_HLIST_HEAD(&pp->owner);
5407 if ((pp->next = head->chain) != NULL)
5408 pp->next->pprev = &pp->next;
5410 pp->pprev = &head->chain;
5415 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5416 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5418 if (pp && hlist_empty(&pp->owner)) {
5420 pp->next->pprev = pp->pprev;
5421 *(pp->pprev) = pp->next;
5422 kmem_cache_free(sctp_bucket_cachep, pp);
5423 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5427 /* Release this socket's reference to a local port. */
5428 static inline void __sctp_put_port(struct sock *sk)
5430 struct sctp_bind_hashbucket *head =
5431 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5432 struct sctp_bind_bucket *pp;
5434 sctp_spin_lock(&head->lock);
5435 pp = sctp_sk(sk)->bind_hash;
5436 __sk_del_bind_node(sk);
5437 sctp_sk(sk)->bind_hash = NULL;
5438 inet_sk(sk)->num = 0;
5439 sctp_bucket_destroy(pp);
5440 sctp_spin_unlock(&head->lock);
5443 void sctp_put_port(struct sock *sk)
5445 sctp_local_bh_disable();
5446 __sctp_put_port(sk);
5447 sctp_local_bh_enable();
5451 * The system picks an ephemeral port and choose an address set equivalent
5452 * to binding with a wildcard address.
5453 * One of those addresses will be the primary address for the association.
5454 * This automatically enables the multihoming capability of SCTP.
5456 static int sctp_autobind(struct sock *sk)
5458 union sctp_addr autoaddr;
5462 /* Initialize a local sockaddr structure to INADDR_ANY. */
5463 af = sctp_sk(sk)->pf->af;
5465 port = htons(inet_sk(sk)->num);
5466 af->inaddr_any(&autoaddr, port);
5468 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5471 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5474 * 4.2 The cmsghdr Structure *
5476 * When ancillary data is sent or received, any number of ancillary data
5477 * objects can be specified by the msg_control and msg_controllen members of
5478 * the msghdr structure, because each object is preceded by
5479 * a cmsghdr structure defining the object's length (the cmsg_len member).
5480 * Historically Berkeley-derived implementations have passed only one object
5481 * at a time, but this API allows multiple objects to be
5482 * passed in a single call to sendmsg() or recvmsg(). The following example
5483 * shows two ancillary data objects in a control buffer.
5485 * |<--------------------------- msg_controllen -------------------------->|
5488 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5490 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5493 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5495 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5498 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5499 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5501 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5503 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5510 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5511 sctp_cmsgs_t *cmsgs)
5513 struct cmsghdr *cmsg;
5515 for (cmsg = CMSG_FIRSTHDR(msg);
5517 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5518 if (!CMSG_OK(msg, cmsg))
5521 /* Should we parse this header or ignore? */
5522 if (cmsg->cmsg_level != IPPROTO_SCTP)
5525 /* Strictly check lengths following example in SCM code. */
5526 switch (cmsg->cmsg_type) {
5528 /* SCTP Socket API Extension
5529 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5531 * This cmsghdr structure provides information for
5532 * initializing new SCTP associations with sendmsg().
5533 * The SCTP_INITMSG socket option uses this same data
5534 * structure. This structure is not used for
5537 * cmsg_level cmsg_type cmsg_data[]
5538 * ------------ ------------ ----------------------
5539 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5541 if (cmsg->cmsg_len !=
5542 CMSG_LEN(sizeof(struct sctp_initmsg)))
5544 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5548 /* SCTP Socket API Extension
5549 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5551 * This cmsghdr structure specifies SCTP options for
5552 * sendmsg() and describes SCTP header information
5553 * about a received message through recvmsg().
5555 * cmsg_level cmsg_type cmsg_data[]
5556 * ------------ ------------ ----------------------
5557 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5559 if (cmsg->cmsg_len !=
5560 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5564 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5566 /* Minimally, validate the sinfo_flags. */
5567 if (cmsgs->info->sinfo_flags &
5568 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5569 SCTP_ABORT | SCTP_EOF))
5581 * Wait for a packet..
5582 * Note: This function is the same function as in core/datagram.c
5583 * with a few modifications to make lksctp work.
5585 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5590 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5592 /* Socket errors? */
5593 error = sock_error(sk);
5597 if (!skb_queue_empty(&sk->sk_receive_queue))
5600 /* Socket shut down? */
5601 if (sk->sk_shutdown & RCV_SHUTDOWN)
5604 /* Sequenced packets can come disconnected. If so we report the
5609 /* Is there a good reason to think that we may receive some data? */
5610 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5613 /* Handle signals. */
5614 if (signal_pending(current))
5617 /* Let another process have a go. Since we are going to sleep
5618 * anyway. Note: This may cause odd behaviors if the message
5619 * does not fit in the user's buffer, but this seems to be the
5620 * only way to honor MSG_DONTWAIT realistically.
5622 sctp_release_sock(sk);
5623 *timeo_p = schedule_timeout(*timeo_p);
5627 finish_wait(sk->sk_sleep, &wait);
5631 error = sock_intr_errno(*timeo_p);
5634 finish_wait(sk->sk_sleep, &wait);
5639 /* Receive a datagram.
5640 * Note: This is pretty much the same routine as in core/datagram.c
5641 * with a few changes to make lksctp work.
5643 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5644 int noblock, int *err)
5647 struct sk_buff *skb;
5650 timeo = sock_rcvtimeo(sk, noblock);
5652 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5653 timeo, MAX_SCHEDULE_TIMEOUT);
5656 /* Again only user level code calls this function,
5657 * so nothing interrupt level
5658 * will suddenly eat the receive_queue.
5660 * Look at current nfs client by the way...
5661 * However, this function was corrent in any case. 8)
5663 if (flags & MSG_PEEK) {
5664 spin_lock_bh(&sk->sk_receive_queue.lock);
5665 skb = skb_peek(&sk->sk_receive_queue);
5667 atomic_inc(&skb->users);
5668 spin_unlock_bh(&sk->sk_receive_queue.lock);
5670 skb = skb_dequeue(&sk->sk_receive_queue);
5676 /* Caller is allowed not to check sk->sk_err before calling. */
5677 error = sock_error(sk);
5681 if (sk->sk_shutdown & RCV_SHUTDOWN)
5684 /* User doesn't want to wait. */
5688 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5697 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5698 static void __sctp_write_space(struct sctp_association *asoc)
5700 struct sock *sk = asoc->base.sk;
5701 struct socket *sock = sk->sk_socket;
5703 if ((sctp_wspace(asoc) > 0) && sock) {
5704 if (waitqueue_active(&asoc->wait))
5705 wake_up_interruptible(&asoc->wait);
5707 if (sctp_writeable(sk)) {
5708 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5709 wake_up_interruptible(sk->sk_sleep);
5711 /* Note that we try to include the Async I/O support
5712 * here by modeling from the current TCP/UDP code.
5713 * We have not tested with it yet.
5715 if (sock->fasync_list &&
5716 !(sk->sk_shutdown & SEND_SHUTDOWN))
5717 sock_wake_async(sock, 2, POLL_OUT);
5722 /* Do accounting for the sndbuf space.
5723 * Decrement the used sndbuf space of the corresponding association by the
5724 * data size which was just transmitted(freed).
5726 static void sctp_wfree(struct sk_buff *skb)
5728 struct sctp_association *asoc;
5729 struct sctp_chunk *chunk;
5732 /* Get the saved chunk pointer. */
5733 chunk = *((struct sctp_chunk **)(skb->cb));
5736 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5737 sizeof(struct sk_buff) +
5738 sizeof(struct sctp_chunk);
5740 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5743 __sctp_write_space(asoc);
5745 sctp_association_put(asoc);
5748 /* Do accounting for the receive space on the socket.
5749 * Accounting for the association is done in ulpevent.c
5750 * We set this as a destructor for the cloned data skbs so that
5751 * accounting is done at the correct time.
5753 void sctp_sock_rfree(struct sk_buff *skb)
5755 struct sock *sk = skb->sk;
5756 struct sctp_ulpevent *event = sctp_skb2event(skb);
5758 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5762 /* Helper function to wait for space in the sndbuf. */
5763 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5766 struct sock *sk = asoc->base.sk;
5768 long current_timeo = *timeo_p;
5771 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5772 asoc, (long)(*timeo_p), msg_len);
5774 /* Increment the association's refcnt. */
5775 sctp_association_hold(asoc);
5777 /* Wait on the association specific sndbuf space. */
5779 prepare_to_wait_exclusive(&asoc->wait, &wait,
5780 TASK_INTERRUPTIBLE);
5783 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5786 if (signal_pending(current))
5787 goto do_interrupted;
5788 if (msg_len <= sctp_wspace(asoc))
5791 /* Let another process have a go. Since we are going
5794 sctp_release_sock(sk);
5795 current_timeo = schedule_timeout(current_timeo);
5796 BUG_ON(sk != asoc->base.sk);
5799 *timeo_p = current_timeo;
5803 finish_wait(&asoc->wait, &wait);
5805 /* Release the association's refcnt. */
5806 sctp_association_put(asoc);
5815 err = sock_intr_errno(*timeo_p);
5823 /* If socket sndbuf has changed, wake up all per association waiters. */
5824 void sctp_write_space(struct sock *sk)
5826 struct sctp_association *asoc;
5827 struct list_head *pos;
5829 /* Wake up the tasks in each wait queue. */
5830 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5831 asoc = list_entry(pos, struct sctp_association, asocs);
5832 __sctp_write_space(asoc);
5836 /* Is there any sndbuf space available on the socket?
5838 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5839 * associations on the same socket. For a UDP-style socket with
5840 * multiple associations, it is possible for it to be "unwriteable"
5841 * prematurely. I assume that this is acceptable because
5842 * a premature "unwriteable" is better than an accidental "writeable" which
5843 * would cause an unwanted block under certain circumstances. For the 1-1
5844 * UDP-style sockets or TCP-style sockets, this code should work.
5847 static int sctp_writeable(struct sock *sk)
5851 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5857 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5858 * returns immediately with EINPROGRESS.
5860 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5862 struct sock *sk = asoc->base.sk;
5864 long current_timeo = *timeo_p;
5867 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5870 /* Increment the association's refcnt. */
5871 sctp_association_hold(asoc);
5874 prepare_to_wait_exclusive(&asoc->wait, &wait,
5875 TASK_INTERRUPTIBLE);
5878 if (sk->sk_shutdown & RCV_SHUTDOWN)
5880 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5883 if (signal_pending(current))
5884 goto do_interrupted;
5886 if (sctp_state(asoc, ESTABLISHED))
5889 /* Let another process have a go. Since we are going
5892 sctp_release_sock(sk);
5893 current_timeo = schedule_timeout(current_timeo);
5896 *timeo_p = current_timeo;
5900 finish_wait(&asoc->wait, &wait);
5902 /* Release the association's refcnt. */
5903 sctp_association_put(asoc);
5908 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5911 err = -ECONNREFUSED;
5915 err = sock_intr_errno(*timeo_p);
5923 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5925 struct sctp_endpoint *ep;
5929 ep = sctp_sk(sk)->ep;
5933 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5934 TASK_INTERRUPTIBLE);
5936 if (list_empty(&ep->asocs)) {
5937 sctp_release_sock(sk);
5938 timeo = schedule_timeout(timeo);
5943 if (!sctp_sstate(sk, LISTENING))
5947 if (!list_empty(&ep->asocs))
5950 err = sock_intr_errno(timeo);
5951 if (signal_pending(current))
5959 finish_wait(sk->sk_sleep, &wait);
5964 static void sctp_wait_for_close(struct sock *sk, long timeout)
5969 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5970 if (list_empty(&sctp_sk(sk)->ep->asocs))
5972 sctp_release_sock(sk);
5973 timeout = schedule_timeout(timeout);
5975 } while (!signal_pending(current) && timeout);
5977 finish_wait(sk->sk_sleep, &wait);
5980 static void sctp_sock_rfree_frag(struct sk_buff *skb)
5982 struct sk_buff *frag;
5987 /* Don't forget the fragments. */
5988 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5989 sctp_sock_rfree_frag(frag);
5992 sctp_sock_rfree(skb);
5995 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
5997 struct sk_buff *frag;
6002 /* Don't forget the fragments. */
6003 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6004 sctp_skb_set_owner_r_frag(frag, sk);
6007 sctp_skb_set_owner_r(skb, sk);
6010 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6011 * and its messages to the newsk.
6013 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6014 struct sctp_association *assoc,
6015 sctp_socket_type_t type)
6017 struct sctp_sock *oldsp = sctp_sk(oldsk);
6018 struct sctp_sock *newsp = sctp_sk(newsk);
6019 struct sctp_bind_bucket *pp; /* hash list port iterator */
6020 struct sctp_endpoint *newep = newsp->ep;
6021 struct sk_buff *skb, *tmp;
6022 struct sctp_ulpevent *event;
6025 /* Migrate socket buffer sizes and all the socket level options to the
6028 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6029 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6030 /* Brute force copy old sctp opt. */
6031 inet_sk_copy_descendant(newsk, oldsk);
6033 /* Restore the ep value that was overwritten with the above structure
6039 /* Hook this new socket in to the bind_hash list. */
6040 pp = sctp_sk(oldsk)->bind_hash;
6041 sk_add_bind_node(newsk, &pp->owner);
6042 sctp_sk(newsk)->bind_hash = pp;
6043 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6045 /* Copy the bind_addr list from the original endpoint to the new
6046 * endpoint so that we can handle restarts properly
6048 if (PF_INET6 == assoc->base.sk->sk_family)
6049 flags = SCTP_ADDR6_ALLOWED;
6050 if (assoc->peer.ipv4_address)
6051 flags |= SCTP_ADDR4_PEERSUPP;
6052 if (assoc->peer.ipv6_address)
6053 flags |= SCTP_ADDR6_PEERSUPP;
6054 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
6055 &oldsp->ep->base.bind_addr,
6056 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
6058 /* Move any messages in the old socket's receive queue that are for the
6059 * peeled off association to the new socket's receive queue.
6061 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6062 event = sctp_skb2event(skb);
6063 if (event->asoc == assoc) {
6064 sctp_sock_rfree_frag(skb);
6065 __skb_unlink(skb, &oldsk->sk_receive_queue);
6066 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6067 sctp_skb_set_owner_r_frag(skb, newsk);
6071 /* Clean up any messages pending delivery due to partial
6072 * delivery. Three cases:
6073 * 1) No partial deliver; no work.
6074 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6075 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6077 skb_queue_head_init(&newsp->pd_lobby);
6078 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6080 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6081 struct sk_buff_head *queue;
6083 /* Decide which queue to move pd_lobby skbs to. */
6084 if (assoc->ulpq.pd_mode) {
6085 queue = &newsp->pd_lobby;
6087 queue = &newsk->sk_receive_queue;
6089 /* Walk through the pd_lobby, looking for skbs that
6090 * need moved to the new socket.
6092 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6093 event = sctp_skb2event(skb);
6094 if (event->asoc == assoc) {
6095 sctp_sock_rfree_frag(skb);
6096 __skb_unlink(skb, &oldsp->pd_lobby);
6097 __skb_queue_tail(queue, skb);
6098 sctp_skb_set_owner_r_frag(skb, newsk);
6102 /* Clear up any skbs waiting for the partial
6103 * delivery to finish.
6105 if (assoc->ulpq.pd_mode)
6106 sctp_clear_pd(oldsk, NULL);
6110 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6111 sctp_sock_rfree_frag(skb);
6112 sctp_skb_set_owner_r_frag(skb, newsk);
6115 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6116 sctp_sock_rfree_frag(skb);
6117 sctp_skb_set_owner_r_frag(skb, newsk);
6120 /* Set the type of socket to indicate that it is peeled off from the
6121 * original UDP-style socket or created with the accept() call on a
6122 * TCP-style socket..
6126 /* Mark the new socket "in-use" by the user so that any packets
6127 * that may arrive on the association after we've moved it are
6128 * queued to the backlog. This prevents a potential race between
6129 * backlog processing on the old socket and new-packet processing
6130 * on the new socket.
6132 * The caller has just allocated newsk so we can guarantee that other
6133 * paths won't try to lock it and then oldsk.
6135 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6136 sctp_assoc_migrate(assoc, newsk);
6138 /* If the association on the newsk is already closed before accept()
6139 * is called, set RCV_SHUTDOWN flag.
6141 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6142 newsk->sk_shutdown |= RCV_SHUTDOWN;
6144 newsk->sk_state = SCTP_SS_ESTABLISHED;
6145 sctp_release_sock(newsk);
6148 /* This proto struct describes the ULP interface for SCTP. */
6149 struct proto sctp_prot = {
6151 .owner = THIS_MODULE,
6152 .close = sctp_close,
6153 .connect = sctp_connect,
6154 .disconnect = sctp_disconnect,
6155 .accept = sctp_accept,
6156 .ioctl = sctp_ioctl,
6157 .init = sctp_init_sock,
6158 .destroy = sctp_destroy_sock,
6159 .shutdown = sctp_shutdown,
6160 .setsockopt = sctp_setsockopt,
6161 .getsockopt = sctp_getsockopt,
6162 .sendmsg = sctp_sendmsg,
6163 .recvmsg = sctp_recvmsg,
6165 .backlog_rcv = sctp_backlog_rcv,
6167 .unhash = sctp_unhash,
6168 .get_port = sctp_get_port,
6169 .obj_size = sizeof(struct sctp_sock),
6172 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6173 struct proto sctpv6_prot = {
6175 .owner = THIS_MODULE,
6176 .close = sctp_close,
6177 .connect = sctp_connect,
6178 .disconnect = sctp_disconnect,
6179 .accept = sctp_accept,
6180 .ioctl = sctp_ioctl,
6181 .init = sctp_init_sock,
6182 .destroy = sctp_destroy_sock,
6183 .shutdown = sctp_shutdown,
6184 .setsockopt = sctp_setsockopt,
6185 .getsockopt = sctp_getsockopt,
6186 .sendmsg = sctp_sendmsg,
6187 .recvmsg = sctp_recvmsg,
6189 .backlog_rcv = sctp_backlog_rcv,
6191 .unhash = sctp_unhash,
6192 .get_port = sctp_get_port,
6193 .obj_size = sizeof(struct sctp6_sock),
6195 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */