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 = sk_wmem_alloc_get(asoc->base.sk);
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 /* In case the user of sctp_connectx() wants an association
1104 * id back, assign one now.
1107 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1112 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1117 /* Initialize sk's dport and daddr for getpeername() */
1118 inet_sk(sk)->dport = htons(asoc->peer.port);
1119 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1120 af->to_sk_daddr(sa_addr, sk);
1123 /* in-kernel sockets don't generally have a file allocated to them
1124 * if all they do is call sock_create_kern().
1126 if (sk->sk_socket->file)
1127 f_flags = sk->sk_socket->file->f_flags;
1129 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1131 err = sctp_wait_for_connect(asoc, &timeo);
1132 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1133 *assoc_id = asoc->assoc_id;
1135 /* Don't free association on exit. */
1140 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1141 " kaddrs: %p err: %d\n",
1144 sctp_association_free(asoc);
1148 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1151 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1152 * sctp_assoc_t *asoc);
1154 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1155 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1156 * or IPv6 addresses.
1158 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1159 * Section 3.1.2 for this usage.
1161 * addrs is a pointer to an array of one or more socket addresses. Each
1162 * address is contained in its appropriate structure (i.e. struct
1163 * sockaddr_in or struct sockaddr_in6) the family of the address type
1164 * must be used to distengish the address length (note that this
1165 * representation is termed a "packed array" of addresses). The caller
1166 * specifies the number of addresses in the array with addrcnt.
1168 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1169 * the association id of the new association. On failure, sctp_connectx()
1170 * returns -1, and sets errno to the appropriate error code. The assoc_id
1171 * is not touched by the kernel.
1173 * For SCTP, the port given in each socket address must be the same, or
1174 * sctp_connectx() will fail, setting errno to EINVAL.
1176 * An application can use sctp_connectx to initiate an association with
1177 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1178 * allows a caller to specify multiple addresses at which a peer can be
1179 * reached. The way the SCTP stack uses the list of addresses to set up
1180 * the association is implementation dependant. This function only
1181 * specifies that the stack will try to make use of all the addresses in
1182 * the list when needed.
1184 * Note that the list of addresses passed in is only used for setting up
1185 * the association. It does not necessarily equal the set of addresses
1186 * the peer uses for the resulting association. If the caller wants to
1187 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1188 * retrieve them after the association has been set up.
1190 * Basically do nothing but copying the addresses from user to kernel
1191 * land and invoking either sctp_connectx(). This is used for tunneling
1192 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1194 * We don't use copy_from_user() for optimization: we first do the
1195 * sanity checks (buffer size -fast- and access check-healthy
1196 * pointer); if all of those succeed, then we can alloc the memory
1197 * (expensive operation) needed to copy the data to kernel. Then we do
1198 * the copying without checking the user space area
1199 * (__copy_from_user()).
1201 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1204 * sk The sk of the socket
1205 * addrs The pointer to the addresses in user land
1206 * addrssize Size of the addrs buffer
1208 * Returns >=0 if ok, <0 errno code on error.
1210 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1211 struct sockaddr __user *addrs,
1213 sctp_assoc_t *assoc_id)
1216 struct sockaddr *kaddrs;
1218 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1219 __func__, sk, addrs, addrs_size);
1221 if (unlikely(addrs_size <= 0))
1224 /* Check the user passed a healthy pointer. */
1225 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1228 /* Alloc space for the address array in kernel memory. */
1229 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1230 if (unlikely(!kaddrs))
1233 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1236 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1245 * This is an older interface. It's kept for backward compatibility
1246 * to the option that doesn't provide association id.
1248 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1249 struct sockaddr __user *addrs,
1252 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1256 * New interface for the API. The since the API is done with a socket
1257 * option, to make it simple we feed back the association id is as a return
1258 * indication to the call. Error is always negative and association id is
1261 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1262 struct sockaddr __user *addrs,
1265 sctp_assoc_t assoc_id = 0;
1268 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1277 * New (hopefully final) interface for the API. The option buffer is used
1278 * both for the returned association id and the addresses.
1280 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1281 char __user *optval,
1284 sctp_assoc_t assoc_id = 0;
1287 if (len < sizeof(assoc_id))
1290 err = __sctp_setsockopt_connectx(sk,
1291 (struct sockaddr __user *)(optval + sizeof(assoc_id)),
1292 len - sizeof(assoc_id), &assoc_id);
1294 if (err == 0 || err == -EINPROGRESS) {
1295 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1297 if (put_user(sizeof(assoc_id), optlen))
1304 /* API 3.1.4 close() - UDP Style Syntax
1305 * Applications use close() to perform graceful shutdown (as described in
1306 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1307 * by a UDP-style socket.
1311 * ret = close(int sd);
1313 * sd - the socket descriptor of the associations to be closed.
1315 * To gracefully shutdown a specific association represented by the
1316 * UDP-style socket, an application should use the sendmsg() call,
1317 * passing no user data, but including the appropriate flag in the
1318 * ancillary data (see Section xxxx).
1320 * If sd in the close() call is a branched-off socket representing only
1321 * one association, the shutdown is performed on that association only.
1323 * 4.1.6 close() - TCP Style Syntax
1325 * Applications use close() to gracefully close down an association.
1329 * int close(int sd);
1331 * sd - the socket descriptor of the association to be closed.
1333 * After an application calls close() on a socket descriptor, no further
1334 * socket operations will succeed on that descriptor.
1336 * API 7.1.4 SO_LINGER
1338 * An application using the TCP-style socket can use this option to
1339 * perform the SCTP ABORT primitive. The linger option structure is:
1342 * int l_onoff; // option on/off
1343 * int l_linger; // linger time
1346 * To enable the option, set l_onoff to 1. If the l_linger value is set
1347 * to 0, calling close() is the same as the ABORT primitive. If the
1348 * value is set to a negative value, the setsockopt() call will return
1349 * an error. If the value is set to a positive value linger_time, the
1350 * close() can be blocked for at most linger_time ms. If the graceful
1351 * shutdown phase does not finish during this period, close() will
1352 * return but the graceful shutdown phase continues in the system.
1354 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1356 struct sctp_endpoint *ep;
1357 struct sctp_association *asoc;
1358 struct list_head *pos, *temp;
1360 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1363 sk->sk_shutdown = SHUTDOWN_MASK;
1364 sk->sk_state = SCTP_SS_CLOSING;
1366 ep = sctp_sk(sk)->ep;
1368 /* Walk all associations on an endpoint. */
1369 list_for_each_safe(pos, temp, &ep->asocs) {
1370 asoc = list_entry(pos, struct sctp_association, asocs);
1372 if (sctp_style(sk, TCP)) {
1373 /* A closed association can still be in the list if
1374 * it belongs to a TCP-style listening socket that is
1375 * not yet accepted. If so, free it. If not, send an
1376 * ABORT or SHUTDOWN based on the linger options.
1378 if (sctp_state(asoc, CLOSED)) {
1379 sctp_unhash_established(asoc);
1380 sctp_association_free(asoc);
1385 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1386 struct sctp_chunk *chunk;
1388 chunk = sctp_make_abort_user(asoc, NULL, 0);
1390 sctp_primitive_ABORT(asoc, chunk);
1392 sctp_primitive_SHUTDOWN(asoc, NULL);
1395 /* Clean up any skbs sitting on the receive queue. */
1396 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1397 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1399 /* On a TCP-style socket, block for at most linger_time if set. */
1400 if (sctp_style(sk, TCP) && timeout)
1401 sctp_wait_for_close(sk, timeout);
1403 /* This will run the backlog queue. */
1404 sctp_release_sock(sk);
1406 /* Supposedly, no process has access to the socket, but
1407 * the net layers still may.
1409 sctp_local_bh_disable();
1410 sctp_bh_lock_sock(sk);
1412 /* Hold the sock, since sk_common_release() will put sock_put()
1413 * and we have just a little more cleanup.
1416 sk_common_release(sk);
1418 sctp_bh_unlock_sock(sk);
1419 sctp_local_bh_enable();
1423 SCTP_DBG_OBJCNT_DEC(sock);
1426 /* Handle EPIPE error. */
1427 static int sctp_error(struct sock *sk, int flags, int err)
1430 err = sock_error(sk) ? : -EPIPE;
1431 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1432 send_sig(SIGPIPE, current, 0);
1436 /* API 3.1.3 sendmsg() - UDP Style Syntax
1438 * An application uses sendmsg() and recvmsg() calls to transmit data to
1439 * and receive data from its peer.
1441 * ssize_t sendmsg(int socket, const struct msghdr *message,
1444 * socket - the socket descriptor of the endpoint.
1445 * message - pointer to the msghdr structure which contains a single
1446 * user message and possibly some ancillary data.
1448 * See Section 5 for complete description of the data
1451 * flags - flags sent or received with the user message, see Section
1452 * 5 for complete description of the flags.
1454 * Note: This function could use a rewrite especially when explicit
1455 * connect support comes in.
1457 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1459 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1461 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1462 struct msghdr *msg, size_t msg_len)
1464 struct sctp_sock *sp;
1465 struct sctp_endpoint *ep;
1466 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1467 struct sctp_transport *transport, *chunk_tp;
1468 struct sctp_chunk *chunk;
1470 struct sockaddr *msg_name = NULL;
1471 struct sctp_sndrcvinfo default_sinfo = { 0 };
1472 struct sctp_sndrcvinfo *sinfo;
1473 struct sctp_initmsg *sinit;
1474 sctp_assoc_t associd = 0;
1475 sctp_cmsgs_t cmsgs = { NULL };
1479 __u16 sinfo_flags = 0;
1480 struct sctp_datamsg *datamsg;
1481 int msg_flags = msg->msg_flags;
1483 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1490 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1492 /* We cannot send a message over a TCP-style listening socket. */
1493 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1498 /* Parse out the SCTP CMSGs. */
1499 err = sctp_msghdr_parse(msg, &cmsgs);
1502 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1506 /* Fetch the destination address for this packet. This
1507 * address only selects the association--it is not necessarily
1508 * the address we will send to.
1509 * For a peeled-off socket, msg_name is ignored.
1511 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1512 int msg_namelen = msg->msg_namelen;
1514 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1519 if (msg_namelen > sizeof(to))
1520 msg_namelen = sizeof(to);
1521 memcpy(&to, msg->msg_name, msg_namelen);
1522 msg_name = msg->msg_name;
1528 /* Did the user specify SNDRCVINFO? */
1530 sinfo_flags = sinfo->sinfo_flags;
1531 associd = sinfo->sinfo_assoc_id;
1534 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1535 msg_len, sinfo_flags);
1537 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1538 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1543 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1544 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1545 * If SCTP_ABORT is set, the message length could be non zero with
1546 * the msg_iov set to the user abort reason.
1548 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1549 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1554 /* If SCTP_ADDR_OVER is set, there must be an address
1555 * specified in msg_name.
1557 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1564 SCTP_DEBUG_PRINTK("About to look up association.\n");
1568 /* If a msg_name has been specified, assume this is to be used. */
1570 /* Look for a matching association on the endpoint. */
1571 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1573 /* If we could not find a matching association on the
1574 * endpoint, make sure that it is not a TCP-style
1575 * socket that already has an association or there is
1576 * no peeled-off association on another socket.
1578 if ((sctp_style(sk, TCP) &&
1579 sctp_sstate(sk, ESTABLISHED)) ||
1580 sctp_endpoint_is_peeled_off(ep, &to)) {
1581 err = -EADDRNOTAVAIL;
1586 asoc = sctp_id2assoc(sk, associd);
1594 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1596 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1597 * socket that has an association in CLOSED state. This can
1598 * happen when an accepted socket has an association that is
1601 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1606 if (sinfo_flags & SCTP_EOF) {
1607 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1609 sctp_primitive_SHUTDOWN(asoc, NULL);
1613 if (sinfo_flags & SCTP_ABORT) {
1615 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1621 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1622 sctp_primitive_ABORT(asoc, chunk);
1628 /* Do we need to create the association? */
1630 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1632 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1637 /* Check for invalid stream against the stream counts,
1638 * either the default or the user specified stream counts.
1641 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1642 /* Check against the defaults. */
1643 if (sinfo->sinfo_stream >=
1644 sp->initmsg.sinit_num_ostreams) {
1649 /* Check against the requested. */
1650 if (sinfo->sinfo_stream >=
1651 sinit->sinit_num_ostreams) {
1659 * API 3.1.2 bind() - UDP Style Syntax
1660 * If a bind() or sctp_bindx() is not called prior to a
1661 * sendmsg() call that initiates a new association, the
1662 * system picks an ephemeral port and will choose an address
1663 * set equivalent to binding with a wildcard address.
1665 if (!ep->base.bind_addr.port) {
1666 if (sctp_autobind(sk)) {
1672 * If an unprivileged user inherits a one-to-many
1673 * style socket with open associations on a privileged
1674 * port, it MAY be permitted to accept new associations,
1675 * but it SHOULD NOT be permitted to open new
1678 if (ep->base.bind_addr.port < PROT_SOCK &&
1679 !capable(CAP_NET_BIND_SERVICE)) {
1685 scope = sctp_scope(&to);
1686 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1693 /* If the SCTP_INIT ancillary data is specified, set all
1694 * the association init values accordingly.
1697 if (sinit->sinit_num_ostreams) {
1698 asoc->c.sinit_num_ostreams =
1699 sinit->sinit_num_ostreams;
1701 if (sinit->sinit_max_instreams) {
1702 asoc->c.sinit_max_instreams =
1703 sinit->sinit_max_instreams;
1705 if (sinit->sinit_max_attempts) {
1706 asoc->max_init_attempts
1707 = sinit->sinit_max_attempts;
1709 if (sinit->sinit_max_init_timeo) {
1710 asoc->max_init_timeo =
1711 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1715 /* Prime the peer's transport structures. */
1716 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1721 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1728 /* ASSERT: we have a valid association at this point. */
1729 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1732 /* If the user didn't specify SNDRCVINFO, make up one with
1735 default_sinfo.sinfo_stream = asoc->default_stream;
1736 default_sinfo.sinfo_flags = asoc->default_flags;
1737 default_sinfo.sinfo_ppid = asoc->default_ppid;
1738 default_sinfo.sinfo_context = asoc->default_context;
1739 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1740 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1741 sinfo = &default_sinfo;
1744 /* API 7.1.7, the sndbuf size per association bounds the
1745 * maximum size of data that can be sent in a single send call.
1747 if (msg_len > sk->sk_sndbuf) {
1752 if (asoc->pmtu_pending)
1753 sctp_assoc_pending_pmtu(asoc);
1755 /* If fragmentation is disabled and the message length exceeds the
1756 * association fragmentation point, return EMSGSIZE. The I-D
1757 * does not specify what this error is, but this looks like
1760 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1766 /* Check for invalid stream. */
1767 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1773 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1774 if (!sctp_wspace(asoc)) {
1775 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1780 /* If an address is passed with the sendto/sendmsg call, it is used
1781 * to override the primary destination address in the TCP model, or
1782 * when SCTP_ADDR_OVER flag is set in the UDP model.
1784 if ((sctp_style(sk, TCP) && msg_name) ||
1785 (sinfo_flags & SCTP_ADDR_OVER)) {
1786 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1794 /* Auto-connect, if we aren't connected already. */
1795 if (sctp_state(asoc, CLOSED)) {
1796 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1799 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1802 /* Break the message into multiple chunks of maximum size. */
1803 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1809 /* Now send the (possibly) fragmented message. */
1810 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1811 sctp_chunk_hold(chunk);
1813 /* Do accounting for the write space. */
1814 sctp_set_owner_w(chunk);
1816 chunk->transport = chunk_tp;
1819 /* Send it to the lower layers. Note: all chunks
1820 * must either fail or succeed. The lower layer
1821 * works that way today. Keep it that way or this
1824 err = sctp_primitive_SEND(asoc, datamsg);
1825 /* Did the lower layer accept the chunk? */
1827 sctp_datamsg_free(datamsg);
1829 sctp_datamsg_put(datamsg);
1831 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1838 /* If we are already past ASSOCIATE, the lower
1839 * layers are responsible for association cleanup.
1845 sctp_association_free(asoc);
1847 sctp_release_sock(sk);
1850 return sctp_error(sk, msg_flags, err);
1857 err = sock_error(sk);
1867 /* This is an extended version of skb_pull() that removes the data from the
1868 * start of a skb even when data is spread across the list of skb's in the
1869 * frag_list. len specifies the total amount of data that needs to be removed.
1870 * when 'len' bytes could be removed from the skb, it returns 0.
1871 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1872 * could not be removed.
1874 static int sctp_skb_pull(struct sk_buff *skb, int len)
1876 struct sk_buff *list;
1877 int skb_len = skb_headlen(skb);
1880 if (len <= skb_len) {
1881 __skb_pull(skb, len);
1885 __skb_pull(skb, skb_len);
1887 skb_walk_frags(skb, list) {
1888 rlen = sctp_skb_pull(list, len);
1889 skb->len -= (len-rlen);
1890 skb->data_len -= (len-rlen);
1901 /* API 3.1.3 recvmsg() - UDP Style Syntax
1903 * ssize_t recvmsg(int socket, struct msghdr *message,
1906 * socket - the socket descriptor of the endpoint.
1907 * message - pointer to the msghdr structure which contains a single
1908 * user message and possibly some ancillary data.
1910 * See Section 5 for complete description of the data
1913 * flags - flags sent or received with the user message, see Section
1914 * 5 for complete description of the flags.
1916 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1918 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1919 struct msghdr *msg, size_t len, int noblock,
1920 int flags, int *addr_len)
1922 struct sctp_ulpevent *event = NULL;
1923 struct sctp_sock *sp = sctp_sk(sk);
1924 struct sk_buff *skb;
1929 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1930 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1931 "len", len, "knoblauch", noblock,
1932 "flags", flags, "addr_len", addr_len);
1936 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1941 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1945 /* Get the total length of the skb including any skb's in the
1954 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1956 event = sctp_skb2event(skb);
1961 sock_recv_timestamp(msg, sk, skb);
1962 if (sctp_ulpevent_is_notification(event)) {
1963 msg->msg_flags |= MSG_NOTIFICATION;
1964 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1966 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1969 /* Check if we allow SCTP_SNDRCVINFO. */
1970 if (sp->subscribe.sctp_data_io_event)
1971 sctp_ulpevent_read_sndrcvinfo(event, msg);
1973 /* FIXME: we should be calling IP/IPv6 layers. */
1974 if (sk->sk_protinfo.af_inet.cmsg_flags)
1975 ip_cmsg_recv(msg, skb);
1980 /* If skb's length exceeds the user's buffer, update the skb and
1981 * push it back to the receive_queue so that the next call to
1982 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1984 if (skb_len > copied) {
1985 msg->msg_flags &= ~MSG_EOR;
1986 if (flags & MSG_PEEK)
1988 sctp_skb_pull(skb, copied);
1989 skb_queue_head(&sk->sk_receive_queue, skb);
1991 /* When only partial message is copied to the user, increase
1992 * rwnd by that amount. If all the data in the skb is read,
1993 * rwnd is updated when the event is freed.
1995 if (!sctp_ulpevent_is_notification(event))
1996 sctp_assoc_rwnd_increase(event->asoc, copied);
1998 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1999 (event->msg_flags & MSG_EOR))
2000 msg->msg_flags |= MSG_EOR;
2002 msg->msg_flags &= ~MSG_EOR;
2005 if (flags & MSG_PEEK) {
2006 /* Release the skb reference acquired after peeking the skb in
2007 * sctp_skb_recv_datagram().
2011 /* Free the event which includes releasing the reference to
2012 * the owner of the skb, freeing the skb and updating the
2015 sctp_ulpevent_free(event);
2018 sctp_release_sock(sk);
2022 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2024 * This option is a on/off flag. If enabled no SCTP message
2025 * fragmentation will be performed. Instead if a message being sent
2026 * exceeds the current PMTU size, the message will NOT be sent and
2027 * instead a error will be indicated to the user.
2029 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2030 char __user *optval,
2031 unsigned int optlen)
2035 if (optlen < sizeof(int))
2038 if (get_user(val, (int __user *)optval))
2041 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2046 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2047 unsigned int optlen)
2049 if (optlen > sizeof(struct sctp_event_subscribe))
2051 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2056 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2058 * This socket option is applicable to the UDP-style socket only. When
2059 * set it will cause associations that are idle for more than the
2060 * specified number of seconds to automatically close. An association
2061 * being idle is defined an association that has NOT sent or received
2062 * user data. The special value of '0' indicates that no automatic
2063 * close of any associations should be performed. The option expects an
2064 * integer defining the number of seconds of idle time before an
2065 * association is closed.
2067 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2068 unsigned int optlen)
2070 struct sctp_sock *sp = sctp_sk(sk);
2072 /* Applicable to UDP-style socket only */
2073 if (sctp_style(sk, TCP))
2075 if (optlen != sizeof(int))
2077 if (copy_from_user(&sp->autoclose, optval, optlen))
2083 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2085 * Applications can enable or disable heartbeats for any peer address of
2086 * an association, modify an address's heartbeat interval, force a
2087 * heartbeat to be sent immediately, and adjust the address's maximum
2088 * number of retransmissions sent before an address is considered
2089 * unreachable. The following structure is used to access and modify an
2090 * address's parameters:
2092 * struct sctp_paddrparams {
2093 * sctp_assoc_t spp_assoc_id;
2094 * struct sockaddr_storage spp_address;
2095 * uint32_t spp_hbinterval;
2096 * uint16_t spp_pathmaxrxt;
2097 * uint32_t spp_pathmtu;
2098 * uint32_t spp_sackdelay;
2099 * uint32_t spp_flags;
2102 * spp_assoc_id - (one-to-many style socket) This is filled in the
2103 * application, and identifies the association for
2105 * spp_address - This specifies which address is of interest.
2106 * spp_hbinterval - This contains the value of the heartbeat interval,
2107 * in milliseconds. If a value of zero
2108 * is present in this field then no changes are to
2109 * be made to this parameter.
2110 * spp_pathmaxrxt - This contains the maximum number of
2111 * retransmissions before this address shall be
2112 * considered unreachable. If a value of zero
2113 * is present in this field then no changes are to
2114 * be made to this parameter.
2115 * spp_pathmtu - When Path MTU discovery is disabled the value
2116 * specified here will be the "fixed" path mtu.
2117 * Note that if the spp_address field is empty
2118 * then all associations on this address will
2119 * have this fixed path mtu set upon them.
2121 * spp_sackdelay - When delayed sack is enabled, this value specifies
2122 * the number of milliseconds that sacks will be delayed
2123 * for. This value will apply to all addresses of an
2124 * association if the spp_address field is empty. Note
2125 * also, that if delayed sack is enabled and this
2126 * value is set to 0, no change is made to the last
2127 * recorded delayed sack timer value.
2129 * spp_flags - These flags are used to control various features
2130 * on an association. The flag field may contain
2131 * zero or more of the following options.
2133 * SPP_HB_ENABLE - Enable heartbeats on the
2134 * specified address. Note that if the address
2135 * field is empty all addresses for the association
2136 * have heartbeats enabled upon them.
2138 * SPP_HB_DISABLE - Disable heartbeats on the
2139 * speicifed address. Note that if the address
2140 * field is empty all addresses for the association
2141 * will have their heartbeats disabled. Note also
2142 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2143 * mutually exclusive, only one of these two should
2144 * be specified. Enabling both fields will have
2145 * undetermined results.
2147 * SPP_HB_DEMAND - Request a user initiated heartbeat
2148 * to be made immediately.
2150 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2151 * heartbeat delayis to be set to the value of 0
2154 * SPP_PMTUD_ENABLE - This field will enable PMTU
2155 * discovery upon the specified address. Note that
2156 * if the address feild is empty then all addresses
2157 * on the association are effected.
2159 * SPP_PMTUD_DISABLE - This field will disable PMTU
2160 * discovery upon the specified address. Note that
2161 * if the address feild is empty then all addresses
2162 * on the association are effected. Not also that
2163 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2164 * exclusive. Enabling both will have undetermined
2167 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2168 * on delayed sack. The time specified in spp_sackdelay
2169 * is used to specify the sack delay for this address. Note
2170 * that if spp_address is empty then all addresses will
2171 * enable delayed sack and take on the sack delay
2172 * value specified in spp_sackdelay.
2173 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2174 * off delayed sack. If the spp_address field is blank then
2175 * delayed sack is disabled for the entire association. Note
2176 * also that this field is mutually exclusive to
2177 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2180 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2181 struct sctp_transport *trans,
2182 struct sctp_association *asoc,
2183 struct sctp_sock *sp,
2186 int sackdelay_change)
2190 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2191 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2196 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2197 * this field is ignored. Note also that a value of zero indicates
2198 * the current setting should be left unchanged.
2200 if (params->spp_flags & SPP_HB_ENABLE) {
2202 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2203 * set. This lets us use 0 value when this flag
2206 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2207 params->spp_hbinterval = 0;
2209 if (params->spp_hbinterval ||
2210 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2213 msecs_to_jiffies(params->spp_hbinterval);
2216 msecs_to_jiffies(params->spp_hbinterval);
2218 sp->hbinterval = params->spp_hbinterval;
2225 trans->param_flags =
2226 (trans->param_flags & ~SPP_HB) | hb_change;
2229 (asoc->param_flags & ~SPP_HB) | hb_change;
2232 (sp->param_flags & ~SPP_HB) | hb_change;
2236 /* When Path MTU discovery is disabled the value specified here will
2237 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2238 * include the flag SPP_PMTUD_DISABLE for this field to have any
2241 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2243 trans->pathmtu = params->spp_pathmtu;
2244 sctp_assoc_sync_pmtu(asoc);
2246 asoc->pathmtu = params->spp_pathmtu;
2247 sctp_frag_point(asoc, params->spp_pathmtu);
2249 sp->pathmtu = params->spp_pathmtu;
2255 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2256 (params->spp_flags & SPP_PMTUD_ENABLE);
2257 trans->param_flags =
2258 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2260 sctp_transport_pmtu(trans);
2261 sctp_assoc_sync_pmtu(asoc);
2265 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2268 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2272 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2273 * value of this field is ignored. Note also that a value of zero
2274 * indicates the current setting should be left unchanged.
2276 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2279 msecs_to_jiffies(params->spp_sackdelay);
2282 msecs_to_jiffies(params->spp_sackdelay);
2284 sp->sackdelay = params->spp_sackdelay;
2288 if (sackdelay_change) {
2290 trans->param_flags =
2291 (trans->param_flags & ~SPP_SACKDELAY) |
2295 (asoc->param_flags & ~SPP_SACKDELAY) |
2299 (sp->param_flags & ~SPP_SACKDELAY) |
2304 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2305 * of this field is ignored. Note also that a value of zero
2306 * indicates the current setting should be left unchanged.
2308 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2310 trans->pathmaxrxt = params->spp_pathmaxrxt;
2312 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2314 sp->pathmaxrxt = params->spp_pathmaxrxt;
2321 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2322 char __user *optval,
2323 unsigned int optlen)
2325 struct sctp_paddrparams params;
2326 struct sctp_transport *trans = NULL;
2327 struct sctp_association *asoc = NULL;
2328 struct sctp_sock *sp = sctp_sk(sk);
2330 int hb_change, pmtud_change, sackdelay_change;
2332 if (optlen != sizeof(struct sctp_paddrparams))
2335 if (copy_from_user(¶ms, optval, optlen))
2338 /* Validate flags and value parameters. */
2339 hb_change = params.spp_flags & SPP_HB;
2340 pmtud_change = params.spp_flags & SPP_PMTUD;
2341 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2343 if (hb_change == SPP_HB ||
2344 pmtud_change == SPP_PMTUD ||
2345 sackdelay_change == SPP_SACKDELAY ||
2346 params.spp_sackdelay > 500 ||
2348 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2351 /* If an address other than INADDR_ANY is specified, and
2352 * no transport is found, then the request is invalid.
2354 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2355 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2356 params.spp_assoc_id);
2361 /* Get association, if assoc_id != 0 and the socket is a one
2362 * to many style socket, and an association was not found, then
2363 * the id was invalid.
2365 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2366 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2369 /* Heartbeat demand can only be sent on a transport or
2370 * association, but not a socket.
2372 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2375 /* Process parameters. */
2376 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2377 hb_change, pmtud_change,
2383 /* If changes are for association, also apply parameters to each
2386 if (!trans && asoc) {
2387 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2389 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2390 hb_change, pmtud_change,
2399 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2401 * This option will effect the way delayed acks are performed. This
2402 * option allows you to get or set the delayed ack time, in
2403 * milliseconds. It also allows changing the delayed ack frequency.
2404 * Changing the frequency to 1 disables the delayed sack algorithm. If
2405 * the assoc_id is 0, then this sets or gets the endpoints default
2406 * values. If the assoc_id field is non-zero, then the set or get
2407 * effects the specified association for the one to many model (the
2408 * assoc_id field is ignored by the one to one model). Note that if
2409 * sack_delay or sack_freq are 0 when setting this option, then the
2410 * current values will remain unchanged.
2412 * struct sctp_sack_info {
2413 * sctp_assoc_t sack_assoc_id;
2414 * uint32_t sack_delay;
2415 * uint32_t sack_freq;
2418 * sack_assoc_id - This parameter, indicates which association the user
2419 * is performing an action upon. Note that if this field's value is
2420 * zero then the endpoints default value is changed (effecting future
2421 * associations only).
2423 * sack_delay - This parameter contains the number of milliseconds that
2424 * the user is requesting the delayed ACK timer be set to. Note that
2425 * this value is defined in the standard to be between 200 and 500
2428 * sack_freq - This parameter contains the number of packets that must
2429 * be received before a sack is sent without waiting for the delay
2430 * timer to expire. The default value for this is 2, setting this
2431 * value to 1 will disable the delayed sack algorithm.
2434 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2435 char __user *optval, unsigned int optlen)
2437 struct sctp_sack_info params;
2438 struct sctp_transport *trans = NULL;
2439 struct sctp_association *asoc = NULL;
2440 struct sctp_sock *sp = sctp_sk(sk);
2442 if (optlen == sizeof(struct sctp_sack_info)) {
2443 if (copy_from_user(¶ms, optval, optlen))
2446 if (params.sack_delay == 0 && params.sack_freq == 0)
2448 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2449 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2450 "in delayed_ack socket option deprecated\n");
2451 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2452 if (copy_from_user(¶ms, optval, optlen))
2455 if (params.sack_delay == 0)
2456 params.sack_freq = 1;
2458 params.sack_freq = 0;
2462 /* Validate value parameter. */
2463 if (params.sack_delay > 500)
2466 /* Get association, if sack_assoc_id != 0 and the socket is a one
2467 * to many style socket, and an association was not found, then
2468 * the id was invalid.
2470 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2471 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2474 if (params.sack_delay) {
2477 msecs_to_jiffies(params.sack_delay);
2479 (asoc->param_flags & ~SPP_SACKDELAY) |
2480 SPP_SACKDELAY_ENABLE;
2482 sp->sackdelay = params.sack_delay;
2484 (sp->param_flags & ~SPP_SACKDELAY) |
2485 SPP_SACKDELAY_ENABLE;
2489 if (params.sack_freq == 1) {
2492 (asoc->param_flags & ~SPP_SACKDELAY) |
2493 SPP_SACKDELAY_DISABLE;
2496 (sp->param_flags & ~SPP_SACKDELAY) |
2497 SPP_SACKDELAY_DISABLE;
2499 } else if (params.sack_freq > 1) {
2501 asoc->sackfreq = params.sack_freq;
2503 (asoc->param_flags & ~SPP_SACKDELAY) |
2504 SPP_SACKDELAY_ENABLE;
2506 sp->sackfreq = params.sack_freq;
2508 (sp->param_flags & ~SPP_SACKDELAY) |
2509 SPP_SACKDELAY_ENABLE;
2513 /* If change is for association, also apply to each transport. */
2515 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2517 if (params.sack_delay) {
2519 msecs_to_jiffies(params.sack_delay);
2520 trans->param_flags =
2521 (trans->param_flags & ~SPP_SACKDELAY) |
2522 SPP_SACKDELAY_ENABLE;
2524 if (params.sack_freq == 1) {
2525 trans->param_flags =
2526 (trans->param_flags & ~SPP_SACKDELAY) |
2527 SPP_SACKDELAY_DISABLE;
2528 } else if (params.sack_freq > 1) {
2529 trans->sackfreq = params.sack_freq;
2530 trans->param_flags =
2531 (trans->param_flags & ~SPP_SACKDELAY) |
2532 SPP_SACKDELAY_ENABLE;
2540 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2542 * Applications can specify protocol parameters for the default association
2543 * initialization. The option name argument to setsockopt() and getsockopt()
2546 * Setting initialization parameters is effective only on an unconnected
2547 * socket (for UDP-style sockets only future associations are effected
2548 * by the change). With TCP-style sockets, this option is inherited by
2549 * sockets derived from a listener socket.
2551 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2553 struct sctp_initmsg sinit;
2554 struct sctp_sock *sp = sctp_sk(sk);
2556 if (optlen != sizeof(struct sctp_initmsg))
2558 if (copy_from_user(&sinit, optval, optlen))
2561 if (sinit.sinit_num_ostreams)
2562 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2563 if (sinit.sinit_max_instreams)
2564 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2565 if (sinit.sinit_max_attempts)
2566 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2567 if (sinit.sinit_max_init_timeo)
2568 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2574 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2576 * Applications that wish to use the sendto() system call may wish to
2577 * specify a default set of parameters that would normally be supplied
2578 * through the inclusion of ancillary data. This socket option allows
2579 * such an application to set the default sctp_sndrcvinfo structure.
2580 * The application that wishes to use this socket option simply passes
2581 * in to this call the sctp_sndrcvinfo structure defined in Section
2582 * 5.2.2) The input parameters accepted by this call include
2583 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2584 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2585 * to this call if the caller is using the UDP model.
2587 static int sctp_setsockopt_default_send_param(struct sock *sk,
2588 char __user *optval,
2589 unsigned int optlen)
2591 struct sctp_sndrcvinfo info;
2592 struct sctp_association *asoc;
2593 struct sctp_sock *sp = sctp_sk(sk);
2595 if (optlen != sizeof(struct sctp_sndrcvinfo))
2597 if (copy_from_user(&info, optval, optlen))
2600 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2601 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2605 asoc->default_stream = info.sinfo_stream;
2606 asoc->default_flags = info.sinfo_flags;
2607 asoc->default_ppid = info.sinfo_ppid;
2608 asoc->default_context = info.sinfo_context;
2609 asoc->default_timetolive = info.sinfo_timetolive;
2611 sp->default_stream = info.sinfo_stream;
2612 sp->default_flags = info.sinfo_flags;
2613 sp->default_ppid = info.sinfo_ppid;
2614 sp->default_context = info.sinfo_context;
2615 sp->default_timetolive = info.sinfo_timetolive;
2621 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2623 * Requests that the local SCTP stack use the enclosed peer address as
2624 * the association primary. The enclosed address must be one of the
2625 * association peer's addresses.
2627 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2628 unsigned int optlen)
2630 struct sctp_prim prim;
2631 struct sctp_transport *trans;
2633 if (optlen != sizeof(struct sctp_prim))
2636 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2639 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2643 sctp_assoc_set_primary(trans->asoc, trans);
2649 * 7.1.5 SCTP_NODELAY
2651 * Turn on/off any Nagle-like algorithm. This means that packets are
2652 * generally sent as soon as possible and no unnecessary delays are
2653 * introduced, at the cost of more packets in the network. Expects an
2654 * integer boolean flag.
2656 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2657 unsigned int optlen)
2661 if (optlen < sizeof(int))
2663 if (get_user(val, (int __user *)optval))
2666 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2672 * 7.1.1 SCTP_RTOINFO
2674 * The protocol parameters used to initialize and bound retransmission
2675 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2676 * and modify these parameters.
2677 * All parameters are time values, in milliseconds. A value of 0, when
2678 * modifying the parameters, indicates that the current value should not
2682 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2684 struct sctp_rtoinfo rtoinfo;
2685 struct sctp_association *asoc;
2687 if (optlen != sizeof (struct sctp_rtoinfo))
2690 if (copy_from_user(&rtoinfo, optval, optlen))
2693 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2695 /* Set the values to the specific association */
2696 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2700 if (rtoinfo.srto_initial != 0)
2702 msecs_to_jiffies(rtoinfo.srto_initial);
2703 if (rtoinfo.srto_max != 0)
2704 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2705 if (rtoinfo.srto_min != 0)
2706 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2708 /* If there is no association or the association-id = 0
2709 * set the values to the endpoint.
2711 struct sctp_sock *sp = sctp_sk(sk);
2713 if (rtoinfo.srto_initial != 0)
2714 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2715 if (rtoinfo.srto_max != 0)
2716 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2717 if (rtoinfo.srto_min != 0)
2718 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2726 * 7.1.2 SCTP_ASSOCINFO
2728 * This option is used to tune the maximum retransmission attempts
2729 * of the association.
2730 * Returns an error if the new association retransmission value is
2731 * greater than the sum of the retransmission value of the peer.
2732 * See [SCTP] for more information.
2735 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2738 struct sctp_assocparams assocparams;
2739 struct sctp_association *asoc;
2741 if (optlen != sizeof(struct sctp_assocparams))
2743 if (copy_from_user(&assocparams, optval, optlen))
2746 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2748 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2751 /* Set the values to the specific association */
2753 if (assocparams.sasoc_asocmaxrxt != 0) {
2756 struct sctp_transport *peer_addr;
2758 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2760 path_sum += peer_addr->pathmaxrxt;
2764 /* Only validate asocmaxrxt if we have more than
2765 * one path/transport. We do this because path
2766 * retransmissions are only counted when we have more
2770 assocparams.sasoc_asocmaxrxt > path_sum)
2773 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2776 if (assocparams.sasoc_cookie_life != 0) {
2777 asoc->cookie_life.tv_sec =
2778 assocparams.sasoc_cookie_life / 1000;
2779 asoc->cookie_life.tv_usec =
2780 (assocparams.sasoc_cookie_life % 1000)
2784 /* Set the values to the endpoint */
2785 struct sctp_sock *sp = sctp_sk(sk);
2787 if (assocparams.sasoc_asocmaxrxt != 0)
2788 sp->assocparams.sasoc_asocmaxrxt =
2789 assocparams.sasoc_asocmaxrxt;
2790 if (assocparams.sasoc_cookie_life != 0)
2791 sp->assocparams.sasoc_cookie_life =
2792 assocparams.sasoc_cookie_life;
2798 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2800 * This socket option is a boolean flag which turns on or off mapped V4
2801 * addresses. If this option is turned on and the socket is type
2802 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2803 * If this option is turned off, then no mapping will be done of V4
2804 * addresses and a user will receive both PF_INET6 and PF_INET type
2805 * addresses on the socket.
2807 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2810 struct sctp_sock *sp = sctp_sk(sk);
2812 if (optlen < sizeof(int))
2814 if (get_user(val, (int __user *)optval))
2825 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2826 * This option will get or set the maximum size to put in any outgoing
2827 * SCTP DATA chunk. If a message is larger than this size it will be
2828 * fragmented by SCTP into the specified size. Note that the underlying
2829 * SCTP implementation may fragment into smaller sized chunks when the
2830 * PMTU of the underlying association is smaller than the value set by
2831 * the user. The default value for this option is '0' which indicates
2832 * the user is NOT limiting fragmentation and only the PMTU will effect
2833 * SCTP's choice of DATA chunk size. Note also that values set larger
2834 * than the maximum size of an IP datagram will effectively let SCTP
2835 * control fragmentation (i.e. the same as setting this option to 0).
2837 * The following structure is used to access and modify this parameter:
2839 * struct sctp_assoc_value {
2840 * sctp_assoc_t assoc_id;
2841 * uint32_t assoc_value;
2844 * assoc_id: This parameter is ignored for one-to-one style sockets.
2845 * For one-to-many style sockets this parameter indicates which
2846 * association the user is performing an action upon. Note that if
2847 * this field's value is zero then the endpoints default value is
2848 * changed (effecting future associations only).
2849 * assoc_value: This parameter specifies the maximum size in bytes.
2851 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2853 struct sctp_assoc_value params;
2854 struct sctp_association *asoc;
2855 struct sctp_sock *sp = sctp_sk(sk);
2858 if (optlen == sizeof(int)) {
2860 "SCTP: Use of int in maxseg socket option deprecated\n");
2862 "SCTP: Use struct sctp_assoc_value instead\n");
2863 if (copy_from_user(&val, optval, optlen))
2865 params.assoc_id = 0;
2866 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2867 if (copy_from_user(¶ms, optval, optlen))
2869 val = params.assoc_value;
2873 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2876 asoc = sctp_id2assoc(sk, params.assoc_id);
2877 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2882 val = asoc->pathmtu;
2883 val -= sp->pf->af->net_header_len;
2884 val -= sizeof(struct sctphdr) +
2885 sizeof(struct sctp_data_chunk);
2887 asoc->user_frag = val;
2888 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2890 sp->user_frag = val;
2898 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2900 * Requests that the peer mark the enclosed address as the association
2901 * primary. The enclosed address must be one of the association's
2902 * locally bound addresses. The following structure is used to make a
2903 * set primary request:
2905 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2906 unsigned int optlen)
2908 struct sctp_sock *sp;
2909 struct sctp_endpoint *ep;
2910 struct sctp_association *asoc = NULL;
2911 struct sctp_setpeerprim prim;
2912 struct sctp_chunk *chunk;
2918 if (!sctp_addip_enable)
2921 if (optlen != sizeof(struct sctp_setpeerprim))
2924 if (copy_from_user(&prim, optval, optlen))
2927 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2931 if (!asoc->peer.asconf_capable)
2934 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2937 if (!sctp_state(asoc, ESTABLISHED))
2940 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2941 return -EADDRNOTAVAIL;
2943 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2944 chunk = sctp_make_asconf_set_prim(asoc,
2945 (union sctp_addr *)&prim.sspp_addr);
2949 err = sctp_send_asconf(asoc, chunk);
2951 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2956 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2957 unsigned int optlen)
2959 struct sctp_setadaptation adaptation;
2961 if (optlen != sizeof(struct sctp_setadaptation))
2963 if (copy_from_user(&adaptation, optval, optlen))
2966 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2972 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2974 * The context field in the sctp_sndrcvinfo structure is normally only
2975 * used when a failed message is retrieved holding the value that was
2976 * sent down on the actual send call. This option allows the setting of
2977 * a default context on an association basis that will be received on
2978 * reading messages from the peer. This is especially helpful in the
2979 * one-2-many model for an application to keep some reference to an
2980 * internal state machine that is processing messages on the
2981 * association. Note that the setting of this value only effects
2982 * received messages from the peer and does not effect the value that is
2983 * saved with outbound messages.
2985 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2986 unsigned int optlen)
2988 struct sctp_assoc_value params;
2989 struct sctp_sock *sp;
2990 struct sctp_association *asoc;
2992 if (optlen != sizeof(struct sctp_assoc_value))
2994 if (copy_from_user(¶ms, optval, optlen))
2999 if (params.assoc_id != 0) {
3000 asoc = sctp_id2assoc(sk, params.assoc_id);
3003 asoc->default_rcv_context = params.assoc_value;
3005 sp->default_rcv_context = params.assoc_value;
3012 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3014 * This options will at a minimum specify if the implementation is doing
3015 * fragmented interleave. Fragmented interleave, for a one to many
3016 * socket, is when subsequent calls to receive a message may return
3017 * parts of messages from different associations. Some implementations
3018 * may allow you to turn this value on or off. If so, when turned off,
3019 * no fragment interleave will occur (which will cause a head of line
3020 * blocking amongst multiple associations sharing the same one to many
3021 * socket). When this option is turned on, then each receive call may
3022 * come from a different association (thus the user must receive data
3023 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3024 * association each receive belongs to.
3026 * This option takes a boolean value. A non-zero value indicates that
3027 * fragmented interleave is on. A value of zero indicates that
3028 * fragmented interleave is off.
3030 * Note that it is important that an implementation that allows this
3031 * option to be turned on, have it off by default. Otherwise an unaware
3032 * application using the one to many model may become confused and act
3035 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3036 char __user *optval,
3037 unsigned int optlen)
3041 if (optlen != sizeof(int))
3043 if (get_user(val, (int __user *)optval))
3046 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3052 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3053 * (SCTP_PARTIAL_DELIVERY_POINT)
3055 * This option will set or get the SCTP partial delivery point. This
3056 * point is the size of a message where the partial delivery API will be
3057 * invoked to help free up rwnd space for the peer. Setting this to a
3058 * lower value will cause partial deliveries to happen more often. The
3059 * calls argument is an integer that sets or gets the partial delivery
3060 * point. Note also that the call will fail if the user attempts to set
3061 * this value larger than the socket receive buffer size.
3063 * Note that any single message having a length smaller than or equal to
3064 * the SCTP partial delivery point will be delivered in one single read
3065 * call as long as the user provided buffer is large enough to hold the
3068 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3069 char __user *optval,
3070 unsigned int optlen)
3074 if (optlen != sizeof(u32))
3076 if (get_user(val, (int __user *)optval))
3079 /* Note: We double the receive buffer from what the user sets
3080 * it to be, also initial rwnd is based on rcvbuf/2.
3082 if (val > (sk->sk_rcvbuf >> 1))
3085 sctp_sk(sk)->pd_point = val;
3087 return 0; /* is this the right error code? */
3091 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3093 * This option will allow a user to change the maximum burst of packets
3094 * that can be emitted by this association. Note that the default value
3095 * is 4, and some implementations may restrict this setting so that it
3096 * can only be lowered.
3098 * NOTE: This text doesn't seem right. Do this on a socket basis with
3099 * future associations inheriting the socket value.
3101 static int sctp_setsockopt_maxburst(struct sock *sk,
3102 char __user *optval,
3103 unsigned int optlen)
3105 struct sctp_assoc_value params;
3106 struct sctp_sock *sp;
3107 struct sctp_association *asoc;
3111 if (optlen == sizeof(int)) {
3113 "SCTP: Use of int in max_burst socket option deprecated\n");
3115 "SCTP: Use struct sctp_assoc_value instead\n");
3116 if (copy_from_user(&val, optval, optlen))
3118 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3119 if (copy_from_user(¶ms, optval, optlen))
3121 val = params.assoc_value;
3122 assoc_id = params.assoc_id;
3128 if (assoc_id != 0) {
3129 asoc = sctp_id2assoc(sk, assoc_id);
3132 asoc->max_burst = val;
3134 sp->max_burst = val;
3140 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3142 * This set option adds a chunk type that the user is requesting to be
3143 * received only in an authenticated way. Changes to the list of chunks
3144 * will only effect future associations on the socket.
3146 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3147 char __user *optval,
3148 unsigned int optlen)
3150 struct sctp_authchunk val;
3152 if (!sctp_auth_enable)
3155 if (optlen != sizeof(struct sctp_authchunk))
3157 if (copy_from_user(&val, optval, optlen))
3160 switch (val.sauth_chunk) {
3162 case SCTP_CID_INIT_ACK:
3163 case SCTP_CID_SHUTDOWN_COMPLETE:
3168 /* add this chunk id to the endpoint */
3169 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3173 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3175 * This option gets or sets the list of HMAC algorithms that the local
3176 * endpoint requires the peer to use.
3178 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3179 char __user *optval,
3180 unsigned int optlen)
3182 struct sctp_hmacalgo *hmacs;
3186 if (!sctp_auth_enable)
3189 if (optlen < sizeof(struct sctp_hmacalgo))
3192 hmacs = kmalloc(optlen, GFP_KERNEL);
3196 if (copy_from_user(hmacs, optval, optlen)) {
3201 idents = hmacs->shmac_num_idents;
3202 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3203 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3208 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3215 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3217 * This option will set a shared secret key which is used to build an
3218 * association shared key.
3220 static int sctp_setsockopt_auth_key(struct sock *sk,
3221 char __user *optval,
3222 unsigned int optlen)
3224 struct sctp_authkey *authkey;
3225 struct sctp_association *asoc;
3228 if (!sctp_auth_enable)
3231 if (optlen <= sizeof(struct sctp_authkey))
3234 authkey = kmalloc(optlen, GFP_KERNEL);
3238 if (copy_from_user(authkey, optval, optlen)) {
3243 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3248 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3249 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3254 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3261 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3263 * This option will get or set the active shared key to be used to build
3264 * the association shared key.
3266 static int sctp_setsockopt_active_key(struct sock *sk,
3267 char __user *optval,
3268 unsigned int optlen)
3270 struct sctp_authkeyid val;
3271 struct sctp_association *asoc;
3273 if (!sctp_auth_enable)
3276 if (optlen != sizeof(struct sctp_authkeyid))
3278 if (copy_from_user(&val, optval, optlen))
3281 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3282 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3285 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3286 val.scact_keynumber);
3290 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3292 * This set option will delete a shared secret key from use.
3294 static int sctp_setsockopt_del_key(struct sock *sk,
3295 char __user *optval,
3296 unsigned int optlen)
3298 struct sctp_authkeyid val;
3299 struct sctp_association *asoc;
3301 if (!sctp_auth_enable)
3304 if (optlen != sizeof(struct sctp_authkeyid))
3306 if (copy_from_user(&val, optval, optlen))
3309 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3310 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3313 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3314 val.scact_keynumber);
3319 /* API 6.2 setsockopt(), getsockopt()
3321 * Applications use setsockopt() and getsockopt() to set or retrieve
3322 * socket options. Socket options are used to change the default
3323 * behavior of sockets calls. They are described in Section 7.
3327 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3328 * int __user *optlen);
3329 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3332 * sd - the socket descript.
3333 * level - set to IPPROTO_SCTP for all SCTP options.
3334 * optname - the option name.
3335 * optval - the buffer to store the value of the option.
3336 * optlen - the size of the buffer.
3338 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3339 char __user *optval, unsigned int optlen)
3343 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3346 /* I can hardly begin to describe how wrong this is. This is
3347 * so broken as to be worse than useless. The API draft
3348 * REALLY is NOT helpful here... I am not convinced that the
3349 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3350 * are at all well-founded.
3352 if (level != SOL_SCTP) {
3353 struct sctp_af *af = sctp_sk(sk)->pf->af;
3354 retval = af->setsockopt(sk, level, optname, optval, optlen);
3361 case SCTP_SOCKOPT_BINDX_ADD:
3362 /* 'optlen' is the size of the addresses buffer. */
3363 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3364 optlen, SCTP_BINDX_ADD_ADDR);
3367 case SCTP_SOCKOPT_BINDX_REM:
3368 /* 'optlen' is the size of the addresses buffer. */
3369 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3370 optlen, SCTP_BINDX_REM_ADDR);
3373 case SCTP_SOCKOPT_CONNECTX_OLD:
3374 /* 'optlen' is the size of the addresses buffer. */
3375 retval = sctp_setsockopt_connectx_old(sk,
3376 (struct sockaddr __user *)optval,
3380 case SCTP_SOCKOPT_CONNECTX:
3381 /* 'optlen' is the size of the addresses buffer. */
3382 retval = sctp_setsockopt_connectx(sk,
3383 (struct sockaddr __user *)optval,
3387 case SCTP_DISABLE_FRAGMENTS:
3388 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3392 retval = sctp_setsockopt_events(sk, optval, optlen);
3395 case SCTP_AUTOCLOSE:
3396 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3399 case SCTP_PEER_ADDR_PARAMS:
3400 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3403 case SCTP_DELAYED_ACK:
3404 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3406 case SCTP_PARTIAL_DELIVERY_POINT:
3407 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3411 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3413 case SCTP_DEFAULT_SEND_PARAM:
3414 retval = sctp_setsockopt_default_send_param(sk, optval,
3417 case SCTP_PRIMARY_ADDR:
3418 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3420 case SCTP_SET_PEER_PRIMARY_ADDR:
3421 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3424 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3427 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3429 case SCTP_ASSOCINFO:
3430 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3432 case SCTP_I_WANT_MAPPED_V4_ADDR:
3433 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3436 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3438 case SCTP_ADAPTATION_LAYER:
3439 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3442 retval = sctp_setsockopt_context(sk, optval, optlen);
3444 case SCTP_FRAGMENT_INTERLEAVE:
3445 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3447 case SCTP_MAX_BURST:
3448 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3450 case SCTP_AUTH_CHUNK:
3451 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3453 case SCTP_HMAC_IDENT:
3454 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3457 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3459 case SCTP_AUTH_ACTIVE_KEY:
3460 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3462 case SCTP_AUTH_DELETE_KEY:
3463 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3466 retval = -ENOPROTOOPT;
3470 sctp_release_sock(sk);
3476 /* API 3.1.6 connect() - UDP Style Syntax
3478 * An application may use the connect() call in the UDP model to initiate an
3479 * association without sending data.
3483 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3485 * sd: the socket descriptor to have a new association added to.
3487 * nam: the address structure (either struct sockaddr_in or struct
3488 * sockaddr_in6 defined in RFC2553 [7]).
3490 * len: the size of the address.
3492 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3500 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3501 __func__, sk, addr, addr_len);
3503 /* Validate addr_len before calling common connect/connectx routine. */
3504 af = sctp_get_af_specific(addr->sa_family);
3505 if (!af || addr_len < af->sockaddr_len) {
3508 /* Pass correct addr len to common routine (so it knows there
3509 * is only one address being passed.
3511 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3514 sctp_release_sock(sk);
3518 /* FIXME: Write comments. */
3519 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3521 return -EOPNOTSUPP; /* STUB */
3524 /* 4.1.4 accept() - TCP Style Syntax
3526 * Applications use accept() call to remove an established SCTP
3527 * association from the accept queue of the endpoint. A new socket
3528 * descriptor will be returned from accept() to represent the newly
3529 * formed association.
3531 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3533 struct sctp_sock *sp;
3534 struct sctp_endpoint *ep;
3535 struct sock *newsk = NULL;
3536 struct sctp_association *asoc;
3545 if (!sctp_style(sk, TCP)) {
3546 error = -EOPNOTSUPP;
3550 if (!sctp_sstate(sk, LISTENING)) {
3555 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3557 error = sctp_wait_for_accept(sk, timeo);
3561 /* We treat the list of associations on the endpoint as the accept
3562 * queue and pick the first association on the list.
3564 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3566 newsk = sp->pf->create_accept_sk(sk, asoc);
3572 /* Populate the fields of the newsk from the oldsk and migrate the
3573 * asoc to the newsk.
3575 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3578 sctp_release_sock(sk);
3583 /* The SCTP ioctl handler. */
3584 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3586 return -ENOIOCTLCMD;
3589 /* This is the function which gets called during socket creation to
3590 * initialized the SCTP-specific portion of the sock.
3591 * The sock structure should already be zero-filled memory.
3593 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3595 struct sctp_endpoint *ep;
3596 struct sctp_sock *sp;
3598 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3602 /* Initialize the SCTP per socket area. */
3603 switch (sk->sk_type) {
3604 case SOCK_SEQPACKET:
3605 sp->type = SCTP_SOCKET_UDP;
3608 sp->type = SCTP_SOCKET_TCP;
3611 return -ESOCKTNOSUPPORT;
3614 /* Initialize default send parameters. These parameters can be
3615 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3617 sp->default_stream = 0;
3618 sp->default_ppid = 0;
3619 sp->default_flags = 0;
3620 sp->default_context = 0;
3621 sp->default_timetolive = 0;
3623 sp->default_rcv_context = 0;
3624 sp->max_burst = sctp_max_burst;
3626 /* Initialize default setup parameters. These parameters
3627 * can be modified with the SCTP_INITMSG socket option or
3628 * overridden by the SCTP_INIT CMSG.
3630 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3631 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3632 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3633 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3635 /* Initialize default RTO related parameters. These parameters can
3636 * be modified for with the SCTP_RTOINFO socket option.
3638 sp->rtoinfo.srto_initial = sctp_rto_initial;
3639 sp->rtoinfo.srto_max = sctp_rto_max;
3640 sp->rtoinfo.srto_min = sctp_rto_min;
3642 /* Initialize default association related parameters. These parameters
3643 * can be modified with the SCTP_ASSOCINFO socket option.
3645 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3646 sp->assocparams.sasoc_number_peer_destinations = 0;
3647 sp->assocparams.sasoc_peer_rwnd = 0;
3648 sp->assocparams.sasoc_local_rwnd = 0;
3649 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3651 /* Initialize default event subscriptions. By default, all the
3654 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3656 /* Default Peer Address Parameters. These defaults can
3657 * be modified via SCTP_PEER_ADDR_PARAMS
3659 sp->hbinterval = sctp_hb_interval;
3660 sp->pathmaxrxt = sctp_max_retrans_path;
3661 sp->pathmtu = 0; // allow default discovery
3662 sp->sackdelay = sctp_sack_timeout;
3664 sp->param_flags = SPP_HB_ENABLE |
3666 SPP_SACKDELAY_ENABLE;
3668 /* If enabled no SCTP message fragmentation will be performed.
3669 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3671 sp->disable_fragments = 0;
3673 /* Enable Nagle algorithm by default. */
3676 /* Enable by default. */
3679 /* Auto-close idle associations after the configured
3680 * number of seconds. A value of 0 disables this
3681 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3682 * for UDP-style sockets only.
3686 /* User specified fragmentation limit. */
3689 sp->adaptation_ind = 0;
3691 sp->pf = sctp_get_pf_specific(sk->sk_family);
3693 /* Control variables for partial data delivery. */
3694 atomic_set(&sp->pd_mode, 0);
3695 skb_queue_head_init(&sp->pd_lobby);
3696 sp->frag_interleave = 0;
3698 /* Create a per socket endpoint structure. Even if we
3699 * change the data structure relationships, this may still
3700 * be useful for storing pre-connect address information.
3702 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3709 SCTP_DBG_OBJCNT_INC(sock);
3710 percpu_counter_inc(&sctp_sockets_allocated);
3713 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3719 /* Cleanup any SCTP per socket resources. */
3720 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3722 struct sctp_endpoint *ep;
3724 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3726 /* Release our hold on the endpoint. */
3727 ep = sctp_sk(sk)->ep;
3728 sctp_endpoint_free(ep);
3729 percpu_counter_dec(&sctp_sockets_allocated);
3731 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3735 /* API 4.1.7 shutdown() - TCP Style Syntax
3736 * int shutdown(int socket, int how);
3738 * sd - the socket descriptor of the association to be closed.
3739 * how - Specifies the type of shutdown. The values are
3742 * Disables further receive operations. No SCTP
3743 * protocol action is taken.
3745 * Disables further send operations, and initiates
3746 * the SCTP shutdown sequence.
3748 * Disables further send and receive operations
3749 * and initiates the SCTP shutdown sequence.
3751 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3753 struct sctp_endpoint *ep;
3754 struct sctp_association *asoc;
3756 if (!sctp_style(sk, TCP))
3759 if (how & SEND_SHUTDOWN) {
3760 ep = sctp_sk(sk)->ep;
3761 if (!list_empty(&ep->asocs)) {
3762 asoc = list_entry(ep->asocs.next,
3763 struct sctp_association, asocs);
3764 sctp_primitive_SHUTDOWN(asoc, NULL);
3769 /* 7.2.1 Association Status (SCTP_STATUS)
3771 * Applications can retrieve current status information about an
3772 * association, including association state, peer receiver window size,
3773 * number of unacked data chunks, and number of data chunks pending
3774 * receipt. This information is read-only.
3776 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3777 char __user *optval,
3780 struct sctp_status status;
3781 struct sctp_association *asoc = NULL;
3782 struct sctp_transport *transport;
3783 sctp_assoc_t associd;
3786 if (len < sizeof(status)) {
3791 len = sizeof(status);
3792 if (copy_from_user(&status, optval, len)) {
3797 associd = status.sstat_assoc_id;
3798 asoc = sctp_id2assoc(sk, associd);
3804 transport = asoc->peer.primary_path;
3806 status.sstat_assoc_id = sctp_assoc2id(asoc);
3807 status.sstat_state = asoc->state;
3808 status.sstat_rwnd = asoc->peer.rwnd;
3809 status.sstat_unackdata = asoc->unack_data;
3811 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3812 status.sstat_instrms = asoc->c.sinit_max_instreams;
3813 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3814 status.sstat_fragmentation_point = asoc->frag_point;
3815 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3816 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3817 transport->af_specific->sockaddr_len);
3818 /* Map ipv4 address into v4-mapped-on-v6 address. */
3819 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3820 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3821 status.sstat_primary.spinfo_state = transport->state;
3822 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3823 status.sstat_primary.spinfo_srtt = transport->srtt;
3824 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3825 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3827 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3828 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3830 if (put_user(len, optlen)) {
3835 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3836 len, status.sstat_state, status.sstat_rwnd,
3837 status.sstat_assoc_id);
3839 if (copy_to_user(optval, &status, len)) {
3849 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3851 * Applications can retrieve information about a specific peer address
3852 * of an association, including its reachability state, congestion
3853 * window, and retransmission timer values. This information is
3856 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3857 char __user *optval,
3860 struct sctp_paddrinfo pinfo;
3861 struct sctp_transport *transport;
3864 if (len < sizeof(pinfo)) {
3869 len = sizeof(pinfo);
3870 if (copy_from_user(&pinfo, optval, len)) {
3875 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3876 pinfo.spinfo_assoc_id);
3880 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3881 pinfo.spinfo_state = transport->state;
3882 pinfo.spinfo_cwnd = transport->cwnd;
3883 pinfo.spinfo_srtt = transport->srtt;
3884 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3885 pinfo.spinfo_mtu = transport->pathmtu;
3887 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3888 pinfo.spinfo_state = SCTP_ACTIVE;
3890 if (put_user(len, optlen)) {
3895 if (copy_to_user(optval, &pinfo, len)) {
3904 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3906 * This option is a on/off flag. If enabled no SCTP message
3907 * fragmentation will be performed. Instead if a message being sent
3908 * exceeds the current PMTU size, the message will NOT be sent and
3909 * instead a error will be indicated to the user.
3911 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3912 char __user *optval, int __user *optlen)
3916 if (len < sizeof(int))
3920 val = (sctp_sk(sk)->disable_fragments == 1);
3921 if (put_user(len, optlen))
3923 if (copy_to_user(optval, &val, len))
3928 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3930 * This socket option is used to specify various notifications and
3931 * ancillary data the user wishes to receive.
3933 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3936 if (len < sizeof(struct sctp_event_subscribe))
3938 len = sizeof(struct sctp_event_subscribe);
3939 if (put_user(len, optlen))
3941 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3946 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3948 * This socket option is applicable to the UDP-style socket only. When
3949 * set it will cause associations that are idle for more than the
3950 * specified number of seconds to automatically close. An association
3951 * being idle is defined an association that has NOT sent or received
3952 * user data. The special value of '0' indicates that no automatic
3953 * close of any associations should be performed. The option expects an
3954 * integer defining the number of seconds of idle time before an
3955 * association is closed.
3957 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3959 /* Applicable to UDP-style socket only */
3960 if (sctp_style(sk, TCP))
3962 if (len < sizeof(int))
3965 if (put_user(len, optlen))
3967 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3972 /* Helper routine to branch off an association to a new socket. */
3973 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3974 struct socket **sockp)
3976 struct sock *sk = asoc->base.sk;
3977 struct socket *sock;
3981 /* An association cannot be branched off from an already peeled-off
3982 * socket, nor is this supported for tcp style sockets.
3984 if (!sctp_style(sk, UDP))
3987 /* Create a new socket. */
3988 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3992 sctp_copy_sock(sock->sk, sk, asoc);
3994 /* Make peeled-off sockets more like 1-1 accepted sockets.
3995 * Set the daddr and initialize id to something more random
3997 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3998 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4000 /* Populate the fields of the newsk from the oldsk and migrate the
4001 * asoc to the newsk.
4003 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4010 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4012 sctp_peeloff_arg_t peeloff;
4013 struct socket *newsock;
4015 struct sctp_association *asoc;
4017 if (len < sizeof(sctp_peeloff_arg_t))
4019 len = sizeof(sctp_peeloff_arg_t);
4020 if (copy_from_user(&peeloff, optval, len))
4023 asoc = sctp_id2assoc(sk, peeloff.associd);
4029 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4031 retval = sctp_do_peeloff(asoc, &newsock);
4035 /* Map the socket to an unused fd that can be returned to the user. */
4036 retval = sock_map_fd(newsock, 0);
4038 sock_release(newsock);
4042 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4043 __func__, sk, asoc, newsock->sk, retval);
4045 /* Return the fd mapped to the new socket. */
4046 peeloff.sd = retval;
4047 if (put_user(len, optlen))
4049 if (copy_to_user(optval, &peeloff, len))
4056 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4058 * Applications can enable or disable heartbeats for any peer address of
4059 * an association, modify an address's heartbeat interval, force a
4060 * heartbeat to be sent immediately, and adjust the address's maximum
4061 * number of retransmissions sent before an address is considered
4062 * unreachable. The following structure is used to access and modify an
4063 * address's parameters:
4065 * struct sctp_paddrparams {
4066 * sctp_assoc_t spp_assoc_id;
4067 * struct sockaddr_storage spp_address;
4068 * uint32_t spp_hbinterval;
4069 * uint16_t spp_pathmaxrxt;
4070 * uint32_t spp_pathmtu;
4071 * uint32_t spp_sackdelay;
4072 * uint32_t spp_flags;
4075 * spp_assoc_id - (one-to-many style socket) This is filled in the
4076 * application, and identifies the association for
4078 * spp_address - This specifies which address is of interest.
4079 * spp_hbinterval - This contains the value of the heartbeat interval,
4080 * in milliseconds. If a value of zero
4081 * is present in this field then no changes are to
4082 * be made to this parameter.
4083 * spp_pathmaxrxt - This contains the maximum number of
4084 * retransmissions before this address shall be
4085 * considered unreachable. If a value of zero
4086 * is present in this field then no changes are to
4087 * be made to this parameter.
4088 * spp_pathmtu - When Path MTU discovery is disabled the value
4089 * specified here will be the "fixed" path mtu.
4090 * Note that if the spp_address field is empty
4091 * then all associations on this address will
4092 * have this fixed path mtu set upon them.
4094 * spp_sackdelay - When delayed sack is enabled, this value specifies
4095 * the number of milliseconds that sacks will be delayed
4096 * for. This value will apply to all addresses of an
4097 * association if the spp_address field is empty. Note
4098 * also, that if delayed sack is enabled and this
4099 * value is set to 0, no change is made to the last
4100 * recorded delayed sack timer value.
4102 * spp_flags - These flags are used to control various features
4103 * on an association. The flag field may contain
4104 * zero or more of the following options.
4106 * SPP_HB_ENABLE - Enable heartbeats on the
4107 * specified address. Note that if the address
4108 * field is empty all addresses for the association
4109 * have heartbeats enabled upon them.
4111 * SPP_HB_DISABLE - Disable heartbeats on the
4112 * speicifed address. Note that if the address
4113 * field is empty all addresses for the association
4114 * will have their heartbeats disabled. Note also
4115 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4116 * mutually exclusive, only one of these two should
4117 * be specified. Enabling both fields will have
4118 * undetermined results.
4120 * SPP_HB_DEMAND - Request a user initiated heartbeat
4121 * to be made immediately.
4123 * SPP_PMTUD_ENABLE - This field will enable PMTU
4124 * discovery upon the specified address. Note that
4125 * if the address feild is empty then all addresses
4126 * on the association are effected.
4128 * SPP_PMTUD_DISABLE - This field will disable PMTU
4129 * discovery upon the specified address. Note that
4130 * if the address feild is empty then all addresses
4131 * on the association are effected. Not also that
4132 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4133 * exclusive. Enabling both will have undetermined
4136 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4137 * on delayed sack. The time specified in spp_sackdelay
4138 * is used to specify the sack delay for this address. Note
4139 * that if spp_address is empty then all addresses will
4140 * enable delayed sack and take on the sack delay
4141 * value specified in spp_sackdelay.
4142 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4143 * off delayed sack. If the spp_address field is blank then
4144 * delayed sack is disabled for the entire association. Note
4145 * also that this field is mutually exclusive to
4146 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4149 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4150 char __user *optval, int __user *optlen)
4152 struct sctp_paddrparams params;
4153 struct sctp_transport *trans = NULL;
4154 struct sctp_association *asoc = NULL;
4155 struct sctp_sock *sp = sctp_sk(sk);
4157 if (len < sizeof(struct sctp_paddrparams))
4159 len = sizeof(struct sctp_paddrparams);
4160 if (copy_from_user(¶ms, optval, len))
4163 /* If an address other than INADDR_ANY is specified, and
4164 * no transport is found, then the request is invalid.
4166 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4167 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4168 params.spp_assoc_id);
4170 SCTP_DEBUG_PRINTK("Failed no transport\n");
4175 /* Get association, if assoc_id != 0 and the socket is a one
4176 * to many style socket, and an association was not found, then
4177 * the id was invalid.
4179 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4180 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4181 SCTP_DEBUG_PRINTK("Failed no association\n");
4186 /* Fetch transport values. */
4187 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4188 params.spp_pathmtu = trans->pathmtu;
4189 params.spp_pathmaxrxt = trans->pathmaxrxt;
4190 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4192 /*draft-11 doesn't say what to return in spp_flags*/
4193 params.spp_flags = trans->param_flags;
4195 /* Fetch association values. */
4196 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4197 params.spp_pathmtu = asoc->pathmtu;
4198 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4199 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4201 /*draft-11 doesn't say what to return in spp_flags*/
4202 params.spp_flags = asoc->param_flags;
4204 /* Fetch socket values. */
4205 params.spp_hbinterval = sp->hbinterval;
4206 params.spp_pathmtu = sp->pathmtu;
4207 params.spp_sackdelay = sp->sackdelay;
4208 params.spp_pathmaxrxt = sp->pathmaxrxt;
4210 /*draft-11 doesn't say what to return in spp_flags*/
4211 params.spp_flags = sp->param_flags;
4214 if (copy_to_user(optval, ¶ms, len))
4217 if (put_user(len, optlen))
4224 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4226 * This option will effect the way delayed acks are performed. This
4227 * option allows you to get or set the delayed ack time, in
4228 * milliseconds. It also allows changing the delayed ack frequency.
4229 * Changing the frequency to 1 disables the delayed sack algorithm. If
4230 * the assoc_id is 0, then this sets or gets the endpoints default
4231 * values. If the assoc_id field is non-zero, then the set or get
4232 * effects the specified association for the one to many model (the
4233 * assoc_id field is ignored by the one to one model). Note that if
4234 * sack_delay or sack_freq are 0 when setting this option, then the
4235 * current values will remain unchanged.
4237 * struct sctp_sack_info {
4238 * sctp_assoc_t sack_assoc_id;
4239 * uint32_t sack_delay;
4240 * uint32_t sack_freq;
4243 * sack_assoc_id - This parameter, indicates which association the user
4244 * is performing an action upon. Note that if this field's value is
4245 * zero then the endpoints default value is changed (effecting future
4246 * associations only).
4248 * sack_delay - This parameter contains the number of milliseconds that
4249 * the user is requesting the delayed ACK timer be set to. Note that
4250 * this value is defined in the standard to be between 200 and 500
4253 * sack_freq - This parameter contains the number of packets that must
4254 * be received before a sack is sent without waiting for the delay
4255 * timer to expire. The default value for this is 2, setting this
4256 * value to 1 will disable the delayed sack algorithm.
4258 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4259 char __user *optval,
4262 struct sctp_sack_info params;
4263 struct sctp_association *asoc = NULL;
4264 struct sctp_sock *sp = sctp_sk(sk);
4266 if (len >= sizeof(struct sctp_sack_info)) {
4267 len = sizeof(struct sctp_sack_info);
4269 if (copy_from_user(¶ms, optval, len))
4271 } else if (len == sizeof(struct sctp_assoc_value)) {
4272 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4273 "in delayed_ack socket option deprecated\n");
4274 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4275 if (copy_from_user(¶ms, optval, len))
4280 /* Get association, if sack_assoc_id != 0 and the socket is a one
4281 * to many style socket, and an association was not found, then
4282 * the id was invalid.
4284 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4285 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4289 /* Fetch association values. */
4290 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4291 params.sack_delay = jiffies_to_msecs(
4293 params.sack_freq = asoc->sackfreq;
4296 params.sack_delay = 0;
4297 params.sack_freq = 1;
4300 /* Fetch socket values. */
4301 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4302 params.sack_delay = sp->sackdelay;
4303 params.sack_freq = sp->sackfreq;
4305 params.sack_delay = 0;
4306 params.sack_freq = 1;
4310 if (copy_to_user(optval, ¶ms, len))
4313 if (put_user(len, optlen))
4319 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4321 * Applications can specify protocol parameters for the default association
4322 * initialization. The option name argument to setsockopt() and getsockopt()
4325 * Setting initialization parameters is effective only on an unconnected
4326 * socket (for UDP-style sockets only future associations are effected
4327 * by the change). With TCP-style sockets, this option is inherited by
4328 * sockets derived from a listener socket.
4330 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4332 if (len < sizeof(struct sctp_initmsg))
4334 len = sizeof(struct sctp_initmsg);
4335 if (put_user(len, optlen))
4337 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4342 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4343 char __user *optval,
4347 struct sctp_association *asoc;
4348 struct list_head *pos;
4351 if (len < sizeof(sctp_assoc_t))
4354 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4357 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4358 "socket option deprecated\n");
4359 /* For UDP-style sockets, id specifies the association to query. */
4360 asoc = sctp_id2assoc(sk, id);
4364 list_for_each(pos, &asoc->peer.transport_addr_list) {
4372 * Old API for getting list of peer addresses. Does not work for 32-bit
4373 * programs running on a 64-bit kernel
4375 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4376 char __user *optval,
4379 struct sctp_association *asoc;
4381 struct sctp_getaddrs_old getaddrs;
4382 struct sctp_transport *from;
4384 union sctp_addr temp;
4385 struct sctp_sock *sp = sctp_sk(sk);
4388 if (len < sizeof(struct sctp_getaddrs_old))
4391 len = sizeof(struct sctp_getaddrs_old);
4393 if (copy_from_user(&getaddrs, optval, len))
4396 if (getaddrs.addr_num <= 0) return -EINVAL;
4398 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4399 "socket option deprecated\n");
4401 /* For UDP-style sockets, id specifies the association to query. */
4402 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4406 to = (void __user *)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 (copy_to_user(to, &temp, addrlen))
4416 if (cnt >= getaddrs.addr_num) break;
4418 getaddrs.addr_num = cnt;
4419 if (put_user(len, optlen))
4421 if (copy_to_user(optval, &getaddrs, len))
4427 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4428 char __user *optval, int __user *optlen)
4430 struct sctp_association *asoc;
4432 struct sctp_getaddrs getaddrs;
4433 struct sctp_transport *from;
4435 union sctp_addr temp;
4436 struct sctp_sock *sp = sctp_sk(sk);
4441 if (len < sizeof(struct sctp_getaddrs))
4444 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4447 /* For UDP-style sockets, id specifies the association to query. */
4448 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4452 to = optval + offsetof(struct sctp_getaddrs,addrs);
4453 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4455 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4457 memcpy(&temp, &from->ipaddr, sizeof(temp));
4458 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4459 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4460 if (space_left < addrlen)
4462 if (copy_to_user(to, &temp, addrlen))
4466 space_left -= addrlen;
4469 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4471 bytes_copied = ((char __user *)to) - optval;
4472 if (put_user(bytes_copied, optlen))
4478 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4479 char __user *optval,
4483 struct sctp_bind_addr *bp;
4484 struct sctp_association *asoc;
4485 struct sctp_sockaddr_entry *addr;
4488 if (len < sizeof(sctp_assoc_t))
4491 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4494 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4495 "socket option deprecated\n");
4498 * For UDP-style sockets, id specifies the association to query.
4499 * If the id field is set to the value '0' then the locally bound
4500 * addresses are returned without regard to any particular
4504 bp = &sctp_sk(sk)->ep->base.bind_addr;
4506 asoc = sctp_id2assoc(sk, id);
4509 bp = &asoc->base.bind_addr;
4512 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4513 * addresses from the global local address list.
4515 if (sctp_list_single_entry(&bp->address_list)) {
4516 addr = list_entry(bp->address_list.next,
4517 struct sctp_sockaddr_entry, list);
4518 if (sctp_is_any(sk, &addr->a)) {
4520 list_for_each_entry_rcu(addr,
4521 &sctp_local_addr_list, list) {
4525 if ((PF_INET == sk->sk_family) &&
4526 (AF_INET6 == addr->a.sa.sa_family))
4529 if ((PF_INET6 == sk->sk_family) &&
4530 inet_v6_ipv6only(sk) &&
4531 (AF_INET == addr->a.sa.sa_family))
4543 /* Protection on the bound address list is not needed,
4544 * since in the socket option context we hold the socket lock,
4545 * so there is no way that the bound address list can change.
4547 list_for_each_entry(addr, &bp->address_list, list) {
4554 /* Helper function that copies local addresses to user and returns the number
4555 * of addresses copied.
4557 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4558 int max_addrs, void *to,
4561 struct sctp_sockaddr_entry *addr;
4562 union sctp_addr temp;
4567 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4571 if ((PF_INET == sk->sk_family) &&
4572 (AF_INET6 == addr->a.sa.sa_family))
4574 if ((PF_INET6 == sk->sk_family) &&
4575 inet_v6_ipv6only(sk) &&
4576 (AF_INET == addr->a.sa.sa_family))
4578 memcpy(&temp, &addr->a, sizeof(temp));
4579 if (!temp.v4.sin_port)
4580 temp.v4.sin_port = htons(port);
4582 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4584 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4585 memcpy(to, &temp, addrlen);
4588 *bytes_copied += addrlen;
4590 if (cnt >= max_addrs) break;
4597 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4598 size_t space_left, int *bytes_copied)
4600 struct sctp_sockaddr_entry *addr;
4601 union sctp_addr temp;
4606 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4610 if ((PF_INET == sk->sk_family) &&
4611 (AF_INET6 == addr->a.sa.sa_family))
4613 if ((PF_INET6 == sk->sk_family) &&
4614 inet_v6_ipv6only(sk) &&
4615 (AF_INET == addr->a.sa.sa_family))
4617 memcpy(&temp, &addr->a, sizeof(temp));
4618 if (!temp.v4.sin_port)
4619 temp.v4.sin_port = htons(port);
4621 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4623 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4624 if (space_left < addrlen) {
4628 memcpy(to, &temp, addrlen);
4632 space_left -= addrlen;
4633 *bytes_copied += addrlen;
4640 /* Old API for getting list of local addresses. Does not work for 32-bit
4641 * programs running on a 64-bit kernel
4643 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4644 char __user *optval, int __user *optlen)
4646 struct sctp_bind_addr *bp;
4647 struct sctp_association *asoc;
4649 struct sctp_getaddrs_old getaddrs;
4650 struct sctp_sockaddr_entry *addr;
4652 union sctp_addr temp;
4653 struct sctp_sock *sp = sctp_sk(sk);
4658 int bytes_copied = 0;
4660 if (len < sizeof(struct sctp_getaddrs_old))
4663 len = sizeof(struct sctp_getaddrs_old);
4664 if (copy_from_user(&getaddrs, optval, len))
4667 if (getaddrs.addr_num <= 0 ||
4668 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4671 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4672 "socket option deprecated\n");
4675 * For UDP-style sockets, id specifies the association to query.
4676 * If the id field is set to the value '0' then the locally bound
4677 * addresses are returned without regard to any particular
4680 if (0 == getaddrs.assoc_id) {
4681 bp = &sctp_sk(sk)->ep->base.bind_addr;
4683 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4686 bp = &asoc->base.bind_addr;
4689 to = getaddrs.addrs;
4691 /* Allocate space for a local instance of packed array to hold all
4692 * the data. We store addresses here first and then put write them
4693 * to the user in one shot.
4695 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4700 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4701 * addresses from the global local address list.
4703 if (sctp_list_single_entry(&bp->address_list)) {
4704 addr = list_entry(bp->address_list.next,
4705 struct sctp_sockaddr_entry, list);
4706 if (sctp_is_any(sk, &addr->a)) {
4707 cnt = sctp_copy_laddrs_old(sk, bp->port,
4709 addrs, &bytes_copied);
4715 /* Protection on the bound address list is not needed since
4716 * in the socket option context we hold a socket lock and
4717 * thus the bound address list can't change.
4719 list_for_each_entry(addr, &bp->address_list, list) {
4720 memcpy(&temp, &addr->a, sizeof(temp));
4721 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4722 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4723 memcpy(buf, &temp, addrlen);
4725 bytes_copied += addrlen;
4727 if (cnt >= getaddrs.addr_num) break;
4731 /* copy the entire address list into the user provided space */
4732 if (copy_to_user(to, addrs, bytes_copied)) {
4737 /* copy the leading structure back to user */
4738 getaddrs.addr_num = cnt;
4739 if (copy_to_user(optval, &getaddrs, len))
4747 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4748 char __user *optval, int __user *optlen)
4750 struct sctp_bind_addr *bp;
4751 struct sctp_association *asoc;
4753 struct sctp_getaddrs getaddrs;
4754 struct sctp_sockaddr_entry *addr;
4756 union sctp_addr temp;
4757 struct sctp_sock *sp = sctp_sk(sk);
4761 int bytes_copied = 0;
4765 if (len < sizeof(struct sctp_getaddrs))
4768 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4772 * For UDP-style sockets, id specifies the association to query.
4773 * If the id field is set to the value '0' then the locally bound
4774 * addresses are returned without regard to any particular
4777 if (0 == getaddrs.assoc_id) {
4778 bp = &sctp_sk(sk)->ep->base.bind_addr;
4780 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4783 bp = &asoc->base.bind_addr;
4786 to = optval + offsetof(struct sctp_getaddrs,addrs);
4787 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4789 addrs = kmalloc(space_left, GFP_KERNEL);
4793 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4794 * addresses from the global local address list.
4796 if (sctp_list_single_entry(&bp->address_list)) {
4797 addr = list_entry(bp->address_list.next,
4798 struct sctp_sockaddr_entry, list);
4799 if (sctp_is_any(sk, &addr->a)) {
4800 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4801 space_left, &bytes_copied);
4811 /* Protection on the bound address list is not needed since
4812 * in the socket option context we hold a socket lock and
4813 * thus the bound address list can't change.
4815 list_for_each_entry(addr, &bp->address_list, list) {
4816 memcpy(&temp, &addr->a, sizeof(temp));
4817 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4818 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4819 if (space_left < addrlen) {
4820 err = -ENOMEM; /*fixme: right error?*/
4823 memcpy(buf, &temp, addrlen);
4825 bytes_copied += addrlen;
4827 space_left -= addrlen;
4831 if (copy_to_user(to, addrs, bytes_copied)) {
4835 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4839 if (put_user(bytes_copied, optlen))
4846 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4848 * Requests that the local SCTP stack use the enclosed peer address as
4849 * the association primary. The enclosed address must be one of the
4850 * association peer's addresses.
4852 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4853 char __user *optval, int __user *optlen)
4855 struct sctp_prim prim;
4856 struct sctp_association *asoc;
4857 struct sctp_sock *sp = sctp_sk(sk);
4859 if (len < sizeof(struct sctp_prim))
4862 len = sizeof(struct sctp_prim);
4864 if (copy_from_user(&prim, optval, len))
4867 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4871 if (!asoc->peer.primary_path)
4874 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4875 asoc->peer.primary_path->af_specific->sockaddr_len);
4877 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4878 (union sctp_addr *)&prim.ssp_addr);
4880 if (put_user(len, optlen))
4882 if (copy_to_user(optval, &prim, len))
4889 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4891 * Requests that the local endpoint set the specified Adaptation Layer
4892 * Indication parameter for all future INIT and INIT-ACK exchanges.
4894 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4895 char __user *optval, int __user *optlen)
4897 struct sctp_setadaptation adaptation;
4899 if (len < sizeof(struct sctp_setadaptation))
4902 len = sizeof(struct sctp_setadaptation);
4904 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4906 if (put_user(len, optlen))
4908 if (copy_to_user(optval, &adaptation, len))
4916 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4918 * Applications that wish to use the sendto() system call may wish to
4919 * specify a default set of parameters that would normally be supplied
4920 * through the inclusion of ancillary data. This socket option allows
4921 * such an application to set the default sctp_sndrcvinfo structure.
4924 * The application that wishes to use this socket option simply passes
4925 * in to this call the sctp_sndrcvinfo structure defined in Section
4926 * 5.2.2) The input parameters accepted by this call include
4927 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4928 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4929 * to this call if the caller is using the UDP model.
4931 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4933 static int sctp_getsockopt_default_send_param(struct sock *sk,
4934 int len, char __user *optval,
4937 struct sctp_sndrcvinfo info;
4938 struct sctp_association *asoc;
4939 struct sctp_sock *sp = sctp_sk(sk);
4941 if (len < sizeof(struct sctp_sndrcvinfo))
4944 len = sizeof(struct sctp_sndrcvinfo);
4946 if (copy_from_user(&info, optval, len))
4949 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4950 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4954 info.sinfo_stream = asoc->default_stream;
4955 info.sinfo_flags = asoc->default_flags;
4956 info.sinfo_ppid = asoc->default_ppid;
4957 info.sinfo_context = asoc->default_context;
4958 info.sinfo_timetolive = asoc->default_timetolive;
4960 info.sinfo_stream = sp->default_stream;
4961 info.sinfo_flags = sp->default_flags;
4962 info.sinfo_ppid = sp->default_ppid;
4963 info.sinfo_context = sp->default_context;
4964 info.sinfo_timetolive = sp->default_timetolive;
4967 if (put_user(len, optlen))
4969 if (copy_to_user(optval, &info, len))
4977 * 7.1.5 SCTP_NODELAY
4979 * Turn on/off any Nagle-like algorithm. This means that packets are
4980 * generally sent as soon as possible and no unnecessary delays are
4981 * introduced, at the cost of more packets in the network. Expects an
4982 * integer boolean flag.
4985 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4986 char __user *optval, int __user *optlen)
4990 if (len < sizeof(int))
4994 val = (sctp_sk(sk)->nodelay == 1);
4995 if (put_user(len, optlen))
4997 if (copy_to_user(optval, &val, len))
5004 * 7.1.1 SCTP_RTOINFO
5006 * The protocol parameters used to initialize and bound retransmission
5007 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5008 * and modify these parameters.
5009 * All parameters are time values, in milliseconds. A value of 0, when
5010 * modifying the parameters, indicates that the current value should not
5014 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5015 char __user *optval,
5016 int __user *optlen) {
5017 struct sctp_rtoinfo rtoinfo;
5018 struct sctp_association *asoc;
5020 if (len < sizeof (struct sctp_rtoinfo))
5023 len = sizeof(struct sctp_rtoinfo);
5025 if (copy_from_user(&rtoinfo, optval, len))
5028 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5030 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5033 /* Values corresponding to the specific association. */
5035 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5036 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5037 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5039 /* Values corresponding to the endpoint. */
5040 struct sctp_sock *sp = sctp_sk(sk);
5042 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5043 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5044 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5047 if (put_user(len, optlen))
5050 if (copy_to_user(optval, &rtoinfo, len))
5058 * 7.1.2 SCTP_ASSOCINFO
5060 * This option is used to tune the maximum retransmission attempts
5061 * of the association.
5062 * Returns an error if the new association retransmission value is
5063 * greater than the sum of the retransmission value of the peer.
5064 * See [SCTP] for more information.
5067 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5068 char __user *optval,
5072 struct sctp_assocparams assocparams;
5073 struct sctp_association *asoc;
5074 struct list_head *pos;
5077 if (len < sizeof (struct sctp_assocparams))
5080 len = sizeof(struct sctp_assocparams);
5082 if (copy_from_user(&assocparams, optval, len))
5085 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5087 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5090 /* Values correspoinding to the specific association */
5092 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5093 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5094 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5095 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5097 (asoc->cookie_life.tv_usec
5100 list_for_each(pos, &asoc->peer.transport_addr_list) {
5104 assocparams.sasoc_number_peer_destinations = cnt;
5106 /* Values corresponding to the endpoint */
5107 struct sctp_sock *sp = sctp_sk(sk);
5109 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5110 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5111 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5112 assocparams.sasoc_cookie_life =
5113 sp->assocparams.sasoc_cookie_life;
5114 assocparams.sasoc_number_peer_destinations =
5116 sasoc_number_peer_destinations;
5119 if (put_user(len, optlen))
5122 if (copy_to_user(optval, &assocparams, len))
5129 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5131 * This socket option is a boolean flag which turns on or off mapped V4
5132 * addresses. If this option is turned on and the socket is type
5133 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5134 * If this option is turned off, then no mapping will be done of V4
5135 * addresses and a user will receive both PF_INET6 and PF_INET type
5136 * addresses on the socket.
5138 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5139 char __user *optval, int __user *optlen)
5142 struct sctp_sock *sp = sctp_sk(sk);
5144 if (len < sizeof(int))
5149 if (put_user(len, optlen))
5151 if (copy_to_user(optval, &val, len))
5158 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5159 * (chapter and verse is quoted at sctp_setsockopt_context())
5161 static int sctp_getsockopt_context(struct sock *sk, int len,
5162 char __user *optval, int __user *optlen)
5164 struct sctp_assoc_value params;
5165 struct sctp_sock *sp;
5166 struct sctp_association *asoc;
5168 if (len < sizeof(struct sctp_assoc_value))
5171 len = sizeof(struct sctp_assoc_value);
5173 if (copy_from_user(¶ms, optval, len))
5178 if (params.assoc_id != 0) {
5179 asoc = sctp_id2assoc(sk, params.assoc_id);
5182 params.assoc_value = asoc->default_rcv_context;
5184 params.assoc_value = sp->default_rcv_context;
5187 if (put_user(len, optlen))
5189 if (copy_to_user(optval, ¶ms, len))
5196 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5197 * This option will get or set the maximum size to put in any outgoing
5198 * SCTP DATA chunk. If a message is larger than this size it will be
5199 * fragmented by SCTP into the specified size. Note that the underlying
5200 * SCTP implementation may fragment into smaller sized chunks when the
5201 * PMTU of the underlying association is smaller than the value set by
5202 * the user. The default value for this option is '0' which indicates
5203 * the user is NOT limiting fragmentation and only the PMTU will effect
5204 * SCTP's choice of DATA chunk size. Note also that values set larger
5205 * than the maximum size of an IP datagram will effectively let SCTP
5206 * control fragmentation (i.e. the same as setting this option to 0).
5208 * The following structure is used to access and modify this parameter:
5210 * struct sctp_assoc_value {
5211 * sctp_assoc_t assoc_id;
5212 * uint32_t assoc_value;
5215 * assoc_id: This parameter is ignored for one-to-one style sockets.
5216 * For one-to-many style sockets this parameter indicates which
5217 * association the user is performing an action upon. Note that if
5218 * this field's value is zero then the endpoints default value is
5219 * changed (effecting future associations only).
5220 * assoc_value: This parameter specifies the maximum size in bytes.
5222 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5223 char __user *optval, int __user *optlen)
5225 struct sctp_assoc_value params;
5226 struct sctp_association *asoc;
5228 if (len == sizeof(int)) {
5230 "SCTP: Use of int in maxseg socket option deprecated\n");
5232 "SCTP: Use struct sctp_assoc_value instead\n");
5233 params.assoc_id = 0;
5234 } else if (len >= sizeof(struct sctp_assoc_value)) {
5235 len = sizeof(struct sctp_assoc_value);
5236 if (copy_from_user(¶ms, optval, sizeof(params)))
5241 asoc = sctp_id2assoc(sk, params.assoc_id);
5242 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5246 params.assoc_value = asoc->frag_point;
5248 params.assoc_value = sctp_sk(sk)->user_frag;
5250 if (put_user(len, optlen))
5252 if (len == sizeof(int)) {
5253 if (copy_to_user(optval, ¶ms.assoc_value, len))
5256 if (copy_to_user(optval, ¶ms, len))
5264 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5265 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5267 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5268 char __user *optval, int __user *optlen)
5272 if (len < sizeof(int))
5277 val = sctp_sk(sk)->frag_interleave;
5278 if (put_user(len, optlen))
5280 if (copy_to_user(optval, &val, len))
5287 * 7.1.25. Set or Get the sctp partial delivery point
5288 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5290 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5291 char __user *optval,
5296 if (len < sizeof(u32))
5301 val = sctp_sk(sk)->pd_point;
5302 if (put_user(len, optlen))
5304 if (copy_to_user(optval, &val, len))
5311 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5312 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5314 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5315 char __user *optval,
5318 struct sctp_assoc_value params;
5319 struct sctp_sock *sp;
5320 struct sctp_association *asoc;
5322 if (len == sizeof(int)) {
5324 "SCTP: Use of int in max_burst socket option deprecated\n");
5326 "SCTP: Use struct sctp_assoc_value instead\n");
5327 params.assoc_id = 0;
5328 } else if (len >= sizeof(struct sctp_assoc_value)) {
5329 len = sizeof(struct sctp_assoc_value);
5330 if (copy_from_user(¶ms, optval, len))
5337 if (params.assoc_id != 0) {
5338 asoc = sctp_id2assoc(sk, params.assoc_id);
5341 params.assoc_value = asoc->max_burst;
5343 params.assoc_value = sp->max_burst;
5345 if (len == sizeof(int)) {
5346 if (copy_to_user(optval, ¶ms.assoc_value, len))
5349 if (copy_to_user(optval, ¶ms, len))
5357 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5358 char __user *optval, int __user *optlen)
5360 struct sctp_hmacalgo __user *p = (void __user *)optval;
5361 struct sctp_hmac_algo_param *hmacs;
5365 if (!sctp_auth_enable)
5368 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5369 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5371 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5374 len = sizeof(struct sctp_hmacalgo) + data_len;
5375 num_idents = data_len / sizeof(u16);
5377 if (put_user(len, optlen))
5379 if (put_user(num_idents, &p->shmac_num_idents))
5381 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5386 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5387 char __user *optval, int __user *optlen)
5389 struct sctp_authkeyid val;
5390 struct sctp_association *asoc;
5392 if (!sctp_auth_enable)
5395 if (len < sizeof(struct sctp_authkeyid))
5397 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5400 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5401 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5405 val.scact_keynumber = asoc->active_key_id;
5407 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5409 len = sizeof(struct sctp_authkeyid);
5410 if (put_user(len, optlen))
5412 if (copy_to_user(optval, &val, len))
5418 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5419 char __user *optval, int __user *optlen)
5421 struct sctp_authchunks __user *p = (void __user *)optval;
5422 struct sctp_authchunks val;
5423 struct sctp_association *asoc;
5424 struct sctp_chunks_param *ch;
5428 if (!sctp_auth_enable)
5431 if (len < sizeof(struct sctp_authchunks))
5434 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5437 to = p->gauth_chunks;
5438 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5442 ch = asoc->peer.peer_chunks;
5446 /* See if the user provided enough room for all the data */
5447 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5448 if (len < 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)) return -EFAULT;
5456 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5461 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5462 char __user *optval, int __user *optlen)
5464 struct sctp_authchunks __user *p = (void __user *)optval;
5465 struct sctp_authchunks val;
5466 struct sctp_association *asoc;
5467 struct sctp_chunks_param *ch;
5471 if (!sctp_auth_enable)
5474 if (len < sizeof(struct sctp_authchunks))
5477 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5480 to = p->gauth_chunks;
5481 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5482 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5486 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5488 ch = sctp_sk(sk)->ep->auth_chunk_list;
5493 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5494 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5497 if (copy_to_user(to, ch->chunks, num_chunks))
5500 len = sizeof(struct sctp_authchunks) + num_chunks;
5501 if (put_user(len, optlen))
5503 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5510 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5511 * This option gets the current number of associations that are attached
5512 * to a one-to-many style socket. The option value is an uint32_t.
5514 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5515 char __user *optval, int __user *optlen)
5517 struct sctp_sock *sp = sctp_sk(sk);
5518 struct sctp_association *asoc;
5521 if (sctp_style(sk, TCP))
5524 if (len < sizeof(u32))
5529 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5533 if (put_user(len, optlen))
5535 if (copy_to_user(optval, &val, len))
5541 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5542 char __user *optval, int __user *optlen)
5547 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5550 /* I can hardly begin to describe how wrong this is. This is
5551 * so broken as to be worse than useless. The API draft
5552 * REALLY is NOT helpful here... I am not convinced that the
5553 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5554 * are at all well-founded.
5556 if (level != SOL_SCTP) {
5557 struct sctp_af *af = sctp_sk(sk)->pf->af;
5559 retval = af->getsockopt(sk, level, optname, optval, optlen);
5563 if (get_user(len, optlen))
5570 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5572 case SCTP_DISABLE_FRAGMENTS:
5573 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5577 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5579 case SCTP_AUTOCLOSE:
5580 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5582 case SCTP_SOCKOPT_PEELOFF:
5583 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5585 case SCTP_PEER_ADDR_PARAMS:
5586 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5589 case SCTP_DELAYED_ACK:
5590 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5594 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5596 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5597 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5600 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5601 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5604 case SCTP_GET_PEER_ADDRS_OLD:
5605 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5608 case SCTP_GET_LOCAL_ADDRS_OLD:
5609 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5612 case SCTP_GET_PEER_ADDRS:
5613 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5616 case SCTP_GET_LOCAL_ADDRS:
5617 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5620 case SCTP_SOCKOPT_CONNECTX3:
5621 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5623 case SCTP_DEFAULT_SEND_PARAM:
5624 retval = sctp_getsockopt_default_send_param(sk, len,
5627 case SCTP_PRIMARY_ADDR:
5628 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5631 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5634 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5636 case SCTP_ASSOCINFO:
5637 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5639 case SCTP_I_WANT_MAPPED_V4_ADDR:
5640 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5643 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5645 case SCTP_GET_PEER_ADDR_INFO:
5646 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5649 case SCTP_ADAPTATION_LAYER:
5650 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5654 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5656 case SCTP_FRAGMENT_INTERLEAVE:
5657 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5660 case SCTP_PARTIAL_DELIVERY_POINT:
5661 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5664 case SCTP_MAX_BURST:
5665 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5668 case SCTP_AUTH_CHUNK:
5669 case SCTP_AUTH_DELETE_KEY:
5670 retval = -EOPNOTSUPP;
5672 case SCTP_HMAC_IDENT:
5673 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5675 case SCTP_AUTH_ACTIVE_KEY:
5676 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5678 case SCTP_PEER_AUTH_CHUNKS:
5679 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5682 case SCTP_LOCAL_AUTH_CHUNKS:
5683 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5686 case SCTP_GET_ASSOC_NUMBER:
5687 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5690 retval = -ENOPROTOOPT;
5694 sctp_release_sock(sk);
5698 static void sctp_hash(struct sock *sk)
5703 static void sctp_unhash(struct sock *sk)
5708 /* Check if port is acceptable. Possibly find first available port.
5710 * The port hash table (contained in the 'global' SCTP protocol storage
5711 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5712 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5713 * list (the list number is the port number hashed out, so as you
5714 * would expect from a hash function, all the ports in a given list have
5715 * such a number that hashes out to the same list number; you were
5716 * expecting that, right?); so each list has a set of ports, with a
5717 * link to the socket (struct sock) that uses it, the port number and
5718 * a fastreuse flag (FIXME: NPI ipg).
5720 static struct sctp_bind_bucket *sctp_bucket_create(
5721 struct sctp_bind_hashbucket *head, unsigned short snum);
5723 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5725 struct sctp_bind_hashbucket *head; /* hash list */
5726 struct sctp_bind_bucket *pp; /* hash list port iterator */
5727 struct hlist_node *node;
5728 unsigned short snum;
5731 snum = ntohs(addr->v4.sin_port);
5733 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5734 sctp_local_bh_disable();
5737 /* Search for an available port. */
5738 int low, high, remaining, index;
5741 inet_get_local_port_range(&low, &high);
5742 remaining = (high - low) + 1;
5743 rover = net_random() % remaining + low;
5747 if ((rover < low) || (rover > high))
5749 index = sctp_phashfn(rover);
5750 head = &sctp_port_hashtable[index];
5751 sctp_spin_lock(&head->lock);
5752 sctp_for_each_hentry(pp, node, &head->chain)
5753 if (pp->port == rover)
5757 sctp_spin_unlock(&head->lock);
5758 } while (--remaining > 0);
5760 /* Exhausted local port range during search? */
5765 /* OK, here is the one we will use. HEAD (the port
5766 * hash table list entry) is non-NULL and we hold it's
5771 /* We are given an specific port number; we verify
5772 * that it is not being used. If it is used, we will
5773 * exahust the search in the hash list corresponding
5774 * to the port number (snum) - we detect that with the
5775 * port iterator, pp being NULL.
5777 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5778 sctp_spin_lock(&head->lock);
5779 sctp_for_each_hentry(pp, node, &head->chain) {
5780 if (pp->port == snum)
5787 if (!hlist_empty(&pp->owner)) {
5788 /* We had a port hash table hit - there is an
5789 * available port (pp != NULL) and it is being
5790 * used by other socket (pp->owner not empty); that other
5791 * socket is going to be sk2.
5793 int reuse = sk->sk_reuse;
5795 struct hlist_node *node;
5797 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5798 if (pp->fastreuse && sk->sk_reuse &&
5799 sk->sk_state != SCTP_SS_LISTENING)
5802 /* Run through the list of sockets bound to the port
5803 * (pp->port) [via the pointers bind_next and
5804 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5805 * we get the endpoint they describe and run through
5806 * the endpoint's list of IP (v4 or v6) addresses,
5807 * comparing each of the addresses with the address of
5808 * the socket sk. If we find a match, then that means
5809 * that this port/socket (sk) combination are already
5812 sk_for_each_bound(sk2, node, &pp->owner) {
5813 struct sctp_endpoint *ep2;
5814 ep2 = sctp_sk(sk2)->ep;
5817 (reuse && sk2->sk_reuse &&
5818 sk2->sk_state != SCTP_SS_LISTENING))
5821 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5822 sctp_sk(sk2), sctp_sk(sk))) {
5827 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5830 /* If there was a hash table miss, create a new port. */
5832 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5835 /* In either case (hit or miss), make sure fastreuse is 1 only
5836 * if sk->sk_reuse is too (that is, if the caller requested
5837 * SO_REUSEADDR on this socket -sk-).
5839 if (hlist_empty(&pp->owner)) {
5840 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5844 } else if (pp->fastreuse &&
5845 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5848 /* We are set, so fill up all the data in the hash table
5849 * entry, tie the socket list information with the rest of the
5850 * sockets FIXME: Blurry, NPI (ipg).
5853 if (!sctp_sk(sk)->bind_hash) {
5854 inet_sk(sk)->num = snum;
5855 sk_add_bind_node(sk, &pp->owner);
5856 sctp_sk(sk)->bind_hash = pp;
5861 sctp_spin_unlock(&head->lock);
5864 sctp_local_bh_enable();
5868 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5869 * port is requested.
5871 static int sctp_get_port(struct sock *sk, unsigned short snum)
5874 union sctp_addr addr;
5875 struct sctp_af *af = sctp_sk(sk)->pf->af;
5877 /* Set up a dummy address struct from the sk. */
5878 af->from_sk(&addr, sk);
5879 addr.v4.sin_port = htons(snum);
5881 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5882 ret = sctp_get_port_local(sk, &addr);
5884 return (ret ? 1 : 0);
5888 * Move a socket to LISTENING state.
5890 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5892 struct sctp_sock *sp = sctp_sk(sk);
5893 struct sctp_endpoint *ep = sp->ep;
5894 struct crypto_hash *tfm = NULL;
5896 /* Allocate HMAC for generating cookie. */
5897 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5898 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5900 if (net_ratelimit()) {
5902 "SCTP: failed to load transform for %s: %ld\n",
5903 sctp_hmac_alg, PTR_ERR(tfm));
5907 sctp_sk(sk)->hmac = tfm;
5911 * If a bind() or sctp_bindx() is not called prior to a listen()
5912 * call that allows new associations to be accepted, the system
5913 * picks an ephemeral port and will choose an address set equivalent
5914 * to binding with a wildcard address.
5916 * This is not currently spelled out in the SCTP sockets
5917 * extensions draft, but follows the practice as seen in TCP
5921 sk->sk_state = SCTP_SS_LISTENING;
5922 if (!ep->base.bind_addr.port) {
5923 if (sctp_autobind(sk))
5926 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5927 sk->sk_state = SCTP_SS_CLOSED;
5932 sk->sk_max_ack_backlog = backlog;
5933 sctp_hash_endpoint(ep);
5938 * 4.1.3 / 5.1.3 listen()
5940 * By default, new associations are not accepted for UDP style sockets.
5941 * An application uses listen() to mark a socket as being able to
5942 * accept new associations.
5944 * On TCP style sockets, applications use listen() to ready the SCTP
5945 * endpoint for accepting inbound associations.
5947 * On both types of endpoints a backlog of '0' disables listening.
5949 * Move a socket to LISTENING state.
5951 int sctp_inet_listen(struct socket *sock, int backlog)
5953 struct sock *sk = sock->sk;
5954 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5957 if (unlikely(backlog < 0))
5962 /* Peeled-off sockets are not allowed to listen(). */
5963 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5966 if (sock->state != SS_UNCONNECTED)
5969 /* If backlog is zero, disable listening. */
5971 if (sctp_sstate(sk, CLOSED))
5975 sctp_unhash_endpoint(ep);
5976 sk->sk_state = SCTP_SS_CLOSED;
5978 sctp_sk(sk)->bind_hash->fastreuse = 1;
5982 /* If we are already listening, just update the backlog */
5983 if (sctp_sstate(sk, LISTENING))
5984 sk->sk_max_ack_backlog = backlog;
5986 err = sctp_listen_start(sk, backlog);
5993 sctp_release_sock(sk);
5998 * This function is done by modeling the current datagram_poll() and the
5999 * tcp_poll(). Note that, based on these implementations, we don't
6000 * lock the socket in this function, even though it seems that,
6001 * ideally, locking or some other mechanisms can be used to ensure
6002 * the integrity of the counters (sndbuf and wmem_alloc) used
6003 * in this place. We assume that we don't need locks either until proven
6006 * Another thing to note is that we include the Async I/O support
6007 * here, again, by modeling the current TCP/UDP code. We don't have
6008 * a good way to test with it yet.
6010 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6012 struct sock *sk = sock->sk;
6013 struct sctp_sock *sp = sctp_sk(sk);
6016 poll_wait(file, sk->sk_sleep, wait);
6018 /* A TCP-style listening socket becomes readable when the accept queue
6021 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6022 return (!list_empty(&sp->ep->asocs)) ?
6023 (POLLIN | POLLRDNORM) : 0;
6027 /* Is there any exceptional events? */
6028 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6030 if (sk->sk_shutdown & RCV_SHUTDOWN)
6032 if (sk->sk_shutdown == SHUTDOWN_MASK)
6035 /* Is it readable? Reconsider this code with TCP-style support. */
6036 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6037 (sk->sk_shutdown & RCV_SHUTDOWN))
6038 mask |= POLLIN | POLLRDNORM;
6040 /* The association is either gone or not ready. */
6041 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6044 /* Is it writable? */
6045 if (sctp_writeable(sk)) {
6046 mask |= POLLOUT | POLLWRNORM;
6048 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6050 * Since the socket is not locked, the buffer
6051 * might be made available after the writeable check and
6052 * before the bit is set. This could cause a lost I/O
6053 * signal. tcp_poll() has a race breaker for this race
6054 * condition. Based on their implementation, we put
6055 * in the following code to cover it as well.
6057 if (sctp_writeable(sk))
6058 mask |= POLLOUT | POLLWRNORM;
6063 /********************************************************************
6064 * 2nd Level Abstractions
6065 ********************************************************************/
6067 static struct sctp_bind_bucket *sctp_bucket_create(
6068 struct sctp_bind_hashbucket *head, unsigned short snum)
6070 struct sctp_bind_bucket *pp;
6072 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6074 SCTP_DBG_OBJCNT_INC(bind_bucket);
6077 INIT_HLIST_HEAD(&pp->owner);
6078 hlist_add_head(&pp->node, &head->chain);
6083 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6084 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6086 if (pp && hlist_empty(&pp->owner)) {
6087 __hlist_del(&pp->node);
6088 kmem_cache_free(sctp_bucket_cachep, pp);
6089 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6093 /* Release this socket's reference to a local port. */
6094 static inline void __sctp_put_port(struct sock *sk)
6096 struct sctp_bind_hashbucket *head =
6097 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6098 struct sctp_bind_bucket *pp;
6100 sctp_spin_lock(&head->lock);
6101 pp = sctp_sk(sk)->bind_hash;
6102 __sk_del_bind_node(sk);
6103 sctp_sk(sk)->bind_hash = NULL;
6104 inet_sk(sk)->num = 0;
6105 sctp_bucket_destroy(pp);
6106 sctp_spin_unlock(&head->lock);
6109 void sctp_put_port(struct sock *sk)
6111 sctp_local_bh_disable();
6112 __sctp_put_port(sk);
6113 sctp_local_bh_enable();
6117 * The system picks an ephemeral port and choose an address set equivalent
6118 * to binding with a wildcard address.
6119 * One of those addresses will be the primary address for the association.
6120 * This automatically enables the multihoming capability of SCTP.
6122 static int sctp_autobind(struct sock *sk)
6124 union sctp_addr autoaddr;
6128 /* Initialize a local sockaddr structure to INADDR_ANY. */
6129 af = sctp_sk(sk)->pf->af;
6131 port = htons(inet_sk(sk)->num);
6132 af->inaddr_any(&autoaddr, port);
6134 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6137 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6140 * 4.2 The cmsghdr Structure *
6142 * When ancillary data is sent or received, any number of ancillary data
6143 * objects can be specified by the msg_control and msg_controllen members of
6144 * the msghdr structure, because each object is preceded by
6145 * a cmsghdr structure defining the object's length (the cmsg_len member).
6146 * Historically Berkeley-derived implementations have passed only one object
6147 * at a time, but this API allows multiple objects to be
6148 * passed in a single call to sendmsg() or recvmsg(). The following example
6149 * shows two ancillary data objects in a control buffer.
6151 * |<--------------------------- msg_controllen -------------------------->|
6154 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6156 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6159 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6161 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6164 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6165 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6167 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6169 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6176 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6177 sctp_cmsgs_t *cmsgs)
6179 struct cmsghdr *cmsg;
6180 struct msghdr *my_msg = (struct msghdr *)msg;
6182 for (cmsg = CMSG_FIRSTHDR(msg);
6184 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6185 if (!CMSG_OK(my_msg, cmsg))
6188 /* Should we parse this header or ignore? */
6189 if (cmsg->cmsg_level != IPPROTO_SCTP)
6192 /* Strictly check lengths following example in SCM code. */
6193 switch (cmsg->cmsg_type) {
6195 /* SCTP Socket API Extension
6196 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6198 * This cmsghdr structure provides information for
6199 * initializing new SCTP associations with sendmsg().
6200 * The SCTP_INITMSG socket option uses this same data
6201 * structure. This structure is not used for
6204 * cmsg_level cmsg_type cmsg_data[]
6205 * ------------ ------------ ----------------------
6206 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6208 if (cmsg->cmsg_len !=
6209 CMSG_LEN(sizeof(struct sctp_initmsg)))
6211 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6215 /* SCTP Socket API Extension
6216 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6218 * This cmsghdr structure specifies SCTP options for
6219 * sendmsg() and describes SCTP header information
6220 * about a received message through recvmsg().
6222 * cmsg_level cmsg_type cmsg_data[]
6223 * ------------ ------------ ----------------------
6224 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6226 if (cmsg->cmsg_len !=
6227 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6231 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6233 /* Minimally, validate the sinfo_flags. */
6234 if (cmsgs->info->sinfo_flags &
6235 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6236 SCTP_ABORT | SCTP_EOF))
6248 * Wait for a packet..
6249 * Note: This function is the same function as in core/datagram.c
6250 * with a few modifications to make lksctp work.
6252 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6257 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6259 /* Socket errors? */
6260 error = sock_error(sk);
6264 if (!skb_queue_empty(&sk->sk_receive_queue))
6267 /* Socket shut down? */
6268 if (sk->sk_shutdown & RCV_SHUTDOWN)
6271 /* Sequenced packets can come disconnected. If so we report the
6276 /* Is there a good reason to think that we may receive some data? */
6277 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6280 /* Handle signals. */
6281 if (signal_pending(current))
6284 /* Let another process have a go. Since we are going to sleep
6285 * anyway. Note: This may cause odd behaviors if the message
6286 * does not fit in the user's buffer, but this seems to be the
6287 * only way to honor MSG_DONTWAIT realistically.
6289 sctp_release_sock(sk);
6290 *timeo_p = schedule_timeout(*timeo_p);
6294 finish_wait(sk->sk_sleep, &wait);
6298 error = sock_intr_errno(*timeo_p);
6301 finish_wait(sk->sk_sleep, &wait);
6306 /* Receive a datagram.
6307 * Note: This is pretty much the same routine as in core/datagram.c
6308 * with a few changes to make lksctp work.
6310 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6311 int noblock, int *err)
6314 struct sk_buff *skb;
6317 timeo = sock_rcvtimeo(sk, noblock);
6319 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6320 timeo, MAX_SCHEDULE_TIMEOUT);
6323 /* Again only user level code calls this function,
6324 * so nothing interrupt level
6325 * will suddenly eat the receive_queue.
6327 * Look at current nfs client by the way...
6328 * However, this function was corrent in any case. 8)
6330 if (flags & MSG_PEEK) {
6331 spin_lock_bh(&sk->sk_receive_queue.lock);
6332 skb = skb_peek(&sk->sk_receive_queue);
6334 atomic_inc(&skb->users);
6335 spin_unlock_bh(&sk->sk_receive_queue.lock);
6337 skb = skb_dequeue(&sk->sk_receive_queue);
6343 /* Caller is allowed not to check sk->sk_err before calling. */
6344 error = sock_error(sk);
6348 if (sk->sk_shutdown & RCV_SHUTDOWN)
6351 /* User doesn't want to wait. */
6355 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6364 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6365 static void __sctp_write_space(struct sctp_association *asoc)
6367 struct sock *sk = asoc->base.sk;
6368 struct socket *sock = sk->sk_socket;
6370 if ((sctp_wspace(asoc) > 0) && sock) {
6371 if (waitqueue_active(&asoc->wait))
6372 wake_up_interruptible(&asoc->wait);
6374 if (sctp_writeable(sk)) {
6375 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6376 wake_up_interruptible(sk->sk_sleep);
6378 /* Note that we try to include the Async I/O support
6379 * here by modeling from the current TCP/UDP code.
6380 * We have not tested with it yet.
6382 if (sock->fasync_list &&
6383 !(sk->sk_shutdown & SEND_SHUTDOWN))
6384 sock_wake_async(sock,
6385 SOCK_WAKE_SPACE, POLL_OUT);
6390 /* Do accounting for the sndbuf space.
6391 * Decrement the used sndbuf space of the corresponding association by the
6392 * data size which was just transmitted(freed).
6394 static void sctp_wfree(struct sk_buff *skb)
6396 struct sctp_association *asoc;
6397 struct sctp_chunk *chunk;
6400 /* Get the saved chunk pointer. */
6401 chunk = *((struct sctp_chunk **)(skb->cb));
6404 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6405 sizeof(struct sk_buff) +
6406 sizeof(struct sctp_chunk);
6408 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6411 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6413 sk->sk_wmem_queued -= skb->truesize;
6414 sk_mem_uncharge(sk, skb->truesize);
6417 __sctp_write_space(asoc);
6419 sctp_association_put(asoc);
6422 /* Do accounting for the receive space on the socket.
6423 * Accounting for the association is done in ulpevent.c
6424 * We set this as a destructor for the cloned data skbs so that
6425 * accounting is done at the correct time.
6427 void sctp_sock_rfree(struct sk_buff *skb)
6429 struct sock *sk = skb->sk;
6430 struct sctp_ulpevent *event = sctp_skb2event(skb);
6432 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6435 * Mimic the behavior of sock_rfree
6437 sk_mem_uncharge(sk, event->rmem_len);
6441 /* Helper function to wait for space in the sndbuf. */
6442 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6445 struct sock *sk = asoc->base.sk;
6447 long current_timeo = *timeo_p;
6450 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6451 asoc, (long)(*timeo_p), msg_len);
6453 /* Increment the association's refcnt. */
6454 sctp_association_hold(asoc);
6456 /* Wait on the association specific sndbuf space. */
6458 prepare_to_wait_exclusive(&asoc->wait, &wait,
6459 TASK_INTERRUPTIBLE);
6462 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6465 if (signal_pending(current))
6466 goto do_interrupted;
6467 if (msg_len <= sctp_wspace(asoc))
6470 /* Let another process have a go. Since we are going
6473 sctp_release_sock(sk);
6474 current_timeo = schedule_timeout(current_timeo);
6475 BUG_ON(sk != asoc->base.sk);
6478 *timeo_p = current_timeo;
6482 finish_wait(&asoc->wait, &wait);
6484 /* Release the association's refcnt. */
6485 sctp_association_put(asoc);
6494 err = sock_intr_errno(*timeo_p);
6502 /* If socket sndbuf has changed, wake up all per association waiters. */
6503 void sctp_write_space(struct sock *sk)
6505 struct sctp_association *asoc;
6507 /* Wake up the tasks in each wait queue. */
6508 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6509 __sctp_write_space(asoc);
6513 /* Is there any sndbuf space available on the socket?
6515 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6516 * associations on the same socket. For a UDP-style socket with
6517 * multiple associations, it is possible for it to be "unwriteable"
6518 * prematurely. I assume that this is acceptable because
6519 * a premature "unwriteable" is better than an accidental "writeable" which
6520 * would cause an unwanted block under certain circumstances. For the 1-1
6521 * UDP-style sockets or TCP-style sockets, this code should work.
6524 static int sctp_writeable(struct sock *sk)
6528 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6534 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6535 * returns immediately with EINPROGRESS.
6537 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6539 struct sock *sk = asoc->base.sk;
6541 long current_timeo = *timeo_p;
6544 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6547 /* Increment the association's refcnt. */
6548 sctp_association_hold(asoc);
6551 prepare_to_wait_exclusive(&asoc->wait, &wait,
6552 TASK_INTERRUPTIBLE);
6555 if (sk->sk_shutdown & RCV_SHUTDOWN)
6557 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6560 if (signal_pending(current))
6561 goto do_interrupted;
6563 if (sctp_state(asoc, ESTABLISHED))
6566 /* Let another process have a go. Since we are going
6569 sctp_release_sock(sk);
6570 current_timeo = schedule_timeout(current_timeo);
6573 *timeo_p = current_timeo;
6577 finish_wait(&asoc->wait, &wait);
6579 /* Release the association's refcnt. */
6580 sctp_association_put(asoc);
6585 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6588 err = -ECONNREFUSED;
6592 err = sock_intr_errno(*timeo_p);
6600 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6602 struct sctp_endpoint *ep;
6606 ep = sctp_sk(sk)->ep;
6610 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6611 TASK_INTERRUPTIBLE);
6613 if (list_empty(&ep->asocs)) {
6614 sctp_release_sock(sk);
6615 timeo = schedule_timeout(timeo);
6620 if (!sctp_sstate(sk, LISTENING))
6624 if (!list_empty(&ep->asocs))
6627 err = sock_intr_errno(timeo);
6628 if (signal_pending(current))
6636 finish_wait(sk->sk_sleep, &wait);
6641 static void sctp_wait_for_close(struct sock *sk, long timeout)
6646 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6647 if (list_empty(&sctp_sk(sk)->ep->asocs))
6649 sctp_release_sock(sk);
6650 timeout = schedule_timeout(timeout);
6652 } while (!signal_pending(current) && timeout);
6654 finish_wait(sk->sk_sleep, &wait);
6657 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6659 struct sk_buff *frag;
6664 /* Don't forget the fragments. */
6665 skb_walk_frags(skb, frag)
6666 sctp_skb_set_owner_r_frag(frag, sk);
6669 sctp_skb_set_owner_r(skb, sk);
6672 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6673 struct sctp_association *asoc)
6675 struct inet_sock *inet = inet_sk(sk);
6676 struct inet_sock *newinet = inet_sk(newsk);
6678 newsk->sk_type = sk->sk_type;
6679 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6680 newsk->sk_flags = sk->sk_flags;
6681 newsk->sk_no_check = sk->sk_no_check;
6682 newsk->sk_reuse = sk->sk_reuse;
6684 newsk->sk_shutdown = sk->sk_shutdown;
6685 newsk->sk_destruct = inet_sock_destruct;
6686 newsk->sk_family = sk->sk_family;
6687 newsk->sk_protocol = IPPROTO_SCTP;
6688 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6689 newsk->sk_sndbuf = sk->sk_sndbuf;
6690 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6691 newsk->sk_lingertime = sk->sk_lingertime;
6692 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6693 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6695 newinet = inet_sk(newsk);
6697 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6698 * getsockname() and getpeername()
6700 newinet->sport = inet->sport;
6701 newinet->saddr = inet->saddr;
6702 newinet->rcv_saddr = inet->rcv_saddr;
6703 newinet->dport = htons(asoc->peer.port);
6704 newinet->pmtudisc = inet->pmtudisc;
6705 newinet->id = asoc->next_tsn ^ jiffies;
6707 newinet->uc_ttl = inet->uc_ttl;
6708 newinet->mc_loop = 1;
6709 newinet->mc_ttl = 1;
6710 newinet->mc_index = 0;
6711 newinet->mc_list = NULL;
6714 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6715 * and its messages to the newsk.
6717 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6718 struct sctp_association *assoc,
6719 sctp_socket_type_t type)
6721 struct sctp_sock *oldsp = sctp_sk(oldsk);
6722 struct sctp_sock *newsp = sctp_sk(newsk);
6723 struct sctp_bind_bucket *pp; /* hash list port iterator */
6724 struct sctp_endpoint *newep = newsp->ep;
6725 struct sk_buff *skb, *tmp;
6726 struct sctp_ulpevent *event;
6727 struct sctp_bind_hashbucket *head;
6729 /* Migrate socket buffer sizes and all the socket level options to the
6732 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6733 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6734 /* Brute force copy old sctp opt. */
6735 inet_sk_copy_descendant(newsk, oldsk);
6737 /* Restore the ep value that was overwritten with the above structure
6743 /* Hook this new socket in to the bind_hash list. */
6744 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6745 sctp_local_bh_disable();
6746 sctp_spin_lock(&head->lock);
6747 pp = sctp_sk(oldsk)->bind_hash;
6748 sk_add_bind_node(newsk, &pp->owner);
6749 sctp_sk(newsk)->bind_hash = pp;
6750 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6751 sctp_spin_unlock(&head->lock);
6752 sctp_local_bh_enable();
6754 /* Copy the bind_addr list from the original endpoint to the new
6755 * endpoint so that we can handle restarts properly
6757 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6758 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6760 /* Move any messages in the old socket's receive queue that are for the
6761 * peeled off association to the new socket's receive queue.
6763 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6764 event = sctp_skb2event(skb);
6765 if (event->asoc == assoc) {
6766 __skb_unlink(skb, &oldsk->sk_receive_queue);
6767 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6768 sctp_skb_set_owner_r_frag(skb, newsk);
6772 /* Clean up any messages pending delivery due to partial
6773 * delivery. Three cases:
6774 * 1) No partial deliver; no work.
6775 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6776 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6778 skb_queue_head_init(&newsp->pd_lobby);
6779 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6781 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6782 struct sk_buff_head *queue;
6784 /* Decide which queue to move pd_lobby skbs to. */
6785 if (assoc->ulpq.pd_mode) {
6786 queue = &newsp->pd_lobby;
6788 queue = &newsk->sk_receive_queue;
6790 /* Walk through the pd_lobby, looking for skbs that
6791 * need moved to the new socket.
6793 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6794 event = sctp_skb2event(skb);
6795 if (event->asoc == assoc) {
6796 __skb_unlink(skb, &oldsp->pd_lobby);
6797 __skb_queue_tail(queue, skb);
6798 sctp_skb_set_owner_r_frag(skb, newsk);
6802 /* Clear up any skbs waiting for the partial
6803 * delivery to finish.
6805 if (assoc->ulpq.pd_mode)
6806 sctp_clear_pd(oldsk, NULL);
6810 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6811 sctp_skb_set_owner_r_frag(skb, newsk);
6813 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6814 sctp_skb_set_owner_r_frag(skb, newsk);
6816 /* Set the type of socket to indicate that it is peeled off from the
6817 * original UDP-style socket or created with the accept() call on a
6818 * TCP-style socket..
6822 /* Mark the new socket "in-use" by the user so that any packets
6823 * that may arrive on the association after we've moved it are
6824 * queued to the backlog. This prevents a potential race between
6825 * backlog processing on the old socket and new-packet processing
6826 * on the new socket.
6828 * The caller has just allocated newsk so we can guarantee that other
6829 * paths won't try to lock it and then oldsk.
6831 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6832 sctp_assoc_migrate(assoc, newsk);
6834 /* If the association on the newsk is already closed before accept()
6835 * is called, set RCV_SHUTDOWN flag.
6837 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6838 newsk->sk_shutdown |= RCV_SHUTDOWN;
6840 newsk->sk_state = SCTP_SS_ESTABLISHED;
6841 sctp_release_sock(newsk);
6845 /* This proto struct describes the ULP interface for SCTP. */
6846 struct proto sctp_prot = {
6848 .owner = THIS_MODULE,
6849 .close = sctp_close,
6850 .connect = sctp_connect,
6851 .disconnect = sctp_disconnect,
6852 .accept = sctp_accept,
6853 .ioctl = sctp_ioctl,
6854 .init = sctp_init_sock,
6855 .destroy = sctp_destroy_sock,
6856 .shutdown = sctp_shutdown,
6857 .setsockopt = sctp_setsockopt,
6858 .getsockopt = sctp_getsockopt,
6859 .sendmsg = sctp_sendmsg,
6860 .recvmsg = sctp_recvmsg,
6862 .backlog_rcv = sctp_backlog_rcv,
6864 .unhash = sctp_unhash,
6865 .get_port = sctp_get_port,
6866 .obj_size = sizeof(struct sctp_sock),
6867 .sysctl_mem = sysctl_sctp_mem,
6868 .sysctl_rmem = sysctl_sctp_rmem,
6869 .sysctl_wmem = sysctl_sctp_wmem,
6870 .memory_pressure = &sctp_memory_pressure,
6871 .enter_memory_pressure = sctp_enter_memory_pressure,
6872 .memory_allocated = &sctp_memory_allocated,
6873 .sockets_allocated = &sctp_sockets_allocated,
6876 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6878 struct proto sctpv6_prot = {
6880 .owner = THIS_MODULE,
6881 .close = sctp_close,
6882 .connect = sctp_connect,
6883 .disconnect = sctp_disconnect,
6884 .accept = sctp_accept,
6885 .ioctl = sctp_ioctl,
6886 .init = sctp_init_sock,
6887 .destroy = sctp_destroy_sock,
6888 .shutdown = sctp_shutdown,
6889 .setsockopt = sctp_setsockopt,
6890 .getsockopt = sctp_getsockopt,
6891 .sendmsg = sctp_sendmsg,
6892 .recvmsg = sctp_recvmsg,
6894 .backlog_rcv = sctp_backlog_rcv,
6896 .unhash = sctp_unhash,
6897 .get_port = sctp_get_port,
6898 .obj_size = sizeof(struct sctp6_sock),
6899 .sysctl_mem = sysctl_sctp_mem,
6900 .sysctl_rmem = sysctl_sctp_rmem,
6901 .sysctl_wmem = sysctl_sctp_wmem,
6902 .memory_pressure = &sctp_memory_pressure,
6903 .enter_memory_pressure = sctp_enter_memory_pressure,
6904 .memory_allocated = &sctp_memory_allocated,
6905 .sockets_allocated = &sctp_sockets_allocated,
6907 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */