1 /* SCTP kernel 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 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 * This SCTP 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 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
73 #include <net/route.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110 extern struct kmem_cache *sctp_bucket_cachep;
111 extern int sysctl_sctp_mem[3];
112 extern int sysctl_sctp_rmem[3];
113 extern int sysctl_sctp_wmem[3];
115 static int sctp_memory_pressure;
116 static atomic_t sctp_memory_allocated;
117 struct percpu_counter sctp_sockets_allocated;
119 static void sctp_enter_memory_pressure(struct sock *sk)
121 sctp_memory_pressure = 1;
125 /* Get the sndbuf space available at the time on the association. */
126 static inline int sctp_wspace(struct sctp_association *asoc)
130 if (asoc->ep->sndbuf_policy)
131 amt = asoc->sndbuf_used;
133 amt = atomic_read(&asoc->base.sk->sk_wmem_alloc);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
139 amt = sk_stream_wspace(asoc->base.sk);
144 amt = asoc->base.sk->sk_sndbuf - amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
284 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
294 sctp_release_sock(sk);
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
311 /* V4 mapped address are really of AF_INET family */
312 if (addr->sa.sa_family == AF_INET6 &&
313 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 if (!opt->pf->af_supported(AF_INET, opt))
317 /* Does this PF support this AF? */
318 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 /* If we get this far, af is valid. */
323 af = sctp_get_af_specific(addr->sa.sa_family);
325 if (len < af->sockaddr_len)
331 /* Bind a local address either to an endpoint or to an association. */
332 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 struct sctp_sock *sp = sctp_sk(sk);
335 struct sctp_endpoint *ep = sp->ep;
336 struct sctp_bind_addr *bp = &ep->base.bind_addr;
341 /* Common sockaddr verification. */
342 af = sctp_sockaddr_af(sp, addr, len);
344 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
349 snum = ntohs(addr->v4.sin_port);
351 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
352 ", port: %d, new port: %d, len: %d)\n",
358 /* PF specific bind() address verification. */
359 if (!sp->pf->bind_verify(sp, addr))
360 return -EADDRNOTAVAIL;
362 /* We must either be unbound, or bind to the same port.
363 * It's OK to allow 0 ports if we are already bound.
364 * We'll just inhert an already bound port in this case
369 else if (snum != bp->port) {
370 SCTP_DEBUG_PRINTK("sctp_do_bind:"
371 " New port %d does not match existing port "
372 "%d.\n", snum, bp->port);
377 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
380 /* See if the address matches any of the addresses we may have
381 * already bound before checking against other endpoints.
383 if (sctp_bind_addr_match(bp, addr, sp))
386 /* Make sure we are allowed to bind here.
387 * The function sctp_get_port_local() does duplicate address
390 addr->v4.sin_port = htons(snum);
391 if ((ret = sctp_get_port_local(sk, addr))) {
395 /* Refresh ephemeral port. */
397 bp->port = inet_sk(sk)->num;
399 /* Add the address to the bind address list.
400 * Use GFP_ATOMIC since BHs will be disabled.
402 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
404 /* Copy back into socket for getsockname() use. */
406 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
407 af->to_sk_saddr(addr, sk);
413 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
415 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
416 * at any one time. If a sender, after sending an ASCONF chunk, decides
417 * it needs to transfer another ASCONF Chunk, it MUST wait until the
418 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
419 * subsequent ASCONF. Note this restriction binds each side, so at any
420 * time two ASCONF may be in-transit on any given association (one sent
421 * from each endpoint).
423 static int sctp_send_asconf(struct sctp_association *asoc,
424 struct sctp_chunk *chunk)
428 /* If there is an outstanding ASCONF chunk, queue it for later
431 if (asoc->addip_last_asconf) {
432 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
436 /* Hold the chunk until an ASCONF_ACK is received. */
437 sctp_chunk_hold(chunk);
438 retval = sctp_primitive_ASCONF(asoc, chunk);
440 sctp_chunk_free(chunk);
442 asoc->addip_last_asconf = chunk;
448 /* Add a list of addresses as bind addresses to local endpoint or
451 * Basically run through each address specified in the addrs/addrcnt
452 * array/length pair, determine if it is IPv6 or IPv4 and call
453 * sctp_do_bind() on it.
455 * If any of them fails, then the operation will be reversed and the
456 * ones that were added will be removed.
458 * Only sctp_setsockopt_bindx() is supposed to call this function.
460 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
465 struct sockaddr *sa_addr;
468 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 for (cnt = 0; cnt < addrcnt; cnt++) {
473 /* The list may contain either IPv4 or IPv6 address;
474 * determine the address length for walking thru the list.
476 sa_addr = (struct sockaddr *)addr_buf;
477 af = sctp_get_af_specific(sa_addr->sa_family);
483 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
486 addr_buf += af->sockaddr_len;
490 /* Failed. Cleanup the ones that have been added */
492 sctp_bindx_rem(sk, addrs, cnt);
500 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
501 * associations that are part of the endpoint indicating that a list of local
502 * addresses are added to the endpoint.
504 * If any of the addresses is already in the bind address list of the
505 * association, we do not send the chunk for that association. But it will not
506 * affect other associations.
508 * Only sctp_setsockopt_bindx() is supposed to call this function.
510 static int sctp_send_asconf_add_ip(struct sock *sk,
511 struct sockaddr *addrs,
514 struct sctp_sock *sp;
515 struct sctp_endpoint *ep;
516 struct sctp_association *asoc;
517 struct sctp_bind_addr *bp;
518 struct sctp_chunk *chunk;
519 struct sctp_sockaddr_entry *laddr;
520 union sctp_addr *addr;
521 union sctp_addr saveaddr;
528 if (!sctp_addip_enable)
534 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
535 __func__, sk, addrs, addrcnt);
537 list_for_each_entry(asoc, &ep->asocs, asocs) {
539 if (!asoc->peer.asconf_capable)
542 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
545 if (!sctp_state(asoc, ESTABLISHED))
548 /* Check if any address in the packed array of addresses is
549 * in the bind address list of the association. If so,
550 * do not send the asconf chunk to its peer, but continue with
551 * other associations.
554 for (i = 0; i < addrcnt; i++) {
555 addr = (union sctp_addr *)addr_buf;
556 af = sctp_get_af_specific(addr->v4.sin_family);
562 if (sctp_assoc_lookup_laddr(asoc, addr))
565 addr_buf += af->sockaddr_len;
570 /* Use the first valid address in bind addr list of
571 * association as Address Parameter of ASCONF CHUNK.
573 bp = &asoc->base.bind_addr;
574 p = bp->address_list.next;
575 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
576 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
577 addrcnt, SCTP_PARAM_ADD_IP);
583 retval = sctp_send_asconf(asoc, chunk);
587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
591 for (i = 0; i < addrcnt; i++) {
592 addr = (union sctp_addr *)addr_buf;
593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
597 addr_buf += af->sockaddr_len;
605 /* Remove a list of addresses from bind addresses list. Do not remove the
608 * Basically run through each address specified in the addrs/addrcnt
609 * array/length pair, determine if it is IPv6 or IPv4 and call
610 * sctp_del_bind() on it.
612 * If any of them fails, then the operation will be reversed and the
613 * ones that were removed will be added back.
615 * At least one address has to be left; if only one address is
616 * available, the operation will return -EBUSY.
618 * Only sctp_setsockopt_bindx() is supposed to call this function.
620 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
622 struct sctp_sock *sp = sctp_sk(sk);
623 struct sctp_endpoint *ep = sp->ep;
625 struct sctp_bind_addr *bp = &ep->base.bind_addr;
628 union sctp_addr *sa_addr;
631 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
635 for (cnt = 0; cnt < addrcnt; cnt++) {
636 /* If the bind address list is empty or if there is only one
637 * bind address, there is nothing more to be removed (we need
638 * at least one address here).
640 if (list_empty(&bp->address_list) ||
641 (sctp_list_single_entry(&bp->address_list))) {
646 sa_addr = (union sctp_addr *)addr_buf;
647 af = sctp_get_af_specific(sa_addr->sa.sa_family);
653 if (!af->addr_valid(sa_addr, sp, NULL)) {
654 retval = -EADDRNOTAVAIL;
658 if (sa_addr->v4.sin_port != htons(bp->port)) {
663 /* FIXME - There is probably a need to check if sk->sk_saddr and
664 * sk->sk_rcv_addr are currently set to one of the addresses to
665 * be removed. This is something which needs to be looked into
666 * when we are fixing the outstanding issues with multi-homing
667 * socket routing and failover schemes. Refer to comments in
668 * sctp_do_bind(). -daisy
670 retval = sctp_del_bind_addr(bp, sa_addr);
672 addr_buf += af->sockaddr_len;
675 /* Failed. Add the ones that has been removed back */
677 sctp_bindx_add(sk, addrs, cnt);
685 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
686 * the associations that are part of the endpoint indicating that a list of
687 * local addresses are removed from the endpoint.
689 * If any of the addresses is already in the bind address list of the
690 * association, we do not send the chunk for that association. But it will not
691 * affect other associations.
693 * Only sctp_setsockopt_bindx() is supposed to call this function.
695 static int sctp_send_asconf_del_ip(struct sock *sk,
696 struct sockaddr *addrs,
699 struct sctp_sock *sp;
700 struct sctp_endpoint *ep;
701 struct sctp_association *asoc;
702 struct sctp_transport *transport;
703 struct sctp_bind_addr *bp;
704 struct sctp_chunk *chunk;
705 union sctp_addr *laddr;
708 struct sctp_sockaddr_entry *saddr;
712 if (!sctp_addip_enable)
718 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
719 __func__, sk, addrs, addrcnt);
721 list_for_each_entry(asoc, &ep->asocs, asocs) {
723 if (!asoc->peer.asconf_capable)
726 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
729 if (!sctp_state(asoc, ESTABLISHED))
732 /* Check if any address in the packed array of addresses is
733 * not present in the bind address list of the association.
734 * If so, do not send the asconf chunk to its peer, but
735 * continue with other associations.
738 for (i = 0; i < addrcnt; i++) {
739 laddr = (union sctp_addr *)addr_buf;
740 af = sctp_get_af_specific(laddr->v4.sin_family);
746 if (!sctp_assoc_lookup_laddr(asoc, laddr))
749 addr_buf += af->sockaddr_len;
754 /* Find one address in the association's bind address list
755 * that is not in the packed array of addresses. This is to
756 * make sure that we do not delete all the addresses in the
759 bp = &asoc->base.bind_addr;
760 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
765 /* We do not need RCU protection throughout this loop
766 * because this is done under a socket lock from the
769 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
776 /* Reset use_as_src flag for the addresses in the bind address
777 * list that are to be deleted.
780 for (i = 0; i < addrcnt; i++) {
781 laddr = (union sctp_addr *)addr_buf;
782 af = sctp_get_af_specific(laddr->v4.sin_family);
783 list_for_each_entry(saddr, &bp->address_list, list) {
784 if (sctp_cmp_addr_exact(&saddr->a, laddr))
785 saddr->state = SCTP_ADDR_DEL;
787 addr_buf += af->sockaddr_len;
790 /* Update the route and saddr entries for all the transports
791 * as some of the addresses in the bind address list are
792 * about to be deleted and cannot be used as source addresses.
794 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
796 dst_release(transport->dst);
797 sctp_transport_route(transport, NULL,
798 sctp_sk(asoc->base.sk));
801 retval = sctp_send_asconf(asoc, chunk);
807 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
810 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
813 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
814 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
817 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
818 * Section 3.1.2 for this usage.
820 * addrs is a pointer to an array of one or more socket addresses. Each
821 * address is contained in its appropriate structure (i.e. struct
822 * sockaddr_in or struct sockaddr_in6) the family of the address type
823 * must be used to distinguish the address length (note that this
824 * representation is termed a "packed array" of addresses). The caller
825 * specifies the number of addresses in the array with addrcnt.
827 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
828 * -1, and sets errno to the appropriate error code.
830 * For SCTP, the port given in each socket address must be the same, or
831 * sctp_bindx() will fail, setting errno to EINVAL.
833 * The flags parameter is formed from the bitwise OR of zero or more of
834 * the following currently defined flags:
836 * SCTP_BINDX_ADD_ADDR
838 * SCTP_BINDX_REM_ADDR
840 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
841 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
842 * addresses from the association. The two flags are mutually exclusive;
843 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
844 * not remove all addresses from an association; sctp_bindx() will
845 * reject such an attempt with EINVAL.
847 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
848 * additional addresses with an endpoint after calling bind(). Or use
849 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
850 * socket is associated with so that no new association accepted will be
851 * associated with those addresses. If the endpoint supports dynamic
852 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
853 * endpoint to send the appropriate message to the peer to change the
854 * peers address lists.
856 * Adding and removing addresses from a connected association is
857 * optional functionality. Implementations that do not support this
858 * functionality should return EOPNOTSUPP.
860 * Basically do nothing but copying the addresses from user to kernel
861 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
862 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
865 * We don't use copy_from_user() for optimization: we first do the
866 * sanity checks (buffer size -fast- and access check-healthy
867 * pointer); if all of those succeed, then we can alloc the memory
868 * (expensive operation) needed to copy the data to kernel. Then we do
869 * the copying without checking the user space area
870 * (__copy_from_user()).
872 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
875 * sk The sk of the socket
876 * addrs The pointer to the addresses in user land
877 * addrssize Size of the addrs buffer
878 * op Operation to perform (add or remove, see the flags of
881 * Returns 0 if ok, <0 errno code on error.
883 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
884 struct sockaddr __user *addrs,
885 int addrs_size, int op)
887 struct sockaddr *kaddrs;
891 struct sockaddr *sa_addr;
895 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
896 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
898 if (unlikely(addrs_size <= 0))
901 /* Check the user passed a healthy pointer. */
902 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
905 /* Alloc space for the address array in kernel memory. */
906 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
907 if (unlikely(!kaddrs))
910 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
915 /* Walk through the addrs buffer and count the number of addresses. */
917 while (walk_size < addrs_size) {
918 sa_addr = (struct sockaddr *)addr_buf;
919 af = sctp_get_af_specific(sa_addr->sa_family);
921 /* If the address family is not supported or if this address
922 * causes the address buffer to overflow return EINVAL.
924 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
929 addr_buf += af->sockaddr_len;
930 walk_size += af->sockaddr_len;
935 case SCTP_BINDX_ADD_ADDR:
936 err = sctp_bindx_add(sk, kaddrs, addrcnt);
939 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
942 case SCTP_BINDX_REM_ADDR:
943 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
946 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
960 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962 * Common routine for handling connect() and sctp_connectx().
963 * Connect will come in with just a single address.
965 static int __sctp_connect(struct sock* sk,
966 struct sockaddr *kaddrs,
968 sctp_assoc_t *assoc_id)
970 struct sctp_sock *sp;
971 struct sctp_endpoint *ep;
972 struct sctp_association *asoc = NULL;
973 struct sctp_association *asoc2;
974 struct sctp_transport *transport;
982 union sctp_addr *sa_addr = NULL;
985 unsigned int f_flags = 0;
990 /* connect() cannot be done on a socket that is already in ESTABLISHED
991 * state - UDP-style peeled off socket or a TCP-style socket that
992 * is already connected.
993 * It cannot be done even on a TCP-style listening socket.
995 if (sctp_sstate(sk, ESTABLISHED) ||
996 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1001 /* Walk through the addrs buffer and count the number of addresses. */
1003 while (walk_size < addrs_size) {
1004 sa_addr = (union sctp_addr *)addr_buf;
1005 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1006 port = ntohs(sa_addr->v4.sin_port);
1008 /* If the address family is not supported or if this address
1009 * causes the address buffer to overflow return EINVAL.
1011 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1016 /* Save current address so we can work with it */
1017 memcpy(&to, sa_addr, af->sockaddr_len);
1019 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1023 /* Make sure the destination port is correctly set
1026 if (asoc && asoc->peer.port && asoc->peer.port != port)
1030 /* Check if there already is a matching association on the
1031 * endpoint (other than the one created here).
1033 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1034 if (asoc2 && asoc2 != asoc) {
1035 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1042 /* If we could not find a matching association on the endpoint,
1043 * make sure that there is no peeled-off association matching
1044 * the peer address even on another socket.
1046 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1047 err = -EADDRNOTAVAIL;
1052 /* If a bind() or sctp_bindx() is not called prior to
1053 * an sctp_connectx() call, the system picks an
1054 * ephemeral port and will choose an address set
1055 * equivalent to binding with a wildcard address.
1057 if (!ep->base.bind_addr.port) {
1058 if (sctp_autobind(sk)) {
1064 * If an unprivileged user inherits a 1-many
1065 * style socket with open associations on a
1066 * privileged port, it MAY be permitted to
1067 * accept new associations, but it SHOULD NOT
1068 * be permitted to open new associations.
1070 if (ep->base.bind_addr.port < PROT_SOCK &&
1071 !capable(CAP_NET_BIND_SERVICE)) {
1077 scope = sctp_scope(&to);
1078 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1085 /* Prime the peer's transport structures. */
1086 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1094 addr_buf += af->sockaddr_len;
1095 walk_size += af->sockaddr_len;
1098 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1103 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1108 /* Initialize sk's dport and daddr for getpeername() */
1109 inet_sk(sk)->dport = htons(asoc->peer.port);
1110 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1111 af->to_sk_daddr(sa_addr, sk);
1114 /* in-kernel sockets don't generally have a file allocated to them
1115 * if all they do is call sock_create_kern().
1117 if (sk->sk_socket->file)
1118 f_flags = sk->sk_socket->file->f_flags;
1120 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1122 err = sctp_wait_for_connect(asoc, &timeo);
1123 if (!err && assoc_id)
1124 *assoc_id = asoc->assoc_id;
1126 /* Don't free association on exit. */
1131 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1132 " kaddrs: %p err: %d\n",
1135 sctp_association_free(asoc);
1139 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1142 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1143 * sctp_assoc_t *asoc);
1145 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1146 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1147 * or IPv6 addresses.
1149 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1150 * Section 3.1.2 for this usage.
1152 * addrs is a pointer to an array of one or more socket addresses. Each
1153 * address is contained in its appropriate structure (i.e. struct
1154 * sockaddr_in or struct sockaddr_in6) the family of the address type
1155 * must be used to distengish the address length (note that this
1156 * representation is termed a "packed array" of addresses). The caller
1157 * specifies the number of addresses in the array with addrcnt.
1159 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1160 * the association id of the new association. On failure, sctp_connectx()
1161 * returns -1, and sets errno to the appropriate error code. The assoc_id
1162 * is not touched by the kernel.
1164 * For SCTP, the port given in each socket address must be the same, or
1165 * sctp_connectx() will fail, setting errno to EINVAL.
1167 * An application can use sctp_connectx to initiate an association with
1168 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1169 * allows a caller to specify multiple addresses at which a peer can be
1170 * reached. The way the SCTP stack uses the list of addresses to set up
1171 * the association is implementation dependant. This function only
1172 * specifies that the stack will try to make use of all the addresses in
1173 * the list when needed.
1175 * Note that the list of addresses passed in is only used for setting up
1176 * the association. It does not necessarily equal the set of addresses
1177 * the peer uses for the resulting association. If the caller wants to
1178 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1179 * retrieve them after the association has been set up.
1181 * Basically do nothing but copying the addresses from user to kernel
1182 * land and invoking either sctp_connectx(). This is used for tunneling
1183 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1185 * We don't use copy_from_user() for optimization: we first do the
1186 * sanity checks (buffer size -fast- and access check-healthy
1187 * pointer); if all of those succeed, then we can alloc the memory
1188 * (expensive operation) needed to copy the data to kernel. Then we do
1189 * the copying without checking the user space area
1190 * (__copy_from_user()).
1192 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1195 * sk The sk of the socket
1196 * addrs The pointer to the addresses in user land
1197 * addrssize Size of the addrs buffer
1199 * Returns >=0 if ok, <0 errno code on error.
1201 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1202 struct sockaddr __user *addrs,
1204 sctp_assoc_t *assoc_id)
1207 struct sockaddr *kaddrs;
1209 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1210 __func__, sk, addrs, addrs_size);
1212 if (unlikely(addrs_size <= 0))
1215 /* Check the user passed a healthy pointer. */
1216 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1219 /* Alloc space for the address array in kernel memory. */
1220 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1221 if (unlikely(!kaddrs))
1224 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1227 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1236 * This is an older interface. It's kept for backward compatibility
1237 * to the option that doesn't provide association id.
1239 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1240 struct sockaddr __user *addrs,
1243 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1247 * New interface for the API. The since the API is done with a socket
1248 * option, to make it simple we feed back the association id is as a return
1249 * indication to the call. Error is always negative and association id is
1252 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1253 struct sockaddr __user *addrs,
1256 sctp_assoc_t assoc_id = 0;
1259 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1267 /* API 3.1.4 close() - UDP Style Syntax
1268 * Applications use close() to perform graceful shutdown (as described in
1269 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1270 * by a UDP-style socket.
1274 * ret = close(int sd);
1276 * sd - the socket descriptor of the associations to be closed.
1278 * To gracefully shutdown a specific association represented by the
1279 * UDP-style socket, an application should use the sendmsg() call,
1280 * passing no user data, but including the appropriate flag in the
1281 * ancillary data (see Section xxxx).
1283 * If sd in the close() call is a branched-off socket representing only
1284 * one association, the shutdown is performed on that association only.
1286 * 4.1.6 close() - TCP Style Syntax
1288 * Applications use close() to gracefully close down an association.
1292 * int close(int sd);
1294 * sd - the socket descriptor of the association to be closed.
1296 * After an application calls close() on a socket descriptor, no further
1297 * socket operations will succeed on that descriptor.
1299 * API 7.1.4 SO_LINGER
1301 * An application using the TCP-style socket can use this option to
1302 * perform the SCTP ABORT primitive. The linger option structure is:
1305 * int l_onoff; // option on/off
1306 * int l_linger; // linger time
1309 * To enable the option, set l_onoff to 1. If the l_linger value is set
1310 * to 0, calling close() is the same as the ABORT primitive. If the
1311 * value is set to a negative value, the setsockopt() call will return
1312 * an error. If the value is set to a positive value linger_time, the
1313 * close() can be blocked for at most linger_time ms. If the graceful
1314 * shutdown phase does not finish during this period, close() will
1315 * return but the graceful shutdown phase continues in the system.
1317 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1319 struct sctp_endpoint *ep;
1320 struct sctp_association *asoc;
1321 struct list_head *pos, *temp;
1323 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1326 sk->sk_shutdown = SHUTDOWN_MASK;
1328 ep = sctp_sk(sk)->ep;
1330 /* Walk all associations on an endpoint. */
1331 list_for_each_safe(pos, temp, &ep->asocs) {
1332 asoc = list_entry(pos, struct sctp_association, asocs);
1334 if (sctp_style(sk, TCP)) {
1335 /* A closed association can still be in the list if
1336 * it belongs to a TCP-style listening socket that is
1337 * not yet accepted. If so, free it. If not, send an
1338 * ABORT or SHUTDOWN based on the linger options.
1340 if (sctp_state(asoc, CLOSED)) {
1341 sctp_unhash_established(asoc);
1342 sctp_association_free(asoc);
1347 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1348 struct sctp_chunk *chunk;
1350 chunk = sctp_make_abort_user(asoc, NULL, 0);
1352 sctp_primitive_ABORT(asoc, chunk);
1354 sctp_primitive_SHUTDOWN(asoc, NULL);
1357 /* Clean up any skbs sitting on the receive queue. */
1358 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1359 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1361 /* On a TCP-style socket, block for at most linger_time if set. */
1362 if (sctp_style(sk, TCP) && timeout)
1363 sctp_wait_for_close(sk, timeout);
1365 /* This will run the backlog queue. */
1366 sctp_release_sock(sk);
1368 /* Supposedly, no process has access to the socket, but
1369 * the net layers still may.
1371 sctp_local_bh_disable();
1372 sctp_bh_lock_sock(sk);
1374 /* Hold the sock, since sk_common_release() will put sock_put()
1375 * and we have just a little more cleanup.
1378 sk_common_release(sk);
1380 sctp_bh_unlock_sock(sk);
1381 sctp_local_bh_enable();
1385 SCTP_DBG_OBJCNT_DEC(sock);
1388 /* Handle EPIPE error. */
1389 static int sctp_error(struct sock *sk, int flags, int err)
1392 err = sock_error(sk) ? : -EPIPE;
1393 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1394 send_sig(SIGPIPE, current, 0);
1398 /* API 3.1.3 sendmsg() - UDP Style Syntax
1400 * An application uses sendmsg() and recvmsg() calls to transmit data to
1401 * and receive data from its peer.
1403 * ssize_t sendmsg(int socket, const struct msghdr *message,
1406 * socket - the socket descriptor of the endpoint.
1407 * message - pointer to the msghdr structure which contains a single
1408 * user message and possibly some ancillary data.
1410 * See Section 5 for complete description of the data
1413 * flags - flags sent or received with the user message, see Section
1414 * 5 for complete description of the flags.
1416 * Note: This function could use a rewrite especially when explicit
1417 * connect support comes in.
1419 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1421 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1423 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1424 struct msghdr *msg, size_t msg_len)
1426 struct sctp_sock *sp;
1427 struct sctp_endpoint *ep;
1428 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1429 struct sctp_transport *transport, *chunk_tp;
1430 struct sctp_chunk *chunk;
1432 struct sockaddr *msg_name = NULL;
1433 struct sctp_sndrcvinfo default_sinfo = { 0 };
1434 struct sctp_sndrcvinfo *sinfo;
1435 struct sctp_initmsg *sinit;
1436 sctp_assoc_t associd = 0;
1437 sctp_cmsgs_t cmsgs = { NULL };
1441 __u16 sinfo_flags = 0;
1442 struct sctp_datamsg *datamsg;
1443 int msg_flags = msg->msg_flags;
1445 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1452 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1454 /* We cannot send a message over a TCP-style listening socket. */
1455 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1460 /* Parse out the SCTP CMSGs. */
1461 err = sctp_msghdr_parse(msg, &cmsgs);
1464 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1468 /* Fetch the destination address for this packet. This
1469 * address only selects the association--it is not necessarily
1470 * the address we will send to.
1471 * For a peeled-off socket, msg_name is ignored.
1473 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1474 int msg_namelen = msg->msg_namelen;
1476 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1481 if (msg_namelen > sizeof(to))
1482 msg_namelen = sizeof(to);
1483 memcpy(&to, msg->msg_name, msg_namelen);
1484 msg_name = msg->msg_name;
1490 /* Did the user specify SNDRCVINFO? */
1492 sinfo_flags = sinfo->sinfo_flags;
1493 associd = sinfo->sinfo_assoc_id;
1496 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1497 msg_len, sinfo_flags);
1499 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1500 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1505 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1506 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1507 * If SCTP_ABORT is set, the message length could be non zero with
1508 * the msg_iov set to the user abort reason.
1510 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1511 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1516 /* If SCTP_ADDR_OVER is set, there must be an address
1517 * specified in msg_name.
1519 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1526 SCTP_DEBUG_PRINTK("About to look up association.\n");
1530 /* If a msg_name has been specified, assume this is to be used. */
1532 /* Look for a matching association on the endpoint. */
1533 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1535 /* If we could not find a matching association on the
1536 * endpoint, make sure that it is not a TCP-style
1537 * socket that already has an association or there is
1538 * no peeled-off association on another socket.
1540 if ((sctp_style(sk, TCP) &&
1541 sctp_sstate(sk, ESTABLISHED)) ||
1542 sctp_endpoint_is_peeled_off(ep, &to)) {
1543 err = -EADDRNOTAVAIL;
1548 asoc = sctp_id2assoc(sk, associd);
1556 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1558 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1559 * socket that has an association in CLOSED state. This can
1560 * happen when an accepted socket has an association that is
1563 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1568 if (sinfo_flags & SCTP_EOF) {
1569 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1571 sctp_primitive_SHUTDOWN(asoc, NULL);
1575 if (sinfo_flags & SCTP_ABORT) {
1577 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1583 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1584 sctp_primitive_ABORT(asoc, chunk);
1590 /* Do we need to create the association? */
1592 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1594 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1599 /* Check for invalid stream against the stream counts,
1600 * either the default or the user specified stream counts.
1603 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1604 /* Check against the defaults. */
1605 if (sinfo->sinfo_stream >=
1606 sp->initmsg.sinit_num_ostreams) {
1611 /* Check against the requested. */
1612 if (sinfo->sinfo_stream >=
1613 sinit->sinit_num_ostreams) {
1621 * API 3.1.2 bind() - UDP Style Syntax
1622 * If a bind() or sctp_bindx() is not called prior to a
1623 * sendmsg() call that initiates a new association, the
1624 * system picks an ephemeral port and will choose an address
1625 * set equivalent to binding with a wildcard address.
1627 if (!ep->base.bind_addr.port) {
1628 if (sctp_autobind(sk)) {
1634 * If an unprivileged user inherits a one-to-many
1635 * style socket with open associations on a privileged
1636 * port, it MAY be permitted to accept new associations,
1637 * but it SHOULD NOT be permitted to open new
1640 if (ep->base.bind_addr.port < PROT_SOCK &&
1641 !capable(CAP_NET_BIND_SERVICE)) {
1647 scope = sctp_scope(&to);
1648 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1655 /* If the SCTP_INIT ancillary data is specified, set all
1656 * the association init values accordingly.
1659 if (sinit->sinit_num_ostreams) {
1660 asoc->c.sinit_num_ostreams =
1661 sinit->sinit_num_ostreams;
1663 if (sinit->sinit_max_instreams) {
1664 asoc->c.sinit_max_instreams =
1665 sinit->sinit_max_instreams;
1667 if (sinit->sinit_max_attempts) {
1668 asoc->max_init_attempts
1669 = sinit->sinit_max_attempts;
1671 if (sinit->sinit_max_init_timeo) {
1672 asoc->max_init_timeo =
1673 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1677 /* Prime the peer's transport structures. */
1678 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1683 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1690 /* ASSERT: we have a valid association at this point. */
1691 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1694 /* If the user didn't specify SNDRCVINFO, make up one with
1697 default_sinfo.sinfo_stream = asoc->default_stream;
1698 default_sinfo.sinfo_flags = asoc->default_flags;
1699 default_sinfo.sinfo_ppid = asoc->default_ppid;
1700 default_sinfo.sinfo_context = asoc->default_context;
1701 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1702 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1703 sinfo = &default_sinfo;
1706 /* API 7.1.7, the sndbuf size per association bounds the
1707 * maximum size of data that can be sent in a single send call.
1709 if (msg_len > sk->sk_sndbuf) {
1714 if (asoc->pmtu_pending)
1715 sctp_assoc_pending_pmtu(asoc);
1717 /* If fragmentation is disabled and the message length exceeds the
1718 * association fragmentation point, return EMSGSIZE. The I-D
1719 * does not specify what this error is, but this looks like
1722 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1728 /* Check for invalid stream. */
1729 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1735 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1736 if (!sctp_wspace(asoc)) {
1737 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1742 /* If an address is passed with the sendto/sendmsg call, it is used
1743 * to override the primary destination address in the TCP model, or
1744 * when SCTP_ADDR_OVER flag is set in the UDP model.
1746 if ((sctp_style(sk, TCP) && msg_name) ||
1747 (sinfo_flags & SCTP_ADDR_OVER)) {
1748 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1756 /* Auto-connect, if we aren't connected already. */
1757 if (sctp_state(asoc, CLOSED)) {
1758 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1761 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1764 /* Break the message into multiple chunks of maximum size. */
1765 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1771 /* Now send the (possibly) fragmented message. */
1772 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1773 sctp_chunk_hold(chunk);
1775 /* Do accounting for the write space. */
1776 sctp_set_owner_w(chunk);
1778 chunk->transport = chunk_tp;
1780 /* Send it to the lower layers. Note: all chunks
1781 * must either fail or succeed. The lower layer
1782 * works that way today. Keep it that way or this
1785 err = sctp_primitive_SEND(asoc, chunk);
1786 /* Did the lower layer accept the chunk? */
1788 sctp_chunk_free(chunk);
1789 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1792 sctp_datamsg_put(datamsg);
1798 /* If we are already past ASSOCIATE, the lower
1799 * layers are responsible for association cleanup.
1805 sctp_association_free(asoc);
1807 sctp_release_sock(sk);
1810 return sctp_error(sk, msg_flags, err);
1817 err = sock_error(sk);
1827 /* This is an extended version of skb_pull() that removes the data from the
1828 * start of a skb even when data is spread across the list of skb's in the
1829 * frag_list. len specifies the total amount of data that needs to be removed.
1830 * when 'len' bytes could be removed from the skb, it returns 0.
1831 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1832 * could not be removed.
1834 static int sctp_skb_pull(struct sk_buff *skb, int len)
1836 struct sk_buff *list;
1837 int skb_len = skb_headlen(skb);
1840 if (len <= skb_len) {
1841 __skb_pull(skb, len);
1845 __skb_pull(skb, skb_len);
1847 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1848 rlen = sctp_skb_pull(list, len);
1849 skb->len -= (len-rlen);
1850 skb->data_len -= (len-rlen);
1861 /* API 3.1.3 recvmsg() - UDP Style Syntax
1863 * ssize_t recvmsg(int socket, struct msghdr *message,
1866 * socket - the socket descriptor of the endpoint.
1867 * message - pointer to the msghdr structure which contains a single
1868 * user message and possibly some ancillary data.
1870 * See Section 5 for complete description of the data
1873 * flags - flags sent or received with the user message, see Section
1874 * 5 for complete description of the flags.
1876 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1878 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1879 struct msghdr *msg, size_t len, int noblock,
1880 int flags, int *addr_len)
1882 struct sctp_ulpevent *event = NULL;
1883 struct sctp_sock *sp = sctp_sk(sk);
1884 struct sk_buff *skb;
1889 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1890 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1891 "len", len, "knoblauch", noblock,
1892 "flags", flags, "addr_len", addr_len);
1896 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1901 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1905 /* Get the total length of the skb including any skb's in the
1914 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1916 event = sctp_skb2event(skb);
1921 sock_recv_timestamp(msg, sk, skb);
1922 if (sctp_ulpevent_is_notification(event)) {
1923 msg->msg_flags |= MSG_NOTIFICATION;
1924 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1926 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1929 /* Check if we allow SCTP_SNDRCVINFO. */
1930 if (sp->subscribe.sctp_data_io_event)
1931 sctp_ulpevent_read_sndrcvinfo(event, msg);
1933 /* FIXME: we should be calling IP/IPv6 layers. */
1934 if (sk->sk_protinfo.af_inet.cmsg_flags)
1935 ip_cmsg_recv(msg, skb);
1940 /* If skb's length exceeds the user's buffer, update the skb and
1941 * push it back to the receive_queue so that the next call to
1942 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1944 if (skb_len > copied) {
1945 msg->msg_flags &= ~MSG_EOR;
1946 if (flags & MSG_PEEK)
1948 sctp_skb_pull(skb, copied);
1949 skb_queue_head(&sk->sk_receive_queue, skb);
1951 /* When only partial message is copied to the user, increase
1952 * rwnd by that amount. If all the data in the skb is read,
1953 * rwnd is updated when the event is freed.
1955 if (!sctp_ulpevent_is_notification(event))
1956 sctp_assoc_rwnd_increase(event->asoc, copied);
1958 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1959 (event->msg_flags & MSG_EOR))
1960 msg->msg_flags |= MSG_EOR;
1962 msg->msg_flags &= ~MSG_EOR;
1965 if (flags & MSG_PEEK) {
1966 /* Release the skb reference acquired after peeking the skb in
1967 * sctp_skb_recv_datagram().
1971 /* Free the event which includes releasing the reference to
1972 * the owner of the skb, freeing the skb and updating the
1975 sctp_ulpevent_free(event);
1978 sctp_release_sock(sk);
1982 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1984 * This option is a on/off flag. If enabled no SCTP message
1985 * fragmentation will be performed. Instead if a message being sent
1986 * exceeds the current PMTU size, the message will NOT be sent and
1987 * instead a error will be indicated to the user.
1989 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1990 char __user *optval, int optlen)
1994 if (optlen < sizeof(int))
1997 if (get_user(val, (int __user *)optval))
2000 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2005 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2008 if (optlen > sizeof(struct sctp_event_subscribe))
2010 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2015 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2017 * This socket option is applicable to the UDP-style socket only. When
2018 * set it will cause associations that are idle for more than the
2019 * specified number of seconds to automatically close. An association
2020 * being idle is defined an association that has NOT sent or received
2021 * user data. The special value of '0' indicates that no automatic
2022 * close of any associations should be performed. The option expects an
2023 * integer defining the number of seconds of idle time before an
2024 * association is closed.
2026 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2029 struct sctp_sock *sp = sctp_sk(sk);
2031 /* Applicable to UDP-style socket only */
2032 if (sctp_style(sk, TCP))
2034 if (optlen != sizeof(int))
2036 if (copy_from_user(&sp->autoclose, optval, optlen))
2042 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2044 * Applications can enable or disable heartbeats for any peer address of
2045 * an association, modify an address's heartbeat interval, force a
2046 * heartbeat to be sent immediately, and adjust the address's maximum
2047 * number of retransmissions sent before an address is considered
2048 * unreachable. The following structure is used to access and modify an
2049 * address's parameters:
2051 * struct sctp_paddrparams {
2052 * sctp_assoc_t spp_assoc_id;
2053 * struct sockaddr_storage spp_address;
2054 * uint32_t spp_hbinterval;
2055 * uint16_t spp_pathmaxrxt;
2056 * uint32_t spp_pathmtu;
2057 * uint32_t spp_sackdelay;
2058 * uint32_t spp_flags;
2061 * spp_assoc_id - (one-to-many style socket) This is filled in the
2062 * application, and identifies the association for
2064 * spp_address - This specifies which address is of interest.
2065 * spp_hbinterval - This contains the value of the heartbeat interval,
2066 * in milliseconds. If a value of zero
2067 * is present in this field then no changes are to
2068 * be made to this parameter.
2069 * spp_pathmaxrxt - This contains the maximum number of
2070 * retransmissions before this address shall be
2071 * considered unreachable. If a value of zero
2072 * is present in this field then no changes are to
2073 * be made to this parameter.
2074 * spp_pathmtu - When Path MTU discovery is disabled the value
2075 * specified here will be the "fixed" path mtu.
2076 * Note that if the spp_address field is empty
2077 * then all associations on this address will
2078 * have this fixed path mtu set upon them.
2080 * spp_sackdelay - When delayed sack is enabled, this value specifies
2081 * the number of milliseconds that sacks will be delayed
2082 * for. This value will apply to all addresses of an
2083 * association if the spp_address field is empty. Note
2084 * also, that if delayed sack is enabled and this
2085 * value is set to 0, no change is made to the last
2086 * recorded delayed sack timer value.
2088 * spp_flags - These flags are used to control various features
2089 * on an association. The flag field may contain
2090 * zero or more of the following options.
2092 * SPP_HB_ENABLE - Enable heartbeats on the
2093 * specified address. Note that if the address
2094 * field is empty all addresses for the association
2095 * have heartbeats enabled upon them.
2097 * SPP_HB_DISABLE - Disable heartbeats on the
2098 * speicifed address. Note that if the address
2099 * field is empty all addresses for the association
2100 * will have their heartbeats disabled. Note also
2101 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2102 * mutually exclusive, only one of these two should
2103 * be specified. Enabling both fields will have
2104 * undetermined results.
2106 * SPP_HB_DEMAND - Request a user initiated heartbeat
2107 * to be made immediately.
2109 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2110 * heartbeat delayis to be set to the value of 0
2113 * SPP_PMTUD_ENABLE - This field will enable PMTU
2114 * discovery upon the specified address. Note that
2115 * if the address feild is empty then all addresses
2116 * on the association are effected.
2118 * SPP_PMTUD_DISABLE - This field will disable PMTU
2119 * discovery upon the specified address. Note that
2120 * if the address feild is empty then all addresses
2121 * on the association are effected. Not also that
2122 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2123 * exclusive. Enabling both will have undetermined
2126 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2127 * on delayed sack. The time specified in spp_sackdelay
2128 * is used to specify the sack delay for this address. Note
2129 * that if spp_address is empty then all addresses will
2130 * enable delayed sack and take on the sack delay
2131 * value specified in spp_sackdelay.
2132 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2133 * off delayed sack. If the spp_address field is blank then
2134 * delayed sack is disabled for the entire association. Note
2135 * also that this field is mutually exclusive to
2136 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2139 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2140 struct sctp_transport *trans,
2141 struct sctp_association *asoc,
2142 struct sctp_sock *sp,
2145 int sackdelay_change)
2149 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2150 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2155 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2156 * this field is ignored. Note also that a value of zero indicates
2157 * the current setting should be left unchanged.
2159 if (params->spp_flags & SPP_HB_ENABLE) {
2161 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2162 * set. This lets us use 0 value when this flag
2165 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2166 params->spp_hbinterval = 0;
2168 if (params->spp_hbinterval ||
2169 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2172 msecs_to_jiffies(params->spp_hbinterval);
2175 msecs_to_jiffies(params->spp_hbinterval);
2177 sp->hbinterval = params->spp_hbinterval;
2184 trans->param_flags =
2185 (trans->param_flags & ~SPP_HB) | hb_change;
2188 (asoc->param_flags & ~SPP_HB) | hb_change;
2191 (sp->param_flags & ~SPP_HB) | hb_change;
2195 /* When Path MTU discovery is disabled the value specified here will
2196 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2197 * include the flag SPP_PMTUD_DISABLE for this field to have any
2200 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2202 trans->pathmtu = params->spp_pathmtu;
2203 sctp_assoc_sync_pmtu(asoc);
2205 asoc->pathmtu = params->spp_pathmtu;
2206 sctp_frag_point(sp, params->spp_pathmtu);
2208 sp->pathmtu = params->spp_pathmtu;
2214 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2215 (params->spp_flags & SPP_PMTUD_ENABLE);
2216 trans->param_flags =
2217 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2219 sctp_transport_pmtu(trans);
2220 sctp_assoc_sync_pmtu(asoc);
2224 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2227 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2231 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2232 * value of this field is ignored. Note also that a value of zero
2233 * indicates the current setting should be left unchanged.
2235 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2238 msecs_to_jiffies(params->spp_sackdelay);
2241 msecs_to_jiffies(params->spp_sackdelay);
2243 sp->sackdelay = params->spp_sackdelay;
2247 if (sackdelay_change) {
2249 trans->param_flags =
2250 (trans->param_flags & ~SPP_SACKDELAY) |
2254 (asoc->param_flags & ~SPP_SACKDELAY) |
2258 (sp->param_flags & ~SPP_SACKDELAY) |
2263 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2264 * of this field is ignored. Note also that a value of zero
2265 * indicates the current setting should be left unchanged.
2267 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2269 trans->pathmaxrxt = params->spp_pathmaxrxt;
2271 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2273 sp->pathmaxrxt = params->spp_pathmaxrxt;
2280 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2281 char __user *optval, int optlen)
2283 struct sctp_paddrparams params;
2284 struct sctp_transport *trans = NULL;
2285 struct sctp_association *asoc = NULL;
2286 struct sctp_sock *sp = sctp_sk(sk);
2288 int hb_change, pmtud_change, sackdelay_change;
2290 if (optlen != sizeof(struct sctp_paddrparams))
2293 if (copy_from_user(¶ms, optval, optlen))
2296 /* Validate flags and value parameters. */
2297 hb_change = params.spp_flags & SPP_HB;
2298 pmtud_change = params.spp_flags & SPP_PMTUD;
2299 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2301 if (hb_change == SPP_HB ||
2302 pmtud_change == SPP_PMTUD ||
2303 sackdelay_change == SPP_SACKDELAY ||
2304 params.spp_sackdelay > 500 ||
2306 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2309 /* If an address other than INADDR_ANY is specified, and
2310 * no transport is found, then the request is invalid.
2312 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2313 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2314 params.spp_assoc_id);
2319 /* Get association, if assoc_id != 0 and the socket is a one
2320 * to many style socket, and an association was not found, then
2321 * the id was invalid.
2323 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2324 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2327 /* Heartbeat demand can only be sent on a transport or
2328 * association, but not a socket.
2330 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2333 /* Process parameters. */
2334 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2335 hb_change, pmtud_change,
2341 /* If changes are for association, also apply parameters to each
2344 if (!trans && asoc) {
2345 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2347 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2348 hb_change, pmtud_change,
2357 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2359 * This option will effect the way delayed acks are performed. This
2360 * option allows you to get or set the delayed ack time, in
2361 * milliseconds. It also allows changing the delayed ack frequency.
2362 * Changing the frequency to 1 disables the delayed sack algorithm. If
2363 * the assoc_id is 0, then this sets or gets the endpoints default
2364 * values. If the assoc_id field is non-zero, then the set or get
2365 * effects the specified association for the one to many model (the
2366 * assoc_id field is ignored by the one to one model). Note that if
2367 * sack_delay or sack_freq are 0 when setting this option, then the
2368 * current values will remain unchanged.
2370 * struct sctp_sack_info {
2371 * sctp_assoc_t sack_assoc_id;
2372 * uint32_t sack_delay;
2373 * uint32_t sack_freq;
2376 * sack_assoc_id - This parameter, indicates which association the user
2377 * is performing an action upon. Note that if this field's value is
2378 * zero then the endpoints default value is changed (effecting future
2379 * associations only).
2381 * sack_delay - This parameter contains the number of milliseconds that
2382 * the user is requesting the delayed ACK timer be set to. Note that
2383 * this value is defined in the standard to be between 200 and 500
2386 * sack_freq - This parameter contains the number of packets that must
2387 * be received before a sack is sent without waiting for the delay
2388 * timer to expire. The default value for this is 2, setting this
2389 * value to 1 will disable the delayed sack algorithm.
2392 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2393 char __user *optval, int optlen)
2395 struct sctp_sack_info params;
2396 struct sctp_transport *trans = NULL;
2397 struct sctp_association *asoc = NULL;
2398 struct sctp_sock *sp = sctp_sk(sk);
2400 if (optlen == sizeof(struct sctp_sack_info)) {
2401 if (copy_from_user(¶ms, optval, optlen))
2404 if (params.sack_delay == 0 && params.sack_freq == 0)
2406 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2407 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2408 "in delayed_ack socket option deprecated\n");
2409 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2410 if (copy_from_user(¶ms, optval, optlen))
2413 if (params.sack_delay == 0)
2414 params.sack_freq = 1;
2416 params.sack_freq = 0;
2420 /* Validate value parameter. */
2421 if (params.sack_delay > 500)
2424 /* Get association, if sack_assoc_id != 0 and the socket is a one
2425 * to many style socket, and an association was not found, then
2426 * the id was invalid.
2428 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2429 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2432 if (params.sack_delay) {
2435 msecs_to_jiffies(params.sack_delay);
2437 (asoc->param_flags & ~SPP_SACKDELAY) |
2438 SPP_SACKDELAY_ENABLE;
2440 sp->sackdelay = params.sack_delay;
2442 (sp->param_flags & ~SPP_SACKDELAY) |
2443 SPP_SACKDELAY_ENABLE;
2447 if (params.sack_freq == 1) {
2450 (asoc->param_flags & ~SPP_SACKDELAY) |
2451 SPP_SACKDELAY_DISABLE;
2454 (sp->param_flags & ~SPP_SACKDELAY) |
2455 SPP_SACKDELAY_DISABLE;
2457 } else if (params.sack_freq > 1) {
2459 asoc->sackfreq = params.sack_freq;
2461 (asoc->param_flags & ~SPP_SACKDELAY) |
2462 SPP_SACKDELAY_ENABLE;
2464 sp->sackfreq = params.sack_freq;
2466 (sp->param_flags & ~SPP_SACKDELAY) |
2467 SPP_SACKDELAY_ENABLE;
2471 /* If change is for association, also apply to each transport. */
2473 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2475 if (params.sack_delay) {
2477 msecs_to_jiffies(params.sack_delay);
2478 trans->param_flags =
2479 (trans->param_flags & ~SPP_SACKDELAY) |
2480 SPP_SACKDELAY_ENABLE;
2482 if (params.sack_freq == 1) {
2483 trans->param_flags =
2484 (trans->param_flags & ~SPP_SACKDELAY) |
2485 SPP_SACKDELAY_DISABLE;
2486 } else if (params.sack_freq > 1) {
2487 trans->sackfreq = params.sack_freq;
2488 trans->param_flags =
2489 (trans->param_flags & ~SPP_SACKDELAY) |
2490 SPP_SACKDELAY_ENABLE;
2498 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2500 * Applications can specify protocol parameters for the default association
2501 * initialization. The option name argument to setsockopt() and getsockopt()
2504 * Setting initialization parameters is effective only on an unconnected
2505 * socket (for UDP-style sockets only future associations are effected
2506 * by the change). With TCP-style sockets, this option is inherited by
2507 * sockets derived from a listener socket.
2509 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2511 struct sctp_initmsg sinit;
2512 struct sctp_sock *sp = sctp_sk(sk);
2514 if (optlen != sizeof(struct sctp_initmsg))
2516 if (copy_from_user(&sinit, optval, optlen))
2519 if (sinit.sinit_num_ostreams)
2520 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2521 if (sinit.sinit_max_instreams)
2522 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2523 if (sinit.sinit_max_attempts)
2524 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2525 if (sinit.sinit_max_init_timeo)
2526 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2532 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2534 * Applications that wish to use the sendto() system call may wish to
2535 * specify a default set of parameters that would normally be supplied
2536 * through the inclusion of ancillary data. This socket option allows
2537 * such an application to set the default sctp_sndrcvinfo structure.
2538 * The application that wishes to use this socket option simply passes
2539 * in to this call the sctp_sndrcvinfo structure defined in Section
2540 * 5.2.2) The input parameters accepted by this call include
2541 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2542 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2543 * to this call if the caller is using the UDP model.
2545 static int sctp_setsockopt_default_send_param(struct sock *sk,
2546 char __user *optval, int optlen)
2548 struct sctp_sndrcvinfo info;
2549 struct sctp_association *asoc;
2550 struct sctp_sock *sp = sctp_sk(sk);
2552 if (optlen != sizeof(struct sctp_sndrcvinfo))
2554 if (copy_from_user(&info, optval, optlen))
2557 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2558 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2562 asoc->default_stream = info.sinfo_stream;
2563 asoc->default_flags = info.sinfo_flags;
2564 asoc->default_ppid = info.sinfo_ppid;
2565 asoc->default_context = info.sinfo_context;
2566 asoc->default_timetolive = info.sinfo_timetolive;
2568 sp->default_stream = info.sinfo_stream;
2569 sp->default_flags = info.sinfo_flags;
2570 sp->default_ppid = info.sinfo_ppid;
2571 sp->default_context = info.sinfo_context;
2572 sp->default_timetolive = info.sinfo_timetolive;
2578 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2580 * Requests that the local SCTP stack use the enclosed peer address as
2581 * the association primary. The enclosed address must be one of the
2582 * association peer's addresses.
2584 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2587 struct sctp_prim prim;
2588 struct sctp_transport *trans;
2590 if (optlen != sizeof(struct sctp_prim))
2593 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2596 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2600 sctp_assoc_set_primary(trans->asoc, trans);
2606 * 7.1.5 SCTP_NODELAY
2608 * Turn on/off any Nagle-like algorithm. This means that packets are
2609 * generally sent as soon as possible and no unnecessary delays are
2610 * introduced, at the cost of more packets in the network. Expects an
2611 * integer boolean flag.
2613 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2618 if (optlen < sizeof(int))
2620 if (get_user(val, (int __user *)optval))
2623 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2629 * 7.1.1 SCTP_RTOINFO
2631 * The protocol parameters used to initialize and bound retransmission
2632 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2633 * and modify these parameters.
2634 * All parameters are time values, in milliseconds. A value of 0, when
2635 * modifying the parameters, indicates that the current value should not
2639 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2640 struct sctp_rtoinfo rtoinfo;
2641 struct sctp_association *asoc;
2643 if (optlen != sizeof (struct sctp_rtoinfo))
2646 if (copy_from_user(&rtoinfo, optval, optlen))
2649 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2651 /* Set the values to the specific association */
2652 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2656 if (rtoinfo.srto_initial != 0)
2658 msecs_to_jiffies(rtoinfo.srto_initial);
2659 if (rtoinfo.srto_max != 0)
2660 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2661 if (rtoinfo.srto_min != 0)
2662 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2664 /* If there is no association or the association-id = 0
2665 * set the values to the endpoint.
2667 struct sctp_sock *sp = sctp_sk(sk);
2669 if (rtoinfo.srto_initial != 0)
2670 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2671 if (rtoinfo.srto_max != 0)
2672 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2673 if (rtoinfo.srto_min != 0)
2674 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2682 * 7.1.2 SCTP_ASSOCINFO
2684 * This option is used to tune the maximum retransmission attempts
2685 * of the association.
2686 * Returns an error if the new association retransmission value is
2687 * greater than the sum of the retransmission value of the peer.
2688 * See [SCTP] for more information.
2691 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2694 struct sctp_assocparams assocparams;
2695 struct sctp_association *asoc;
2697 if (optlen != sizeof(struct sctp_assocparams))
2699 if (copy_from_user(&assocparams, optval, optlen))
2702 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2704 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2707 /* Set the values to the specific association */
2709 if (assocparams.sasoc_asocmaxrxt != 0) {
2712 struct sctp_transport *peer_addr;
2714 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2716 path_sum += peer_addr->pathmaxrxt;
2720 /* Only validate asocmaxrxt if we have more then
2721 * one path/transport. We do this because path
2722 * retransmissions are only counted when we have more
2726 assocparams.sasoc_asocmaxrxt > path_sum)
2729 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2732 if (assocparams.sasoc_cookie_life != 0) {
2733 asoc->cookie_life.tv_sec =
2734 assocparams.sasoc_cookie_life / 1000;
2735 asoc->cookie_life.tv_usec =
2736 (assocparams.sasoc_cookie_life % 1000)
2740 /* Set the values to the endpoint */
2741 struct sctp_sock *sp = sctp_sk(sk);
2743 if (assocparams.sasoc_asocmaxrxt != 0)
2744 sp->assocparams.sasoc_asocmaxrxt =
2745 assocparams.sasoc_asocmaxrxt;
2746 if (assocparams.sasoc_cookie_life != 0)
2747 sp->assocparams.sasoc_cookie_life =
2748 assocparams.sasoc_cookie_life;
2754 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2756 * This socket option is a boolean flag which turns on or off mapped V4
2757 * addresses. If this option is turned on and the socket is type
2758 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2759 * If this option is turned off, then no mapping will be done of V4
2760 * addresses and a user will receive both PF_INET6 and PF_INET type
2761 * addresses on the socket.
2763 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2766 struct sctp_sock *sp = sctp_sk(sk);
2768 if (optlen < sizeof(int))
2770 if (get_user(val, (int __user *)optval))
2781 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2782 * This option will get or set the maximum size to put in any outgoing
2783 * SCTP DATA chunk. If a message is larger than this size it will be
2784 * fragmented by SCTP into the specified size. Note that the underlying
2785 * SCTP implementation may fragment into smaller sized chunks when the
2786 * PMTU of the underlying association is smaller than the value set by
2787 * the user. The default value for this option is '0' which indicates
2788 * the user is NOT limiting fragmentation and only the PMTU will effect
2789 * SCTP's choice of DATA chunk size. Note also that values set larger
2790 * than the maximum size of an IP datagram will effectively let SCTP
2791 * control fragmentation (i.e. the same as setting this option to 0).
2793 * The following structure is used to access and modify this parameter:
2795 * struct sctp_assoc_value {
2796 * sctp_assoc_t assoc_id;
2797 * uint32_t assoc_value;
2800 * assoc_id: This parameter is ignored for one-to-one style sockets.
2801 * For one-to-many style sockets this parameter indicates which
2802 * association the user is performing an action upon. Note that if
2803 * this field's value is zero then the endpoints default value is
2804 * changed (effecting future associations only).
2805 * assoc_value: This parameter specifies the maximum size in bytes.
2807 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2809 struct sctp_assoc_value params;
2810 struct sctp_association *asoc;
2811 struct sctp_sock *sp = sctp_sk(sk);
2814 if (optlen == sizeof(int)) {
2816 "SCTP: Use of int in maxseg socket option deprecated\n");
2818 "SCTP: Use struct sctp_assoc_value instead\n");
2819 if (copy_from_user(&val, optval, optlen))
2821 params.assoc_id = 0;
2822 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2823 if (copy_from_user(¶ms, optval, optlen))
2825 val = params.assoc_value;
2829 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2832 asoc = sctp_id2assoc(sk, params.assoc_id);
2833 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2838 val = asoc->pathmtu;
2839 val -= sp->pf->af->net_header_len;
2840 val -= sizeof(struct sctphdr) +
2841 sizeof(struct sctp_data_chunk);
2844 asoc->frag_point = val;
2846 sp->user_frag = val;
2848 /* Update the frag_point of the existing associations. */
2849 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2850 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2859 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2861 * Requests that the peer mark the enclosed address as the association
2862 * primary. The enclosed address must be one of the association's
2863 * locally bound addresses. The following structure is used to make a
2864 * set primary request:
2866 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2869 struct sctp_sock *sp;
2870 struct sctp_endpoint *ep;
2871 struct sctp_association *asoc = NULL;
2872 struct sctp_setpeerprim prim;
2873 struct sctp_chunk *chunk;
2879 if (!sctp_addip_enable)
2882 if (optlen != sizeof(struct sctp_setpeerprim))
2885 if (copy_from_user(&prim, optval, optlen))
2888 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2892 if (!asoc->peer.asconf_capable)
2895 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2898 if (!sctp_state(asoc, ESTABLISHED))
2901 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2902 return -EADDRNOTAVAIL;
2904 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2905 chunk = sctp_make_asconf_set_prim(asoc,
2906 (union sctp_addr *)&prim.sspp_addr);
2910 err = sctp_send_asconf(asoc, chunk);
2912 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2917 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2920 struct sctp_setadaptation adaptation;
2922 if (optlen != sizeof(struct sctp_setadaptation))
2924 if (copy_from_user(&adaptation, optval, optlen))
2927 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2933 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2935 * The context field in the sctp_sndrcvinfo structure is normally only
2936 * used when a failed message is retrieved holding the value that was
2937 * sent down on the actual send call. This option allows the setting of
2938 * a default context on an association basis that will be received on
2939 * reading messages from the peer. This is especially helpful in the
2940 * one-2-many model for an application to keep some reference to an
2941 * internal state machine that is processing messages on the
2942 * association. Note that the setting of this value only effects
2943 * received messages from the peer and does not effect the value that is
2944 * saved with outbound messages.
2946 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2949 struct sctp_assoc_value params;
2950 struct sctp_sock *sp;
2951 struct sctp_association *asoc;
2953 if (optlen != sizeof(struct sctp_assoc_value))
2955 if (copy_from_user(¶ms, optval, optlen))
2960 if (params.assoc_id != 0) {
2961 asoc = sctp_id2assoc(sk, params.assoc_id);
2964 asoc->default_rcv_context = params.assoc_value;
2966 sp->default_rcv_context = params.assoc_value;
2973 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2975 * This options will at a minimum specify if the implementation is doing
2976 * fragmented interleave. Fragmented interleave, for a one to many
2977 * socket, is when subsequent calls to receive a message may return
2978 * parts of messages from different associations. Some implementations
2979 * may allow you to turn this value on or off. If so, when turned off,
2980 * no fragment interleave will occur (which will cause a head of line
2981 * blocking amongst multiple associations sharing the same one to many
2982 * socket). When this option is turned on, then each receive call may
2983 * come from a different association (thus the user must receive data
2984 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2985 * association each receive belongs to.
2987 * This option takes a boolean value. A non-zero value indicates that
2988 * fragmented interleave is on. A value of zero indicates that
2989 * fragmented interleave is off.
2991 * Note that it is important that an implementation that allows this
2992 * option to be turned on, have it off by default. Otherwise an unaware
2993 * application using the one to many model may become confused and act
2996 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2997 char __user *optval,
3002 if (optlen != sizeof(int))
3004 if (get_user(val, (int __user *)optval))
3007 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3013 * 7.1.25. Set or Get the sctp partial delivery point
3014 * (SCTP_PARTIAL_DELIVERY_POINT)
3015 * This option will set or get the SCTP partial delivery point. This
3016 * point is the size of a message where the partial delivery API will be
3017 * invoked to help free up rwnd space for the peer. Setting this to a
3018 * lower value will cause partial delivery's to happen more often. The
3019 * calls argument is an integer that sets or gets the partial delivery
3022 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3023 char __user *optval,
3028 if (optlen != sizeof(u32))
3030 if (get_user(val, (int __user *)optval))
3033 sctp_sk(sk)->pd_point = val;
3035 return 0; /* is this the right error code? */
3039 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3041 * This option will allow a user to change the maximum burst of packets
3042 * that can be emitted by this association. Note that the default value
3043 * is 4, and some implementations may restrict this setting so that it
3044 * can only be lowered.
3046 * NOTE: This text doesn't seem right. Do this on a socket basis with
3047 * future associations inheriting the socket value.
3049 static int sctp_setsockopt_maxburst(struct sock *sk,
3050 char __user *optval,
3053 struct sctp_assoc_value params;
3054 struct sctp_sock *sp;
3055 struct sctp_association *asoc;
3059 if (optlen < sizeof(int))
3062 if (optlen == sizeof(int)) {
3064 "SCTP: Use of int in max_burst socket option deprecated\n");
3066 "SCTP: Use struct sctp_assoc_value instead\n");
3067 if (copy_from_user(&val, optval, optlen))
3069 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3070 if (copy_from_user(¶ms, optval, optlen))
3072 val = params.assoc_value;
3073 assoc_id = params.assoc_id;
3079 if (assoc_id != 0) {
3080 asoc = sctp_id2assoc(sk, assoc_id);
3083 asoc->max_burst = val;
3085 sp->max_burst = val;
3091 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3093 * This set option adds a chunk type that the user is requesting to be
3094 * received only in an authenticated way. Changes to the list of chunks
3095 * will only effect future associations on the socket.
3097 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3098 char __user *optval,
3101 struct sctp_authchunk val;
3103 if (!sctp_auth_enable)
3106 if (optlen != sizeof(struct sctp_authchunk))
3108 if (copy_from_user(&val, optval, optlen))
3111 switch (val.sauth_chunk) {
3113 case SCTP_CID_INIT_ACK:
3114 case SCTP_CID_SHUTDOWN_COMPLETE:
3119 /* add this chunk id to the endpoint */
3120 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3124 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3126 * This option gets or sets the list of HMAC algorithms that the local
3127 * endpoint requires the peer to use.
3129 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3130 char __user *optval,
3133 struct sctp_hmacalgo *hmacs;
3137 if (!sctp_auth_enable)
3140 if (optlen < sizeof(struct sctp_hmacalgo))
3143 hmacs = kmalloc(optlen, GFP_KERNEL);
3147 if (copy_from_user(hmacs, optval, optlen)) {
3152 idents = hmacs->shmac_num_idents;
3153 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3154 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3159 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3166 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3168 * This option will set a shared secret key which is used to build an
3169 * association shared key.
3171 static int sctp_setsockopt_auth_key(struct sock *sk,
3172 char __user *optval,
3175 struct sctp_authkey *authkey;
3176 struct sctp_association *asoc;
3179 if (!sctp_auth_enable)
3182 if (optlen <= sizeof(struct sctp_authkey))
3185 authkey = kmalloc(optlen, GFP_KERNEL);
3189 if (copy_from_user(authkey, optval, optlen)) {
3194 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3199 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3200 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3205 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3212 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3214 * This option will get or set the active shared key to be used to build
3215 * the association shared key.
3217 static int sctp_setsockopt_active_key(struct sock *sk,
3218 char __user *optval,
3221 struct sctp_authkeyid val;
3222 struct sctp_association *asoc;
3224 if (!sctp_auth_enable)
3227 if (optlen != sizeof(struct sctp_authkeyid))
3229 if (copy_from_user(&val, optval, optlen))
3232 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3233 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3236 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3237 val.scact_keynumber);
3241 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3243 * This set option will delete a shared secret key from use.
3245 static int sctp_setsockopt_del_key(struct sock *sk,
3246 char __user *optval,
3249 struct sctp_authkeyid val;
3250 struct sctp_association *asoc;
3252 if (!sctp_auth_enable)
3255 if (optlen != sizeof(struct sctp_authkeyid))
3257 if (copy_from_user(&val, optval, optlen))
3260 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3261 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3264 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3265 val.scact_keynumber);
3270 /* API 6.2 setsockopt(), getsockopt()
3272 * Applications use setsockopt() and getsockopt() to set or retrieve
3273 * socket options. Socket options are used to change the default
3274 * behavior of sockets calls. They are described in Section 7.
3278 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3279 * int __user *optlen);
3280 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3283 * sd - the socket descript.
3284 * level - set to IPPROTO_SCTP for all SCTP options.
3285 * optname - the option name.
3286 * optval - the buffer to store the value of the option.
3287 * optlen - the size of the buffer.
3289 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3290 char __user *optval, int optlen)
3294 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3297 /* I can hardly begin to describe how wrong this is. This is
3298 * so broken as to be worse than useless. The API draft
3299 * REALLY is NOT helpful here... I am not convinced that the
3300 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3301 * are at all well-founded.
3303 if (level != SOL_SCTP) {
3304 struct sctp_af *af = sctp_sk(sk)->pf->af;
3305 retval = af->setsockopt(sk, level, optname, optval, optlen);
3312 case SCTP_SOCKOPT_BINDX_ADD:
3313 /* 'optlen' is the size of the addresses buffer. */
3314 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3315 optlen, SCTP_BINDX_ADD_ADDR);
3318 case SCTP_SOCKOPT_BINDX_REM:
3319 /* 'optlen' is the size of the addresses buffer. */
3320 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3321 optlen, SCTP_BINDX_REM_ADDR);
3324 case SCTP_SOCKOPT_CONNECTX_OLD:
3325 /* 'optlen' is the size of the addresses buffer. */
3326 retval = sctp_setsockopt_connectx_old(sk,
3327 (struct sockaddr __user *)optval,
3331 case SCTP_SOCKOPT_CONNECTX:
3332 /* 'optlen' is the size of the addresses buffer. */
3333 retval = sctp_setsockopt_connectx(sk,
3334 (struct sockaddr __user *)optval,
3338 case SCTP_DISABLE_FRAGMENTS:
3339 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3343 retval = sctp_setsockopt_events(sk, optval, optlen);
3346 case SCTP_AUTOCLOSE:
3347 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3350 case SCTP_PEER_ADDR_PARAMS:
3351 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3354 case SCTP_DELAYED_ACK:
3355 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3357 case SCTP_PARTIAL_DELIVERY_POINT:
3358 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3362 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3364 case SCTP_DEFAULT_SEND_PARAM:
3365 retval = sctp_setsockopt_default_send_param(sk, optval,
3368 case SCTP_PRIMARY_ADDR:
3369 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3371 case SCTP_SET_PEER_PRIMARY_ADDR:
3372 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3375 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3378 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3380 case SCTP_ASSOCINFO:
3381 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3383 case SCTP_I_WANT_MAPPED_V4_ADDR:
3384 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3387 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3389 case SCTP_ADAPTATION_LAYER:
3390 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3393 retval = sctp_setsockopt_context(sk, optval, optlen);
3395 case SCTP_FRAGMENT_INTERLEAVE:
3396 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3398 case SCTP_MAX_BURST:
3399 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3401 case SCTP_AUTH_CHUNK:
3402 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3404 case SCTP_HMAC_IDENT:
3405 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3408 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3410 case SCTP_AUTH_ACTIVE_KEY:
3411 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3413 case SCTP_AUTH_DELETE_KEY:
3414 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3417 retval = -ENOPROTOOPT;
3421 sctp_release_sock(sk);
3427 /* API 3.1.6 connect() - UDP Style Syntax
3429 * An application may use the connect() call in the UDP model to initiate an
3430 * association without sending data.
3434 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3436 * sd: the socket descriptor to have a new association added to.
3438 * nam: the address structure (either struct sockaddr_in or struct
3439 * sockaddr_in6 defined in RFC2553 [7]).
3441 * len: the size of the address.
3443 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3451 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3452 __func__, sk, addr, addr_len);
3454 /* Validate addr_len before calling common connect/connectx routine. */
3455 af = sctp_get_af_specific(addr->sa_family);
3456 if (!af || addr_len < af->sockaddr_len) {
3459 /* Pass correct addr len to common routine (so it knows there
3460 * is only one address being passed.
3462 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3465 sctp_release_sock(sk);
3469 /* FIXME: Write comments. */
3470 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3472 return -EOPNOTSUPP; /* STUB */
3475 /* 4.1.4 accept() - TCP Style Syntax
3477 * Applications use accept() call to remove an established SCTP
3478 * association from the accept queue of the endpoint. A new socket
3479 * descriptor will be returned from accept() to represent the newly
3480 * formed association.
3482 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3484 struct sctp_sock *sp;
3485 struct sctp_endpoint *ep;
3486 struct sock *newsk = NULL;
3487 struct sctp_association *asoc;
3496 if (!sctp_style(sk, TCP)) {
3497 error = -EOPNOTSUPP;
3501 if (!sctp_sstate(sk, LISTENING)) {
3506 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3508 error = sctp_wait_for_accept(sk, timeo);
3512 /* We treat the list of associations on the endpoint as the accept
3513 * queue and pick the first association on the list.
3515 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3517 newsk = sp->pf->create_accept_sk(sk, asoc);
3523 /* Populate the fields of the newsk from the oldsk and migrate the
3524 * asoc to the newsk.
3526 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3529 sctp_release_sock(sk);
3534 /* The SCTP ioctl handler. */
3535 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3537 return -ENOIOCTLCMD;
3540 /* This is the function which gets called during socket creation to
3541 * initialized the SCTP-specific portion of the sock.
3542 * The sock structure should already be zero-filled memory.
3544 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3546 struct sctp_endpoint *ep;
3547 struct sctp_sock *sp;
3549 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3553 /* Initialize the SCTP per socket area. */
3554 switch (sk->sk_type) {
3555 case SOCK_SEQPACKET:
3556 sp->type = SCTP_SOCKET_UDP;
3559 sp->type = SCTP_SOCKET_TCP;
3562 return -ESOCKTNOSUPPORT;
3565 /* Initialize default send parameters. These parameters can be
3566 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3568 sp->default_stream = 0;
3569 sp->default_ppid = 0;
3570 sp->default_flags = 0;
3571 sp->default_context = 0;
3572 sp->default_timetolive = 0;
3574 sp->default_rcv_context = 0;
3575 sp->max_burst = sctp_max_burst;
3577 /* Initialize default setup parameters. These parameters
3578 * can be modified with the SCTP_INITMSG socket option or
3579 * overridden by the SCTP_INIT CMSG.
3581 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3582 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3583 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3584 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3586 /* Initialize default RTO related parameters. These parameters can
3587 * be modified for with the SCTP_RTOINFO socket option.
3589 sp->rtoinfo.srto_initial = sctp_rto_initial;
3590 sp->rtoinfo.srto_max = sctp_rto_max;
3591 sp->rtoinfo.srto_min = sctp_rto_min;
3593 /* Initialize default association related parameters. These parameters
3594 * can be modified with the SCTP_ASSOCINFO socket option.
3596 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3597 sp->assocparams.sasoc_number_peer_destinations = 0;
3598 sp->assocparams.sasoc_peer_rwnd = 0;
3599 sp->assocparams.sasoc_local_rwnd = 0;
3600 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3602 /* Initialize default event subscriptions. By default, all the
3605 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3607 /* Default Peer Address Parameters. These defaults can
3608 * be modified via SCTP_PEER_ADDR_PARAMS
3610 sp->hbinterval = sctp_hb_interval;
3611 sp->pathmaxrxt = sctp_max_retrans_path;
3612 sp->pathmtu = 0; // allow default discovery
3613 sp->sackdelay = sctp_sack_timeout;
3615 sp->param_flags = SPP_HB_ENABLE |
3617 SPP_SACKDELAY_ENABLE;
3619 /* If enabled no SCTP message fragmentation will be performed.
3620 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3622 sp->disable_fragments = 0;
3624 /* Enable Nagle algorithm by default. */
3627 /* Enable by default. */
3630 /* Auto-close idle associations after the configured
3631 * number of seconds. A value of 0 disables this
3632 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3633 * for UDP-style sockets only.
3637 /* User specified fragmentation limit. */
3640 sp->adaptation_ind = 0;
3642 sp->pf = sctp_get_pf_specific(sk->sk_family);
3644 /* Control variables for partial data delivery. */
3645 atomic_set(&sp->pd_mode, 0);
3646 skb_queue_head_init(&sp->pd_lobby);
3647 sp->frag_interleave = 0;
3649 /* Create a per socket endpoint structure. Even if we
3650 * change the data structure relationships, this may still
3651 * be useful for storing pre-connect address information.
3653 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3660 SCTP_DBG_OBJCNT_INC(sock);
3661 percpu_counter_inc(&sctp_sockets_allocated);
3664 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3670 /* Cleanup any SCTP per socket resources. */
3671 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3673 struct sctp_endpoint *ep;
3675 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3677 /* Release our hold on the endpoint. */
3678 ep = sctp_sk(sk)->ep;
3679 sctp_endpoint_free(ep);
3680 percpu_counter_dec(&sctp_sockets_allocated);
3682 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3686 /* API 4.1.7 shutdown() - TCP Style Syntax
3687 * int shutdown(int socket, int how);
3689 * sd - the socket descriptor of the association to be closed.
3690 * how - Specifies the type of shutdown. The values are
3693 * Disables further receive operations. No SCTP
3694 * protocol action is taken.
3696 * Disables further send operations, and initiates
3697 * the SCTP shutdown sequence.
3699 * Disables further send and receive operations
3700 * and initiates the SCTP shutdown sequence.
3702 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3704 struct sctp_endpoint *ep;
3705 struct sctp_association *asoc;
3707 if (!sctp_style(sk, TCP))
3710 if (how & SEND_SHUTDOWN) {
3711 ep = sctp_sk(sk)->ep;
3712 if (!list_empty(&ep->asocs)) {
3713 asoc = list_entry(ep->asocs.next,
3714 struct sctp_association, asocs);
3715 sctp_primitive_SHUTDOWN(asoc, NULL);
3720 /* 7.2.1 Association Status (SCTP_STATUS)
3722 * Applications can retrieve current status information about an
3723 * association, including association state, peer receiver window size,
3724 * number of unacked data chunks, and number of data chunks pending
3725 * receipt. This information is read-only.
3727 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3728 char __user *optval,
3731 struct sctp_status status;
3732 struct sctp_association *asoc = NULL;
3733 struct sctp_transport *transport;
3734 sctp_assoc_t associd;
3737 if (len < sizeof(status)) {
3742 len = sizeof(status);
3743 if (copy_from_user(&status, optval, len)) {
3748 associd = status.sstat_assoc_id;
3749 asoc = sctp_id2assoc(sk, associd);
3755 transport = asoc->peer.primary_path;
3757 status.sstat_assoc_id = sctp_assoc2id(asoc);
3758 status.sstat_state = asoc->state;
3759 status.sstat_rwnd = asoc->peer.rwnd;
3760 status.sstat_unackdata = asoc->unack_data;
3762 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3763 status.sstat_instrms = asoc->c.sinit_max_instreams;
3764 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3765 status.sstat_fragmentation_point = asoc->frag_point;
3766 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3767 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3768 transport->af_specific->sockaddr_len);
3769 /* Map ipv4 address into v4-mapped-on-v6 address. */
3770 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3771 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3772 status.sstat_primary.spinfo_state = transport->state;
3773 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3774 status.sstat_primary.spinfo_srtt = transport->srtt;
3775 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3776 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3778 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3779 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3781 if (put_user(len, optlen)) {
3786 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3787 len, status.sstat_state, status.sstat_rwnd,
3788 status.sstat_assoc_id);
3790 if (copy_to_user(optval, &status, len)) {
3800 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3802 * Applications can retrieve information about a specific peer address
3803 * of an association, including its reachability state, congestion
3804 * window, and retransmission timer values. This information is
3807 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3808 char __user *optval,
3811 struct sctp_paddrinfo pinfo;
3812 struct sctp_transport *transport;
3815 if (len < sizeof(pinfo)) {
3820 len = sizeof(pinfo);
3821 if (copy_from_user(&pinfo, optval, len)) {
3826 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3827 pinfo.spinfo_assoc_id);
3831 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3832 pinfo.spinfo_state = transport->state;
3833 pinfo.spinfo_cwnd = transport->cwnd;
3834 pinfo.spinfo_srtt = transport->srtt;
3835 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3836 pinfo.spinfo_mtu = transport->pathmtu;
3838 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3839 pinfo.spinfo_state = SCTP_ACTIVE;
3841 if (put_user(len, optlen)) {
3846 if (copy_to_user(optval, &pinfo, len)) {
3855 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3857 * This option is a on/off flag. If enabled no SCTP message
3858 * fragmentation will be performed. Instead if a message being sent
3859 * exceeds the current PMTU size, the message will NOT be sent and
3860 * instead a error will be indicated to the user.
3862 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3863 char __user *optval, int __user *optlen)
3867 if (len < sizeof(int))
3871 val = (sctp_sk(sk)->disable_fragments == 1);
3872 if (put_user(len, optlen))
3874 if (copy_to_user(optval, &val, len))
3879 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3881 * This socket option is used to specify various notifications and
3882 * ancillary data the user wishes to receive.
3884 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3887 if (len < sizeof(struct sctp_event_subscribe))
3889 len = sizeof(struct sctp_event_subscribe);
3890 if (put_user(len, optlen))
3892 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3897 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3899 * This socket option is applicable to the UDP-style socket only. When
3900 * set it will cause associations that are idle for more than the
3901 * specified number of seconds to automatically close. An association
3902 * being idle is defined an association that has NOT sent or received
3903 * user data. The special value of '0' indicates that no automatic
3904 * close of any associations should be performed. The option expects an
3905 * integer defining the number of seconds of idle time before an
3906 * association is closed.
3908 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3910 /* Applicable to UDP-style socket only */
3911 if (sctp_style(sk, TCP))
3913 if (len < sizeof(int))
3916 if (put_user(len, optlen))
3918 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3923 /* Helper routine to branch off an association to a new socket. */
3924 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3925 struct socket **sockp)
3927 struct sock *sk = asoc->base.sk;
3928 struct socket *sock;
3929 struct inet_sock *inetsk;
3933 /* An association cannot be branched off from an already peeled-off
3934 * socket, nor is this supported for tcp style sockets.
3936 if (!sctp_style(sk, UDP))
3939 /* Create a new socket. */
3940 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3944 /* Populate the fields of the newsk from the oldsk and migrate the
3945 * asoc to the newsk.
3947 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3949 /* Make peeled-off sockets more like 1-1 accepted sockets.
3950 * Set the daddr and initialize id to something more random
3952 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3953 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3954 inetsk = inet_sk(sock->sk);
3955 inetsk->id = asoc->next_tsn ^ jiffies;
3962 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3964 sctp_peeloff_arg_t peeloff;
3965 struct socket *newsock;
3967 struct sctp_association *asoc;
3969 if (len < sizeof(sctp_peeloff_arg_t))
3971 len = sizeof(sctp_peeloff_arg_t);
3972 if (copy_from_user(&peeloff, optval, len))
3975 asoc = sctp_id2assoc(sk, peeloff.associd);
3981 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3983 retval = sctp_do_peeloff(asoc, &newsock);
3987 /* Map the socket to an unused fd that can be returned to the user. */
3988 retval = sock_map_fd(newsock, 0);
3990 sock_release(newsock);
3994 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3995 __func__, sk, asoc, newsock->sk, retval);
3997 /* Return the fd mapped to the new socket. */
3998 peeloff.sd = retval;
3999 if (put_user(len, optlen))
4001 if (copy_to_user(optval, &peeloff, len))
4008 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4010 * Applications can enable or disable heartbeats for any peer address of
4011 * an association, modify an address's heartbeat interval, force a
4012 * heartbeat to be sent immediately, and adjust the address's maximum
4013 * number of retransmissions sent before an address is considered
4014 * unreachable. The following structure is used to access and modify an
4015 * address's parameters:
4017 * struct sctp_paddrparams {
4018 * sctp_assoc_t spp_assoc_id;
4019 * struct sockaddr_storage spp_address;
4020 * uint32_t spp_hbinterval;
4021 * uint16_t spp_pathmaxrxt;
4022 * uint32_t spp_pathmtu;
4023 * uint32_t spp_sackdelay;
4024 * uint32_t spp_flags;
4027 * spp_assoc_id - (one-to-many style socket) This is filled in the
4028 * application, and identifies the association for
4030 * spp_address - This specifies which address is of interest.
4031 * spp_hbinterval - This contains the value of the heartbeat interval,
4032 * in milliseconds. If a value of zero
4033 * is present in this field then no changes are to
4034 * be made to this parameter.
4035 * spp_pathmaxrxt - This contains the maximum number of
4036 * retransmissions before this address shall be
4037 * considered unreachable. If a value of zero
4038 * is present in this field then no changes are to
4039 * be made to this parameter.
4040 * spp_pathmtu - When Path MTU discovery is disabled the value
4041 * specified here will be the "fixed" path mtu.
4042 * Note that if the spp_address field is empty
4043 * then all associations on this address will
4044 * have this fixed path mtu set upon them.
4046 * spp_sackdelay - When delayed sack is enabled, this value specifies
4047 * the number of milliseconds that sacks will be delayed
4048 * for. This value will apply to all addresses of an
4049 * association if the spp_address field is empty. Note
4050 * also, that if delayed sack is enabled and this
4051 * value is set to 0, no change is made to the last
4052 * recorded delayed sack timer value.
4054 * spp_flags - These flags are used to control various features
4055 * on an association. The flag field may contain
4056 * zero or more of the following options.
4058 * SPP_HB_ENABLE - Enable heartbeats on the
4059 * specified address. Note that if the address
4060 * field is empty all addresses for the association
4061 * have heartbeats enabled upon them.
4063 * SPP_HB_DISABLE - Disable heartbeats on the
4064 * speicifed address. Note that if the address
4065 * field is empty all addresses for the association
4066 * will have their heartbeats disabled. Note also
4067 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4068 * mutually exclusive, only one of these two should
4069 * be specified. Enabling both fields will have
4070 * undetermined results.
4072 * SPP_HB_DEMAND - Request a user initiated heartbeat
4073 * to be made immediately.
4075 * SPP_PMTUD_ENABLE - This field will enable PMTU
4076 * discovery upon the specified address. Note that
4077 * if the address feild is empty then all addresses
4078 * on the association are effected.
4080 * SPP_PMTUD_DISABLE - This field will disable PMTU
4081 * discovery upon the specified address. Note that
4082 * if the address feild is empty then all addresses
4083 * on the association are effected. Not also that
4084 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4085 * exclusive. Enabling both will have undetermined
4088 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4089 * on delayed sack. The time specified in spp_sackdelay
4090 * is used to specify the sack delay for this address. Note
4091 * that if spp_address is empty then all addresses will
4092 * enable delayed sack and take on the sack delay
4093 * value specified in spp_sackdelay.
4094 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4095 * off delayed sack. If the spp_address field is blank then
4096 * delayed sack is disabled for the entire association. Note
4097 * also that this field is mutually exclusive to
4098 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4101 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4102 char __user *optval, int __user *optlen)
4104 struct sctp_paddrparams params;
4105 struct sctp_transport *trans = NULL;
4106 struct sctp_association *asoc = NULL;
4107 struct sctp_sock *sp = sctp_sk(sk);
4109 if (len < sizeof(struct sctp_paddrparams))
4111 len = sizeof(struct sctp_paddrparams);
4112 if (copy_from_user(¶ms, optval, len))
4115 /* If an address other than INADDR_ANY is specified, and
4116 * no transport is found, then the request is invalid.
4118 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4119 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4120 params.spp_assoc_id);
4122 SCTP_DEBUG_PRINTK("Failed no transport\n");
4127 /* Get association, if assoc_id != 0 and the socket is a one
4128 * to many style socket, and an association was not found, then
4129 * the id was invalid.
4131 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4132 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4133 SCTP_DEBUG_PRINTK("Failed no association\n");
4138 /* Fetch transport values. */
4139 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4140 params.spp_pathmtu = trans->pathmtu;
4141 params.spp_pathmaxrxt = trans->pathmaxrxt;
4142 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4144 /*draft-11 doesn't say what to return in spp_flags*/
4145 params.spp_flags = trans->param_flags;
4147 /* Fetch association values. */
4148 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4149 params.spp_pathmtu = asoc->pathmtu;
4150 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4151 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4153 /*draft-11 doesn't say what to return in spp_flags*/
4154 params.spp_flags = asoc->param_flags;
4156 /* Fetch socket values. */
4157 params.spp_hbinterval = sp->hbinterval;
4158 params.spp_pathmtu = sp->pathmtu;
4159 params.spp_sackdelay = sp->sackdelay;
4160 params.spp_pathmaxrxt = sp->pathmaxrxt;
4162 /*draft-11 doesn't say what to return in spp_flags*/
4163 params.spp_flags = sp->param_flags;
4166 if (copy_to_user(optval, ¶ms, len))
4169 if (put_user(len, optlen))
4176 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4178 * This option will effect the way delayed acks are performed. This
4179 * option allows you to get or set the delayed ack time, in
4180 * milliseconds. It also allows changing the delayed ack frequency.
4181 * Changing the frequency to 1 disables the delayed sack algorithm. If
4182 * the assoc_id is 0, then this sets or gets the endpoints default
4183 * values. If the assoc_id field is non-zero, then the set or get
4184 * effects the specified association for the one to many model (the
4185 * assoc_id field is ignored by the one to one model). Note that if
4186 * sack_delay or sack_freq are 0 when setting this option, then the
4187 * current values will remain unchanged.
4189 * struct sctp_sack_info {
4190 * sctp_assoc_t sack_assoc_id;
4191 * uint32_t sack_delay;
4192 * uint32_t sack_freq;
4195 * sack_assoc_id - This parameter, indicates which association the user
4196 * is performing an action upon. Note that if this field's value is
4197 * zero then the endpoints default value is changed (effecting future
4198 * associations only).
4200 * sack_delay - This parameter contains the number of milliseconds that
4201 * the user is requesting the delayed ACK timer be set to. Note that
4202 * this value is defined in the standard to be between 200 and 500
4205 * sack_freq - This parameter contains the number of packets that must
4206 * be received before a sack is sent without waiting for the delay
4207 * timer to expire. The default value for this is 2, setting this
4208 * value to 1 will disable the delayed sack algorithm.
4210 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4211 char __user *optval,
4214 struct sctp_sack_info params;
4215 struct sctp_association *asoc = NULL;
4216 struct sctp_sock *sp = sctp_sk(sk);
4218 if (len >= sizeof(struct sctp_sack_info)) {
4219 len = sizeof(struct sctp_sack_info);
4221 if (copy_from_user(¶ms, optval, len))
4223 } else if (len == sizeof(struct sctp_assoc_value)) {
4224 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4225 "in delayed_ack socket option deprecated\n");
4226 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4227 if (copy_from_user(¶ms, optval, len))
4232 /* Get association, if sack_assoc_id != 0 and the socket is a one
4233 * to many style socket, and an association was not found, then
4234 * the id was invalid.
4236 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4237 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4241 /* Fetch association values. */
4242 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4243 params.sack_delay = jiffies_to_msecs(
4245 params.sack_freq = asoc->sackfreq;
4248 params.sack_delay = 0;
4249 params.sack_freq = 1;
4252 /* Fetch socket values. */
4253 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4254 params.sack_delay = sp->sackdelay;
4255 params.sack_freq = sp->sackfreq;
4257 params.sack_delay = 0;
4258 params.sack_freq = 1;
4262 if (copy_to_user(optval, ¶ms, len))
4265 if (put_user(len, optlen))
4271 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4273 * Applications can specify protocol parameters for the default association
4274 * initialization. The option name argument to setsockopt() and getsockopt()
4277 * Setting initialization parameters is effective only on an unconnected
4278 * socket (for UDP-style sockets only future associations are effected
4279 * by the change). With TCP-style sockets, this option is inherited by
4280 * sockets derived from a listener socket.
4282 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4284 if (len < sizeof(struct sctp_initmsg))
4286 len = sizeof(struct sctp_initmsg);
4287 if (put_user(len, optlen))
4289 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4294 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4295 char __user *optval,
4299 struct sctp_association *asoc;
4300 struct list_head *pos;
4303 if (len < sizeof(sctp_assoc_t))
4306 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4309 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4310 "socket option deprecated\n");
4311 /* For UDP-style sockets, id specifies the association to query. */
4312 asoc = sctp_id2assoc(sk, id);
4316 list_for_each(pos, &asoc->peer.transport_addr_list) {
4324 * Old API for getting list of peer addresses. Does not work for 32-bit
4325 * programs running on a 64-bit kernel
4327 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4328 char __user *optval,
4331 struct sctp_association *asoc;
4333 struct sctp_getaddrs_old getaddrs;
4334 struct sctp_transport *from;
4336 union sctp_addr temp;
4337 struct sctp_sock *sp = sctp_sk(sk);
4340 if (len < sizeof(struct sctp_getaddrs_old))
4343 len = sizeof(struct sctp_getaddrs_old);
4345 if (copy_from_user(&getaddrs, optval, len))
4348 if (getaddrs.addr_num <= 0) return -EINVAL;
4350 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4351 "socket option deprecated\n");
4353 /* For UDP-style sockets, id specifies the association to query. */
4354 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4358 to = (void __user *)getaddrs.addrs;
4359 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4361 memcpy(&temp, &from->ipaddr, sizeof(temp));
4362 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4363 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4364 if (copy_to_user(to, &temp, addrlen))
4368 if (cnt >= getaddrs.addr_num) break;
4370 getaddrs.addr_num = cnt;
4371 if (put_user(len, optlen))
4373 if (copy_to_user(optval, &getaddrs, len))
4379 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4380 char __user *optval, int __user *optlen)
4382 struct sctp_association *asoc;
4384 struct sctp_getaddrs getaddrs;
4385 struct sctp_transport *from;
4387 union sctp_addr temp;
4388 struct sctp_sock *sp = sctp_sk(sk);
4393 if (len < sizeof(struct sctp_getaddrs))
4396 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4399 /* For UDP-style sockets, id specifies the association to query. */
4400 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4404 to = optval + offsetof(struct sctp_getaddrs,addrs);
4405 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4407 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4409 memcpy(&temp, &from->ipaddr, sizeof(temp));
4410 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4411 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4412 if (space_left < addrlen)
4414 if (copy_to_user(to, &temp, addrlen))
4418 space_left -= addrlen;
4421 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4423 bytes_copied = ((char __user *)to) - optval;
4424 if (put_user(bytes_copied, optlen))
4430 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4431 char __user *optval,
4435 struct sctp_bind_addr *bp;
4436 struct sctp_association *asoc;
4437 struct sctp_sockaddr_entry *addr;
4440 if (len < sizeof(sctp_assoc_t))
4443 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4446 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4447 "socket option deprecated\n");
4450 * For UDP-style sockets, id specifies the association to query.
4451 * If the id field is set to the value '0' then the locally bound
4452 * addresses are returned without regard to any particular
4456 bp = &sctp_sk(sk)->ep->base.bind_addr;
4458 asoc = sctp_id2assoc(sk, id);
4461 bp = &asoc->base.bind_addr;
4464 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4465 * addresses from the global local address list.
4467 if (sctp_list_single_entry(&bp->address_list)) {
4468 addr = list_entry(bp->address_list.next,
4469 struct sctp_sockaddr_entry, list);
4470 if (sctp_is_any(sk, &addr->a)) {
4472 list_for_each_entry_rcu(addr,
4473 &sctp_local_addr_list, list) {
4477 if ((PF_INET == sk->sk_family) &&
4478 (AF_INET6 == addr->a.sa.sa_family))
4481 if ((PF_INET6 == sk->sk_family) &&
4482 inet_v6_ipv6only(sk) &&
4483 (AF_INET == addr->a.sa.sa_family))
4495 /* Protection on the bound address list is not needed,
4496 * since in the socket option context we hold the socket lock,
4497 * so there is no way that the bound address list can change.
4499 list_for_each_entry(addr, &bp->address_list, list) {
4506 /* Helper function that copies local addresses to user and returns the number
4507 * of addresses copied.
4509 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4510 int max_addrs, void *to,
4513 struct sctp_sockaddr_entry *addr;
4514 union sctp_addr temp;
4519 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4523 if ((PF_INET == sk->sk_family) &&
4524 (AF_INET6 == addr->a.sa.sa_family))
4526 if ((PF_INET6 == sk->sk_family) &&
4527 inet_v6_ipv6only(sk) &&
4528 (AF_INET == addr->a.sa.sa_family))
4530 memcpy(&temp, &addr->a, sizeof(temp));
4531 if (!temp.v4.sin_port)
4532 temp.v4.sin_port = htons(port);
4534 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4536 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4537 memcpy(to, &temp, addrlen);
4540 *bytes_copied += addrlen;
4542 if (cnt >= max_addrs) break;
4549 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4550 size_t space_left, int *bytes_copied)
4552 struct sctp_sockaddr_entry *addr;
4553 union sctp_addr temp;
4558 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4562 if ((PF_INET == sk->sk_family) &&
4563 (AF_INET6 == addr->a.sa.sa_family))
4565 if ((PF_INET6 == sk->sk_family) &&
4566 inet_v6_ipv6only(sk) &&
4567 (AF_INET == addr->a.sa.sa_family))
4569 memcpy(&temp, &addr->a, sizeof(temp));
4570 if (!temp.v4.sin_port)
4571 temp.v4.sin_port = htons(port);
4573 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4575 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4576 if (space_left < addrlen) {
4580 memcpy(to, &temp, addrlen);
4584 space_left -= addrlen;
4585 *bytes_copied += addrlen;
4592 /* Old API for getting list of local addresses. Does not work for 32-bit
4593 * programs running on a 64-bit kernel
4595 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4596 char __user *optval, int __user *optlen)
4598 struct sctp_bind_addr *bp;
4599 struct sctp_association *asoc;
4601 struct sctp_getaddrs_old getaddrs;
4602 struct sctp_sockaddr_entry *addr;
4604 union sctp_addr temp;
4605 struct sctp_sock *sp = sctp_sk(sk);
4610 int bytes_copied = 0;
4612 if (len < sizeof(struct sctp_getaddrs_old))
4615 len = sizeof(struct sctp_getaddrs_old);
4616 if (copy_from_user(&getaddrs, optval, len))
4619 if (getaddrs.addr_num <= 0 ||
4620 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4623 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4624 "socket option deprecated\n");
4627 * For UDP-style sockets, id specifies the association to query.
4628 * If the id field is set to the value '0' then the locally bound
4629 * addresses are returned without regard to any particular
4632 if (0 == getaddrs.assoc_id) {
4633 bp = &sctp_sk(sk)->ep->base.bind_addr;
4635 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4638 bp = &asoc->base.bind_addr;
4641 to = getaddrs.addrs;
4643 /* Allocate space for a local instance of packed array to hold all
4644 * the data. We store addresses here first and then put write them
4645 * to the user in one shot.
4647 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4652 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4653 * addresses from the global local address list.
4655 if (sctp_list_single_entry(&bp->address_list)) {
4656 addr = list_entry(bp->address_list.next,
4657 struct sctp_sockaddr_entry, list);
4658 if (sctp_is_any(sk, &addr->a)) {
4659 cnt = sctp_copy_laddrs_old(sk, bp->port,
4661 addrs, &bytes_copied);
4667 /* Protection on the bound address list is not needed since
4668 * in the socket option context we hold a socket lock and
4669 * thus the bound address list can't change.
4671 list_for_each_entry(addr, &bp->address_list, list) {
4672 memcpy(&temp, &addr->a, sizeof(temp));
4673 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4674 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4675 memcpy(buf, &temp, addrlen);
4677 bytes_copied += addrlen;
4679 if (cnt >= getaddrs.addr_num) break;
4683 /* copy the entire address list into the user provided space */
4684 if (copy_to_user(to, addrs, bytes_copied)) {
4689 /* copy the leading structure back to user */
4690 getaddrs.addr_num = cnt;
4691 if (copy_to_user(optval, &getaddrs, len))
4699 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4700 char __user *optval, int __user *optlen)
4702 struct sctp_bind_addr *bp;
4703 struct sctp_association *asoc;
4705 struct sctp_getaddrs getaddrs;
4706 struct sctp_sockaddr_entry *addr;
4708 union sctp_addr temp;
4709 struct sctp_sock *sp = sctp_sk(sk);
4713 int bytes_copied = 0;
4717 if (len < sizeof(struct sctp_getaddrs))
4720 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4724 * For UDP-style sockets, id specifies the association to query.
4725 * If the id field is set to the value '0' then the locally bound
4726 * addresses are returned without regard to any particular
4729 if (0 == getaddrs.assoc_id) {
4730 bp = &sctp_sk(sk)->ep->base.bind_addr;
4732 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4735 bp = &asoc->base.bind_addr;
4738 to = optval + offsetof(struct sctp_getaddrs,addrs);
4739 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4741 addrs = kmalloc(space_left, GFP_KERNEL);
4745 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4746 * addresses from the global local address list.
4748 if (sctp_list_single_entry(&bp->address_list)) {
4749 addr = list_entry(bp->address_list.next,
4750 struct sctp_sockaddr_entry, list);
4751 if (sctp_is_any(sk, &addr->a)) {
4752 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4753 space_left, &bytes_copied);
4763 /* Protection on the bound address list is not needed since
4764 * in the socket option context we hold a socket lock and
4765 * thus the bound address list can't change.
4767 list_for_each_entry(addr, &bp->address_list, list) {
4768 memcpy(&temp, &addr->a, sizeof(temp));
4769 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4770 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4771 if (space_left < addrlen) {
4772 err = -ENOMEM; /*fixme: right error?*/
4775 memcpy(buf, &temp, addrlen);
4777 bytes_copied += addrlen;
4779 space_left -= addrlen;
4783 if (copy_to_user(to, addrs, bytes_copied)) {
4787 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4791 if (put_user(bytes_copied, optlen))
4798 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4800 * Requests that the local SCTP stack use the enclosed peer address as
4801 * the association primary. The enclosed address must be one of the
4802 * association peer's addresses.
4804 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4805 char __user *optval, int __user *optlen)
4807 struct sctp_prim prim;
4808 struct sctp_association *asoc;
4809 struct sctp_sock *sp = sctp_sk(sk);
4811 if (len < sizeof(struct sctp_prim))
4814 len = sizeof(struct sctp_prim);
4816 if (copy_from_user(&prim, optval, len))
4819 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4823 if (!asoc->peer.primary_path)
4826 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4827 asoc->peer.primary_path->af_specific->sockaddr_len);
4829 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4830 (union sctp_addr *)&prim.ssp_addr);
4832 if (put_user(len, optlen))
4834 if (copy_to_user(optval, &prim, len))
4841 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4843 * Requests that the local endpoint set the specified Adaptation Layer
4844 * Indication parameter for all future INIT and INIT-ACK exchanges.
4846 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4847 char __user *optval, int __user *optlen)
4849 struct sctp_setadaptation adaptation;
4851 if (len < sizeof(struct sctp_setadaptation))
4854 len = sizeof(struct sctp_setadaptation);
4856 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4858 if (put_user(len, optlen))
4860 if (copy_to_user(optval, &adaptation, len))
4868 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4870 * Applications that wish to use the sendto() system call may wish to
4871 * specify a default set of parameters that would normally be supplied
4872 * through the inclusion of ancillary data. This socket option allows
4873 * such an application to set the default sctp_sndrcvinfo structure.
4876 * The application that wishes to use this socket option simply passes
4877 * in to this call the sctp_sndrcvinfo structure defined in Section
4878 * 5.2.2) The input parameters accepted by this call include
4879 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4880 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4881 * to this call if the caller is using the UDP model.
4883 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4885 static int sctp_getsockopt_default_send_param(struct sock *sk,
4886 int len, char __user *optval,
4889 struct sctp_sndrcvinfo info;
4890 struct sctp_association *asoc;
4891 struct sctp_sock *sp = sctp_sk(sk);
4893 if (len < sizeof(struct sctp_sndrcvinfo))
4896 len = sizeof(struct sctp_sndrcvinfo);
4898 if (copy_from_user(&info, optval, len))
4901 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4902 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4906 info.sinfo_stream = asoc->default_stream;
4907 info.sinfo_flags = asoc->default_flags;
4908 info.sinfo_ppid = asoc->default_ppid;
4909 info.sinfo_context = asoc->default_context;
4910 info.sinfo_timetolive = asoc->default_timetolive;
4912 info.sinfo_stream = sp->default_stream;
4913 info.sinfo_flags = sp->default_flags;
4914 info.sinfo_ppid = sp->default_ppid;
4915 info.sinfo_context = sp->default_context;
4916 info.sinfo_timetolive = sp->default_timetolive;
4919 if (put_user(len, optlen))
4921 if (copy_to_user(optval, &info, len))
4929 * 7.1.5 SCTP_NODELAY
4931 * Turn on/off any Nagle-like algorithm. This means that packets are
4932 * generally sent as soon as possible and no unnecessary delays are
4933 * introduced, at the cost of more packets in the network. Expects an
4934 * integer boolean flag.
4937 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4938 char __user *optval, int __user *optlen)
4942 if (len < sizeof(int))
4946 val = (sctp_sk(sk)->nodelay == 1);
4947 if (put_user(len, optlen))
4949 if (copy_to_user(optval, &val, len))
4956 * 7.1.1 SCTP_RTOINFO
4958 * The protocol parameters used to initialize and bound retransmission
4959 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4960 * and modify these parameters.
4961 * All parameters are time values, in milliseconds. A value of 0, when
4962 * modifying the parameters, indicates that the current value should not
4966 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4967 char __user *optval,
4968 int __user *optlen) {
4969 struct sctp_rtoinfo rtoinfo;
4970 struct sctp_association *asoc;
4972 if (len < sizeof (struct sctp_rtoinfo))
4975 len = sizeof(struct sctp_rtoinfo);
4977 if (copy_from_user(&rtoinfo, optval, len))
4980 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4982 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4985 /* Values corresponding to the specific association. */
4987 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4988 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4989 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4991 /* Values corresponding to the endpoint. */
4992 struct sctp_sock *sp = sctp_sk(sk);
4994 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4995 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4996 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4999 if (put_user(len, optlen))
5002 if (copy_to_user(optval, &rtoinfo, len))
5010 * 7.1.2 SCTP_ASSOCINFO
5012 * This option is used to tune the maximum retransmission attempts
5013 * of the association.
5014 * Returns an error if the new association retransmission value is
5015 * greater than the sum of the retransmission value of the peer.
5016 * See [SCTP] for more information.
5019 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5020 char __user *optval,
5024 struct sctp_assocparams assocparams;
5025 struct sctp_association *asoc;
5026 struct list_head *pos;
5029 if (len < sizeof (struct sctp_assocparams))
5032 len = sizeof(struct sctp_assocparams);
5034 if (copy_from_user(&assocparams, optval, len))
5037 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5039 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5042 /* Values correspoinding to the specific association */
5044 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5045 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5046 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5047 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5049 (asoc->cookie_life.tv_usec
5052 list_for_each(pos, &asoc->peer.transport_addr_list) {
5056 assocparams.sasoc_number_peer_destinations = cnt;
5058 /* Values corresponding to the endpoint */
5059 struct sctp_sock *sp = sctp_sk(sk);
5061 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5062 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5063 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5064 assocparams.sasoc_cookie_life =
5065 sp->assocparams.sasoc_cookie_life;
5066 assocparams.sasoc_number_peer_destinations =
5068 sasoc_number_peer_destinations;
5071 if (put_user(len, optlen))
5074 if (copy_to_user(optval, &assocparams, len))
5081 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5083 * This socket option is a boolean flag which turns on or off mapped V4
5084 * addresses. If this option is turned on and the socket is type
5085 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5086 * If this option is turned off, then no mapping will be done of V4
5087 * addresses and a user will receive both PF_INET6 and PF_INET type
5088 * addresses on the socket.
5090 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5091 char __user *optval, int __user *optlen)
5094 struct sctp_sock *sp = sctp_sk(sk);
5096 if (len < sizeof(int))
5101 if (put_user(len, optlen))
5103 if (copy_to_user(optval, &val, len))
5110 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5111 * (chapter and verse is quoted at sctp_setsockopt_context())
5113 static int sctp_getsockopt_context(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5116 struct sctp_assoc_value params;
5117 struct sctp_sock *sp;
5118 struct sctp_association *asoc;
5120 if (len < sizeof(struct sctp_assoc_value))
5123 len = sizeof(struct sctp_assoc_value);
5125 if (copy_from_user(¶ms, optval, len))
5130 if (params.assoc_id != 0) {
5131 asoc = sctp_id2assoc(sk, params.assoc_id);
5134 params.assoc_value = asoc->default_rcv_context;
5136 params.assoc_value = sp->default_rcv_context;
5139 if (put_user(len, optlen))
5141 if (copy_to_user(optval, ¶ms, len))
5148 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5149 * This option will get or set the maximum size to put in any outgoing
5150 * SCTP DATA chunk. If a message is larger than this size it will be
5151 * fragmented by SCTP into the specified size. Note that the underlying
5152 * SCTP implementation may fragment into smaller sized chunks when the
5153 * PMTU of the underlying association is smaller than the value set by
5154 * the user. The default value for this option is '0' which indicates
5155 * the user is NOT limiting fragmentation and only the PMTU will effect
5156 * SCTP's choice of DATA chunk size. Note also that values set larger
5157 * than the maximum size of an IP datagram will effectively let SCTP
5158 * control fragmentation (i.e. the same as setting this option to 0).
5160 * The following structure is used to access and modify this parameter:
5162 * struct sctp_assoc_value {
5163 * sctp_assoc_t assoc_id;
5164 * uint32_t assoc_value;
5167 * assoc_id: This parameter is ignored for one-to-one style sockets.
5168 * For one-to-many style sockets this parameter indicates which
5169 * association the user is performing an action upon. Note that if
5170 * this field's value is zero then the endpoints default value is
5171 * changed (effecting future associations only).
5172 * assoc_value: This parameter specifies the maximum size in bytes.
5174 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5175 char __user *optval, int __user *optlen)
5177 struct sctp_assoc_value params;
5178 struct sctp_association *asoc;
5180 if (len == sizeof(int)) {
5182 "SCTP: Use of int in maxseg socket option deprecated\n");
5184 "SCTP: Use struct sctp_assoc_value instead\n");
5185 params.assoc_id = 0;
5186 } else if (len >= sizeof(struct sctp_assoc_value)) {
5187 len = sizeof(struct sctp_assoc_value);
5188 if (copy_from_user(¶ms, optval, sizeof(params)))
5193 asoc = sctp_id2assoc(sk, params.assoc_id);
5194 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5198 params.assoc_value = asoc->frag_point;
5200 params.assoc_value = sctp_sk(sk)->user_frag;
5202 if (put_user(len, optlen))
5204 if (len == sizeof(int)) {
5205 if (copy_to_user(optval, ¶ms.assoc_value, len))
5208 if (copy_to_user(optval, ¶ms, len))
5216 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5217 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5219 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5220 char __user *optval, int __user *optlen)
5224 if (len < sizeof(int))
5229 val = sctp_sk(sk)->frag_interleave;
5230 if (put_user(len, optlen))
5232 if (copy_to_user(optval, &val, len))
5239 * 7.1.25. Set or Get the sctp partial delivery point
5240 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5242 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5243 char __user *optval,
5248 if (len < sizeof(u32))
5253 val = sctp_sk(sk)->pd_point;
5254 if (put_user(len, optlen))
5256 if (copy_to_user(optval, &val, len))
5263 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5264 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5266 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5267 char __user *optval,
5270 struct sctp_assoc_value params;
5271 struct sctp_sock *sp;
5272 struct sctp_association *asoc;
5274 if (len < sizeof(int))
5277 if (len == sizeof(int)) {
5279 "SCTP: Use of int in max_burst socket option deprecated\n");
5281 "SCTP: Use struct sctp_assoc_value instead\n");
5282 params.assoc_id = 0;
5283 } else if (len == sizeof (struct sctp_assoc_value)) {
5284 if (copy_from_user(¶ms, optval, len))
5291 if (params.assoc_id != 0) {
5292 asoc = sctp_id2assoc(sk, params.assoc_id);
5295 params.assoc_value = asoc->max_burst;
5297 params.assoc_value = sp->max_burst;
5299 if (len == sizeof(int)) {
5300 if (copy_to_user(optval, ¶ms.assoc_value, len))
5303 if (copy_to_user(optval, ¶ms, len))
5311 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5312 char __user *optval, int __user *optlen)
5314 struct sctp_hmacalgo __user *p = (void __user *)optval;
5315 struct sctp_hmac_algo_param *hmacs;
5319 if (!sctp_auth_enable)
5322 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5323 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5325 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5328 len = sizeof(struct sctp_hmacalgo) + data_len;
5329 num_idents = data_len / sizeof(u16);
5331 if (put_user(len, optlen))
5333 if (put_user(num_idents, &p->shmac_num_idents))
5335 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5340 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5341 char __user *optval, int __user *optlen)
5343 struct sctp_authkeyid val;
5344 struct sctp_association *asoc;
5346 if (!sctp_auth_enable)
5349 if (len < sizeof(struct sctp_authkeyid))
5351 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5354 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5355 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5359 val.scact_keynumber = asoc->active_key_id;
5361 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5363 len = sizeof(struct sctp_authkeyid);
5364 if (put_user(len, optlen))
5366 if (copy_to_user(optval, &val, len))
5372 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5373 char __user *optval, int __user *optlen)
5375 struct sctp_authchunks __user *p = (void __user *)optval;
5376 struct sctp_authchunks val;
5377 struct sctp_association *asoc;
5378 struct sctp_chunks_param *ch;
5382 if (!sctp_auth_enable)
5385 if (len < sizeof(struct sctp_authchunks))
5388 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5391 to = p->gauth_chunks;
5392 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5396 ch = asoc->peer.peer_chunks;
5400 /* See if the user provided enough room for all the data */
5401 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5402 if (len < num_chunks)
5405 if (copy_to_user(to, ch->chunks, num_chunks))
5408 len = sizeof(struct sctp_authchunks) + num_chunks;
5409 if (put_user(len, optlen)) return -EFAULT;
5410 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5415 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5416 char __user *optval, int __user *optlen)
5418 struct sctp_authchunks __user *p = (void __user *)optval;
5419 struct sctp_authchunks val;
5420 struct sctp_association *asoc;
5421 struct sctp_chunks_param *ch;
5425 if (!sctp_auth_enable)
5428 if (len < sizeof(struct sctp_authchunks))
5431 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5434 to = p->gauth_chunks;
5435 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5436 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5440 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5442 ch = sctp_sk(sk)->ep->auth_chunk_list;
5447 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5448 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5451 if (copy_to_user(to, ch->chunks, num_chunks))
5454 len = sizeof(struct sctp_authchunks) + num_chunks;
5455 if (put_user(len, optlen))
5457 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5463 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5464 char __user *optval, int __user *optlen)
5469 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5472 /* I can hardly begin to describe how wrong this is. This is
5473 * so broken as to be worse than useless. The API draft
5474 * REALLY is NOT helpful here... I am not convinced that the
5475 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5476 * are at all well-founded.
5478 if (level != SOL_SCTP) {
5479 struct sctp_af *af = sctp_sk(sk)->pf->af;
5481 retval = af->getsockopt(sk, level, optname, optval, optlen);
5485 if (get_user(len, optlen))
5492 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5494 case SCTP_DISABLE_FRAGMENTS:
5495 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5499 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5501 case SCTP_AUTOCLOSE:
5502 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5504 case SCTP_SOCKOPT_PEELOFF:
5505 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5507 case SCTP_PEER_ADDR_PARAMS:
5508 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5511 case SCTP_DELAYED_ACK:
5512 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5516 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5518 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5519 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5522 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5523 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5526 case SCTP_GET_PEER_ADDRS_OLD:
5527 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5530 case SCTP_GET_LOCAL_ADDRS_OLD:
5531 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5534 case SCTP_GET_PEER_ADDRS:
5535 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5538 case SCTP_GET_LOCAL_ADDRS:
5539 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5542 case SCTP_DEFAULT_SEND_PARAM:
5543 retval = sctp_getsockopt_default_send_param(sk, len,
5546 case SCTP_PRIMARY_ADDR:
5547 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5550 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5553 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5555 case SCTP_ASSOCINFO:
5556 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5558 case SCTP_I_WANT_MAPPED_V4_ADDR:
5559 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5562 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5564 case SCTP_GET_PEER_ADDR_INFO:
5565 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5568 case SCTP_ADAPTATION_LAYER:
5569 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5573 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5575 case SCTP_FRAGMENT_INTERLEAVE:
5576 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5579 case SCTP_PARTIAL_DELIVERY_POINT:
5580 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5583 case SCTP_MAX_BURST:
5584 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5587 case SCTP_AUTH_CHUNK:
5588 case SCTP_AUTH_DELETE_KEY:
5589 retval = -EOPNOTSUPP;
5591 case SCTP_HMAC_IDENT:
5592 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5594 case SCTP_AUTH_ACTIVE_KEY:
5595 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5597 case SCTP_PEER_AUTH_CHUNKS:
5598 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5601 case SCTP_LOCAL_AUTH_CHUNKS:
5602 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5606 retval = -ENOPROTOOPT;
5610 sctp_release_sock(sk);
5614 static void sctp_hash(struct sock *sk)
5619 static void sctp_unhash(struct sock *sk)
5624 /* Check if port is acceptable. Possibly find first available port.
5626 * The port hash table (contained in the 'global' SCTP protocol storage
5627 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5628 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5629 * list (the list number is the port number hashed out, so as you
5630 * would expect from a hash function, all the ports in a given list have
5631 * such a number that hashes out to the same list number; you were
5632 * expecting that, right?); so each list has a set of ports, with a
5633 * link to the socket (struct sock) that uses it, the port number and
5634 * a fastreuse flag (FIXME: NPI ipg).
5636 static struct sctp_bind_bucket *sctp_bucket_create(
5637 struct sctp_bind_hashbucket *head, unsigned short snum);
5639 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5641 struct sctp_bind_hashbucket *head; /* hash list */
5642 struct sctp_bind_bucket *pp; /* hash list port iterator */
5643 struct hlist_node *node;
5644 unsigned short snum;
5647 snum = ntohs(addr->v4.sin_port);
5649 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5650 sctp_local_bh_disable();
5653 /* Search for an available port. */
5654 int low, high, remaining, index;
5657 inet_get_local_port_range(&low, &high);
5658 remaining = (high - low) + 1;
5659 rover = net_random() % remaining + low;
5663 if ((rover < low) || (rover > high))
5665 index = sctp_phashfn(rover);
5666 head = &sctp_port_hashtable[index];
5667 sctp_spin_lock(&head->lock);
5668 sctp_for_each_hentry(pp, node, &head->chain)
5669 if (pp->port == rover)
5673 sctp_spin_unlock(&head->lock);
5674 } while (--remaining > 0);
5676 /* Exhausted local port range during search? */
5681 /* OK, here is the one we will use. HEAD (the port
5682 * hash table list entry) is non-NULL and we hold it's
5687 /* We are given an specific port number; we verify
5688 * that it is not being used. If it is used, we will
5689 * exahust the search in the hash list corresponding
5690 * to the port number (snum) - we detect that with the
5691 * port iterator, pp being NULL.
5693 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5694 sctp_spin_lock(&head->lock);
5695 sctp_for_each_hentry(pp, node, &head->chain) {
5696 if (pp->port == snum)
5703 if (!hlist_empty(&pp->owner)) {
5704 /* We had a port hash table hit - there is an
5705 * available port (pp != NULL) and it is being
5706 * used by other socket (pp->owner not empty); that other
5707 * socket is going to be sk2.
5709 int reuse = sk->sk_reuse;
5711 struct hlist_node *node;
5713 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5714 if (pp->fastreuse && sk->sk_reuse &&
5715 sk->sk_state != SCTP_SS_LISTENING)
5718 /* Run through the list of sockets bound to the port
5719 * (pp->port) [via the pointers bind_next and
5720 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5721 * we get the endpoint they describe and run through
5722 * the endpoint's list of IP (v4 or v6) addresses,
5723 * comparing each of the addresses with the address of
5724 * the socket sk. If we find a match, then that means
5725 * that this port/socket (sk) combination are already
5728 sk_for_each_bound(sk2, node, &pp->owner) {
5729 struct sctp_endpoint *ep2;
5730 ep2 = sctp_sk(sk2)->ep;
5733 (reuse && sk2->sk_reuse &&
5734 sk2->sk_state != SCTP_SS_LISTENING))
5737 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5738 sctp_sk(sk2), sctp_sk(sk))) {
5743 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5746 /* If there was a hash table miss, create a new port. */
5748 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5751 /* In either case (hit or miss), make sure fastreuse is 1 only
5752 * if sk->sk_reuse is too (that is, if the caller requested
5753 * SO_REUSEADDR on this socket -sk-).
5755 if (hlist_empty(&pp->owner)) {
5756 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5760 } else if (pp->fastreuse &&
5761 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5764 /* We are set, so fill up all the data in the hash table
5765 * entry, tie the socket list information with the rest of the
5766 * sockets FIXME: Blurry, NPI (ipg).
5769 if (!sctp_sk(sk)->bind_hash) {
5770 inet_sk(sk)->num = snum;
5771 sk_add_bind_node(sk, &pp->owner);
5772 sctp_sk(sk)->bind_hash = pp;
5777 sctp_spin_unlock(&head->lock);
5780 sctp_local_bh_enable();
5784 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5785 * port is requested.
5787 static int sctp_get_port(struct sock *sk, unsigned short snum)
5790 union sctp_addr addr;
5791 struct sctp_af *af = sctp_sk(sk)->pf->af;
5793 /* Set up a dummy address struct from the sk. */
5794 af->from_sk(&addr, sk);
5795 addr.v4.sin_port = htons(snum);
5797 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5798 ret = sctp_get_port_local(sk, &addr);
5800 return (ret ? 1 : 0);
5804 * 3.1.3 listen() - UDP Style Syntax
5806 * By default, new associations are not accepted for UDP style sockets.
5807 * An application uses listen() to mark a socket as being able to
5808 * accept new associations.
5810 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5812 struct sctp_sock *sp = sctp_sk(sk);
5813 struct sctp_endpoint *ep = sp->ep;
5815 /* Only UDP style sockets that are not peeled off are allowed to
5818 if (!sctp_style(sk, UDP))
5821 /* If backlog is zero, disable listening. */
5823 if (sctp_sstate(sk, CLOSED))
5826 sctp_unhash_endpoint(ep);
5827 sk->sk_state = SCTP_SS_CLOSED;
5831 /* Return if we are already listening. */
5832 if (sctp_sstate(sk, LISTENING))
5836 * If a bind() or sctp_bindx() is not called prior to a listen()
5837 * call that allows new associations to be accepted, the system
5838 * picks an ephemeral port and will choose an address set equivalent
5839 * to binding with a wildcard address.
5841 * This is not currently spelled out in the SCTP sockets
5842 * extensions draft, but follows the practice as seen in TCP
5845 * Additionally, turn off fastreuse flag since we are not listening
5847 sk->sk_state = SCTP_SS_LISTENING;
5848 if (!ep->base.bind_addr.port) {
5849 if (sctp_autobind(sk))
5852 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5853 sk->sk_state = SCTP_SS_CLOSED;
5856 sctp_sk(sk)->bind_hash->fastreuse = 0;
5859 sctp_hash_endpoint(ep);
5864 * 4.1.3 listen() - TCP Style Syntax
5866 * Applications uses listen() to ready the SCTP endpoint for accepting
5867 * inbound associations.
5869 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5871 struct sctp_sock *sp = sctp_sk(sk);
5872 struct sctp_endpoint *ep = sp->ep;
5874 /* If backlog is zero, disable listening. */
5876 if (sctp_sstate(sk, CLOSED))
5879 sctp_unhash_endpoint(ep);
5880 sk->sk_state = SCTP_SS_CLOSED;
5884 if (sctp_sstate(sk, LISTENING))
5888 * If a bind() or sctp_bindx() is not called prior to a listen()
5889 * call that allows new associations to be accepted, the system
5890 * picks an ephemeral port and will choose an address set equivalent
5891 * to binding with a wildcard address.
5893 * This is not currently spelled out in the SCTP sockets
5894 * extensions draft, but follows the practice as seen in TCP
5897 sk->sk_state = SCTP_SS_LISTENING;
5898 if (!ep->base.bind_addr.port) {
5899 if (sctp_autobind(sk))
5902 sctp_sk(sk)->bind_hash->fastreuse = 0;
5904 sk->sk_max_ack_backlog = backlog;
5905 sctp_hash_endpoint(ep);
5910 * Move a socket to LISTENING state.
5912 int sctp_inet_listen(struct socket *sock, int backlog)
5914 struct sock *sk = sock->sk;
5915 struct crypto_hash *tfm = NULL;
5918 if (unlikely(backlog < 0))
5923 if (sock->state != SS_UNCONNECTED)
5926 /* Allocate HMAC for generating cookie. */
5927 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5928 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5930 if (net_ratelimit()) {
5932 "SCTP: failed to load transform for %s: %ld\n",
5933 sctp_hmac_alg, PTR_ERR(tfm));
5940 switch (sock->type) {
5941 case SOCK_SEQPACKET:
5942 err = sctp_seqpacket_listen(sk, backlog);
5945 err = sctp_stream_listen(sk, backlog);
5954 /* Store away the transform reference. */
5955 if (!sctp_sk(sk)->hmac)
5956 sctp_sk(sk)->hmac = tfm;
5958 sctp_release_sock(sk);
5961 crypto_free_hash(tfm);
5966 * This function is done by modeling the current datagram_poll() and the
5967 * tcp_poll(). Note that, based on these implementations, we don't
5968 * lock the socket in this function, even though it seems that,
5969 * ideally, locking or some other mechanisms can be used to ensure
5970 * the integrity of the counters (sndbuf and wmem_alloc) used
5971 * in this place. We assume that we don't need locks either until proven
5974 * Another thing to note is that we include the Async I/O support
5975 * here, again, by modeling the current TCP/UDP code. We don't have
5976 * a good way to test with it yet.
5978 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5980 struct sock *sk = sock->sk;
5981 struct sctp_sock *sp = sctp_sk(sk);
5984 poll_wait(file, sk->sk_sleep, wait);
5986 /* A TCP-style listening socket becomes readable when the accept queue
5989 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5990 return (!list_empty(&sp->ep->asocs)) ?
5991 (POLLIN | POLLRDNORM) : 0;
5995 /* Is there any exceptional events? */
5996 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5998 if (sk->sk_shutdown & RCV_SHUTDOWN)
6000 if (sk->sk_shutdown == SHUTDOWN_MASK)
6003 /* Is it readable? Reconsider this code with TCP-style support. */
6004 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6005 (sk->sk_shutdown & RCV_SHUTDOWN))
6006 mask |= POLLIN | POLLRDNORM;
6008 /* The association is either gone or not ready. */
6009 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6012 /* Is it writable? */
6013 if (sctp_writeable(sk)) {
6014 mask |= POLLOUT | POLLWRNORM;
6016 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6018 * Since the socket is not locked, the buffer
6019 * might be made available after the writeable check and
6020 * before the bit is set. This could cause a lost I/O
6021 * signal. tcp_poll() has a race breaker for this race
6022 * condition. Based on their implementation, we put
6023 * in the following code to cover it as well.
6025 if (sctp_writeable(sk))
6026 mask |= POLLOUT | POLLWRNORM;
6031 /********************************************************************
6032 * 2nd Level Abstractions
6033 ********************************************************************/
6035 static struct sctp_bind_bucket *sctp_bucket_create(
6036 struct sctp_bind_hashbucket *head, unsigned short snum)
6038 struct sctp_bind_bucket *pp;
6040 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6042 SCTP_DBG_OBJCNT_INC(bind_bucket);
6045 INIT_HLIST_HEAD(&pp->owner);
6046 hlist_add_head(&pp->node, &head->chain);
6051 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6052 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6054 if (pp && hlist_empty(&pp->owner)) {
6055 __hlist_del(&pp->node);
6056 kmem_cache_free(sctp_bucket_cachep, pp);
6057 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6061 /* Release this socket's reference to a local port. */
6062 static inline void __sctp_put_port(struct sock *sk)
6064 struct sctp_bind_hashbucket *head =
6065 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6066 struct sctp_bind_bucket *pp;
6068 sctp_spin_lock(&head->lock);
6069 pp = sctp_sk(sk)->bind_hash;
6070 __sk_del_bind_node(sk);
6071 sctp_sk(sk)->bind_hash = NULL;
6072 inet_sk(sk)->num = 0;
6073 sctp_bucket_destroy(pp);
6074 sctp_spin_unlock(&head->lock);
6077 void sctp_put_port(struct sock *sk)
6079 sctp_local_bh_disable();
6080 __sctp_put_port(sk);
6081 sctp_local_bh_enable();
6085 * The system picks an ephemeral port and choose an address set equivalent
6086 * to binding with a wildcard address.
6087 * One of those addresses will be the primary address for the association.
6088 * This automatically enables the multihoming capability of SCTP.
6090 static int sctp_autobind(struct sock *sk)
6092 union sctp_addr autoaddr;
6096 /* Initialize a local sockaddr structure to INADDR_ANY. */
6097 af = sctp_sk(sk)->pf->af;
6099 port = htons(inet_sk(sk)->num);
6100 af->inaddr_any(&autoaddr, port);
6102 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6105 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6108 * 4.2 The cmsghdr Structure *
6110 * When ancillary data is sent or received, any number of ancillary data
6111 * objects can be specified by the msg_control and msg_controllen members of
6112 * the msghdr structure, because each object is preceded by
6113 * a cmsghdr structure defining the object's length (the cmsg_len member).
6114 * Historically Berkeley-derived implementations have passed only one object
6115 * at a time, but this API allows multiple objects to be
6116 * passed in a single call to sendmsg() or recvmsg(). The following example
6117 * shows two ancillary data objects in a control buffer.
6119 * |<--------------------------- msg_controllen -------------------------->|
6122 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6124 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6127 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6129 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6132 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6133 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6135 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6137 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6144 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6145 sctp_cmsgs_t *cmsgs)
6147 struct cmsghdr *cmsg;
6148 struct msghdr *my_msg = (struct msghdr *)msg;
6150 for (cmsg = CMSG_FIRSTHDR(msg);
6152 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6153 if (!CMSG_OK(my_msg, cmsg))
6156 /* Should we parse this header or ignore? */
6157 if (cmsg->cmsg_level != IPPROTO_SCTP)
6160 /* Strictly check lengths following example in SCM code. */
6161 switch (cmsg->cmsg_type) {
6163 /* SCTP Socket API Extension
6164 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6166 * This cmsghdr structure provides information for
6167 * initializing new SCTP associations with sendmsg().
6168 * The SCTP_INITMSG socket option uses this same data
6169 * structure. This structure is not used for
6172 * cmsg_level cmsg_type cmsg_data[]
6173 * ------------ ------------ ----------------------
6174 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6176 if (cmsg->cmsg_len !=
6177 CMSG_LEN(sizeof(struct sctp_initmsg)))
6179 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6183 /* SCTP Socket API Extension
6184 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6186 * This cmsghdr structure specifies SCTP options for
6187 * sendmsg() and describes SCTP header information
6188 * about a received message through recvmsg().
6190 * cmsg_level cmsg_type cmsg_data[]
6191 * ------------ ------------ ----------------------
6192 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6194 if (cmsg->cmsg_len !=
6195 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6199 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6201 /* Minimally, validate the sinfo_flags. */
6202 if (cmsgs->info->sinfo_flags &
6203 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6204 SCTP_ABORT | SCTP_EOF))
6216 * Wait for a packet..
6217 * Note: This function is the same function as in core/datagram.c
6218 * with a few modifications to make lksctp work.
6220 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6225 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6227 /* Socket errors? */
6228 error = sock_error(sk);
6232 if (!skb_queue_empty(&sk->sk_receive_queue))
6235 /* Socket shut down? */
6236 if (sk->sk_shutdown & RCV_SHUTDOWN)
6239 /* Sequenced packets can come disconnected. If so we report the
6244 /* Is there a good reason to think that we may receive some data? */
6245 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6248 /* Handle signals. */
6249 if (signal_pending(current))
6252 /* Let another process have a go. Since we are going to sleep
6253 * anyway. Note: This may cause odd behaviors if the message
6254 * does not fit in the user's buffer, but this seems to be the
6255 * only way to honor MSG_DONTWAIT realistically.
6257 sctp_release_sock(sk);
6258 *timeo_p = schedule_timeout(*timeo_p);
6262 finish_wait(sk->sk_sleep, &wait);
6266 error = sock_intr_errno(*timeo_p);
6269 finish_wait(sk->sk_sleep, &wait);
6274 /* Receive a datagram.
6275 * Note: This is pretty much the same routine as in core/datagram.c
6276 * with a few changes to make lksctp work.
6278 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6279 int noblock, int *err)
6282 struct sk_buff *skb;
6285 timeo = sock_rcvtimeo(sk, noblock);
6287 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6288 timeo, MAX_SCHEDULE_TIMEOUT);
6291 /* Again only user level code calls this function,
6292 * so nothing interrupt level
6293 * will suddenly eat the receive_queue.
6295 * Look at current nfs client by the way...
6296 * However, this function was corrent in any case. 8)
6298 if (flags & MSG_PEEK) {
6299 spin_lock_bh(&sk->sk_receive_queue.lock);
6300 skb = skb_peek(&sk->sk_receive_queue);
6302 atomic_inc(&skb->users);
6303 spin_unlock_bh(&sk->sk_receive_queue.lock);
6305 skb = skb_dequeue(&sk->sk_receive_queue);
6311 /* Caller is allowed not to check sk->sk_err before calling. */
6312 error = sock_error(sk);
6316 if (sk->sk_shutdown & RCV_SHUTDOWN)
6319 /* User doesn't want to wait. */
6323 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6332 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6333 static void __sctp_write_space(struct sctp_association *asoc)
6335 struct sock *sk = asoc->base.sk;
6336 struct socket *sock = sk->sk_socket;
6338 if ((sctp_wspace(asoc) > 0) && sock) {
6339 if (waitqueue_active(&asoc->wait))
6340 wake_up_interruptible(&asoc->wait);
6342 if (sctp_writeable(sk)) {
6343 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6344 wake_up_interruptible(sk->sk_sleep);
6346 /* Note that we try to include the Async I/O support
6347 * here by modeling from the current TCP/UDP code.
6348 * We have not tested with it yet.
6350 if (sock->fasync_list &&
6351 !(sk->sk_shutdown & SEND_SHUTDOWN))
6352 sock_wake_async(sock,
6353 SOCK_WAKE_SPACE, POLL_OUT);
6358 /* Do accounting for the sndbuf space.
6359 * Decrement the used sndbuf space of the corresponding association by the
6360 * data size which was just transmitted(freed).
6362 static void sctp_wfree(struct sk_buff *skb)
6364 struct sctp_association *asoc;
6365 struct sctp_chunk *chunk;
6368 /* Get the saved chunk pointer. */
6369 chunk = *((struct sctp_chunk **)(skb->cb));
6372 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6373 sizeof(struct sk_buff) +
6374 sizeof(struct sctp_chunk);
6376 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6379 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6381 sk->sk_wmem_queued -= skb->truesize;
6382 sk_mem_uncharge(sk, skb->truesize);
6385 __sctp_write_space(asoc);
6387 sctp_association_put(asoc);
6390 /* Do accounting for the receive space on the socket.
6391 * Accounting for the association is done in ulpevent.c
6392 * We set this as a destructor for the cloned data skbs so that
6393 * accounting is done at the correct time.
6395 void sctp_sock_rfree(struct sk_buff *skb)
6397 struct sock *sk = skb->sk;
6398 struct sctp_ulpevent *event = sctp_skb2event(skb);
6400 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6403 * Mimic the behavior of sock_rfree
6405 sk_mem_uncharge(sk, event->rmem_len);
6409 /* Helper function to wait for space in the sndbuf. */
6410 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6413 struct sock *sk = asoc->base.sk;
6415 long current_timeo = *timeo_p;
6418 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6419 asoc, (long)(*timeo_p), msg_len);
6421 /* Increment the association's refcnt. */
6422 sctp_association_hold(asoc);
6424 /* Wait on the association specific sndbuf space. */
6426 prepare_to_wait_exclusive(&asoc->wait, &wait,
6427 TASK_INTERRUPTIBLE);
6430 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6433 if (signal_pending(current))
6434 goto do_interrupted;
6435 if (msg_len <= sctp_wspace(asoc))
6438 /* Let another process have a go. Since we are going
6441 sctp_release_sock(sk);
6442 current_timeo = schedule_timeout(current_timeo);
6443 BUG_ON(sk != asoc->base.sk);
6446 *timeo_p = current_timeo;
6450 finish_wait(&asoc->wait, &wait);
6452 /* Release the association's refcnt. */
6453 sctp_association_put(asoc);
6462 err = sock_intr_errno(*timeo_p);
6470 /* If socket sndbuf has changed, wake up all per association waiters. */
6471 void sctp_write_space(struct sock *sk)
6473 struct sctp_association *asoc;
6475 /* Wake up the tasks in each wait queue. */
6476 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6477 __sctp_write_space(asoc);
6481 /* Is there any sndbuf space available on the socket?
6483 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6484 * associations on the same socket. For a UDP-style socket with
6485 * multiple associations, it is possible for it to be "unwriteable"
6486 * prematurely. I assume that this is acceptable because
6487 * a premature "unwriteable" is better than an accidental "writeable" which
6488 * would cause an unwanted block under certain circumstances. For the 1-1
6489 * UDP-style sockets or TCP-style sockets, this code should work.
6492 static int sctp_writeable(struct sock *sk)
6496 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6502 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6503 * returns immediately with EINPROGRESS.
6505 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6507 struct sock *sk = asoc->base.sk;
6509 long current_timeo = *timeo_p;
6512 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6515 /* Increment the association's refcnt. */
6516 sctp_association_hold(asoc);
6519 prepare_to_wait_exclusive(&asoc->wait, &wait,
6520 TASK_INTERRUPTIBLE);
6523 if (sk->sk_shutdown & RCV_SHUTDOWN)
6525 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6528 if (signal_pending(current))
6529 goto do_interrupted;
6531 if (sctp_state(asoc, ESTABLISHED))
6534 /* Let another process have a go. Since we are going
6537 sctp_release_sock(sk);
6538 current_timeo = schedule_timeout(current_timeo);
6541 *timeo_p = current_timeo;
6545 finish_wait(&asoc->wait, &wait);
6547 /* Release the association's refcnt. */
6548 sctp_association_put(asoc);
6553 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6556 err = -ECONNREFUSED;
6560 err = sock_intr_errno(*timeo_p);
6568 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6570 struct sctp_endpoint *ep;
6574 ep = sctp_sk(sk)->ep;
6578 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6579 TASK_INTERRUPTIBLE);
6581 if (list_empty(&ep->asocs)) {
6582 sctp_release_sock(sk);
6583 timeo = schedule_timeout(timeo);
6588 if (!sctp_sstate(sk, LISTENING))
6592 if (!list_empty(&ep->asocs))
6595 err = sock_intr_errno(timeo);
6596 if (signal_pending(current))
6604 finish_wait(sk->sk_sleep, &wait);
6609 static void sctp_wait_for_close(struct sock *sk, long timeout)
6614 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6615 if (list_empty(&sctp_sk(sk)->ep->asocs))
6617 sctp_release_sock(sk);
6618 timeout = schedule_timeout(timeout);
6620 } while (!signal_pending(current) && timeout);
6622 finish_wait(sk->sk_sleep, &wait);
6625 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6627 struct sk_buff *frag;
6632 /* Don't forget the fragments. */
6633 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6634 sctp_sock_rfree_frag(frag);
6637 sctp_sock_rfree(skb);
6640 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6642 struct sk_buff *frag;
6647 /* Don't forget the fragments. */
6648 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6649 sctp_skb_set_owner_r_frag(frag, sk);
6652 sctp_skb_set_owner_r(skb, sk);
6655 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6656 * and its messages to the newsk.
6658 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6659 struct sctp_association *assoc,
6660 sctp_socket_type_t type)
6662 struct sctp_sock *oldsp = sctp_sk(oldsk);
6663 struct sctp_sock *newsp = sctp_sk(newsk);
6664 struct sctp_bind_bucket *pp; /* hash list port iterator */
6665 struct sctp_endpoint *newep = newsp->ep;
6666 struct sk_buff *skb, *tmp;
6667 struct sctp_ulpevent *event;
6668 struct sctp_bind_hashbucket *head;
6670 /* Migrate socket buffer sizes and all the socket level options to the
6673 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6674 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6675 /* Brute force copy old sctp opt. */
6676 inet_sk_copy_descendant(newsk, oldsk);
6678 /* Restore the ep value that was overwritten with the above structure
6684 /* Hook this new socket in to the bind_hash list. */
6685 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6686 sctp_local_bh_disable();
6687 sctp_spin_lock(&head->lock);
6688 pp = sctp_sk(oldsk)->bind_hash;
6689 sk_add_bind_node(newsk, &pp->owner);
6690 sctp_sk(newsk)->bind_hash = pp;
6691 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6692 sctp_spin_unlock(&head->lock);
6693 sctp_local_bh_enable();
6695 /* Copy the bind_addr list from the original endpoint to the new
6696 * endpoint so that we can handle restarts properly
6698 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6699 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6701 /* Move any messages in the old socket's receive queue that are for the
6702 * peeled off association to the new socket's receive queue.
6704 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6705 event = sctp_skb2event(skb);
6706 if (event->asoc == assoc) {
6707 sctp_sock_rfree_frag(skb);
6708 __skb_unlink(skb, &oldsk->sk_receive_queue);
6709 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6710 sctp_skb_set_owner_r_frag(skb, newsk);
6714 /* Clean up any messages pending delivery due to partial
6715 * delivery. Three cases:
6716 * 1) No partial deliver; no work.
6717 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6718 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6720 skb_queue_head_init(&newsp->pd_lobby);
6721 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6723 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6724 struct sk_buff_head *queue;
6726 /* Decide which queue to move pd_lobby skbs to. */
6727 if (assoc->ulpq.pd_mode) {
6728 queue = &newsp->pd_lobby;
6730 queue = &newsk->sk_receive_queue;
6732 /* Walk through the pd_lobby, looking for skbs that
6733 * need moved to the new socket.
6735 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6736 event = sctp_skb2event(skb);
6737 if (event->asoc == assoc) {
6738 sctp_sock_rfree_frag(skb);
6739 __skb_unlink(skb, &oldsp->pd_lobby);
6740 __skb_queue_tail(queue, skb);
6741 sctp_skb_set_owner_r_frag(skb, newsk);
6745 /* Clear up any skbs waiting for the partial
6746 * delivery to finish.
6748 if (assoc->ulpq.pd_mode)
6749 sctp_clear_pd(oldsk, NULL);
6753 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6754 sctp_sock_rfree_frag(skb);
6755 sctp_skb_set_owner_r_frag(skb, newsk);
6758 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6759 sctp_sock_rfree_frag(skb);
6760 sctp_skb_set_owner_r_frag(skb, newsk);
6763 /* Set the type of socket to indicate that it is peeled off from the
6764 * original UDP-style socket or created with the accept() call on a
6765 * TCP-style socket..
6769 /* Mark the new socket "in-use" by the user so that any packets
6770 * that may arrive on the association after we've moved it are
6771 * queued to the backlog. This prevents a potential race between
6772 * backlog processing on the old socket and new-packet processing
6773 * on the new socket.
6775 * The caller has just allocated newsk so we can guarantee that other
6776 * paths won't try to lock it and then oldsk.
6778 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6779 sctp_assoc_migrate(assoc, newsk);
6781 /* If the association on the newsk is already closed before accept()
6782 * is called, set RCV_SHUTDOWN flag.
6784 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6785 newsk->sk_shutdown |= RCV_SHUTDOWN;
6787 newsk->sk_state = SCTP_SS_ESTABLISHED;
6788 sctp_release_sock(newsk);
6792 /* This proto struct describes the ULP interface for SCTP. */
6793 struct proto sctp_prot = {
6795 .owner = THIS_MODULE,
6796 .close = sctp_close,
6797 .connect = sctp_connect,
6798 .disconnect = sctp_disconnect,
6799 .accept = sctp_accept,
6800 .ioctl = sctp_ioctl,
6801 .init = sctp_init_sock,
6802 .destroy = sctp_destroy_sock,
6803 .shutdown = sctp_shutdown,
6804 .setsockopt = sctp_setsockopt,
6805 .getsockopt = sctp_getsockopt,
6806 .sendmsg = sctp_sendmsg,
6807 .recvmsg = sctp_recvmsg,
6809 .backlog_rcv = sctp_backlog_rcv,
6811 .unhash = sctp_unhash,
6812 .get_port = sctp_get_port,
6813 .obj_size = sizeof(struct sctp_sock),
6814 .sysctl_mem = sysctl_sctp_mem,
6815 .sysctl_rmem = sysctl_sctp_rmem,
6816 .sysctl_wmem = sysctl_sctp_wmem,
6817 .memory_pressure = &sctp_memory_pressure,
6818 .enter_memory_pressure = sctp_enter_memory_pressure,
6819 .memory_allocated = &sctp_memory_allocated,
6820 .sockets_allocated = &sctp_sockets_allocated,
6823 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6825 struct proto sctpv6_prot = {
6827 .owner = THIS_MODULE,
6828 .close = sctp_close,
6829 .connect = sctp_connect,
6830 .disconnect = sctp_disconnect,
6831 .accept = sctp_accept,
6832 .ioctl = sctp_ioctl,
6833 .init = sctp_init_sock,
6834 .destroy = sctp_destroy_sock,
6835 .shutdown = sctp_shutdown,
6836 .setsockopt = sctp_setsockopt,
6837 .getsockopt = sctp_getsockopt,
6838 .sendmsg = sctp_sendmsg,
6839 .recvmsg = sctp_recvmsg,
6841 .backlog_rcv = sctp_backlog_rcv,
6843 .unhash = sctp_unhash,
6844 .get_port = sctp_get_port,
6845 .obj_size = sizeof(struct sctp6_sock),
6846 .sysctl_mem = sysctl_sctp_mem,
6847 .sysctl_rmem = sysctl_sctp_rmem,
6848 .sysctl_wmem = sysctl_sctp_wmem,
6849 .memory_pressure = &sctp_memory_pressure,
6850 .enter_memory_pressure = sctp_enter_memory_pressure,
6851 .memory_allocated = &sctp_memory_allocated,
6852 .sockets_allocated = &sctp_sockets_allocated,
6854 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */